1 the PREVENTION of ORAL SIDE EFFECTS of CANCER TREATMENT VOLUME I of II a Thesis Submitted to the University of Manchester

THE PREVENTION OF ORAL SIDE EFFECTS OF CANCER TREATMENT

VOLUME I of II

A thesis submitted to The University of Manchester for the degree of Doctor of Philosophy in the Faculty of Biology, Medicine and Health

2017

PHILIP RILEY

SCHOOL OF MEDICAL SCIENCES

Division of Dentistry

1 CONTENTS

VOLUME I OF II:

ABSTRACT ...... 6

DECLARATION ...... 7

ACKNOWLEDGEMENTS ...... 8

THE AUTHOR ...... 8

1 INTRODUCTION ...... 9 1.1 Overview of side effects caused by cancer treatment ...... 9 1.2 Oral mucositis ...... 9 1.3 Salivary gland dysfunction ...... 19 1.4 Importance of systematic reviews ...... 21 1.5 Importance of clinical guidelines ...... 24 1.6 What are the gaps? ...... 27 2 AIMS ...... 29

3 INTERVENTIONS FOR PREVENTING ORAL MUCOSITIS IN PATIENTS WITH CANCER RECEIVING TREATMENT: ORAL CRYOTHERAPY ...... 30 3.1 Abstract ...... 30 3.2 Background ...... 33 3.3 Objectives ...... 36 3.4 Methods ...... 36 3.5 Results ...... 44 3.5.1 Description of studies ...... 44 3.5.2 Risk of bias in included studies ...... 53 3.5.3 Effects of interventions ...... 57 3.6 Summary of findings tables ...... 79 1 Cryotherapy versus control for preventing oral mucositis in adults receiving fluorouracil-based treatment for solid cancers ...... 79 2 Cryotherapy versus control for preventing oral mucositis in adults receiving high- dose melphalan-based treatment prior to haematopoietic stem cell transplantation for haematological cancers ...... 81 3.7 Discussion ...... 84 3.8 Authors' conclusions ...... 89 4 INTERVENTIONS FOR PREVENTING ORAL MUCOSITIS IN PATIENTS WITH CANCER RECEIVING TREATMENT: CYTOKINES AND GROWTH FACTORS ...... 91 2 4.1 Abstract ...... 91 4.2 Background ...... 93 4.3 Objectives ...... 97 4.4 Methods ...... 97 4.5 Results ...... 105 4.5.1 Description of studies ...... 105 Excluded studies...... 118 4.5.2 Risk of bias in included studies ...... 119 4.5.3 Effects of interventions ...... 123 4.6 Summary of findings tables ...... 157 1 KGF compared to placebo for preventing oral mucositis in adults with cancer receiving treatment ...... 157 2 GM-CSF compared to placebo/no treatment for preventing oral mucositis in adults with cancer receiving treatment ...... 160 3 G-CSF compared to placebo/no treatment for preventing oral mucositis in adults with cancer receiving treatment ...... 163 4.7 Discussion ...... 165 4.8 Authors' conclusions ...... 171 5 PHARMACOLOGICAL INTERVENTIONS FOR PREVENTING DRY MOUTH AND SALIVARY GLAND DYSFUNCTION FOLLOWING RADIOTHERAPY ...... 173 5.1 Abstract ...... 173 5.2 Background ...... 175 5.3 Objectives ...... 177 5.4 Methods ...... 177 5.5 Results ...... 184 5.5.1 Description of studies ...... 184 5.5.2 Risk of bias in included studies ...... 189 5.5.3 Effects of interventions ...... 193 5.6 Summary of findings tables ...... 245 1 Pilocarpine compared to no treatment/placebo for preventing salivary gland dysfunction following radiotherapy ...... 245 2 Amifostine compared to no treatment/placebo for preventing salivary gland dysfunction following radiotherapy ...... 248 3 Palifermin compared to placebo for preventing salivary gland dysfunction following radiotherapy...... 251 5.7 Discussion ...... 252 5.8 Authors' conclusions ...... 256 6 THE QUALITY OF CLINICAL GUIDELINES ON MOUTH CARE FOR CANCER PATIENTS ...... 258 6.1 Abstract ...... 258 3 6.2 Background ...... 258 6.3 Objective ...... 259 6.4 Methods ...... 259 6.5 Results ...... 265 6.6 Discussion ...... 276 6.7 Conclusion ...... 278 7 DISCUSSION ...... 279

8 REFERENCES ...... 288

VOLUME II OF II

APPENDIX 1: Oral cryotherapy review - electronic search strategies ...... 306

APPENDIX 2: Oral cryotherapy review - characteristics of studies and risk of bias tables ...... 310 Characteristics of included studies ...... 310 Characteristics of excluded studies...... 333 Characteristics of studies awaiting classification ...... 335 Characteristics of ongoing studies ...... 338 APPENDIX 3: Oral mucositis measurement scales ...... 340

APPENDIX 4: Oral cryotherapy review - study data not included in analyses 342

APPENDIX 5: Oral cryotherapy review - secondary outcome analyses ...... 344

APPENDIX 6: Cytokines and growth factors review - electronic search strategies ...... 347

APPENDIX 7: Cytokines and growth factors review - characteristics of studies and risk of bias tables ...... 357 Characteristics of included studies ...... 357 Characteristics of excluded studies...... 421 Characteristics of studies awaiting classification ...... 424 APPENDIX 8: Cytokines and growth factors review - secondary outcome analyses ...... 430

APPENDIX 9: Cytokines and growth factors review - adverse events tables ... 448

APPENDIX 10: Salivary gland dysfunction review - electronic search strategies453

APPENDIX 11: Salivary gland dysfunction review - characteristics of studies and risk of bias tables ...... 457 Characteristics of included studies ...... 457 Characteristics of excluded studies...... 511

4 Characteristics of studies awaiting classification ...... 515 Characteristics of ongoing studies ...... 515 APPENDIX 12: Salivary gland dysfunction review - study data not included in analyses ...... 516

APPENDIX 13: Salivary gland dysfunction review - secondary outcome analyses537

APPENDIX 14: Results of individual assessors’ appraisal of included guidelines548

APPENDIX 15: Protocol for future research on the development of a core outcome set for oral mucositis prevention in adult cancer patients ...... 552

LIST OF FIGURES Figure 1: The hierarchy of evidence…………………………………………………….22 Figure 2: Total number of prescriptions for antibiotic prophylaxis………………….26 Figure 3: Oral cryotherapy review: PRISMA flow diagram………………………….45 Figure 4: Oral cryotherapy review: Risk of bias summary…………………………...56 Figure 5: Cytokines and growth factors review: PRISMA flow diagram…………..106 Figure 6: Cytokines and growth factors review: Risk of bias summary………….122-3 Figure 7: Salivary gland dysfunction review: PRISMA flow diagram……………...185 Figure 8: Salivary gland dysfunction review: Risk of bias summary……………..192-3

LIST OF TABLES Table 1: Quality scores for each included guideline by AGREE II domain………266-7 Table 2: Summary of AGREE II domain scores……………………………………...268 Table 3: Appraisal of guideline recommendations………………………………..270-5

WORD COUNT: 65,845

5 ABSTRACT

Background: Treatments for cancer can cause a range of oral side effects which can compromise patients’ quality of life and even their chances of survival. The evidence for the prevention of some of these side-effects, primarily oral mucositis and salivary gland dysfunction, is often scarce or conflicting. Such evidence is needed to support the recommendations given in clinical guidelines. Aims: (i) To assess the effects of interventions for preventing oral mucositis in patients with cancer who are receiving treatment; (ii) To assess the effects of pharmacological interventions for the prevention of radiation-induced salivary gland dysfunction; (iii) To assess the quality of guidelines on the management of oral complications caused by cancer treatment. Methods: (i & ii) Systematic reviews of randomised controlled trials, using Cochrane methodology were carried out to assess the effects of interventions (oral cryotherapy and cytokines and growth factors for (i)); (iii) Guidelines published from 2005 onwards were appraised using the AGREE II instrument. The guidelines were also assessed to determine whether or not their recommendations were supported by high quality systematic reviews. Results: (i) Both oral cryotherapy and keratinocyte growth factor were found to prevent oral mucositis in patients receiving certain types of treatment. They also appeared to be relatively safe interventions; (ii) Amifostine was found to be beneficial for preventing dry mouth in the short- to medium-term in patients receiving radiotherapy to the head and neck. However, amifostine is associated with side effects; (iii) The quality of guidelines on the management of oral complications caused by cancer treatment is very poor. They are generally poorly reported, methodologically deficient, and the recommendations are rarely supported by systematic review evidence. Conclusions: More high quality randomised controlled trials are needed of interventions for preventing oral mucositis and salivary gland dysfunction. Clinical guidelines in this field should not be used without a rigorous appraisal incorporating an assessment of the evidence base for their recommendations. Different stakeholders, especially systematic reviewers and guideline developers, need to work together to improve the currently low standard of clinical guidelines.

6 DECLARATION

No portion of the work referred to in the thesis has been submitted in support of an application for another degree or qualification of this or any other university or other institute of learning.

i. The author of this thesis (including any appendices and/or schedules to this thesis) owns certain copyright or related rights in it (the “Copyright”) and s/he has given The University of Manchester certain rights to use such Copyright, including for administrative purposes. ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as amended) and regulations issued under it or, where appropriate, in accordance with licensing agreements which the University has from time to time. This page must form part of any such copies made. iii. The ownership of certain Copyright, patents, designs, trademarks and other intellectual property (the “Intellectual Property”) and any reproductions of copyright works in the thesis, for example graphs and tables (“Reproductions”), which may be described in this thesis, may not be owned by the author and may be owned by third parties. Such Intellectual Property and Reproductions cannot and must not be made available for use without the prior written permission of the owner(s) of the relevant Intellectual Property and/or Reproductions. iv. Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property and/or Reproductions described in it may take place is available in the University IP Policy (see http://documents.manchester.ac.uk/DocuInfo.aspx?DocID=24420), in any relevant Thesis restriction declarations deposited in the University Library, The University Library’s regulations (see http://www.library.manchester.ac.uk/about/regulations/) and in The University’s policy on Presentation of Theses.

7 ACKNOWLEDGEMENTS

There is a lot to be said for being in the right place, at the right time and with the right people. I honestly do not believe I would have been in the position I am in today, having been afforded so many opportunities, including doing a PhD, if it was not for the belief, support and kindness shown to me by Professors Helen Worthington, Anne-Marie Glenny, Jan Clarkson and Tanya Walsh, and also by Luisa Fernandez and Anne Littlewood. A special mention goes to Richard Macey for the encouragement, laughs and flapjacks.

Thanks to my mum and dad for all your love and support, for giving me such a happy upbringing, and for making me believe I could do anything.

All my work is dedicated to my girls: my wife Laura, my daughters Beatrice and Olive, and little she-bump who is yet to be named. You are everything to me.

THE AUTHOR

I am a Research Fellow in the Division of Dentistry, School of Medical Sciences, The University of Manchester. I have worked at Cochrane Oral Health’s editorial base since 2006 when I started in a junior editorial role, progressing to Deputy Managing Editor in 2010. At this point, after becoming familiar with research methodology, I was encouraged by Professors Helen Worthington and Anne-Marie Glenny to study part time for a Masters degree in Public Health at The University of Manchester, which I completed in 2013 gaining a distinction. I was then made a Systematic Reviewer and Editor for Cochrane Oral Health and began studying for a PhD. I have published systematic reviews regularly since 2012 and assisted many others with their systematic reviews. Over the last few years, I have been lucky enough to become involved with the teaching of systematic review methodology and wider research methods internationally. I hope to continue learning throughout my career, to produce high quality research that is useful to clinicians and ultimately benefits patients and improves their lives, and to be able to continue collaborating with such amazing people as those who I have been so fortunate to work with thus far on my journey.

8 1 INTRODUCTION

1.1 Overview of side effects caused by cancer treatment

Treatments for cancer, such as chemotherapy, radiotherapy of the head and neck, and targeted therapy, commonly cause toxic oral side effects (1-3). One of the most widely researched of these side effects is oral mucositis (1). Other major oral complications arising from cancer treatment include salivary gland dysfunction, oral candidiasis, herpes simplex virus and dental caries. This introduction concentrates on two of these side effects, oral mucositis and salivary gland dysfunction, as the thesis is mainly concerned with these conditions.

1.2 Oral mucositis

Introduction Mucositis can occur anywhere along the gastrointestinal tract, affecting the mucous membranes (moist tissue lining the tract) and causing lesions to develop (2). Oral mucositis causes pain and difficulties with eating, swallowing, and even talking (2). These morbidities can even further reduce a cancer patient’s quality of life and lead to the requirement of pain relief medication, nutritional support (e.g. a feeding tube), and specialist oral hygiene care (4-6). Furthermore, in immunosuppressed patients (those with myelosuppression or neutropenia), there is an increased risk of bacteraemia, fungemia, and sepsis, resulting from secondary infection of the mucositis lesions, which can necessitate antibiotics and increased hospitalisation, and even cause death (2, 6, 7). Therefore, oral mucositis can be a dose-limiting condition, disrupting a patient’s optimal cancer treatment plan and consequentially decreasing the chances of survival (3, 6, 7).

Incidence The literature reports a range of incidence and prevalence figures for oral mucositis. However, this is understandable when considering that the risk of developing mucositis varies with the mode of treatment or combinations of treatments received, which will be determined by the type and severity of cancer. Sonis states that around 8% of new cancer patients have a greater than 50% chance of developing ulcerative mucositis, whilst 43% are at low risk having undergone surgery or less intense radiotherapy and/or chemotherapy regimens, with the remaining majority being at 20% to 49% risk (8). Scully et al report that mucositis occurs in at least 75% of high risk patients (e.g. those receiving radiotherapy for 9 head and neck cancers or any high dose chemotherapy), but may be under-reported in lower risk groups due to them being more likely to be outpatients with less observation, or due to less reporting of moderate mucositis, or alternatively due to hesitation on the part of patients or clinicians to disrupt optimal cancer treatment (2). Further features of mucositis risk associated with therapy are, firstly, that the risk increases when chemotherapy is combined with either radiotherapy or other chemotherapy, and secondly that risk greatly increases as patients receive further cycles of treatment (e.g. from 20% to over 60% in breast cancer patients receiving standard therapy) (2, 8).

Clinical course The first symptom of oral mucositis typically occurs 3 to 5 days after chemotherapy or at 10 Gy of cumulative doses of head and neck radiation, and presents as erythema (reddening of the mucosa) (2, 8). Ulceration typically begins at 7 to 10 days after chemotherapy or at 30 GY of cumulative radiotherapy, and is associated with the start of severe pain, necessitating strong pain-killing interventions such as opioids, and disrupting the patient’s diet (2). Chemotherapy-induced oral mucositis is regarded as an acute condition, generally resolving within 14 days of treatment according to some reports (8), and 21 days in others (2). As standard radiation therapy lasts for 7 weeks, radiotherapy- induced oral mucositis is chronic in nature, and may only resolve 3 to 4 weeks after treatment has ended (8).

Risk factors Risk factors for oral mucositis can be categorised as being either patient-associated or treatment-associated (8). The latter category includes factors such as treatment type, dose, schedule, route of administration, formulation, and concomitant treatment (1). The underlying mechanisms explaining the differing incidence and severity relating to some of these risk factors are not well described in the literature (9). However, it is not difficult to comprehend that a patient with oral cancer receiving chemoradiotherapy would be at greater risk of developing oral mucositis than a patient with cancer of the hypopharynx, as the oral cancer patient would incur direct damage to the tissues of the oral cavity (8). Patient-associated variables are also not fully explained, but they include age, gender, body mass/weight, nutritional status, oral microbiota, oral health and hygiene, salivary potential, neutrophil counts, pre-existing conditions (e.g. psoriasis and Addison’s disease), renal and hepatic function, genetics, and the ability to metabolise and eliminate drugs (1, 8, 9). It has been postulated that the lack of a clear understanding of oral mucositis risk factors, along with a lack of stratification by these risk factors in clinical studies, has to a large extent

10 contributed to an inconsistent and sometimes contradictory body of evidence on effective treatments (9).

Economic burden As oral mucositis requires preventative therapy, to pre-emptively stop or reduce its consequences, or to treat it once it develops, and also because of the above discussed problems and high incidence and prevalence rates, oral mucositis is recognised as a considerable burden on healthcare resources (1). For example, in a retrospective cohort study of patients with head and neck primary cancers conducted in the USA, cancer patients with oral mucositis were hospitalised more (23% versus 6%) and visited the emergency department more (40% versus 6%) than cancer patients without oral mucositis (10). This translated into an increase in incremental costs associated with oral mucositis ranging from $1700 to $6000 per patient, depending on the severity of the condition (10). Such additional costs clearly demonstrate the economic burden caused by oral mucositis.

Pathogenesis A sound understanding of the pathogenesis of mucositis is essential in order to find and develop treatments that will disrupt the underlying biological processes; an example of this being the US Food and Drug Administration’s (FDA) approval of palifermin in 2004 (8).

Until recently it was thought that there was a fairly simple and linear process leading to mucositis, whereby chemotherapy or radiotherapy would damage the basal epithelial cell layer, resulting in the reduced regenerative ability of rapidly dividing epithelial stem cells, and subsequently tissue wasting (atrophy), and finally ulceration (3, 8). It is now thought that mucositis is not only mediated by the epithelium (the external layers of cells lining the mucosal glands), rather it is likely to be a much more complex and multifactorial process (3). Much of our knowledge of the pathobiology of mucositis is derived from animal models (1). Such studies have suggested that, triggered by pro-inflammatory cytokines (e.g. interleukin-1 beta and tumour necrosis factor alpha - TNF-α), damage to the endothelium (the thin layer of cells that line the inside of blood and lymphatic vessels) and connective tissues beneath the oral mucosa (the submucosa) occurs prior to epithelial degeneration (11). Furthermore, damage to the epithelium may lead to invasion of the submucosa by bacterial cell wall products which results in further stimulation of destructive cytokine release from macrophages (large infection-fighting cells) (11). Cancer treatment damages the bone marrow, reducing the number of neutrophils (a type of white blood cell that responds to injury and infection) (12), and such neutropenic patients are at

11 risk of potentially fatal sepsis if whole bacteria penetrate the submucosa (3). Therefore it might be expected that an intervention aimed at reducing the expression of pro- inflammatory cytokines might impede the development of mucositis.

Whilst it has been reported that the composition of oral bacteria is dramatically altered by cancer treatment, for example one small pilot study found that 61% of the species present after chemotherapy were not present beforehand (13), the fact that oral mucositis occurs before such changes in oral flora indicates that, whilst detrimental to the severity, they do not have a primary causative role (1). This conclusion is backed up by rigorous evidence of a lack of effect for antimicrobials, such as chlorhexidine, in the management of oral mucositis (14).

The pathogenesis involved in producing the mucosal toxicities seen as a result of targeted therapies for cancer treatment is currently under investigation due to limited understanding, although it is thought to be different from chemotherapy and radiotherapy- induced mucositis (1, 7). Ulcers arising after targeted therapy are similar to those resulting from aphthous stomatitis (regular mouth ulcers) and tend to exist in conjunction with a skin rash, and therefore have typically been referred to as stomatitis in order to differentiate from oral mucositis (1).

The following five-phase model of oral mucositis put forward by Sonis is the currently accepted explanation for the sequence of events underlying the condition (3, 8).

1. Initiation: Clonogenic cell death of basal epithelial cells occurs due to damage to DNA caused by chemotherapy and/or radiation. This produces reactive oxygen species (ROS). 2. Primary damage response: Radiation, chemotherapy, ROS and DNA strand breaks all contribute to the activation of transcription factors such as nuclear factor kappa beta (NF-κB) and sphingomyelinases. All of the above leads to the upregulation of pro- inflammatory cytokines (such as TNF-α), nitric oxide (NO), ceramide and matrix metalloproteinases, resulting in tissue injury and cell death, and eventually the destruction of the oral mucosa. 3. Signal amplification: Local tissue response is altered by positive or negative feedback from some of the above molecules, and this results in the amplification and prolonging of tissue injury. For example TNF-α can positively feedback on NF-κB thus inducing more pro-inflammatory cytokine production.

12 4. Ulceration: Bacteria colonise the ulcers and their cell wall products infiltrate the submucosa, activating tissue macrophages which results in further production of pro- inflammatory cytokines, inflammation and pain. This is the stage where, as mentioned above, neutropenic patients are at risk of sepsis or bacteraemia if whole bacteria are able to infiltrate the submucosa. 5. Healing: Signalling from the extracellular matrix of the submucosa results in epithelial proliferation and differentiation. The local oral flora are reinstated.

Interventions and their evidence base The above biological events 1 to 3 will occur as a result of cancer treatment, but if interventions aimed at prevention are successful then ulceration may be avoided or attenuated. This in turn may reduce pain, hospitalisations, cancer treatment interruptions, and other morbidities previously discussed.

There are a multitude of interventions that have been used in clinical studies and in practice, yet despite the immense research resources which have been consumed with the intention of improving outcomes for cancer patients, the treatments that are commonly utilised are of limited effectiveness, and some have already been shown to be ineffective. It has been stated that the management of mucositis has changed little over time (8), and it has been postulated that this may be due to the reliance in some healthcare professionals’ clinical practice of following tradition and adopting subjective approaches (6). There is also evidence that children and adolescents across the UK receive an inappropriately diverse array of interventions (15). As the effects of these interventions are also variable, with some having been shown to lack any effect, this has produced undesirable inequalities in terms of the health care received by cancer patients (16). However, a caveat to such a negative overview, is that recent advances in the understanding of the pathobiology of oral mucositis, as described above, should provide the basis for advances in the clinical management of the condition (7). Furthermore, the identification of the lack of effect of certain interventions, which has occurred through systematically reviewing the evidence over recent years, is of the utmost importance both for patient outcomes, and societally in terms of cost-effectiveness/cost-utility in an era of scarce resources (6).

Prevention The latest version of a Cochrane systematic review of randomised controlled trials of interventions to prevent oral mucositis (14), which was published in 2011, found the strongest body of evidence (i.e. meta-analyses of more than 550 participants) supporting

13 the use of cryotherapy (ice chips) or keratinocyte growth factor. When patients received cryotherapy, their risk of developing oral mucositis of any severity was 26% less than patients receiving no treatment at a median follow-up of 28 days. The risk of developing severe oral mucositis was 64% less. However, the body of evidence that the results were based upon was assessed as being of low quality due to risk of bias in the studies and due to heterogeneity (variability in clinical and methodological aspects of the studies). When patients received keratinocyte growth factor, their risk of developing oral mucositis of any severity was 18% less than patients receiving a placebo at a median follow-up of 28 days. The risk of developing severe oral mucositis was 28% less. Again, the quality of the body of evidence was assessed as being low.

The Cochrane review found weaker and less reliable bodies of evidence (i.e. meta-analyses of between 90 and 350 participants) supporting the use of aloe vera, amifositine, intravenous glutamine, granulocyte colony-stimulating factor (G-CSF or GCSF), honey, laser, polymixin/tobramycin/amphotericin (PTA) lozenges/paste, and sucralfate. It is important that those using such evidence to inform clinical practice have the necessary skills to interpret the evidence. For example, the effect estimate for laser suggests that the risk of severe oral mucositis would be reduced by 80%, and this may be falsely interpreted to be a better prevention option than keratinocyte growth factor, the effect estimate of which suggests only a 28% reduction of severe oral mucositis. However, it would be important to take into account the fact that the evidence for lasers is based upon two small heterogeneous studies, both of which were assessed as being at high risk of bias, and with a combined sample size of 97 participants.

Perhaps one of the most important findings of the review is that there is no evidence of a benefit for chlorhexidine. This could potentially benefit many patients who might otherwise have been treated with this intervention and not gained any preventive effects, and help to improve the cost-effectiveness of cancer care.

Clinical guidelines go further by making recommendations after interpretation, by a range of experts, of the best available evidence and relating it to specific instances. As discussed above, there is empirical evidence of inappropriate diversity in clinical practice, and this can to some extent be alleviated by the adoption of guidelines, which can in turn help to reduce inequalities in health care (16). The Mucositis Study Group (MSG) of the Multinational Association of Supportive Care in Cancer/International Society of Oral Oncology (MASCC/ISOO) is a group which was set up in 1998 for the purpose of

14 producing evidence-based clinical practice guidelines for managing mucositis (both oral and gastrointestinal), which they first published in 2004, with the latest update being published in 2013 (17). A systematic review such as the Cochrane one discussed above may limit the inclusion criteria to include only randomised controlled trials and, for some interventions, there may be insufficient evidence to make recommendations/suggestions, or to perform subgroup analyses in order to see the differences the interventions may have in different patients (i.e. those with different cancers and/or receiving different treatment regimens). The MASCC/ISOO guidelines incorporate evidence from a series of systematic reviews undertaken by their various members, and these reviews are not limited to randomised controlled trials because much of the literature on oral mucositis is in the form of lower levels of evidence (18). This allows the guideline group to utilise their expert knowledge and experience to make suggestions based on supposedly less internally valid evidence, thus including interventions for which there is no current evidence from randomised controlled trials (18). One such example is that of a basic oral care protocol to prevent oral mucositis, of which the Cochrane review concludes that there is no evidence of a benefit based on two randomised controlled trials, but the MASCC/ISOO panel suggests its use based on level III evidence (19, 20). Level III evidence equates to “well designed, quasi-experimental studies, such as nonrandomised, controlled single-group, pretest-posttest comparison, cohort, time or matched case-control series” (17). This demonstrates the importance of including lower levels of evidence in guidelines, as many cancer patients have been shown to benefit from oral care protocols (19), and thus by making this suggestion, many more patients may benefit as a result of the guidelines.

Much like the Cochrane review presents distinct stronger and weaker bodies of evidence for different interventions, the guidelines make a similar distinction. Recommendations are the strongest statements and are only made if the guidelines are based on level I or II evidence (meta-analyses and high powered randomised controlled trials) (17, 18). Suggestions are made based on evidence of levels III, IV and V and must be accompanied by panel consensus (17, 18). The alternative is that no guideline is possible and this occurs when there is insufficient evidence or when there is a lack of panel consensus on the interpretation of the available evidence (17, 18).

There are too many specific recommendations and suggestions both in favour of and against interventions in the current MASCC/ISOO guidelines to discuss here.

Treatment

15 As the incidence figures discussed above demonstrate, many cancer patients do go on to develop oral mucositis, and owing to the potential effects of oral mucositis on patient morbidity and the course of cancer therapy, treating established oral mucositis is vital. However, the body of research is much smaller for treatment than it is for prevention. To demonstrate this, the 2011 update of the Cochrane review on prevention included 131 randomised controlled trials, whereas the 2010 update of the Cochrane review on treatment included only 32 randomised controlled trials (20).

Unfortunately, the latest version of the Cochrane review pre-dates the routine inclusion of ‘summary of findings’ tables using the GRADE approach (GRADE will be discussed in a later section), and therefore there is no formal assessment of the body of evidence for each treatment comparison. However, the review concluded that limited evidence shows that low level laser reduces oral mucositis severity. As can be expected, for treatment the focus shifts to managing pain, and the review also concluded that less morphine is used and the duration of pain is slightly less when patients control their analgesia than when they are on a continuous drip, but that there is no evidence that either method is better for reducing pain. The evidence on the method of morphine delivery was deemed to be unreliable. The review included 27 different interventions but most comparisons were only assessed by a single study.

The MASCC/ISOO guidelines also includes suggestions that: transdermal fentanyl may be effective for reducing oral mucositis-related pain in patients chemotherapy with or without total body irradiation; 2% morphine mouthwash may be effective for pain reduction in head and neck cancer patients receiving chemoradiation; and 0.5% doxepin mouthwash may reduce pain (no specific cancer type or treatment stated). The first two suggestions were based on level III evidence, whilst the suggestion for doxepin was based on level IV evidence (case reports and clinical examples). Furthermore, there is a recommendation, based on level I and II evidence, that sucralfate mouthwash should not be used to treat oral mucositis in patients receiving chemotherapy or radiotherapy for head and neck cancers.

How the interventions might work Basic oral care This includes oral care protocols, dental care before and during cancer treatment, bland rinses containing saline or sodium bicarbonate, and other mouthrinses such as chlorhexidine, mixed medication (e.g. containing topical anaesthetics and mucosal coating agents), and calcium phosphate (19). Most of these interventions are aimed at oral

16 decontamination, with the assumption that this will reduce the severity of oral mucositis, due to the fact that colonisation of ulcerative lesions by bacteria is known to worsen the condition (21). It is also hypothesised that immunosuppressed patients may benefit from the control of opportunistic bacteria which could otherwise potentially lead to sepsis, and possibly death (21). Such interventions do not have a direct effect on the primary biological mechanisms which cause mucosal tissue injury, and rather are for the purpose of damage limitation (19).

Growth factors and cytokines As discussed in the section on pathogenesis, oral epithelial cells divide rapidly, and thus such cells are vulnerable to cancer treatment, which is aimed at destroying rapidly dividing cancer cells. Growth factors (e.g. keratinocyte growth factor, colony-stimulating factors, epidermal growth factor, transforming growth factor-beta, and whey-derived growth factor) are proteins which bind to receptors of target cells and increase proliferation of the epithelium, with the intention of increasing its thickness and integrity (22). Anti- inflammatory cytokines (e.g. interleukin-11) are also proteins or glycoproteins which bind to receptors of target cells, and they are thought to reduce expression of pro-inflammatory cytokines that are so important in oral mucositis pathogenesis (22). Currently, growth factors are only recommended for patients with haematological cancers undergoing high- dose chemotherapy and total body irradiation because, in theory, they may also encourage the proliferation of cancer cells in solid tumours (21).

Anti-inflammatory agents As discussed in the section on pathogenesis, the inflammatory response after a patient receives cancer treatment causes the release of pro-inflammatory cytokines which cause tissue injury. Bacteria then colonise the mucositis ulcers and this leads to further production of pro-inflammatory cytokines. Interventions in this category include benzydamine mouthwash, diphenhydramine mouthwash, prostaglandin E2 lozenges, misoprostol, immunoglobulin, corticosteroids, indomethacin PO tablets, azelastine PO tablets, mesalazine topical gel, disprin PO tablets, orgotein intramuscular injections, flurbiprofen tooth patches, histamine topical gel, colchicine mouthwash, and placentrex (human placental extract) intramuscular injections. The rationale behind the use of anti- inflammatory agents is to attenuate pro-inflammatory cytokine production (23).

Antimicrobials, coating agents, anaesthetics and analgesics

17 Antimicrobials (e.g. acyclovir, clarithromycin, nystatin, triclosan, kefir, iseganan, povidone-iodine, combination antimicrobial lozenges/pastes, and fluconazole) are intended to decrease the bacterial load in the oral cavity and thus reduce the colonisation of mucosal lesions, which could otherwise lead to systemic infection or/and more severe mucositis (24). It is hypothesised that anaesthetics (e.g. tetracaine, amethocaine, dyclonine, MGI-209 with benzocaine, and cocaine) and analgesics (e.g. capsaicin, methadone, ketamine, fentanyl, topical morphine, nortriptyline, gabapentin, and doxepin) would reduce pain, whilst coating agents (e.g. sucralfate and Gelclair, a polyvinylpyrrolidone-sodium hyaluronate gel) would provide a barrier over the ulcers which would protect exposed nerve endings, thus reducing pain, and also helping the healing process (24). Therefore, intuitively, these agents would be more suited to the treatment of oral mucositis.

Low-level laser therapy Although not fully understood, it has been postulated that this intervention might affect the pathogenesis of oral mucositis by attenuating the production of reactive oxygen species and/or pro-inflammatory cytokines (21, 25).

Cryotherapy The mechanism by which this intervention is thought to work is vasoconstriction, which reduces the blood flow to the oral tissues, and therefore reduces the amount of chemotherapy drugs which reach the area, which in turn should avoid the sequence of biological events leading to the development of oral mucositis (26). Therefore cryotherapy is not relevant in the prevention of radiotherapy-induced oral mucositis (21).

Amifostine As discussed in the section on pathogenesis, radiotherapy and chemotherapy cause direct damage to the basal epithelial cells, leading to the generation of reactive oxygen species, which in turn lead to the upregulation of inflammatory pathways. Amifositine is an organic thiophosphate which is hypothesised to scavenge reactive oxygen species, thus in theory preventing the progression of the inflammatory response (27).

Natural agents There are too many natural agents to list but the following are perhaps the main (best known) ones.  Glutamine is an amino acid which is thought to reduce the damage produced by reactive oxygen species (28).

18  Vitamin A is postulated to encourage cell replication in the mucosa, whilst vitamin E may alter the inflammatory process and offer protection of the mucosa as a result of its properties of membrane stabilisation and being an antioxidant (28).  Honey is thought to have anti-inflammatory effects (28).  Zinc is an essential element required for a range of functions such as growth, healing and the health of the immune system (28).  Aloe vera is thought to help the healing of wounds (28).

Miscellaneous agents The main agents in this category are as follows.  Allopurinol may attenuate the effects and toxicity of certain chemotherapy drugs, such as 5-FU and methotrexate, in normal tissue (29).  Interventions to increase saliva secretion (pilocarpine, chewing gum, and bethanechol) could potentially help by lubricating and cleaning the mucosa of food debris and bacteria, whilst propantheline decreases saliva secretion with the aim of reducing the amount of chemotherapy drugs delivered in saliva to directly to the mucosa, thus reducing toxicity (29).  Pentoxifylline downregulates the production of TNF-α and thus may attenuate the inflammatory response (29).

1.3 Salivary gland dysfunction

Introduction Salivary gland dysfunction is an umbrella term that encompasses xerostomia or salivary gland hypofunction, or both. Xerostomia is the term used to describe the subjective sensation of having a dry mouth, whilst salivary gland hypofunction refers to a decrease in saliva flow (30). Such hyposalivation has been defined as an unstimulated whole saliva flow rate 0.1 ml per minute or lower (31). Normally, healthy individuals will experience xerostomia when their unstimulated whole saliva flow rate falls below 50% of their normal level (32). However, a reduction in saliva flow is not completely necessary in order for xerostomia to occur (33), as it may be due to changes in the composition of the saliva (34).

Salivary gland dysfunction is an almost inevitable side effect of radiotherapy to the head and neck (35). The typical total dose of a course of radiotherapy given to treat head and neck cancers is 50 Gy to 70 Gy (36), yet doses over 52 Gy will cause severe salivary gland dysfunction (33). A significant decrease in saliva flow occurs within one week of the start 19 of a course of radiotherapy, and continues to worsen throughout treatment, resulting in permanent salivary gland dysfunction (36). It has been reported that even a dose of 20 Gy is sufficient to cause permanent damage to saliva flow if received as a single dose (33). The prevalence of xerostomia has been reported as 93% (95% confidence interval 83% to 100%) during radiotherapy and 85% (95% confidence interval 78% to 93%) more than 2 years after (35). In addition to the discomfort of xerostomia, a reduction in saliva flow can result in oral infections, tooth decay, oral discomfort and pain, and problems with swallowing, eating, drinking and speaking (35). All of these issues will negatively affect the patient’s quality of life.

Unlike oral mucositis, which can be life threatening and can interrupt the normal course of cancer treatment, salivary gland dysfunction does not have such risks, and is therefore much less researched (37). Therefore such issues as economic burden are much less clear, and it is difficult to attribute resource utilisation to salivary gland dysfunction in the context of cancer treatment and its many other side effects, although it is acknowledged that direct and indirect impacts are likely (30).

Despite the permanent damage to the major salivary glands (parotid, submandibular and sublingual), some patients’ xerostomia may improve over time and it is thought that this could be due to a compensatory increase in flow from the minor glands or the patient getting used to xerostomia, or both of these reasons (38).

Interventions and their evidence base The evidence base for prophylactic pharmacological agents for salivary gland dysfunction is weak, and some guidelines do not currently make any recommendations, with preventative measures focusing more on non-pharmacological interventions such as parotid-sparing radiotherapy (39). Prior to 1990, only palliative treatments such as sialogogues (for promoting saliva secretion in surviving salivary gland tissues), oral hygiene, symptom relief, and saliva substitutes were used due to a lack of effective prophylactic drugs (30). Some of the main pharmacological agents that have been used for prevention can be categorised as follows.

Parasympathomimetic drugs These drugs stimulate the parasympathetic nervous system, promoting the secretion of saliva (37). The most widely known of these is pilocarpine, which has been approved in some countries, including the FDA in the USA (36). However, due to stimulation of other

20 bodily secretions such as tears, gastric juices, sweat, and other unwanted biological effects, some clinicians do not consider its use warranted (33).

Cevimeline and bethanechol are other newer drugs of this class, but which have been researched much less than pilocarpine.

Cytoprotective drugs Amifostine is a radioprotector that has also been assessed as a prophylactic agent against oral mucositis, as described above. It has been shown to accumulate in the salivary glands, protecting the tissue from damage by radiotherapy (30). However, amifostine has shown conflicting results (36). Furthermore, it is not currently recommended in this setting due to its side effects and high cost (39).

1.4 Importance of systematic reviews

As reported results have often been highly variable across studies, this makes systematic reviews all the more important in summarising and attempting to clarify the overall picture. Systematic reviews are one of the most important tools in evidence-based practice, which has been defined as “the conscientious, explicit, and judicious use of current best evidence in making decisions about the care of individual patients” (40). Due to the sheer volume of healthcare literature which is published, clinicians and others involved in the healthcare profession would need to continually read and critically appraise an overwhelming amount in order to make evidence-based decisions (41). Systematic reviews help to alleviate this problem, summarising and appraising the literature by using explicit and systematic methodology aimed at minimising bias in order that conclusions and resulting decisions are likely to be more valid (42). Systematic reviews may contain meta- analyses which pool the data from multiple studies, thus increasing power (i.e. overcoming the problem of small sample size) and allowing a more precise effect estimate to be reported (42). An example of the value of meta-analysis can be shown using the Cochrane review on the prevention of mucositis (14). If a clinician searched for randomised controlled trials to show whether or not to use chlorhexidine to prevent oral mucositis in their patients, they may be misled if they only looked at the results of either of the two studies in that comparison which favoured chlorhexidine over the placebo. However, the overall meta-analysis of four studies shows that there is no evidence that chlorhexidine is more effective than placebo, and as such chlorhexidine is not recommended in the MASCC/ISOO clinical guidelines (20).

In evidence-based practice there is a hierarchy of evidence as shown below in Figure 1.

Figure 1: The hierarchy of evidence (43)

It is vital that the Cochrane systematic reviews on the treatment and prevention of oral mucositis are of the highest standard and are kept up-to-date as they are considered to be the gold standard in terms of demonstrating the effects of interventions, and they also feed into clinical guidelines such as those produced by MASCC/ISOO. The Cochrane reviews are restricted to including randomised controlled trials (RCTs), which are the second highest level of evidence in the above hierarchy. This is because a well-conducted randomised controlled trial has the potential to avoid biases, the most significant being selection bias, whereby adequate randomisation, and concealment of the process, should ensure that known and unknown prognostic factors are equally distributed between treatment groups, thus ensuring that any differences in outcomes between the groups are therefore due to the intervention (44).

Bias is “a systematic error, or deviation from the truth, in results and inferences” and can lead to misinterpretation of the true treatment effect estimate, in terms of either an underestimation or overestimation, and there is empirical evidence to show that the way in which a randomised controlled trial is designed, conducted and analysed can lead to biases (42). Therefore, considering the risk of bias in a study is about the internal validity and whether or not the results of the study can be trusted. The Cochrane risk of bias tool sets Cochrane reviews apart from other types of systematic review which may only give

22 cursory consideration to risk of bias, or may use less rigorous methods. It considers bias as being distinct from quality (methodological quality) because it is possible that researchers could have conducted a study to a high standard, but it is inherently at high risk of bias due to a factor beyond their control (e.g. it may be impossible to avoid performance bias in a trial of surgery versus medicine, or there may be unavoidable attrition) (42). For each included study in a Cochrane review, a risk of bias table is constructed. This is a two-part tool typically assessing the following seven domains of risk of bias.

 Random sequence generation (selection bias)  Allocation concealment (selection bias)  Blinding of participants and personnel (performance bias)  Blinding of outcome assessment (detection bias)  Incomplete outcome data (attrition bias)  Selective reporting (reporting bias)  Other bias

For each domain, the first part of the process involves describing what was reported to have happened in the study, thus providing a rationale for the second part, in which a judgement of low, high or unclear risk of bias is assigned. This is a transparent process which allows the readers of the review to clearly understand the reason for any judgements, and permitting them to disagree and make their own interpretation. If all domains are rated as low risk, then the study can be said to have a low risk of bias overall. However, if there is at least one domain at unclear risk of bias, the study would be judged to be at unclear overall risk of bias. Furthermore, if at least one domain was at high risk of bias, then the study would be deemed to be at high risk of bias overall.

The Methodological Expectations of Cochrane Intervention Reviews (MECIR) (www.editorial-unit.cochrane.org/mecir) states that it is highly desirable for Cochrane reviews to contain summary of findings tables. These tables are intended to help assess the quality of the body of evidence for the main comparisons and outcomes of the review using the GRADE (Grading of Recommendations, Assessment, Development and Evaluation) approach (45). The tables also clearly display: details about population, settings, and intervention/comparison; details about the scale of the outcome measure (e.g. if it was a 0 to 10 scale and whether a lower/higher score indicated an improved/worse outcome) and the follow-up times; illustrative information on the magnitude of effect of the interventions; and a summary of the available data (42). The quality of the body of 23 evidence is assessed by considering the risk of bias of the included studies, the directness of the evidence (i.e. are the participants, interventions and outcomes relevant to the research question), inconsistency of the results (i.e. heterogeneity), precision of the estimates, and the risk of publication bias. The rating is downgraded for any flaws in the above five factors and the table is accompanied by explanatory footnotes explaining the reasons for decisions to downgrade (45). As in the risk of bias assessments, this methodology is transparent and allows the reader to understand the reasons for the ratings, and to make their own judgements. The rating of the body of evidence can be assessed as high (“further research is very unlikely to change our confidence in the estimate of effect”), moderate (“further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate”), low (“further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate”), or very low (“any estimate of effect is very uncertain”) (45).

1.5 Importance of clinical guidelines

What are clinical guidelines? Clinical guidelines were defined in 1990 by the Institute of Medicine (IoM): “Practice guidelines are systematically developed statements to assist practitioner and patient decisions about appropriate health care for specific clinical circumstances” (46). In 2011, the IoM updated the definition: “Clinical Practice Guidelines are statements that include recommendations intended to optimize patient care. They are informed by a systematic review of evidence and an assessment of the benefits and harms of alternative care options” (47). Clarifying the term ‘clinical practice guideline’, this revision was intended to distinguish between the preferable evidence-based form of guideline, developed in a transparent and methodologically rigorous manner, and other forms of guidelines which are perhaps less rigorously developed (47).

The IoM also identify some key indicators of quality, stating that “To be trustworthy, guidelines should  be based on a systematic review of the existing evidence;  be developed by a knowledgeable, multidisciplinary panel of experts and representatives from key affected groups;  consider important patient subgroups and patient preferences, as appropriate;

24  be based on an explicit and transparent process that minimizes distortions, biases, and conflicts of interest;  provide a clear explanation of the logical relationships between alternative care options and health outcomes, and provide ratings of both the quality of evidence and the strength of recommendations; and  be reconsidered and revised as appropriate when important new evidence warrants modifications of recommendations” (47).

Two of the major bodies developing clinical guidelines in the UK are the National Institute for health and Care Excellence (NICE), and the Scottish Intercollegiate Guidelines Network (SIGN). Both of these bodies have developed highly rigorous processes for the production of their guidelines and have amassed huge resources on their websites, including their own manuals. SIGN present a simplified summary of the guideline development process as (48): 1. Identify questions 2. Search for evidence 3. Look at the evidence 4. Make judgements and recommendations 5. Publish 6. Let everybody know about our guidelines.

Why are guidelines needed? Clinical practice can vary internationally, nationally, regionally, or even by individual doctor in the same location (49). One example of national variation is that of oral care for children being treated for cancer in the UK (15). In theory, the development and use of clinical guidelines, particularly when based on rigorous systematic reviews, should reduce such inappropriate variation, not only by promoting the use of effective practices, but also by discouraging the use of ineffective practices (49). One of the most commonly heard complaints about Cochrane reviews is the perceived overuse of phrases such as ‘there was insufficient evidence of a difference’, but this can be equally as important as a statement that an intervention is effective. For example, past Cochrane reviews on mouth care for cancer patients have shown a lack of evidence to support the use of chlorhexidine for preventing oral mucositis (14) and of chlorhexidine and nystatin for preventing oral candidiasis (50). These reviews fed into UK guidelines on mouth care for children, teenagers and young adults receiving cancer treatment carried out by the Children's Cancer and Leukaemia Group and Paediatric Oncology Nurses Forum (CCLG-PONF) Mouth Care 25 Group (16). Subsequently, an audit to monitor the uptake of national guidelines in the UK found evidence of large changes in prescribing patterns as a result of the guidelines, resulting in a 40% decrease in the routine use of prophylactic nystatin and a 34% decrease in the use of prophylactic chlorhexidine (51). Eradicating the use of ineffective and possibly even harmful interventions or practices could mean that there are more resources to invest in those interventions that have been proven to be beneficial.

Another example of the impact of guidelines is in the case of antibiotics given to moderate and high-risk patients prior to invasive dental treatment to prevent bacterial endocarditis. A Cochrane review on this topic, published in 2004 and updated in 2008 and 2013, concluded that there was no evidence about whether this practice was effective or ineffective against bacterial endocarditis (52). This created a lot of debate around the topic, due to the global problem of the overprescription of antibiotics and the emergence of antibiotic-resistant bacteria. Many dentists were inevitably reluctant to change their practice and put their patients at risk. The review was used to inform NICE’s 2008 guidelines, which recommended a stop to the routine prescription of antibiotics for high risk patients undergoing dental and other procedures (53). A study published in 2011 collected data on antibiotic prescribing from 2004 and up to 2 years following the publication of the NICE guidelines (54). Figure 2 below shows the dramatic impact of the NICE guidelines, which resulted in a 78.6% reduction in antibiotic prescribing, the large majority being due to dentists changing their practice.

Figure 2: Total number of prescriptions for antibiotic prophylaxis (amoxicillin 3 g or clindamycin 600 mg) dispensed each month by type of prescriber. Red lines represent average figure for prescriptions every three months (54) 26

However, there are also instances where guidelines may be perceived to be unhelpful. Whilst inappropriate variation in practice can be avoided, conversely, rigid adherence to guidelines may mean that one or more of the key elements of evidence-based practice is overlooked. That is to say patient preference may not be taken into account and the clinician’s judgement may be diminished as ‘cookbook medicine’ becomes the norm and clinicians do not feel that they are free to tailor treatment to the individual patient (49). Furthermore, in many areas of healthcare the evidence base is insufficient to make recommendations and this may lead to reliance on expert opinion, yet often those experts have financial conflicts of interest linked to industry (55).

The solution could be to ensure that guidelines are developed in the most transparent way possible, using rigorous methodology and the Appraisal of Guidelines for Research and Evaluation (AGREE) instrument can help with this endeavour. Guidelines should not be followed blindly, just as primary studies and systematic reviews should not be trusted at face value, as the quality, rigour and methodology can vary greatly. They should be critically appraised using an appropriate tool. The AGREE instrument is a checklist that can be used both to guide the reporting of guidelines and to evaluate their methodological quality, and it has become the internationally accepted standard (56). Indeed, in SIGN’s manual, which is another resource to aid guideline development, the AGREE instrument is reproduced and cited as the most evidence-based example of a published standard for guideline development methodology (57).

1.6 What are the gaps?

As discussed, Cochrane systematic reviews are arguably the best form of evidence when assessing the effects of interventions. This notion, along with the fact that oral mucositis and salivary gland dysfunction are such debilitating problems for cancer patients, implies that the Cochrane reviews on prevention these conditions should be kept up-to-date and should be made usable and accessible to end-users, including clinicians, patients and guideline developers. The Mucositis Study Group of the MASCC/ISOO consists of a large group of experts who carry out important work producing and updating clinical guidelines for the management of oral mucositis. As the Cochrane reviews feed into their guideline work, and the prevention review has become so large in terms of included studies, it may be more practical to split the current reviews into the following treatment categories currently used by MASCC/ISOO.

 Basic oral care/good clinical practice  Growth factors and cytokines  Anti-inflammatory agents  Antimicrobials, mucosal coating agents, anaesthetics and analgesics  Laser and other light therapy  Cryotherapy  Natural and miscellaneous agents

This could potentially make the reviews more user-friendly and facilitate ease of regular updating, as the task is currently a daunting one for any systematic reviewer.

There is currently no high quality systematic review evaluating all pharmacological interventions for preventing radiotherapy-induced salivary gland dysfunction. This inevitably means that guidelines are very variable and inconsistent.

Finally, it would be beneficial to carry out a critical appraisal of guidelines on mouth care for cancer patients to give an indication of their quality and to assess whether or not they are based on appropriately rigorous and methodologically sound evidence.

28 2 AIMS

The aims of this thesis are to:

 assess the effects of interventions for preventing oral mucositis in patients with cancer who are receiving treatment;  assess the effects of pharmacological interventions for the prevention of radiation- induced salivary gland dysfunction;  assess the quality of guidelines on the management of oral complications caused by cancer treatment

Chapters 3 and 4 address the first aim through systematic reviews following Cochrane methodology. It was not considered feasible to continue updating the original Cochrane review of all interventions for preventing oral mucositis and at the same time achieve the required clinical relevance. Therefore the review is to be split into separate reviews based on interventions or categories of interventions used by the Mucositis Study Group of the Multinational Association of Supportive Care in Cancer/International Society of Oral Oncology as listed in Chapters 1.6, 3.2 and 4.2. Two promising areas were chosen for assessment: 1) oral cryotherapy, and 2) cytokines and growth factors.

Chapter 5 addresses the second aim, again through a Cochrane systematic review. This topic has long been a priority of Cochrane Oral Health’s editorial team to publish. As the review matches the topic of the thesis on mouth care for cancer patients, the opportunity was taken to complete this review.

Chapter 6 addresses the final aim using the AGREE II instrument and ad hoc methods. After assessing the evidence for some interventions for preventing oral complications due to cancer treatment, it seemed appropriate to look at how such evidence is being used to inform clinical practice.

The thesis follows the University of Manchester’s alternate format where chapters are presented in the style of a journal publication. Chapters 3 and 5 have been peer reviewed and published on the Cochrane Database of Systematic Reviews, whilst Chapter 4 has been peer reviewed and is awaiting publication. Chapter 6 will also be submitted to a suitable journal for peer review and publication. An overall discussion of issues that influence the usefulness and interpretation/uptake of the evidence is presented in Chapter 7.

29 3 INTERVENTIONS FOR PREVENTING ORAL MUCOSITIS IN PATIENTS WITH CANCER RECEIVING TREATMENT: ORAL CRYOTHERAPY

3.1 Abstract

Background

Oral mucositis is a side effect of chemotherapy, head and neck radiotherapy, and targeted therapy, affecting over 75% of high risk patients. Ulceration can lead to severe pain and difficulty eating and drinking, which may necessitate opioid analgesics, hospitalisation and nasogastric or intravenous nutrition. These complications may lead to interruptions or alterations to cancer therapy, which may reduce survival. There is also a risk of death from sepsis if pathogens enter the ulcers of immunocompromised patients. Ulcerative oral mucositis can be costly to healthcare systems, yet there are few preventive interventions proven to be beneficial. Oral cryotherapy is a low-cost, simple intervention which is unlikely to cause side-effects. It has shown promise in clinical trials and warrants an up-to- date Cochrane review to assess and summarise the international evidence.

Objectives

To assess the effects of oral cryotherapy for preventing oral mucositis in patients with cancer who are receiving treatment.

Search methods

We searched the following databases: the Cochrane Oral Health Group Trials Register (to 17 June 2015), the Cochrane Central Register of Controlled Trials (CENTRAL) (Cochrane Library 2015, Issue 5), MEDLINE via Ovid (1946 to 17 June 2015), EMBASE via Ovid (1980 to 17 June 2015), CANCERLIT via PubMed (1950 to 17 June 2015) and CINAHL via EBSCO (1937 to 17 June 2015). We searched the US National Institutes of Health Trials Registry, and the WHO Clinical Trials Registry Platform for ongoing trials. No restrictions were placed on the language or date of publication when searching databases.

Selection criteria

We included parallel-design randomised controlled trials (RCTs) assessing the effects of oral cryotherapy in patients with cancer receiving treatment. We used outcomes from a published core outcome set registered on the COMET website.

30 Data collection and analysis

Two review authors independently screened the results of electronic searches, extracted data and assessed risk of bias. We contacted study authors for information where feasible. For dichotomous outcomes, we reported risk ratios (RR) and 95% confidence intervals (CI). For continuous outcomes, we reported mean differences (MD) and 95% CIs. We pooled similar studies in random-effects meta-analyses. We reported adverse effects in a narrative format.

Main results

We included 14 RCTs analysing 1280 participants. The vast majority of participants did not receive radiotherapy to the head and neck, so this review primarily assesses prevention of chemotherapy-induced oral mucositis. All studies were at high risk of bias. The following results are for the main comparison: oral cryotherapy versus control (standard care or no treatment).

Adults receiving fluorouracil-based (5FU) chemotherapy for solid cancers

Oral cryotherapy probably reduces oral mucositis of any severity (RR 0.61, 95% CI 0.52 to 0.72, 5 studies, 444 analysed, moderate quality evidence). In a population where 728 per 1000 would develop oral mucositis, oral cryotherapy would reduce this to 444 (95% CI 379 to 524). The number needed to treat to benefit one additional person (NNTB), i.e. to prevent them from developing oral mucositis, is 4 people (95% CI 3 to 5).

The results were similar for moderate to severe oral mucositis (RR 0.52, 95% CI 0.41 to 0.65, 5 studies, 444 analysed, moderate quality evidence). NNTB 4 (95% CI 4 to 6).

Severe oral mucositis is probably reduced (RR 0.40, 95% CI 0.27 to 0.61, 5 studies, 444 analysed, moderate quality evidence). Where 300 per 1000 would develop severe oral mucositis, oral cryotherapy would reduce this to 120 (95% CI 81 to 183), NNTB 6 (95% CI 5 to 9).

Adults receiving high-dose melphalan-based chemotherapy before haematopoietic stem cell transplantation (HSCT)

Oral cryotherapy may reduce oral mucositis of any severity (RR 0.59, 95% CI 0.35 to 1.01, 5 studies, 270 analysed, low quality evidence). Where 824 per 1000 would develop oral mucositis, oral cryotherapy would reduce this to 486 (95% CI reduced to 289 to increased

31 to 833). The NNTB is 3, although the uncertainty surrounding the effect estimate means that the 95% CI ranges from 2 NNTB, to 111 NNTH (number needed to treat in order to harm one additional person, i.e. for one additional person to develop oral mucositis).

The results were similar for moderate to severe oral mucositis (RR 0.43, 95% CI 0.17 to 1.09, 5 studies, 270 analysed, low quality evidence). NNTB 3 (95% CI 2 NNTB to 17 NNTH).

Severe oral mucositis is probably reduced (RR 0.38, 95% CI 0.20 to 0.72, 5 studies, 270 analysed, moderate quality evidence). Where 427 per 1000 would develop severe oral mucositis, oral cryotherapy would reduce this to 162 (95% CI 85 to 308), NNTB 4 (95% CI 3 to 9).

Oral cryotherapy was shown to be safe, with very low rates of minor adverse effects, such as headaches, chills, numbness/taste disturbance, and tooth pain. This appears to contribute to the high rates of compliance seen in the included studies.

There was limited or no evidence on the secondary outcomes of this review, or on patients undergoing other chemotherapies, radiotherapy, targeted therapy, or on comparisons of oral cryotherapy with other interventions or different oral cryotherapy regimens. Therefore no further robust conclusions can be made. There was also no evidence on the effects of oral cryotherapy in children undergoing cancer treatment.

Authors' conclusions

We are confident that oral cryotherapy leads to large reductions in oral mucositis of all severities in adults receiving 5FU for solid cancers. We are less confident in the ability of oral cryotherapy to reduce oral mucositis in adults receiving high-dose melphalan before HSCT. Evidence suggests that it does reduce oral mucositis in these adults, but we are less certain about the size of the reduction, which could be large or small. However, we are confident that there is an appreciable reduction in severe oral mucositis in these adults.

This Cochrane review includes some very recent and currently unpublished data, and strengthens international guideline statements for adults receiving the above cancer treatments.

32 3.2 Background

Description of the condition

Treating cancer with chemotherapy, radiotherapy of the head and neck, or targeted therapy can cause toxic oral side effects (1-3). Perhaps the most widely researched of these side effects is oral mucositis (1), which affects at least 75% of high risk patients (those receiving head and neck radiotherapy or high-dose chemotherapy) (2). Oral mucositis may be under-reported in lower risk groups for various reasons: their tendency to be outpatients with less observation; less reporting of moderate mucositis; or patients and clinicians wishing to avoid any disruption to optimal cancer treatment (2).

Simply put, oral mucositis affects the oral mucosa (the mucous membrane of moist tissue lining the oral cavity) and can lead to the development of lesions (ulcers). However, the process that leads to oral mucositis is complex and multifactorial, with Sonis' five phase model being the currently accepted explanation for the sequence of events underlying the condition (3, 8).

1. Initiation: DNA damage caused by chemotherapy or radiotherapy results in the loss of ability to proliferate in the basal cells of the epithelium (the external layers of cells lining the oral mucosa). This produces reactive oxygen species (ROS). 2. Primary damage response: Radiotherapy, chemotherapy, ROS, and DNA strand breaks all contribute to the activation of transcription factors such as nuclear factor kappa beta (NF-Kβ), and sphingomyelinases. All this leads to the upregulation of pro-inflammatory cytokines (e.g. tumour necrosis factor alpha - TNF-α), nitric oxide, ceramide, and matrix metalloproteinases, resulting in the thinning of the epithelium through tissue injury and cell death, culminating with the destruction of the oral mucosa. 3. Signal amplification: Some of the molecules in the previous phase can lead to the exacerbation and prolonging of tissue injury through positive or negative feedback (e.g. TNF-α can positively feedback on NF-Kβ thus inducing more pro- inflammatory cytokine production). 4. Ulceration: Bacteria colonise ulcers and their cell wall products infiltrate the submucosa (the connective tissues beneath the oral mucosa), activating tissue macrophages (white blood cells that respond to infection or damaged/dead cells), which results in further production of pro-inflammatory cytokines, inflammation, and pain.

33 5. Healing: Signalling from the extracellular matrix of the submucosa results in epithelial proliferation and differentiation, and thus a thickening of the epithelium. The local oral flora are reinstated.

Understanding of the pathobiology leading to mucosal toxicity as a result of targeted therapies (e.g. mammalian target of rapamycin (mTOR) inhibitor-associated stomatitis - mIAS) is currently limited, but it is thought to differ from chemotherapy- and radiotherapy-induced mucositis, and the clinical presentation of the ulcers is more similar to aphthous stomatitis (1, 7, 58).

Chemotherapy-induced oral mucositis is regarded as an acute condition, with ulceration normally occurring one week after treatment, and resolving within three weeks of treatment (8). Radiotherapy-induced oral mucositis is chronic in nature, with ulceration normally occurring around two weeks into a seven-week treatment cycle, and resolving three to four weeks after treatment has ended (8).

Ulceration is the most significant phase as it leads to pain of varying severity, and difficulties with eating, swallowing, and talking (2). This in turn leads to the consumption of pain relief medication, requirement for nutritional support (e.g. a feeding tube), treatment of the oral mucositis, specialist oral hygiene care, increased medical appointments and use of staff and resources, and, in some instances, hospitalisation (4-6). Thus the negative impact on the quality of life of cancer patients, when they are already suffering, is severe (59, 60). Further problems can occur in immunosuppressed patients if whole bacteria on the ulcer surface cross into the underlying submucosa, potentially leading to bacteraemia and sepsis, which require antibiotics and hospitalisation, and can cause death (2, 6, 7).

Therefore, oral mucositis can be a dose-limiting condition, disrupting a patient's optimal cancer treatment plan and consequentially decreasing their chances of survival (3, 6, 7). The additional costs associated with oral mucositis can be significant, with one study reporting a median incremental cost of USD 18,515 per patient (61). These costs have been reported to be as much as USD 42,749 more per patient when ulcerative oral mucositis is present (62).

Description of the intervention

Fluorouracil (5FU) is a common chemotherapy treatment for solid cancers and, in this setting, oral cryotherapy typically involves placing ice chips in the mouth five minutes

34 prior to chemotherapy and continuing for 30 minutes (21). In other settings, such as patients with haematological cancers receiving high-dose melphalan prior to stem cell transplantation, oral cryotherapy is administered for longer periods of time, even as long as seven hours (63). The ice chips are typically rounded to avoid any sharp edges or corners that may cause irritation to the patient, and also so that they can be easily moved around in the mouth (64).

The advantages of using cryotherapy over other interventions are its availability, cost- effectiveness, ease of administration, and safety (in terms of lack of side-effects), and that it is well tolerated by patients (26).

How the intervention might work

The use of ice chips in the mouth cools the oral tissues and causes the blood vessels to narrow (vasoconstriction), thus reducing blood flow to the area and therefore also restricting the amounts of the chemotherapy drugs delivered to the tissues (2, 21, 26). Cryotherapy may only be effective in the prevention of oral mucositis in patients receiving chemotherapy drugs that have a short half-life, such as bolus 5-FU, bolus edatrexate, and high-dose melphalan (2, 21, 26). Considering the mechanism by which cryotherapy can prevent oral mucositis caused by chemotherapy, it is unclear whether or not it could have any effect on oral mucositis caused by radiotherapy (21). It is also unclear whether or not cryotherapy could have any role in the prevention of targeted therapy-induced stomatitis.

Why it is important to do this review

This review is the first of a series that will replace the previously published Cochrane review covering all interventions for the prevention of oral mucositis in patients with cancer receiving treatment (14). The Mucositis Study Group (MSG) of the Multinational Association of Supportive Care in Cancer/International Society of Oral Oncology (MASCC/ISOO) is a group that was set up in 1998 for the purpose of producing international evidence-based clinical practice guidelines for managing mucositis (both oral and gastrointestinal), which they first published in 2004, with the latest update published in 2014 (65). In order to facilitate the future updating of Cochrane reviews on this topic, and also to make them more usable to clinicians, guideline developers, and consumers, we have decided to divide the original Cochrane review into the same intervention categories as those used by MASCC/ISOO, which are as follows:

 basic oral care/good clinical practice;

35  growth factors and cytokines;  anti-inflammatory agents;  antimicrobials, mucosal coating agents, anaesthetics, and analgesics;  laser and other light therapy;  cryotherapy;  natural and miscellaneous agents;  amifostine.

We believe that running in tandem with the MASCC/ISOO categories will enable the Cochrane reviews to more easily feed into such guidelines. We will also be able to be more thorough and rigorous in our assessment and summarising of the evidence in each of the categories, which was not feasible in a single Cochrane review approaching 150 included studies.

It is also important to do this review as it is consistently shown to be the most used review produced by the Cochrane Oral Health Group (in terms of full-text downloads). It was also ranked by an expert panel of oral medicine specialists as being the most important topic in the field of oral medicine in an international prioritisation exercise carried out by the Cochrane Oral Health Group in 2014 (66).

3.3 Objectives

To assess the effects of oral cryotherapy for preventing oral mucositis in patients with cancer who are receiving treatment.

3.4 Methods

Criteria for considering studies for this review Types of studies

We included all randomised controlled trials (RCTs) of parallel design. It is possible to conduct cross-over studies in this area as patients may receive several treatment sessions, with any mucositis completely healing in the periods between the sessions. However, we did not include cross-over data as we cannot discount any period effects, with mucositis risk increasing as patients receive further cycles of treatment (2, 8). Instead, we used the first-period data only and treated such studies as parallel studies.

36 Types of participants

We included all patients with cancer who are receiving treatment.

Types of interventions

We included studies comparing oral cryotherapy for the prevention of oral mucositis against usual care, no treatment, or any other treatment to prevent oral mucositis. We also included studies comparing different regimens of oral cryotherapy against each other (head-to-head studies). We planned to include studies of oral mucositis caused by chemotherapy, radiotherapy, and targeted therapy.

We excluded studies with 'complex' interventions for the prevention of mucositis, such as lasers plus cryotherapy versus lasers. We excluded studies assessing different cancer treatments where the primary outcome is survival/cure, with mucositis as a toxicity.

Types of outcome measures

We are in agreement with Williamson et al that, if clinical trials and systematic reviews are to be utilised, the outcomes assessed should be those considered important to patients, healthcare professionals, and other key stakeholders (67). If outcomes and outcome measures are inconsistent across studies, it will not be possible to compare and summarise research, and there is potential for outcome reporting bias, with the selective reporting of results based on statistical significance and favourability (68-70). This can lead to exaggerated estimates of effect in systematic reviews of interventions, leading to an incorrect belief that an intervention is more beneficial that it truly is (68). It is thought that the way to address this problem is to develop disease- or condition-specific core outcome sets to be used as a minimum when conducting and reporting clinical trials (67, 68).

Therefore we used the core outcome set produced by Bellm et al, which is registered on the COMET (Core Outcome Measures in Effectiveness Trials) Initiative's website (www.comet-initiative.org), and is the only core outcome set for oral mucositis known to us (71).

Primary outcomes

Mucositis incidence of any severity. We used mucositis measured on a 0 to 4 point scale (none to severe) and dichotomised it as any mucositis (0 versus 1+), moderate to severe mucositis (0 to 1 versus 2+), and severe mucositis (0 to 2 versus 3+).

37 Some studies measure mucositis using a composite scale. If it was possible to extract the 'mucositis only' data from the total score, we would have included the data in the analyses. If it was not possible, we would have recorded the composite data in an additional table.

Secondary outcomes

 Interruptions to cancer treatment.  Oral pain.  Quality of life.  Normalcy of diet (including use of percutaneous endoscopic gastrostomy (PEG) feeding tubes or total parenteral nutrition (TPN)).  Adverse events.  Number of days in hospital.  Number of days of treatment with opioid analgesics.  Number of days unable to take medicine orally.

Search methods for identification of studies

For the identification of studies included or considered for this review, we developed detailed search strategies for each database searched. These were based on the search strategy we developed for MEDLINE (Ovid), which we revised appropriately for each database. The study design filter used by the Cochrane Oral Health Group was added to the search of the EMBASE database to limit the search to randomised controlled trials. The search strategies are presented in full in Appendix 1.

Electronic searches

We searched the following electronic databases:

 the Cochrane Oral Health Group Trials Register (to 17 June 2015);  the Cochrane Central Register of Controlled Trials (CENTRAL) (Cochrane Library 2015, Issue 5);  MEDLINE via Ovid (1946 to 17 June 2015);  EMBASE via Ovid (1980 to 17 June 2015);  CANCERLIT via PubMed (1950 to 17 June 2015);  CINAHL via EBSCO (1937 to 17 June 2015).

No restrictions were placed on the language or date of publication when searching the electronic databases. 38 Searching other resources

We searched the following databases for ongoing trials:

 US National Institutes of Health Trials Registry (http://clinicaltrials.gov) (to 17 June 2015);  World Health Organization (WHO) International Clinical Trials Registry Platform (http://apps.who.int/trialsearch/default.aspx) (to 17 June 2015).

We only included handsearching done as part of the Cochrane Worldwide Handsearching Programme and uploaded to CENTRAL.

Data collection and analysis Selection of studies

Two review authors independently screened the titles and abstracts retrieved from the electronic searches. We obtained full-text copies of all studies that appeared to meet the inclusion criteria of the review, or where there was insufficient information in the title or abstract to make a clear judgement. Two review authors independently assessed the full text copies for eligibility and attempted to resolve any disagreements through discussion. We consulted a third review author if we could not resolve disagreements.

On assessing the full text article, we discarded any studies that clearly did not meet the inclusion criteria. We recorded all other studies that did not meet the inclusion criteria, along with reasons for exclusion, in the ‘Characteristics of excluded studies table’ (see Appendix 2).

Data extraction and management

Two review authors independently extracted the data from each included study using a specially designed data extraction form, which we first piloted on a small sample of studies. We contacted study authors for clarification or missing data where necessary and feasible. We resolved any disagreements through discussion, consulting a third review author to achieve consensus when necessary.

We recorded the following data for each included study in the ‘Characteristics of included studies’ table (see Appendix 2).

 Trial design, location, number of centres, recruitment period.

39  Inclusion/exclusion criteria, age and gender of participants, number randomised/analysed, any other potentially important prognostic factors (e.g. cancer type, cancer treatment, etc.).  Detailed description of the intervention and comparator, including timing and duration. Information on compliance with the cryotherapy regimen.  Details of the outcomes reported, including method of assessment and time(s) assessed.  Details of sample size calculations, adverse effects, funding sources, declarations/conflicts of interest.

Assessment of risk of bias in included studies

Two review authors independently assessed the risk of bias of each included study using the Cochrane domain-based, two-part tool as described in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (42). We contacted study authors for clarification or missing information where necessary and feasible. We resolved any disagreements through discussion, consulting a third review author to achieve consensus when necessary.

We completed a 'Risk of bias' table for each included study (see Appendix 2). For each domain of risk of bias, we first described what was reported to have happened in the study. This provided the rationale for our judgement of whether that domain was at low, high, or unclear risk of bias.

We assessed the following domains:

1. sequence generation (selection bias); 2. allocation concealment (selection bias); 3. blinding of participants and personnel (performance bias); 4. blinding of outcome assessment (detection bias); 5. incomplete outcome data (attrition bias); 6. selective outcome reporting (reporting bias); 7. other bias.

We categorised the overall risk of bias of individual studies. Studies were categorised as being at low, high, or unclear risk of bias according to the following criteria:

40  low risk of bias (plausible bias unlikely to seriously alter the results) if all domains were at low risk of bias;  high risk of bias (plausible bias that seriously weakens confidence in the results) if one or more domains were at high risk of bias; or  unclear risk of bias (plausible bias that raises some doubt about the results) if one or more domains were at unclear risk of bias.

We also presented the 'Risk of bias' summary graphically.

Measures of treatment effect

For continuous outcomes (e.g. oral pain on a visual analogue scale) where studies used the same scale, we used the mean values and standard deviations (SDs) reported in the studies in order to express the estimate of effect as mean difference (MD) with 95% confidence interval (CI). Where different scales were used, we would have considered expressing the treatment effect as standardised mean difference (SMD) with 95% CI.

For dichotomous outcomes (e.g. mucositis of any severity/no mucositis), we expressed the estimate of effect as a risk ratio (RR) with 95% CI.

Unit of analysis issues

The participant was the unit of analysis.

Dealing with missing data

We attempted to contact the author(s) of all included studies, where feasible, for clarification, missing data, and details of any other outcomes that may have been measured but not reported. We used the methods described in Section 7.7.3 of the Cochrane Handbook for Systematic Reviews of Interventions to estimate missing SDs (42). We did not use any other statistical methods or perform any further imputation to account for missing data.

Assessment of heterogeneity

41 significant heterogeneity. We quantified heterogeneity using the I2 statistic. A guide to interpretation of the I2 statistic given in Section 9.5.2 of the Cochrane Handbook for Systematic Reviews of Interventions is as follows (42):

 0% to 40%: might not be important;  30% to 60%: may represent moderate heterogeneity;  50% to 90%: may represent substantial heterogeneity;  75% to 100%: considerable heterogeneity.

Assessment of reporting biases

If at least 10 studies were included in a meta-analysis, we planned to assess publication bias according to the recommendations on testing for funnel plot asymmetry (72), as described in Section 10.4 of the Cochrane Handbook for Systematic Reviews of Interventions (42). If asymmetry was identified, we would examine possible causes. We were not able to assess publication bias in this way because, although we had a sufficient number of studies in our meta-analyses for the primary outcome, they were split into subgroups containing less than 10 studies, with no pooling of the subgroup totals.

Data synthesis

We only carried out meta-analyses where there were studies of similar comparisons reporting the same outcomes. We combined MDs for continuous data, and RRs for dichotomous data. Our general approach was to use a random-effects model. With this approach, the CIs for the average intervention effect were wider than those that would have been obtained using a fixed-effect approach, leading to a more conservative interpretation.

We used an additional table to report the results from studies not suitable for inclusion in a meta-analysis.

Subgroup analysis and investigation of heterogeneity

We carried out subgroup analyses according to type of cancer treatment. We stated in the protocol that we would also use both type of cancer treatment and cancer type as categories for subgroup analyses. However, these categories are very closely related, so it did not make sense to do so. We also would have considered age group (children versus adults) as a category for subgroup analyses, if there had been sufficient numbers of studies with these differing populations.

42 Sensitivity analysis

As all studies were at high risk of both performance and detection bias, it was not possible to test the robustness of our results by performing sensitivity analyses based on excluding the studies at unclear or high risk of bias from the analyses.

If any meta-analyses had included several small studies and a single very large study, we would have undertaken a sensitivity analysis comparing the effect estimates from both random-effects and fixed-effect models. If these were different we would have reported on both analyses as part of the results section, and we would have considered possible interpretation.

Presentation of main results

We produced a 'Summary of findings' table for each comparison that included more than one study, and for the main outcomes (listed below). We produced a separate table for each of the two main cancer treatment types in this review: treatment of solid cancers and treatment of haematological cancers. We used GRADE methods (45), and the GRADEpro online tool for developing 'Summary of findings' tables (www.guidelinedevelopment.org). We assessed the quality of the body of evidence for each comparison and outcome by considering the overall risk of bias of the included studies, the directness of the evidence, the inconsistency of the results, the precision of the estimates, and the risk of publication bias. We categorised the quality of each body of evidence as high, moderate, low, or very low.

Main outcomes:

 mucositis incidence;  interruptions to cancer treatment;  oral pain;  quality of life;  normalcy of diet;  adverse events;  number of days in hospital.

3.5 Results

3.5.1 Description of studies

Results of the search

The electronic searches retrieved 745 references to studies. After removing duplicates, this figure was reduced to 426. We examined the titles and abstracts of these references and discarded all but 40 with no further assessment. Where possible, we obtained full-text copies of these 40 potentially relevant references, and we linked any multiple references to the same study under a single study ID, resulting in a total of 32 studies. We excluded 11 studies (11 references) at this stage. The remaining 21 studies (29 references) appeared to meet our inclusion criteria and we were able to include 14 of these studies (20 references). Of the remaining 7 studies (9 references), 5 studies (7 references) were awaiting assessment due to insufficient information in the abstract or trials registry record to allow inclusion in the review, and 2 studies (2 references) were ongoing. We presented this process as a flow chart in Figure 3.

Figure 3: PRISMA flow diagram

45 Included studies

Fourteen studies were included in this review (see ‘Characteristics of included studies’ tables in Appendix 2):

 Askarifar 2015 (73)  Cascinu 1994 (74)  Gori 2007 (75)  Heydari 2012 (76)  Kakoei 2013 (77)  Katranci 2012 (78)  Lilleby 2006 (63)  Mahood 1991 (79)  Rocke 1993 (80)  Salvador 2012 (81)  Sorensen 2008 (82)  Svanberg 2007 (83)  Toro 2013 (84)  Zhang 2011 (85)

Characteristics of the trial designs and settings

Twelve studies were of parallel design, with the remaining two studies using a cross-over design (Mahood 1991; Rocke 1993) . We treated the cross-over studies as parallel studies by using only the first-period data. Eleven studies had two arms, whilst two studies had three arms (Sorensen 2008; Toro 2013), one of which was excluded from Toro 2013 as we felt that the comparison of the third group, using supersaturated calcium phosphate rinse (Caphosol®), with the co-intervention of cryotherapy plus saline may be confounded. Another study had four arms, with one of those arms being excluded from this review because it involved treating oral mucositis rather than preventing it (Zhang 2011). Four studies were conducted in the USA (Lilleby 2006; Mahood 1991; Rocke 1993; Toro 2013), three in Iran (Askarifar 2015; Heydari 2012; Kakoei 2013), two in Italy (Cascinu 1994; Gori 2007), and one in each of Turkey (Katranci 2012), Canada (Salvador 2012), Denmark (Sorensen 2008), Sweden (Svanberg 2007), and China (Zhang 2011). There were nine single-centre studies, one with two centres (Heydari 2012), and four that were multicentre

46 but were unclear about how many centres were involved (Gori 2007; Mahood 1991; Rocke 1993; Sorensen 2008). Five studies did not report the duration of the trial from start to finish, but the remaining studies ranged in total duration from six months to five years.

Eight studies reported details of a sample size calculation: two of these studies achieved their required sample size (Kakoei 2013; Salvador 2012), two did not (Sorensen 2008; Toro 2013), three were unclear whether or not the sample size requirements were met (Askarifar 2015; Gori 2007; Lilleby 2006), and the remaining study did not use the primary outcome of oral mucositis incidence/severity to calculate the required sample size (Svanberg 2007). Six studies did not mention sample size calculation (Cascinu 1994; Heydari 2012; Katranci 2012; Mahood 1991; Rocke 1993; Zhang 2011).

Eleven studies reported on funding sources, all of which were in the form of independent funding from government, charities or universities. The remaining three studies did not report any funding sources (Cascinu 1994; Katranci 2012; Zhang 2011). Four studies declared that there were no conflicts of interest (Askarifar 2015; Katranci 2012; Salvador 2012; Toro 2013), whilst the other ten studies did not mention conflicts of interest.

Characteristics of the participants

There were 1316 participants randomised to interventions (including only the intervention groups relevant to this review), of which 1280 were included in the studies' analyses. Age ranged from 8 to 85 years across the studies, with mean or median ages ranging between 36 to 63 years. However, only one study reported the inclusion of children, although this was a small minority (Gori 2007). In general, there were more males than females in the studies. Only one study included more females than males, but this was because 50% of the participants had breast cancer (Heydari 2012). One study involved participants undergoing radiotherapy to the head and neck, with the remaining studies involving participants undergoing chemotherapy. No studies involved targeted therapy.

Solid cancers

In eight studies, the participants had solid cancers. In four of these studies, the majority of cancers were gastrointestinal, colorectal, and breast (Cascinu 1994; Heydari 2012; Katranci 2012; Sorensen 2008), whilst two studies did not state the types of cancers involved (Mahood 1991; Rocke 1993), one study included only head and neck cancers (Kakoei 2013), and one study included only bone cancer (osteosarcoma) (Zhang 2011). The cancer treatment in these studies of solid cancers mostly involved fluorouracil (5FU),

47 normally in conjunction with leucovorin (Cascinu 1994; Heydari 2012; Katranci 2012; Mahood 1991; Rocke 1993; Sorensen 2008). The study on participants with head and neck cancers was the only study included in this review that assessed the effects of oral cryotherapy for preventing oral mucositis in people receiving radiotherapy-only (Kakoei 2013). In the remaining study including participants with osteosarcoma (Zhang 2011), the cancer treatment was high-dose methotrexate plus vincristine and leucovorin.

Haematological cancers

In five studies, the participants had haematological cancers: multiple myeloma (Lilleby 2006; Salvador 2012; Toro 2013), Hodgkin lymphoma, non-Hodgkin lymphoma, and multiple myeloma (Askarifar 2015), and mixed (Gori 2007). The cancer treatment in these studies involved high-dose melphalan (Askarifar 2015; Lilleby 2006; Salvador 2012; Toro 2013), or low-dose methotrexate for preventing graft-versus-host disease (GVHD), which is the rejection of donor cells, after allogeneic (cells from a donor) stem cell transplantation (Gori 2007).

One study included one participant in each group (2.6%) with solid cancer (testicular), with the remaining participants all having a mixture of haematological cancers (Svanberg 2007). The cancer treatment in this study was mixed, with the majority (73%) of participants receiving either high-dose melphalan or BEAC regimen (carmustine, etoposide, cytarabine and cyclophosphamide).

In these six studies involving haematological cancers, participants also had haematopoietic stem cell transplantation (HSCT) following chemotherapy. However, in Gori 2007, the authors assessed the effects of oral cryotherapy for preventing oral mucositis due to further chemotherapy given after HSCT for the purpose of preventing GVHD.

Characteristics of the interventions and comparisons

One study compared different durations (30 minutes versus 60 minutes) of oral cryotherapy (Rocke 1993). Eight studies compared oral cryotherapy against no treatment (Cascinu 1994; Gori 2007; Heydari 2012; Kakoei 2013; Mahood 1991; Salvador 2012; Svanberg 2007; Toro 2013). In four of these studies, both the oral cryotherapy group and control group received standard oral care (Kakoei 2013; Salvador 2012; Svanberg 2007; Toro 2013), so the comparison could be thought of as being oral cryotherapy versus no treatment (no extra treatment). One of these four studies also had a supersaturated calcium phosphate rinse (Caphosol®) group which we excluded from this review because we did

48 not consider this intervention versus cryotherapy plus saline rinse to be an eligible comparison, due to potential confounding from the latter co-intervention (Toro 2013). Another four studies compared oral cryotherapy against some form of standard oral care: saline rinse (Askarifar 2015; Lilleby 2006; Sorensen 2008), unspecified (Katranci 2012). One of these studies had a third intervention group that received chlorhexidine rinse (Sorensen 2008). The remaining study had three eligible intervention groups and compared oral cryotherapy against leucovorin rinse and high-dose leucovorin rinse (Zhang 2011).

Most studies used ice chips for cooling the oral cavity, with one study also allowing an alternative option of using popsicles (a flavoured ice lolly) (Gori 2007), and another study allowed the alternative option of ice-cold water (Svanberg 2007). Two studies used ice cubes (Askarifar 2015; Kakoei 2013), possibly suggesting a larger size of ice pieces, whereas another two studies used crushed ice (Sorensen 2008; Toro 2013), possibly suggesting smaller ice pieces. One study exclusively used ice water (Zhang 2011).

The duration of treatment with oral cryotherapy varied widely according to chemotherapy regimen, and was unclear in five studies (Askarifar 2015; Cascinu 1994; Svanberg 2007; Toro 2013; Zhang 2011). The most consistent cryotherapy schedule was in those participants receiving fluorouracil (5FU) and leucovorin, and typically consisted of 30 minutes (45 minutes in Sorensen 2008; 60 minutes in half the participants in Rocke 1993) per day for five consecutive days. The longest duration of oral cryotherapy was seven hours (Lilleby 2006).

Characteristics of the outcomes

We wrote to authors of 11 of the included studies to ask if they had measured any other outcomes than those mentioned in the study reports. We did not write to authors of three of the included studies as they were published before the year 2000 and we thought it was unfeasible to obtain any extra data (Cascinu 1994; Mahood 1991; Rocke 1993). The authors of one study since provided us with data for two outcomes not reported in their study: 'interruptions to cancer treatment' and 'oral pain' (Heydari 2012).

Primary outcome

For the primary outcome of oral mucositis, we were interested in both the presence/absence of oral mucositis, and also different levels of severity. All 14 included studies measured oral mucositis. Eleven studies used the WHO 0 to 4 scale, or a scale based on this. Three studies used the National Cancer Institute common toxicity criteria

49 (NCI-CTC) 0 to 4 scale (Lilleby 2006; Sorensen 2008; Zhang 2011). The WHO and NCI- CTC scales include both subjective and objective elements, and are highly comparable (see Appendix 3), such that it is not necessary to use standardised mean difference when including both types of measurement in a meta-analysis. One study used a modified version of the Oral Mucositis Assessment Scale (OMAS) as the primary tool for measurement of oral mucositis, but provided us with full data according to the WHO scale (Svanberg 2007). The OMAS is an objective scale, measuring ulceration (0 to 3 scale) and erythema (0 to 2 scale) at nine different sites in the oral cavity (see Appendix 8).

Eight studies reported the data in our preferred format which was the maximum oral mucositis score, on a 0 to 4 scale, experienced by each participant over the length of the study (Cascinu 1994; Gori 2007; Heydari 2012; Lilleby 2006; Mahood 1991; Rocke 1993; Sorensen 2008; Toro 2013). One study reported the incidence of any oral mucositis (i.e. grades 1 to 4), but appeared to selectively report the incidence of each grade on a single day quite early in the study (i.e. not the maximum score experienced by each participant over the length of the study) (Zhang 2011). Four studies reported the mean oral mucositis score on multiple assessment days (Askarifar 2015; Kakoei 2013; Salvador 2012; Svanberg 2007), whilst the remaining study reported the incidence of each oral mucositis grade on multiple assessment days (Katranci 2012).

We wrote to all authors who reported means or on multiple assessment days or both, to request incidence data in the form of a single table reporting the maximum oral mucositis score (on a 0 to 4 scale) experienced by each participant over the length of the study. The authors of two studies supplied us with the data in this format (Salvador 2012; Svanberg 2007), and another author supplied us with incidence data, but on multiple assessment days (Askarifar 2015). We decided to use the data on the day with the highest incidence of grades > 0 (day 7) because, in this study, this probably most closely equated to the maximum score experienced per participant (as reported in the majority of other studies in the meta-analyses). For the other study reporting incidence data (Katranci 2012), but on multiple assessment days, we received no response from the authors, so we again used the data on the day with the highest incidence of grades > 0 (day 14), although we were not sure how valid this was because there was still a high incidence of severe oral mucositis at day 21. For mean scores on multiple assessment days, we used the day with the highest control group mean and recorded the results in an additional table.

The majority of studies assessed oral mucositis at multiple timepoints over approximately four weeks (Cascinu 1994; Gori 2007; Lilleby 2006; Mahood 1991; Rocke 1993; Sorensen

50 2008; Toro 2013), with some assessing it at multiple timepoints over three weeks (Askarifar 2015; Heydari 2012; Katranci 2012; Svanberg 2007), and the remaining studies assessing it at multiple timepoints over 10 to 14 days (Kakoei 2013; Salvador 2012; Zhang 2011). Where participants had multiple cycles of treatment, we only reported the results for the first cycle.

When studies reported oral mucositis data assessed separately by physicians and participants, we generally chose to use the physician-judgement as we felt that this may have been the more objective of the two, and therefore potentially less biased. There was one exception where only the participant-judged data was reported in the study as there was no significant difference between that and the physician-judgement (Sorensen 2008).

To summarise, we were able to include the primary outcome data for 13 of the included studies: 11 in the comparison of cryotherapy versus control (Askarifar 2015; Cascinu 1994; Gori 2007; Heydari 2012; Katranci 2012; Lilleby 2006; Mahood 1991; Salvador 2012; Sorensen 2008; Svanberg 2007; Toro 2013); one in the comparison of different durations of cryotherapy (Rocke 1993); and one in the comparison of cryotherapy versus leucovorin rinses (Zhang 2011).

Secondary outcomes Interruptions to cancer treatment

No studies reported this important outcome, but the authors of one study responded to our email request and provided us with both dichotomous (event) data and continuous data, the latter in the form of 'days of interruption' (Heydari 2012).

Oral pain

Four studies reported oral pain (Kakoei 2013; Lilleby 2006; Salvador 2012; Svanberg 2007), another study measured oral pain but the study report was currently being written and the data were undergoing analysis (Toro 2013), and the authors of a further study responded to our email request and provided us with oral pain data (Heydari 2012). Only two of these five studies reported data that we were able to combine in a meta-analysis (Lilleby 2006; Salvador 2012), which was the mean oral pain score on a 0 to 10 scale for the study period. Another study also measured oral pain on a 0 to 10 scale, but did not report any usable data in the form of mean and standard deviation for the study period (Svanberg 2007). One study reported mean pain scores on multiple assessment days but did not describe the scale used, and therefore we were unable to use the data for meta-

51 analysis (Kakoei 2013). The authors of one study provided us with data measured on a 1 to 5 scale representing the duration of time for which pain was experienced, unlike the other studies where the score represented pain intensity (Heydari 2012). We decided to present these data in an additional table. It was not clear what scale was used in the remaining study for which the data were not yet available (Toro 2013).

Quality of life

No studies reported this outcome. The authors of one study have confirmed that quality of life was measured using the Patient-Reported Oral Mucositis Symptom (PROMS) scale, but the data were currently unavailable as the study report and analysis had not yet been completed (Toro 2013).

Normalcy of diet

Two studies reported the duration, measured in days, of total parenteral nutrition (TPN) (Lilleby 2006; Svanberg 2007). However, one of these studies reported a median and range for each group, rather than mean and standard deviation, so we were unable to use the data in our analysis (Lilleby 2006). Another study reported the functional intake of food and fluids on a 1 to 5 scale, but we were unable to use the data that were presented in a mixed- effect regression model (Salvador 2012).

Adverse events

It was difficult to assess adverse effects in many cases due to the difficulty to distinguish between effects caused by oral cryotherapy and those caused by various cancer treatments. It also did not make sense to formally meta-analyse data from comparisons of oral cryotherapy with no treatment. Furthermore, in the context of cancer treatment, and with no serious effects being expected from sucking ice, we considered this outcome to be of limited importance in this particular review (whereas it may be more important in our reviews of other potentially more harmful interventions for preventing oral mucositis e.g. growth factors and cytokines). We therefore decided to report this outcome in a narrative format.

Four studies did not mention adverse effects (Askarifar 2015; Gori 2007; Kakoei 2013; Zhang 2011), whilst the quality of reporting for this outcome varied between the remaining nine studies.

Number of days in hospital 52 Three studies reported this outcome (Lilleby 2006; Salvador 2012; Svanberg 2007). We were able to combine the data for two studies, but the third study reported the median and range for each group, and we were therefore unable to use the data (Lilleby 2006).

Number of days of treatment with opioid analgesics

Two studies reported this outcome (Lilleby 2006; Svanberg 2007), but one reported the median and range for each group, and we were therefore unable to use the data (Lilleby 2006). A further study had measured 'narcotic use' but the data were currently unavailable as the study report and analysis had not yet been completed, and we did not know how this outcome was measured (Toro 2013).

Number of days unable to take medicine orally

No studies reported this outcome.

Excluded studies

We excluded 11 studies from this review for the following reasons.

 Not a randomised controlled trial (RCT): non-random allocation to groups (Karagozoglu 2005 (64); Papadeas 2007 (86); Sato 2006 (87)); used historical controls (Aisa 2005 (88); Mori 2006 (89)); translation to English confirmed it was case series (Ohyama 1994 (90)).  Unclear if RCT and authors have not provided clarification (Sato 1997 (91)).  Cross-over study with no reporting of first-period data (Baydar 2005 (92); Castelino 2011 (93); Nikoletti 2005 (94)).  Confounding due to co-interventions and therefore not possible to attribute effect to cryotherapy (de Paula Eduardo 2014 (95)).

3.5.2 Risk of bias in included studies

(A risk of bias table is presented for each included study in Appendix 2.) Allocation (selection bias) Random sequence generation

Eight studies described an adequate method of random sequence generation (Askarifar 2015; Gori 2007; Heydari 2012; Kakoei 2013; Katranci 2012; Salvador 2012; Svanberg 2007; Zhang 2011), and the authors of three further studies responded to our requests for further information (Cascinu 1994; Lilleby 2006; Toro 2013), which clarified that their 53 methods were adequate. Therefore we assessed 11 studies as being at low risk of bias for this domain. The remaining three studies only stated that participants were randomised but did not describe their methods, so they were assessed as being at unclear risk of bias for this domain (Mahood 1991; Rocke 1993; Sorensen 2008).

Allocation concealment

Two studies provided details of how the random sequence was concealed from those involved in the study (Salvador 2012; Svanberg 2007), with a further five studies providing details through correspondence (Askarifar 2015; Cascinu 1994; Heydari 2012; Lilleby 2006; Toro 2013). Therefore we assessed seven studies as being at low risk of bias for this domain. The remaining seven studies did not mention any methods used to conceal the random sequence, and we assessed them as being at unclear risk of bias (Gori 2007; Kakoei 2013; Katranci 2012; Mahood 1991; Rocke 1993; Sorensen 2008; Zhang 2011).

Overall, seven studies were at low risk of selection bias, meaning that we assessed both of the above domains as being at low risk of bias (Askarifar 2015; Cascinu 1994; Heydari 2012; Lilleby 2006; Salvador 2012; Svanberg 2007; Toro 2013). The remaining seven studies were at unclear risk of selection bias because one or both of the above domains were rated as unclear (Gori 2007; Kakoei 2013; Katranci 2012; Mahood 1991; Rocke 1993; Sorensen 2008; Zhang 2011).

Blinding (performance bias and detection bias) Blinding of participants and personnel (performance bias)

It is not possible to blind participants or personnel to whether or not oral cryotherapy has been allocated. Knowledge of treatment allocation could affect expectations and behaviours (e.g. control group participants may take other interventions). Therefore we judged all 14 studies to be at high risk of performance bias.

Blinding of outcome assessment (detection bias)

The subjective elements involved in the World Health Organisation (WHO) and National Cancer Institute common toxicity criteria (NCI-CTC) tools, which were used to measure oral mucositis in the studies included in this review (Appendix 3), required the patient's assessment of pain/soreness and their ability to swallow. Even if blinding of outcome assessment was attempted, we did not believe it was feasible. Therefore we judged all 14 studies to be at high risk of detection bias.

54 Incomplete outcome data (attrition bias)

Only two studies were at high risk of attrition bias (Askarifar 2015; Sorensen 2008), both due to differential attrition between groups, which may have been linked to prognosis. The remaining 12 studies had negligible or no attrition and we assessed them as being at low risk of attrition bias.

Selective reporting (reporting bias)

Three studies were at high risk of selective reporting bias. One of these studies reported oral pain, which was not stated in the trials registry record or the methods section of the study report (Kakoei 2013). It was possible that the decision to report this outcome was based on statistical significance. Another study stated only that there was no significant differences for oral pain, but reported no data (Svanberg 2007). The remaining study only reported the incidence of each grade of oral mucositis on day 4, despite stating that it was measured on the day of chemotherapy and day 10 (Zhang 2011). We assessed the other 11 studies as being at low risk of selective reporting bias as we could detect no obvious problems.

Other potential sources of bias

We did not consider there to be any issues arising from other potential sources of bias in any of the studies and we therefore assessed them all as being at low risk of bias for this domain.

Overall risk of bias

All fourteen studies were assessed as being at high overall risk of bias (Figure 4).

Figure 4: Risk of bias summary: review authors’ judgements about each risk of bias item for each included study

56 3.5.3 Effects of interventions

Comparison 1: Cryotherapy versus control (standard care or no treatment) Oral mucositis (any) Fluorouracil (5FU) treatment for solid cancers

Five studies (Cascinu 1994; Heydari 2012; Katranci 2012; Mahood 1991; Sorensen 2008), all at high risk of bias and analysing 444 participants, were combined in a meta-analysis which showed that oral cryotherapy reduced the risk of developing oral mucositis (risk ratio (RR) 0.61, 95% confidence interval (CI) 0.52 to 0.72, P < 0.00001) when compared to control (Analysis 1.1). There was no heterogeneity in this effect estimate (I2 = 0%). This can be interpreted as a 39% (95% CI 28% to 48%) reduction in the risk of developing oral mucositis associated with oral cryotherapy. Or in other words, 4 people (95% CI 3 to 5 people) would need to receive oral cryotherapy in order to prevent one additional person from developing oral mucositis.

High-dose melphalan-based treatment prior to stem cell transplantation

Five studies (Askarifar 2015; Lilleby 2006; Salvador 2012; Svanberg 2007; Toro 2013), all at high risk of bias and analysing 270 participants, were combined in a meta-analysis. There is evidence to show that oral cryotherapy reduced the risk of developing oral mucositis (RR 0.59, 95% CI 0.35 to 1.01, P = 0.05) when compared to control (Analysis 1.1). Oral cryotherapy reduced the risk of developing oral mucositis by 41% although the uncertainty surrounding this effect estimate means that the 95% CI ranges from a 65% reduction to a 1% increase in risk. We would need to treat 3 people with oral cryotherapy to prevent one additional person from developing oral mucositis. The confidence interval ranges from 2 people needing to receive oral cryotherapy to prevent one additional person from developing oral mucositis, to 111 people needing to receive oral cryotherapy for one additional person to develop oral mucositis. There was considerable heterogeneity associated with this effect estimate (I2 = 95%), although the reason/s for this inconsistency is/are not clear on investigation of the characteristics of the studies, in terms of the participants, settings, methods, and interventions.

Methotrexate treatment post-stem cell transplantation to prevent graft versus host disease

One study (Gori 2007), at high risk of bias and analysing 122 participants, showed that there was insufficient evidence to determine whether or not oral cryotherapy reduced the

57 risk of developing oral mucositis (RR 0.98, 95% CI 0.90 to 1.07, P = 0.73) when compared to control (Analysis 1.1).

Radiotherapy to the head and neck

One study (Kakoei 2013), at high risk of bias and analysing 40 participants, showed that there was insufficient evidence to determine whether or not oral cryotherapy reduced the mean severity of oral mucositis 14 days after radiotherapy (mean difference (MD) -0.25, 95% CI -0.72 to 0.22, P = 0.29) when compared to control (Appendix 4).

Analysis 1.1: Cryotherapy versus control; Outcome: Oral mucositis (any) 59 Oral mucositis (moderate to severe) Fluorouracil (5FU) treatment for solid cancers

The same five studies (Cascinu 1994; Heydari 2012; Katranci 2012; Mahood 1991; Sorensen 2008), when combined, showed that oral cryotherapy reduced the risk of developing moderate to severe oral mucositis (RR 0.52, 95% CI 0.41 to 0.65, P < 0.00001) when compared to control (Analysis 1.2). There was no heterogeneity (I2 = 0%). Oral cryotherapy reduced the risk of developing moderate to severe oral mucositis by 48% (95% CI 35% to 59%). We would need to treat 4 people (95% CI 4 to 6 people) with oral cryotherapy in order to prevent one additional person from developing this outcome.

High-dose melphalan-based treatment prior to stem cell transplantation

The same five studies (Askarifar 2015; Lilleby 2006; Salvador 2012; Svanberg 2007; Toro 2013), when combined, showed that there was evidence to show that oral cryotherapy reduced the risk of developing moderate to severe oral mucositis (RR 0.43, 95% CI 0.17 to 1.09, P = 0.07) when compared to control (Analysis 1.2). Oral cryotherapy reduced the risk of developing moderate to severe oral mucositis by 57% although the uncertainty surrounding this effect estimate means that the 95% CI ranges from a 83% reduction to a 9% increase in risk. We would need to treat 3 people with oral cryotherapy to prevent one additional person from developing moderate to severe oral mucositis. The confidence interval ranges from 2 people needing to receive oral cryotherapy to prevent one additional person from developing moderate to severe oral mucositis, to 17 people needing to receive oral cryotherapy for one additional person to develop moderate to severe oral mucositis. There was considerable heterogeneity associated with this effect estimate (I2 = 92%), although the reason/s for this inconsistency is/are not clear on investigation of the characteristics of the studies.

Methotrexate treatment post-stem cell transplantation to prevent graft versus host disease

The same study (Gori 2007) showed that there was insufficient evidence to determine whether or not oral cryotherapy reduced the risk of developing moderate to severe oral mucositis (RR 1.01, 95% CI 0.85 to 1.20, P = 0.93) when compared to control (Analysis 1.2).

Radiotherapy to the head and neck

The study in this subgroup did not assess this outcome.

Analysis 1.2: Cryotherapy versus control; Outcome: Oral mucositis (moderate to severe) 61 Oral mucositis (severe) Fluorouracil (5FU) treatment for solid cancers

The same five studies (Cascinu 1994; Heydari 2012; Katranci 2012; Mahood 1991; Sorensen 2008), when combined, showed that oral cryotherapy reduced the risk of developing severe oral mucositis (RR 0.40, 95% CI 0.27 to 0.61, P < 0.0001) when compared to control (Analysis 1.3). There was no heterogeneity (I2 = 0%). Oral cryotherapy reduced the risk of developing severe oral mucositis by 60% (39% to 73%). We would need to treat 6 people (95% CI 5 to 9 people) with oral cryotherapy to prevent one additional person from developing severe oral mucositis.

High-dose melphalan-based treatment prior to stem cell transplantation

The same five studies (Askarifar 2015; Lilleby 2006; Salvador 2012; Svanberg 2007; Toro 2013), when combined, showed that oral cryotherapy reduced the risk of developing severe oral mucositis (RR 0.38, 95% CI 0.20 to 0.72, P = 0.003) when compared to control (Analysis 1.3). Oral cryotherapy reduced the risk of developing severe oral mucositis by 62% (95% CI 28% to 80%). We would need to treat 4 people (95% CI 3 to 9 people) with oral cryotherapy to prevent one additional person from developing severe oral mucositis. There was moderate amount of heterogeneity associated with this effect estimate (I2 = 42%), although the reason/s for this inconsistency is/are not clear on investigation of the characteristics of the studies, in terms of the participants, settings, methods, and interventions.

Methotrexate treatment post-stem cell transplantation to prevent graft versus host disease

The same study (Gori 2007) showed that there was insufficient evidence to determine whether or not oral cryotherapy reduced the risk of developing moderate to severe oral mucositis (RR 0.88, 95% CI 0.61 to 1.25, P = 0.47) when compared to control (Analysis 1.3).

Radiotherapy to the head and neck

The study in this subgroup did not assess this outcome.

Analysis 1.3: Cryotherapy versus control; Outcome: Oral mucositis (severe) 63 Interruptions to cancer treatment Fluorouracil (5FU) treatment for solid cancers

One study (Heydari 2012), at high risk of bias and analysing 80 participants, showed that oral cryotherapy reduced the risk of treatment interruption (RR 0.44, 95% CI 0.20 to 0.95, P = 0.04) when compared to control (Appendix 5: Analysis 1.4). In other words, oral cryotherapy reduced the risk of treatment interruption by 56% (95% CI 5% to 80%). We would need to treat 5 people (95% 4 to 50 people) with oral cryotherapy in order to prevent one additional person from having a treatment interruption. The uncertainty, due to the single study with a small number of participants, surrounding this effect estimate was reflected by the wide range in the confidence interval.

The same study showed that oral cryotherapy reduced the duration of treatment interruption by eight days (95% CI 9.26 to 6.74 days, P < 0.00001) when compared to control (Appendix 5: Analysis 1.5).

High-dose melphalan-based treatment prior to stem cell transplantation

No studies in this subgroup assessed this outcome.

Graft versus host disease prophylaxis

The study in this subgroup did not assess this outcome.

Radiotherapy to the head and neck

The study in this subgroup did not assess this outcome

Oral pain

Fluorouracil (5FU) treatment for solid cancers

One study (Heydari 2012), at high risk of bias and analysing 80 participants, showed that oral cryotherapy reduced the duration of oral pain experience (MD -1.93, 95% CI -2.37 to - 1.49, P < 0.00001) when compared to control (Appendix 4). The 1 to 5 scale used to measure this outcome made it impossible to accurately interpret this result in any meaningful way.

64 High-dose melphalan-based treatment prior to stem cell transplantation

Two studies (Lilleby 2006; Salvador 2012), both at high risk of bias and analysing 85 participants, showed that oral cryotherapy reduced the intensity of oral pain (MD -1.50, 95% CI -2.11 to -0.89, P < 0.00001) when compared to control (Appendix 5: Analysis 1.6). There was a negligible amount of heterogeneity associated with this effect estimate (I2 = 4%). This result represents a 70% reduction in oral pain but this figure may be misleading as there was only a 1.5 point reduction on a 0 to 10 scale.

Graft versus host disease prophylaxis

The study in this subgroup did not assess this outcome.

Radiotherapy to the head and neck

The study in this subgroup assessed oral pain but did not report details of the scale used, so we were unable to use the data

Quality of life

Fluorouracil (5FU) treatment for solid cancers

No studies in this subgroup assessed this outcome.

High-dose melphalan-based treatment prior to stem cell transplantation

One study assessed this outcome but the data are currently unavailable as the study report and analysis have not yet been completed (Toro 2013).

Graft versus host disease prophylaxis

The study in this subgroup did not assess this outcome.

Radiotherapy to the head and neck

The study in this subgroup did not assess this outcome.

Normalcy of diet

65 Fluorouracil (5FU) treatment for solid cancers

No studies in this subgroup assessed this outcome.

High-dose melphalan-based treatment prior to stem cell transplantation

One study, at high risk of bias and analysing 78 participants (Svanberg 2007), showed that oral cryotherapy reduced the duration of total parenteral nutrition by 2.18 days (95% CI 0.03 to 4.33 days, P = 0.05) when compared to control (Appendix 5: Analysis 1.7).

A further study, at high risk of bias and analysing 40 participants (Lilleby 2006), reported that oral cryotherapy reduced the number of days of total parenteral nutrition (Appendix 4).

Graft versus host disease prophylaxis

The study in this subgroup did not assess this outcome.

Radiotherapy to the head and neck

The study in this subgroup did not assess this outcome

Adverse events

Fluorouracil (5FU) treatment for solid cancers

One study reported that there were no adverse effects ("problems") (Katranci 2012), whilst the other four studies reported low rates of very minor adverse events (headache, chills, numbness/taste disturbance).

High-dose melphalan-based treatment prior to stem cell transplantation

One study did not report adverse events (Askarifar 2015), whilst another informed us by email that there were "no serious adverse events" (Toro 2013). The remaining two studies reported only low rates of shooting pain from the teeth (Svanberg 2007), and coldness (Lilleby 2006). In the latter study, this stopped some participants from continuing the cryotherapy regimen, although this is not surprising as they were required to hold ice in their mouths continuously for a total of seven hours.

66 Graft versus host disease prophylaxis

The study in this subgroup did not report on adverse events.

Radiotherapy to the head and neck

The study in this subgroup did not report on adverse events. However, we assume that there were none because the participants only had to hold ice in their mouth for 5-minute periods.

Number of days in hospital Fluorouracil (5FU) treatment for solid cancers

No studies in this subgroup assessed this outcome.

High-dose melphalan-based treatment prior to stem cell transplantation

Two studies (Salvador 2012; Svanberg 2007), both at high risk of bias and analysing 123 participants, were combined in a meta-analysis with no heterogeneity (I2 = 0%). There was insufficient evidence to show that oral cryotherapy reduced the duration of hospitalisation (MD -1.39 days, 95% CI -2.97 to 0.19 days, P = 0.09) when compared to control (Appendix 5: Analysis 1.8).

A further study, at high risk of bias and analysing 40 participants (Lilleby 2006), reported that there was insufficient evidence to show a reduction in the duration of hospitalisation (Appendix 4).

Graft versus host disease prophylaxis

The study in this subgroup did not assess this outcome.

Radiotherapy to the head and neck

The study in this subgroup did not assess this outcome

Number of days of treatment with opioid analgesics

Fluorouracil (5FU) treatment for solid cancers

67 No studies in this subgroup assessed this outcome.

High-dose melphalan-based treatment prior to stem cell transplantation

One study (Svanberg 2007), at high risk of bias and analysing 78 participants, showed that there was insufficient evidence to determine whether or not oral cryotherapy reduced the duration of opioid use (MD -2.28 days, 95% CI -5.33 to 0.77 days, P = 0.14) when compared to control (Appendix 5: Analysis 1.9).

A further study (Lilleby 2006), at high risk of bias and analysing 40 participants, reported that oral cryotherapy reduced the duration of opioid use (Appendix 4).

Graft versus host disease prophylaxis

The study in this subgroup did not assess this outcome.

Radiotherapy to the head and neck

The study in this subgroup did not assess this outcome

Number of days unable to take medicine orally

Fluorouracil (5FU) treatment for solid cancers

No studies in this subgroup assessed this outcome.

High-dose melphalan-based treatment prior to stem cell transplantation

No studies in this subgroup assessed this outcome.

Graft versus host disease prophylaxis

The study in this subgroup did not assess this outcome.

Radiotherapy to the head and neck

The study in this subgroup did not assess this outcome.

68 Comparison 2: Different oral cryotherapy regimens

One study (Rocke 1993), at high risk of bias and analysing 178 participants having treatment for solid cancer (fluorouracil and leucovorin), showed that there was insufficient evidence to determine whether or not the risk of developing oral mucositis was different when using 30 minutes or 60 minutes of oral cryotherapy (RR 0.89, 95% CI 0.62 to 1.29, P = 0.54) (Analysis 2.1).

The results were similar for the risk of developing moderate to severe oral mucositis (RR 0.68, 95% CI 0.36 to 1.30, P = 0.25) (Analysis 2.2), and severe oral mucositis (RR 0.60, 95% CI 0.23 to 1.58, P = 0.30) (Analysis 2.3).

The study in this comparison did not assess any other outcomes of this review.

Analysis 2.1: Different oral cryotherapy regimes; Outcome: Oral mucositis (any)

Analysis 2.2: Different oral cryotherapy regimes; Outcome: Oral mucositis (moderate to severe)

Analysis 2.3: Different oral cryotherapy regimes; Outcome: Oral mucositis (severe)

71 Comparison 3: Cryotherapy versus chlorhexidine

One study (Sorensen 2008), at high risk of bias and analysing 133 participants having treatment for solid cancer (fluorouracil and leucovorin), showed that there was insufficient evidence to determine whether or not the risk of developing oral mucositis was different when using oral cryotherapy or chlorhexidine rinse (RR 0.97, 95% CI 0.71 to 1.32, P = 0.84) (Analysis 3.1).

The results were similar for the risk of developing moderate to severe oral mucositis (RR 0.89, 95% CI 0.51 to 1.56, P = 0.68) (Analysis 3.2), and severe oral mucositis (RR 0.86, 95% CI 0.34 to 2.18, P = 0.76) (Analysis 3.3).

The study in this comparison did not assess any other outcomes of this review.

Analysis 3.1: Cryotherapy versus chlorhexidine; Outcome: Oral mucositis (any)

Analysis 3.2: Cryotherapy versus chlorhexidine; Outcome: Oral mucositis (moderate to severe)

Analysis 3.3: Cryotherapy versus chlorhexidine; Outcome: Oral mucositis (severe)

74 Comparison 4: Cryotherapy versus low-dose leucovorin versus high-dose leucovorin

One study (Zhang 2011), at high risk of bias and analysing 147 participants having treatment for solid (osteosarcoma) cancer (high-dose methotrexate, vincristine and leucovorin), showed that oral cryotherapy reduced the risk of developing oral mucositis when compared to both low-dose leucovorin rinse (RR 0.67, 95% CI 0.50 to 0.90, P = 0.008) and high-dose leucovorin rinse (RR 0.65, 95% CI 0.47 to 0.90, P = 0.01) (Analysis 4.1).

There was insufficient evidence to determine whether or not the risk of developing moderate to severe oral mucositis was different when using oral cryotherapy or low-dose leucovorin rinse (RR 0.18, 95% CI 0.01 to 3.42, P = 0.25), or when using oral cryotherapy compared to high-dose leucovorin rinse (not estimable - no participants had developed moderate or severe oral mucositis by the fourth day after chemotherapy when the data were reported) (Analysis 4.2).

There was insufficient evidence to determine whether or not the risk of developing severe oral mucositis was different when using oral cryotherapy compared to either low-dose or high-dose leucovorin rinse. No participants had developed severe oral mucositis by the fourth day after chemotherapy when the data were reported.

The study in this comparison did not assess any other outcomes of this review.

Analysis 4.1: Cryotherapy versus leucovorin; Outcome: Oral mucositis (any)

Analysis 4.2: Cryotherapy versus leucovorin; Outcome: Oral mucositis (moderate to severe)

77 Compliance

Two studies did not report on compliance (Askarifar 2015; Zhang 2011). The remaining studies reported a high degree of compliance, with the large majority of participants managing to keep their mouths constantly cooled.

78 3.6 Summary of findings tables

1 Cryotherapy versus control for preventing oral mucositis in adults receiving fluorouracil-based treatment for solid cancers

Cryotherapy versus control for preventing oral mucositis in adults receiving fluorouracil-based treatment for solid cancers Patient or population: adults** with solid cancers receiving fluorouracil-based cancer treatment Setting: hospital Intervention: cryotherapy Comparison: control (no treatment or routine care) Outcomes Anticipated absolute effects* (95% Relative No of Quality of the Comments CI) effect participants evidence Risk with Risk with (95% (studies) (GRADE) control cryotherapy CI) Oral mucositis (any) 728 per 444 per 1000 RR 0.61 444 ⊕⊕⊕⊝ Oral cryotherapy reduced the risk of 1000 (379 to 524) (0.52 to (5 RCTs) MODERATE developing oral mucositis by 39% (95% 0.72) 1 CI 28% to 48%). We would need to treat (NNTB) 4 people (95% CI 3 to 5 people) with oral cryotherapy to prevent 1 additional person from developing oral mucositis Oral mucositis (moderate + 530 per 276 per 1000 RR 0.52 444 ⊕⊕⊕⊝ Oral cryotherapy reduced the risk of severe) 1000 (217 to 344) (0.41 to (5 RCTs) MODERATE developing moderate to severe oral 0.65) 1 mucositis by 48% (95% CI 35% to 59%). NNTB = 4 (95% CI 4 to 6) Oral mucositis (severe) 300 per 120 per 1000 RR 0.40 444 ⊕⊕⊕⊝ Oral cryotherapy reduced the risk of 1000 (81 to 183) (0.27 to (5 RCTs) MODERATE severe oral mucositis by 60% (95% CI 0.61) 1 39% to 73%). NNTB = 6 (95% CI 5 to 9)

79 Interruptions to cancer 400 per 176 per 1000 RR 0.44 80 ⊕⊝⊝⊝ Oral cryotherapy reduced the risk of treatment 1000 (80 to 380) (0.20 to (1 RCT) VERY LOW 2 treatment interruption by 56% (95% CI 0.95) 3 4 5% to 80%). NNTB 5 (95% CI 4 to 50) Oral pain (1 to 5 scale: 1 = The mean The mean oral pain (1 - 80 ⊕⊝⊝⊝ The duration of oral pain experienced was never, 2 = 1 day of week, 3 oral pain (1 to 5 scale) in the (1 RCT) VERY LOW 2 less in the oral cryotherapy group (see = 2 to 3 days of week, 4 = to 5 scale) cryotherapy group was 3 4 Appendix 4) most of week, 5 = 7 days of was 3.64 1.93 lower (2.37 to week) 1.49 lower) Quality of life No studies assessed this outcome Normalcy of diet No studies assessed this outcome Duration of hospitalisation No studies assessed this outcome *The risk in the intervention group (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; NNTB: number needed to treat to benefit; RCT: randomised controlled trial; RR: risk ratio GRADE Working Group grades of evidence High quality: we are very confident that the true effect lies close to that of the estimate of the effect Moderate quality: we are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low quality: our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low quality: we have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

Footnotes 1 All 5 studies at high risk of performance and detection bias 2 Single study at high risk of performance and detection bias 3 Low sample size and wide confidence interval 4 Indirect due to single study in 1 setting (in terms of country, healthcare system, and participants potentially differing from other countries) ** There were no studies conducted on children

80 2 Cryotherapy versus control for preventing oral mucositis in adults receiving high-dose melphalan-based treatment prior to haematopoietic stem cell transplantation for haematological cancers

Cryotherapy versus control for preventing oral mucositis in adults receiving high-dose melphalan-based treatment prior to haematopoietic stem cell transplantation for haematological cancers Patient or population: adults** with haematological cancers receiving high-dose melphalan-based treatment prior to haematopoietic stem cell transplantation Setting: hospital Intervention: cryotherapy Comparison: control (no treatment or routine care) Outcomes Anticipated absolute effects* (95% CI) Relative No of Quality of the Comments Risk with Risk with cryotherapy effect participants evidence control (95% (studies) (GRADE) CI) Oral mucositis 824 per 1000 486 per 1000 RR 0.59 270 ⊕⊕⊝⊝ Oral cryotherapy appears to reduce the risk (any) (289 to 833) (0.35 to (5 RCTs) LOW 1 2 of developing oral mucositis. However, there 1.01) is great uncertainty about our estimate, and there is an extremely small chance of a 1% increase in the risk of developing oral mucositis compared to control. The point estimate suggests a 41% reduction in the risk of developing oral mucositis, with the confidence interval ranging from a 65% reduction to a 1% increase in risk. NNTB = 3 (95% CI 2 NNTB to 111 NNTH) Oral mucositis 679 per 1000 292 per 1000 RR 0.43 270 ⊕⊕⊝⊝ Oral cryotherapy appears to reduce the risk (moderate + (115 to 741) (0.17 to (5 RCTs) LOW 1 2 of developing moderate to severe oral severe) 1.09) mucositis. However, there is great

81 uncertainty about our estimate, and there is a very small chance of a 9% increase in the risk of developing moderate to severe oral mucositis compared to control. The point estimate suggests a 57% reduction in the risk of developing moderate to severe oral mucositis, with the confidence interval ranging from an 83% reduction to a 9% increase in risk. NNTB = 3 (95% CI 2 NNTB to 17 NNTH) Oral mucositis 427 per 1000 162 per 1000 RR 0.38 270 ⊕⊕⊕⊝ Oral cryotherapy reduced the risk of (severe) (85 to 308) (0.20 to (5 RCTs) MODERATE developing severe oral mucositis by 62% 0.72) 1 (95% CI 28% to 80%). NNTB = 4 (95% CI 3 to 9) Interruptions to No studies assessed this outcome cancer treatment Oral pain on a 0 The weighted The mean oral pain (0 to - 85 ⊕⊕⊝⊝ Oral cryotherapy reduced oral pain by 70% (no pain) to 10 mean oral pain (0 10 scale) in the (2 RCTs) LOW 3 4 although the clinical importance of a 1.5 (maximum pain) to 10 scale) was cryotherapy group was point-reduction on a 0 to 10 scale is scale 2.13 1.5 lower (2.11 to 0.89 questionable lower) Quality of life 1 study assessed this outcome but the data are currently unavailable as the study report and analysis have not yet been completed Normalcy of diet The mean number The mean number of - 78 ⊕⊝⊝⊝ Oral cryotherapy reduced the duration of (days of total of days of TPN days of TPN in the (1 RCT) VERY LOW 5 TPN by 2.18 days. There was some parenteral was 7 intervention group was 6 7 additional very low quality evidence, from a nutrition - TPN) 2.18 days fewer (4.33 to single small study at high risk of bias, 0.03 fewer) reporting only median, range and P value, that oral cryotherapy reduced the number of

82 days of TPN (see Appendix 4) Duration of The mean The mean duration of - 123 ⊕⊕⊝⊝ There is insufficient evidence to show that hospitalisation duration of hospitalisation (days) in (2 RCTs) LOW 3 4 oral cryotherapy reduces the duration of (days) hospitalisation the intervention group hospitalisation. This is supported by (days) was 0 was 1.39 undefined additional very low quality evidence, from a fewer (2.97 fewer to single small study at high risk of bias, 0.19 more) reporting only median, range and P value, that there is insufficient evidence to show a reduction in this outcome (see Appendix 4) *The risk in the intervention group (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; NNTB: number needed to treat to benefit; NNTH: number needed to treat to harm; RCT: randomised controlled trial; RR: risk ratio; TPN: total parenteral nutrition GRADE Working Group grades of evidence High quality: we are very confident that the true effect lies close to that of the estimate of the effect Moderate quality: we are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low quality: our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low quality: we have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

Footnotes 1 All 5 studies at risk of performance and detection bias, with 1 study at added risk of attrition bias 2 The I2 value indicates that a considerable amount (> 90%) of the variability in effect estimates is due to heterogeneity rather than chance 3 Both studies at high risk of performance and detection bias 4 Low sample size from 2 small studies 5 Single study at high risk of performance and detection bias 6 Low sample size and wide confidence interval 7 Indirect due to single study in 1 setting (in terms of country, healthcare system, and participants potentially differing from other countries) ** There were no studies conducted on children 83 3.7 Discussion

Summary of main results

Fourteen randomised controlled trials (RCTs) met our eligibility criteria and were included in this review. We assessed the body of evidence for each comparison and outcome using GRADE methodology (45). Most of the evidence we found relates to the comparison of oral cryotherapy versus a control group of standard care or no treatment in adult patients. We only present a 'Summary of findings' table where we were able to perform a meta- analysis for the main outcomes, thus we produced two separate tables for: 1) evidence for participants receiving fluorouracil-based (5FU) treatment for solid cancers (Summary of findings table 1); and 2) evidence for participants receiving high-dose melphalan-based cancer treatment prior to haematopoietic stem cell transplantation (HSCT) (Summary of findings table 2).

Our main findings are as follows.

Adults receiving 5FU-based treatment for solid cancers

 Oral cryotherapy probably reduces oral mucositis of all severities (moderate quality evidence).  There is some weak evidence that oral cryotherapy reduces both the incidence of treatment interruptions and the duration of treatment interruptions (very low quality evidence).  There is some weak evidence that oral cryotherapy reduces the duration of pain experience (very low quality evidence).

Adults receiving high-dose melphalan prior to HSCT

 Oral cryotherapy may reduce the incidence of both oral mucositis (any severity versus none) and moderate to severe oral mucositis (low quality evidence), and probably does reduce the incidence of severe oral mucositis (moderate quality evidence).  Oral cryotherapy may lead to a small reduction in oral pain (low quality evidence).  There is some weak evidence that oral cryotherapy reduces the duration of total parenteral nutrition (TPN) (very low quality evidence).  There is insufficient evidence to determine whether or not oral cryotherapy reduces the number of days in hospital (low quality evidence).

84  There is insufficient evidence to determine whether or not oral cryotherapy reduces the number of days of treatment with opioid analgesics (very low quality evidence).

There is insufficient evidence to determine the effects of oral cryotherapy in: a) people receiving low-dose methotrexate for preventing graft-versus-host disease (GVHD) after HSCT; and b) people receiving head and neck radiotherapy (both very low quality evidence).

For the other comparisons included in this review, there was only very low quality evidence provided by single studies. In participants receiving 5FU-based treatment for solid cancers, there is insufficient evidence to determine whether or not there is a difference between: a) 30 minutes and 60 minutes of oral cryotherapy; or b) oral cryotherapy and chlorhexidine rinse, for reducing the risk of developing any severity of oral mucositis. In participants receiving high-dose methotrexate, vincristine and leucovorin for bone cancer (osteosarcoma), there is some weak evidence that oral cryotherapy reduces the risk of developing oral mucositis when compared to leucovorin rinses.

Oral cryotherapy is safe, with low rates of very minor adverse effects, which ensures that compliance with this therapy is generally high.

Overall completeness and applicability of evidence

We have found evidence which partially answers the review question of whether or not oral cryotherapy can prevent oral mucositis in cancer patients who are receiving treatment. We have enough evidence to answer this question for adults receiving 5FU-based treatment for solid cancers in a satisfactory manner. However, for adults with haematological cancers receiving melphalan-based cancer treatment prior to HSCT, the evidence is not complete and may be difficult to apply to different settings due to conflicting results amongst the individual studies. We would welcome more evidence in this group of patients and in a variety of settings. There is also very limited evidence on people receiving chemotherapy as GVHD prophylaxis, and also people receiving radiotherapy, whilst there is no RCT evidence on people receiving targeted therapy. The rationale behind the biological plausibility for using oral cryotherapy in cancer patients receiving chemotherapy has been explained in the literature. However, this is not the case for patients receiving radiotherapy to the head and neck or targeted therapy, which makes it difficult to determine the procedure (i.e. when and for how long to perform cryotherapy), yet it is still worth further investigation if there is a chance that this simple intervention can help these patients. 85 There is no evidence on the effects of this intervention in children, perhaps because it may be considered unfeasible to expect children to hold ice in their mouths for long periods of time, due to the discomfort caused. However, as demonstrated in some of the studies included in this review, there are alternatives to holding crushed ice or ice chips or ice cubes in the mouth, such as the use of iced water/drinks or ice lollies/popsicles. It is also possible that ice chips or cubes may be a potential choking hazard in children.

There is also a lack of evidence on the effects of oral cryotherapy in low-income and lower-middle-income countries.

Quality of the evidence

We included 14 RCTs that analysed 1280 participants. The body of evidence that we identified on adults receiving 5FU-based treatment for solid cancers allows a fairly robust conclusion to be made about the effects of oral cryotherapy for reducing the risk of oral mucositis in that group of patients. That is to say we can be quite confident in the results and we therefore assessed the evidence as being of moderate quality. We downgraded the rating of the evidence by one level due to the lack of blinding of participants, personnel and outcome assessors. Lack of blinding of participants and personnel in these studies is an unavoidable problem and does not necessarily reflect the 'quality' in terms of whether or not the studies were well conducted, although it does affect the risk of bias, which we must account for in our assessment of the quality of the body of evidence. Future studies may avoid the risk of detection bias by using a more objective measurement tool for the assessment of oral mucositis. However, this would affect our ability to combine the new studies in meta-analyses with the current published studies that all use either the WHO or NCI-CTC tools, which incorporate subjective elements.

The body of evidence that we identified on adults receiving high-dose melphalan-based cancer treatment prior to HSCT does not allow for such robust conclusions for the outcomes 'any mucositis' and 'moderate to severe mucositis'. There were the same problems with risk of bias as discussed above, and also inconsistency in the results of the individual studies, and we therefore downgraded our rating of the quality of the body of evidence by two levels, resulting in low quality evidence. Although we would conclude that oral cryotherapy may be beneficial in this group of patients, the results were very uncertain, with wide confidence intervals, and even the possibility of a slight increase in the risk of developing any or moderate to severe mucositis. However, we did not downgrade for precision as the upper confidence limits did not include appreciable harm,

86 and also because we do not consider it to be very low quality evidence. The evidence for oral cryotherapy reducing the risk of severe oral mucositis in this group of patients was more robust, resulting in moderate quality evidence.

We did not downgrade either of the above bodies of evidence for indirectness resulting from the studies only being conducted on adults. However, there is some evidence that, with some diseases, children have more toxicity than adults who receive the same chemotherapy regimens (96). Therefore, it is possible that results in adults are not directly applicable to children, who may metabolise drugs differently and therefore experience different toxicity severity.

There was insufficient evidence, of very low quality, on the effects of oral cryotherapy for reducing the risk of oral mucositis to allow any conclusions to be made for people receiving chemotherapy as GVHD prophylaxis, and also people receiving radiotherapy. This was downgraded for risk of bias, imprecision and indirectness, the latter because they were single study subgroups and were not replicated in any other settings. There was no evidence relating to children-only.

It was unfortunate that there was such limited evidence on the secondary outcomes of this review. These outcomes, and others, are likely to be important to patients and clinicians, and this serves to highlight the importance of core outcomes sets such as the one published by Bellm et al (71). We would urge future trialists to utilise such outcome sets in order to standardise what is measured and how it is measured, so as to maximise the usefulness of data produced by the randomisation of people, and thus reduce research waste.

Potential biases in the review process

We attempted to minimise biases in the review process but we cannot discount the fact that it is affected by other biases. For example, there were two studies that presented oral mucositis incidence data on multiple days rather than our preferred format of presenting a maximum score experienced per participant over the whole study period (Askarifar 2015; Katranci 2012). Rather than lose important data, we decided to present the data from the day on which the incidence of grades higher than 0 was highest. We chose this method with the intention of equating to the maximum score experience, though we acknowledge this is not ideal and may not equate accurately enough. Another decision we made was to exclude studies where the participants received oral cryotherapy plus another intervention, even if the control group also received the other intervention. For example, one study that we excluded compared oral cryotherapy plus laser therapy versus laser therapy (de Paula 87 Eduardo 2014). We could not exclude the possibility of interaction between the interventions and therefore we would not be confident in stating that any effect was due to the cryotherapy. However, in some of the studies, both groups received standard care, with the intervention group also receiving oral cryotherapy. We felt that this was a different situation because the standard care involved things like advice and education on mouth care, or mouth rinsing with saline, none of which would interact with the cryotherapy in a way that could bias the results.

Obtaining data from trial authors is also essential to reduce outcome reporting bias and we will attempt to encourage authors to reply to emails requesting further information in future updates in various ways. This will include providing copies of the previous version of the review and explaining the benefits of having their research included in a Cochrane review.

Agreements and disagreements with other studies or reviews

A recent meta-analysis conducted in China on oral cryotherapy for preventing oral mucositis in patients with haematological cancers undergoing HSCT (97), concluded that oral cryotherapy reduces the incidence of severe oral mucositis (risk ratio (RR) 0.52, 95% confidence interval 0.27 to 0.99). This was a smaller reduction than we calculated (RR 0.38, 95% CI 0.20 to 0.72), but they included one study that we excluded due to the possibility of confounding (de Paula Eduardo 2014), another study that we did not include in this subgroup due to treatment differences (Gori 2007), and another study that we have been unable to obtain the data for, which is currently listed in ‘Characteristics of studies awaiting classification’ in Appendix 2 (98). We included two studies that the Chinese authors did not include, for which we managed to obtain data through correspondence with the study authors (Askarifar 2015; Salvador 2012). The Chinese authors were also in agreement with our results on 'duration of opioid use' and 'days of TPN', although they included data from Lilleby 2006, which reported medians and ranges for some of our outcomes, and it was not clear what methods they used to include this study in the meta- analyses. The inclusion of Lilleby 2006 data in their meta-analysis for the outcome 'duration of hospitalisation' resulted in a significant result in favour of oral cryotherapy, in contrast to our result which did not achieve significance.

Our results strongly support the conclusions of another systematic review of oral cryotherapy that was carried out by the Mucositis Study Group (MSG) of the Multinational Association of Supportive Care in Cancer/International Society of Oral Oncology

88 (MASCC/ISOO), and that is not limited to RCTs (26). This review feeds into the 'MASCC/ISOO Clinical Practice Guidelines for the Management of Mucositis Secondary to Cancer Therapy (65). Specifically, the guidance from this group is currently as follows.

 Recommendations in favour of an intervention (i.e. strong evidence supporting effectiveness): The panel recommends that 30 minutes of oral cryotherapy be used to prevent oral mucositis in patients receiving bolus 5FU chemotherapy (level II evidence).  Suggestions in favour of an intervention (i.e. weaker evidence supporting effectiveness): The panel suggests that oral cryotherapy be used to prevent oral mucositis in patients receiving high-dose melphalan, with or without total body irradiation, as conditioning for HSCT (level III evidence).

This guidance is perhaps further strengthened by our Cochrane systematic review, as we have included new data/evidence not included in the MASCC/ISOO review or guidelines, yet the conclusions remain the same. It could even be argued that, due to the new data included in our two subgroups of patients matching those above, the level of evidence would now be level I for patients receiving bolus 5FU chemotherapy, and level I or II for patients receiving high-dose melphalan, with or without total body irradiation, as conditioning for HSCT. Level I evidence is "evidence obtained from meta-analysis of multiple, well-designed, controlled studies; randomized trials with low false-positive and false-negative errors (high power)" (65). There are now five RCTs in our 5FU subgroup with sufficient power to give a clear and fairly precise result. Our melphalan subgroup also now has five RCTs, but with fewer participants, giving a less precise result. However, the evidence for prevention of severe oral mucositis in this subgroup is fairly robust.

3.8 Authors' conclusions

Implications for practice

We are confident that oral cryotherapy leads to large reductions in the incidence of oral mucositis of all severities in adults receiving fluorouracil-based (5FU) treatment for solid cancers. We are less confident in the ability of oral cryotherapy to reduce the incidence of oral mucositis in adults receiving high-dose melphalan-based cancer treatment prior to haematopoietic stem cell transplantation (HSCT). Evidence suggests that it does reduce oral mucositis in these adults, but we are less certain about the size of the reduction, which

89 could be large or small. However, we are confident that there is an appreciable reduction in severe oral mucositis in these adults.

This Cochrane systematic review has included some very recent and currently unpublished data, and strengthens international guideline statements for adults receiving the above cancer treatments.

Implications for research

It is fairly clear that oral cryotherapy is beneficial for adults receiving 5FU chemotherapy, and therefore further randomised controlled trials (RCTs) are probably not warranted. As cryotherapy is non-invasive, instead it may be better to conduct new RCTs looking at using cryotherapy alongside other treatments.

We need more studies assessing oral cryotherapy in people receiving high-dose melphalan- based cancer treatment prior to HSCT in order to further strengthen the body of evidence reported in this review. Further investigation of the optimum cryotherapy regimen may be warranted in these patients and, as mentioned above, RCTs of cryotherapy alongside other treatments. We also need more studies looking at the effects of oral cryotherapy in people receiving other cancer treatments, such as different chemotherapy drugs (including those given as prophylaxis for graft-versus-host disease after HSCT), radiotherapy, and even targeted therapies. Studies of children may also be warranted if this intervention can help them complete their optimum cancer treatment regimens, whilst minimising their pain, discomfort and ability to eat and drink properly.

We urge trialists conducting future RCTs to measure and report all the outcomes, as a minimum, listed in the core outcome set produced by Bellm et al (71). For our primary outcome of oral mucositis, it would be beneficial to use both a measurement tool such as the WHO or NCI-NCT scale (see Appendix 3), to allow us to combine the data with those already included in this review, and also an objective scale with blinded outcome assessment, in order to reduce the bias inherent in these studies.

90 4 INTERVENTIONS FOR PREVENTING ORAL MUCOSITIS IN PATIENTS WITH CANCER RECEIVING TREATMENT: CYTOKINES AND GROWTH FACTORS

4.1 Abstract

Background

Oral mucositis is a side effect of chemotherapy, head and neck radiotherapy, and targeted therapy, affecting over 75% of high risk patients. Ulceration can lead to severe pain and difficulty eating and drinking, which may necessitate opioid analgesics, hospitalisation and supplemental nutrition. These complications may disrupt cancer therapy, which may reduce survival. There is also a risk of death from sepsis if pathogens enter the ulcers of immunocompromised patients. Ulcerative oral mucositis can be costly to healthcare systems, yet there are few preventive interventions proven to be beneficial. Cytokines and growth factors may help the regeneration of cells lining of the mouth, thus preventing or reducing oral mucositis and its negative effects.

Objectives

To assess the effects of cytokines and growth factors for preventing oral mucositis in patients with cancer who are receiving treatment.

Search methods

Cochrane Oral Health's information specialist searched the following databases: Cochrane Oral Health’s Trials Register (to 10 May 2017), the Cochrane Central Register of Controlled Trials (CENTRAL) (the Cochrane Library, 2017, Issue 4), MEDLINE Ovid (1946 to 10 May 2017), Embase Ovid (7 December 2015 to 10 May 2017), CINAHL EBSCO (Cumulative Index to Nursing and Allied Health Literature; 1937 to 10 May 2017) and CANCERLIT PubMed (1950 to 10 May 2017). The US National Institute of Health Ongoing Trials Register and the World Health Organization International Clinical Trials Registry Platform were searched for ongoing trials. No restrictions were placed on language or publication status when searching the electronic databases, however the search of Embase was restricted to the last 18 months due to the Cochrane Embase Project to identify all clinical trials and add them to CENTRAL.

Selection criteria 91 We included parallel-design randomised controlled trials (RCTs) assessing the effects of cytokines and growth factors in patients with cancer receiving treatment.

Data collection and analysis

Two review authors independently screened the results of electronic searches, extracted data and assessed risk of bias. For dichotomous outcomes, we reported risk ratios (RR) and 95% confidence intervals (CI). For continuous outcomes, we reported mean differences (MD) and 95% CIs. We pooled similar studies in random-effects meta-analyses. We reported adverse effects in a narrative format.

Main results

We included 35 RCTs analysing 3102 participants. Thirteen studies were at low risk of bias, 12 studies were at unclear risk of bias, and 10 studies were at high risk of bias.

Our main findings were regarding keratinocyte growth factor (KGF) and are summarised as follows.

There is probably a reduction in the risk of moderate to severe oral mucositis in adults receiving bone marrow/stem cell transplantation after conditioning therapy for haematological cancers (RR 0.89, 95% CI 0.80 to 0.99; 6 studies; 852 participants; moderate quality evidence). We would need to treat 11 adults with KGF in order to prevent one additional adult from developing this outcome (95% CI 6 to 112). There might be a reduction in the risk of severe oral mucositis in this population, but there is also some possibility of harm (RR 0.85, 95% CI 0.65 to 1.11; 6 studies; 852 participants; moderate quality evidence). We would need to treat 10 adults with KGF in order to prevent one additional adult from developing this outcome (95% CI 5 to prevent the outcome to 14 to cause the outcome).

There is probably a reduction in the risk of moderate to severe oral mucositis in adults receiving radiotherapy to the head and neck with cisplatin or fluorouracil (RR 0.91, 95% CI 0.83 to 1.00; 3 studies; 471 participants; moderate quality evidence). We would need to treat 12 adults with KGF in order to prevent one additional adult from developing this outcome (95% CI 7 to infinity). It is very likely that there is a reduction in the risk of severe oral mucositis in this population (RR 0.79, 95% CI 0.69 to 0.90; 3 studies; 471 participants; high quality evidence). We would need to treat 7 adults with KGF in order to prevent one additional adult from developing this outcome (95% CI 5 to 15).

92 It is very likely that there is a reduction in the risk of moderate to severe oral mucositis in adults receiving chemotherapy alone for mixed solid and haematological cancers (RR 0.56, 95% CI 0.45 to 0.70; 4 studies; 344 participants; high quality evidence). We would need to treat 4 adults with KGF in order to prevent one additional adult from developing this outcome (95% CI 3 to 6). There is probably a reduction in the risk of severe oral mucositis in this population (RR 0.30, 95% CI 0.14 to 0.65; 3 studies; 263 participants; moderate quality evidence). We would need to treat 10 adults with KGF in order to prevent one additional adult from developing this outcome (95% CI 8 to 19).

Due to the low volume of evidence, single-study comparisons and insufficient sample sizes, we found no compelling evidence of a benefit for any other cytokines or growth factors and there was no evidence on children. There did not appear to be any serious adverse effects of any of the interventions assessed in this review.

Authors' conclusions

We are confident that KGF reduces the risk of developing moderate to severe and severe oral mucositis in adults who are receiving: a) radiotherapy to the head and neck with cisplatin or fluorouracil; or b) chemotherapy alone for mixed solid and haematological cancers. We are less confident about a benefit for KGF in adults receiving bone marrow/stem cell transplant after conditioning therapy for haematological cancers because of multiple factors involved in that population, such as whether or not they received total body irradiation (TBI) and whether the transplant was autologous (the patients' own cells) or allogeneic (cells from a donor). KGF appears to be a relatively safe intervention.

Due to limited research, we are not confident that there are any beneficial effects of other cytokines and growth factors. There is currently insufficient evidence to draw any conclusions about the use of cytokines and growth factors in children.

4.2 Background

Description of the condition

Treating cancer with chemotherapy, radiotherapy of the head and neck, or targeted therapy can cause toxic oral side effects (1-3). Perhaps the most widely researched of these side effects is oral mucositis (1), which affects at least 75% of high risk patients (those receiving head and neck radiotherapy or high-dose chemotherapy) (2). Oral mucositis may 93 be under-reported in lower risk groups for various reasons: their tendency to be outpatients with less observation; less reporting of moderate mucositis; or patients and clinicians wishing to avoid any disruption to optimal cancer treatment (2).

94 radiotherapy-induced mucositis, and the clinical presentation of the ulcers is more similar to aphthous stomatitis (1, 7, 58).

Ulceration is the most significant phase as it leads to pain of varying severity, and difficulties with eating, swallowing, and talking (2). This in turn leads to the consumption of pain relief medication, nutritional support (i.e. nasogastric or intravenous feeding), treatment of the oral mucositis, specialist oral hygiene care, increased medical appointments and use of staff and resources, and, in some instances, hospitalisation (4-6). Thus the negative impact on the quality of life of cancer patients, when they are already suffering, is severe (59, 60). Further problems can occur in immunosuppressed patients if whole bacteria on the ulcer surface cross into the underlying submucosa, potentially leading to bacteraemia and sepsis, which require antibiotics and hospitalisation, and can cause death (2, 6, 7).

Therefore, oral mucositis can be a dose-limiting condition, disrupting a patient's optimal cancer treatment plan (3, 6, 7). The additional costs associated with oral mucositis are significant, with one study reporting a median incremental cost of USD 18,515 per patient (61). These costs have been reported to be as much as USD 42,749 more per patient when ulcerative oral mucositis is present (62).

Description of the intervention

As described above, oral mucositis occurs partly as result of the loss of regenerative ability of the oral epithelial cells. Growth factors and anti-inflammatory cytokines are used to counteract the biological processes leading to this loss of proliferative ability. Growth factors and anti-inflammatory cytokines include (22):

 keratinocyte growth factor;  colony-stimulating factors;  epidermal growth factor;  transforming growth factor-beta;  whey-derived growth factor;

95  interleukin-11;  ATL-104;  trefoil factor.

How the intervention might work

The growth factors described here are proteins that bind to receptors of target cells and either increase the proliferation of the epithelial cells that form the mucous membrane lining of the oral cavity, or promote the recovery of the white blood cells that contribute to the maintenance of oral health following conventional or high dose chemotherapy (with or without radiotherapy) (22). Anti-inflammatory cytokines are also proteins or glycoproteins that bind to receptors of target cells, and are thought to alter the complex balance of pro- and anti-inflammatory cytokines involved in the pathogenesis of oral mucositis (22).

Currently, evidence-based guidelines recommend growth factors for the prevention of oral mucositis in patients with haematological cancers undergoing high-dose chemotherapy and total body irradiation prior to haematopoietic stem cell transplantation (65). It has been postulated that tumour cells may also have receptors accommodating cytokines and growth factors, thus encouraging the proliferation of cancer cells in solid tumours (21, 99). A 2010 systematic review suggested that the risk of acute myeloid leukemia (AML) or myelodysplastic syndrome (MDS) is increased in people with various cancers receiving chemotherapy with granulocyte colony-stimulating factor (G-CSF) when compared to those receiving chemotherapy without G-CSF (100). The authors concluded that it was not clear whether the increased AML/MDS risk was due to G-CSF or due to the increased chemotherapy dose-intensity in those patients. However, the review also reported a reduction in overall mortality for those receiving G-CSF.

As there is currently no clear consensus on this issue, the use of cytokines and growth factors for the prevention of oral mucositis is not limited to haematological cancer patients in practice.

Why it is important to do this review

This review is part of a series that will replace the previously published Cochrane review covering all interventions for the prevention of oral mucositis in patients with cancer receiving treatment (14). The Mucositis Study Group (MSG) of the Multinational Association of Supportive Care in Cancer/International Society of Oral Oncology (MASCC/ISOO) is a group that was set up in 1998 for the purpose of producing

96 international evidence-based clinical practice guidelines for managing mucositis (both oral and gastrointestinal), which they first published in 2004, with the latest update published in 2014 (65). In order to facilitate the future updating of Cochrane reviews on this topic, and also to make them more usable to clinicians, guideline developers, and consumers, we have decided to divide the original Cochrane review into the same intervention categories as those used by MASCC/ISOO, which are as follows:

 basic oral care/good clinical practice;  growth factors and cytokines;  anti-inflammatory agents;  antimicrobials, mucosal coating agents, anaesthetics, and analgesics;  laser and other light therapy;  cryotherapy;  natural and miscellaneous agents;  amifostine.

We believe that following the MASCC/ISOO structure will better enable the Cochrane reviews to feed into such guidelines. We can also be more thorough and rigorous in our assessment and summarising of the evidence in each of the categories, which was not feasible in a single Cochrane review approaching 150 included studies.

4.3 Objectives

To assess the effects of cytokines and growth factors for preventing oral mucositis in patients with cancer who are receiving treatment.

4.4 Methods

Criteria for considering studies for this review

97 Types of studies

We included all randomised controlled trials (RCTs) of parallel design. It is possible to conduct cross-over studies in this area as patients may receive several treatment sessions/cycles, with any mucositis completely healing in the periods between the sessions. However, we did not include cross-over data as we could not discount any period effects, with mucositis risk increasing as patients receive further cycles of treatment (2, 8). Instead, we used the first-period data only and treated such studies as parallel studies.

Types of participants

We included all patients with cancer who were receiving treatment.

Types of interventions

We included studies comparing growth factors and cytokines for the prevention of oral mucositis (we would also have included targeted therapy-induced stomatitis) against usual care, no treatment, or any other treatment to prevent oral mucositis. We also included studies comparing different growth factors and cytokines or different regimens of growth factors and cytokines against each other (head-to-head studies).

We excluded studies with 'complex' interventions for the prevention of mucositis, such as lasers plus growth factors and cytokines versus lasers. We excluded studies assessing different cancer treatments where the primary outcome was survival/cure, with mucositis as a toxicity.

Types of outcome measures

We are in agreement with Williamson 2012 that, if clinical trials and systematic reviews are to be utilised, the outcomes assessed should be those considered important to patients, healthcare professionals, and other key stakeholders. If outcomes and outcome measures are inconsistent across studies, it will not be possible to compare and summarise research, and there is potential for outcome reporting bias, with the selective reporting of results based on statistical significance and favourability (68-70). This can lead to exaggerated estimates of effect in systematic reviews of interventions, leading to an incorrect belief that an intervention is more beneficial that it truly is (68). It is thought that the way to address this problem is to develop disease- or condition-specific core outcome sets to be used as a minimum when conducting and reporting clinical trials (67, 68).

98 Therefore we used the core outcome set produced by Bellm 2002, which is registered on the COMET (Core Outcome Measures in Effectiveness Trials) Initiative's website (www.comet-initiative.org), and is the only core outcome set for oral mucositis known to us (71).

Primary outcomes

Some studies measure mucositis using a composite scale. If it was possible to extract the 'mucositis only' data from the total score, we would have included the data in the analyses. If it was not possible, we would have recorded the composite data in an additional table.

Secondary outcomes

Search methods for identification of studies Electronic searches

Cochrane Oral Health’s Information Specialist conducted systematic searches in the following databases for randomised controlled trials and controlled clinical trials without language or publication status restrictions:

 Cochrane Oral Health’s Trials Register (searched 10 May 2017);  Cochrane Central Register of Controlled Trials (CENTRAL; 2017, Issue 4) in the Cochrane Library (searched 10 May 2017);  MEDLINE Ovid (1946 to 10 May 2017);  Embase Ovid (7 December 2015 to 10 May 2017); 99  CINAHL EBSCO (Cumulative Index to Nursing and Allied Health Literature; 1937 to 10 May 2017);  CANCERLIT (Cancer subset within PubMed; 1950 to 10 May 2017).

Subject strategies were modelled on the search strategy designed for MEDLINE Ovid. Where appropriate, they were combined with subject strategy adaptations of the highly sensitive search strategy designed by Cochrane for identifying randomised controlled trials and controlled clinical trials as described in the Cochrane Handbook for Systematic Reviews of Interventions Chapter 6 (101). The search strategies are presented in full in Appendix 6.

Due to the Cochrane Embase Project to identify all clinical trials in the database and add them to CENTRAL, only most recent months of the Embase database were searched. See the searching page on the Cochrane Oral Health website for more information. No other restrictions were placed on the date of publication when searching the electronic databases.

Searching other resources

We searched the following trial registries for ongoing studies (see Appendix 6):

 US National Institutes of Health Ongoing Trials Register ClinicalTrials.gov (http://clinicaltrials.gov; searched 10 May 2017);  World Health Organization International Clinical Trials Registry Platform (http://apps.who.int/trialsearch/default.aspx; searched 10 May 2017).

Only handsearching done as part of the Cochrane Worldwide Handsearching Programme and uploaded to CENTRAL was included.

We searched the reference lists of included studies and relevant systematic reviews for further studies.

We did not perform a separate search for adverse effects of interventions used, we considered adverse effects described in included studies only.

Data collection and analysis Selection of studies

Two review authors independently screened the titles and abstracts retrieved from the electronic searches. We obtained full text copies of all studies that appeared to meet the

100 inclusion criteria of the review, or where there was insufficient information in the title or abstract to make a clear judgement. Two review authors independently assessed the full text copies for eligibility and attempted to resolve any disagreements through discussion. We consulted a third review author when we were unable to resolve disagreements.

On assessing the full text article, we discarded any studies that clearly did not meet the inclusion criteria. We recorded all other studies that did not meet the inclusion criteria, along with reasons for exclusion, in the 'Characteristics of excluded studies' table (see Appendix 7).

Data extraction and management

We recorded the following data for each included study in the 'Characteristics of included studies' table (see Appendix 7).

 Trial design, location, number of centres, recruitment period.  Inclusion/exclusion criteria, age and gender of participants, number randomised/analysed, any other potentially important prognostic factors (e.g. cancer type, cancer treatment, etc.).  Detailed description of the intervention and comparator, including timing and duration. Information on compliance with the intervention.  Details of the outcomes reported, including method of assessment and time(s) assessed.  Details of sample size calculations, adverse effects, funding sources, declarations/conflicts of interest.

Assessment of risk of bias in included studies

101 disagreements through discussion, consulting a third review author to achieve consensus when necessary.

We completed a 'Risk of bias' table for each included study (see Appendix 7). For each domain of risk of bias, we first described what was reported to have happened in the study. This provided the rationale for our judgement of whether that domain was at low, high, or unclear risk of bias.

We assessed the following domains:

We categorised the overall risk of bias of individual studies. Studies were categorised as being at low, high, or unclear risk of bias according to the following criteria:

 low risk of bias (plausible bias unlikely to seriously alter the results) if all domains were at low risk of bias;  high risk of bias (plausible bias that seriously weakens confidence in the results) if one or more domains were at high risk of bias; or  unclear risk of bias (plausible bias that raises some doubt about the results) if one or more domains were at unclear risk of bias.

We also presented the 'Risk of bias' summary graphically.

Measures of treatment effect

For continuous outcomes (e.g. oral pain on a visual analogue scale) where studies used the same scale, we used the mean values and standard deviations (SDs) reported in the studies in order to express the estimate of effect as mean difference (MD) with 95% confidence interval (CI). Where different scales were used, we expressed the treatment effect as standardised mean difference (SMD) with 95% CI.

102 For dichotomous outcomes (e.g. mucositis of any severity/no mucositis), we expressed the estimate of effect as a risk ratio (RR) with 95% CI.

We did not use area under the curve (AUC) data due to variation in length of follow-up for outcome assessment, variation in the length of the scale used to measure the outcome and also variation or lack of clarity whether the results were reported in terms of total area under the curve or average over the time period.

Unit of analysis issues

The participant was the unit of analysis.

Dealing with missing data

We attempted to contact the author(s) of all included studies, where feasible, for clarification, and missing data. We would have used the methods described in Section 7.7.3 of the Cochrane Handbook for Systematic Reviews of Interventions to estimate missing SDs (42). We did not use any other statistical methods or perform any further imputation to account for missing data.

Assessment of heterogeneity

When a sufficient number of studies were included in any meta-analyses, we assessed clinical heterogeneity by examining the characteristics of the studies, the similarity between the types of participants, the interventions, and the outcomes. We also assessed heterogeneity statistically using a Chi² test, where a P value < 0.1 indicates statistically significant heterogeneity. We quantified heterogeneity using the I² statistic. A guide to interpretation of the I² statistic given in Section 9.5.2 of the Cochrane Handbook for Systematic Reviews of Interventions is as follows (42):

 0% to 40%: might not be important;  30% to 60%: may represent moderate heterogeneity;  50% to 90%: may represent substantial heterogeneity;  75% to 100%: considerable heterogeneity.

Assessment of reporting biases

If at least 10 studies were included in a meta-analysis, we planned to assess publication bias according to the recommendations on testing for funnel plot asymmetry (72), as

103 described in Section 10.4 of the Cochrane Handbook for Systematic Reviews of Interventions (42). If asymmetry was identified, we would have examined possible causes. We were not able to assess publication bias in this way because, although we had a sufficient number of studies in our meta-analyses for the primary outcome in one comparison, they were split into subgroups containing less than 10 studies, with no pooling of the subgroup totals.

Data synthesis

We used an additional table to report the results from studies not suitable for inclusion in a meta-analysis, but only for the primary outcome.

Subgroup analysis and investigation of heterogeneity

We carried out subgroup analyses according to type of cancer treatment. We also would have considered age group (children versus adults) as a category for subgroup analyses, if there had been sufficient numbers of studies with these differing populations.

Sensitivity analysis

If there had been sufficient numbers of studies in the meta-analyses, we would have tested the robustness of our results by performing sensitivity analyses based on excluding the studies at unclear or high risk of bias from the analyses.

If any meta-analyses had included several small studies and a single very large study, we would have carried out a sensitivity analysis comparing the effect estimates from both random-effects and fixed-effect models. If these were different we would have reported on both analyses as part of the results section, and considered possible interpretation.

Presentation of main results

We produced a 'Summary of findings' table for each comparison in which there was more than one study in at least one of the subgroups based of cancer treatment. We included the

104 incidence of moderate to severe oral mucositis, the incidence of severe oral mucositis and adverse events. We used GRADE methods (45), and the GRADEpro online tool for developing 'Summary of findings' tables (www.guidelinedevelopment.org). We assessed the quality of the body of evidence for each comparison and outcome by considering the overall risk of bias of the included studies, the directness of the evidence, the inconsistency of the results, the precision of the estimates, and the risk of publication bias. We categorised the quality of each body of evidence as high, moderate, low, or very low.

4.5 Results

4.5.1 Description of studies

Results of the search

Our electronic searches identified 5125 records. After removing duplicates, this number was reduced to 3145. We examined the titles and abstracts of these records and discarded 3042, leaving 103 records to examine in more detail. Where possible, we obtained full-text copies of these potentially relevant records and linked any references pertaining to the same study under a single study ID. These 103 records represented 73 studies. We excluded 24 studies at this stage. The remaining 49 studies met our inclusion criteria and we included 35 of these studies in the review. The remaining 14 studies were awaiting assessment because we did not have enough information to be able to include them in the review. We presented this process as a study flow chart in Figure 5.

105

Figure 5: PRISMA flow diagram

106

Included studies

We included 35 studies in this review. For further information see the ‘Characteristics of included studies’ tables in Appendix 7.

 Antoun 2009 (102)  Linch 1993 (121)  Blazar 2006 (103)  Lucchese 2016a (122)  Blijlevens 2013 (104)  Lucchese 2016b (123)  Bradstock 2014 (105)  Makkonen 2000 (124)  Brizel 2008 (106)  McAleese 2006 (125)  Cartee 1995 (107)  Meropol 2003 (126)  Cesaro 2013 (108)  Nemunaitis 1995 (127)  Chi 1995 (109)  Peterson 2009 (128)  Crawford 1999 (110)  Rosen 2006 (129)  Dazzi 2003 (111)  Saarilahti 2002 (130)  Fink 2011 (112)  Schneider 1999 (131)  Freytes 2004 (113)  Spielberger 2004 (132)  Gholizadeh 2016 (114)  Su 2006 (133)  Henke 2011 (115)  Vadhan-Raj 2010 (134)  Hosseinjani 2017 (116)  van der Lelie 2001 (135)  Jagasia 2012 (117)  Wu 2009 (136)  Katano 1995 (118)

 Kim 2017 (119)

 Le 2011 (120)

Characteristics of the trials Study design

One study was a cross-over design that reported the first-period data separately (Chi 1995), whilst the remaining studies all used a parallel design.

Number of arms

Twenty-seven studies had two arms, three studies had three arms (Blijlevens 2013; Freytes 2004; Peterson 2009), one study had four arms (Wu 2009), two studies had five arms

107 (Cartee 1995; Linch 1993), and two studies had seven arms (Blazar 2006; Meropol 2003). Where studies had more than two arms, this was because they tested a range of doses of the cytokine/growth factor. In such instances we combined the arms testing different doses to make pairwise comparisons against the control group. Where possible, we also made head-to-head comparisons of doses (Blijlevens 2013; Cartee 1995; Freytes 2004; Meropol 2003; Peterson 2009).

Country

Nine studies were conducted in the USA (Blazar 2006; Cartee 1995; Crawford 1999; Freytes 2004; Meropol 2003; Schneider 1999; Spielberger 2004; Su 2006; Vadhan-Raj 2010), four in Italy (Cesaro 2013; Dazzi 2003; Lucchese 2016a; Lucchese 2016b), two in each of South Korea (Kim 2017; Wu 2009), the UK (Linch 1993; McAleese 2006), Iran (Gholizadeh 2016; Hosseinjani 2017), Finland (Makkonen 2000; Saarilahti 2002), and one in each of the Netherlands (van der Lelie 2001), Russia (Peterson 2009), Japan (Katano 1995), Germany (Fink 2011), China (Chi 1995), Australia (Bradstock 2014), and France (Antoun 2009). The remaining seven studies were conducted across more than one country: USA and Australia (Jagasia 2012; Rosen 2006); USA and Canada (Nemunaitis 1995); Australia, Canada and the USA (Brizel 2008); Australia, Canada and Europe (Henke 2011); Canada, USA and Europe (Le 2011); 14 European countries (Blijlevens 2013).

Number of centres

Fifteen studies were conducted at a single-centre (Antoun 2009; Cartee 1995; Chi 1995; Dazzi 2003; Fink 2011; Hosseinjani 2017; Katano 1995; Kim 2017; Lucchese 2016a; Lucchese 2016b; McAleese 2006; Saarilahti 2002; Su 2006; Vadhan-Raj 2010; van der Lelie 2001). Eighteen studies were multi-centric, ranging from two sites (Blazar 2006; Makkonen 2000) to 46 sites (Le 2011). It was unclear how many centres were involved in the remaining two studies (Gholizadeh 2016; Schneider 1999).

Trials registries

We were able to find a trials registry number for 13 studies (Blijlevens 2013; Bradstock 2014; Cesaro 2013; Fink 2011; Gholizadeh 2016; Henke 2011; Hosseinjani 2017; Jagasia 2012; Kim 2017; Le 2011; McAleese 2006; Spielberger 2004; Vadhan-Raj 2010), although only six studies mentioned it in the study report (Bradstock 2014; Cesaro 2013;

108 Gholizadeh 2016; Hosseinjani 2017; Kim 2017; Vadhan-Raj 2010), whilst a further study mentioned an obsolete number (Jagasia 2012).

Sample size calculation

Twenty-one studies reported details of sample size calculations, but four of these were not based on oral mucositis (Cesaro 2013; Crawford 1999; Jagasia 2012; Su 2006). One further study stated that 36 participants "should be enough to demonstrate a clinically significant difference", with no details reported (van der Lelie 2001).

Funding and conflicts of interest

This information is difficult to summarise as it was not always adequately reported.

Nineteen studies appeared to be funded by industry alone i.e. it was explicitly stated that they received industry funding or that industry supplied the interventions or both. Five studies appeared to be funded by government/public sector alone and did not state whether or not the interventions were supplied by industry (Cartee 1995; Lucchese 2016a; Lucchese 2016b; Su 2006; Wu 2009). Four studies reported both government and industry funding, three of which stated that industry provided the interventions (Bradstock 2014; Chi 1995; Kim 2017), and one of which was not clear (Blazar 2006). Two studies stated that there was no funding for the study (Cesaro 2013; Hosseinjani 2017). The remaining five studies did not mention funding (Dazzi 2003; Freytes 2004; Gholizadeh 2016; McAleese 2006; Saarilahti 2002).

Ten studies, all industry funded, declared conflicts of interest for reasons such as board membership of the funder, employment or leadership roles with the funder, receipt of lecture fees or consultancy fees or research funding or honoraria from the funder (Antoun 2009; Blijlevens 2013; Brizel 2008; Henke 2011; Jagasia 2012; Le 2011; Peterson 2009; Rosen 2006; Spielberger 2004; Vadhan-Raj 2010). Six of those studies also declared that some authors owned equity/stock with the funder (Brizel 2008; Henke 2011; Jagasia 2012; Le 2011; Rosen 2006; Spielberger 2004). Three studies did not explicitly declare conflicts of interest, but some authors were employed by the funder (Crawford 1999; Linch 1993; Nemunaitis 1995). Eight studies stated that there were no conflicts of interest (Bradstock 2014; Cesaro 2013; Gholizadeh 2016; Hosseinjani 2017; Kim 2017; Lucchese 2016a; Lucchese 2016b; Su 2006). The remaining 14 studies did not mention conflicts of interest.

Characteristics of the participants

109 Number randomised/analysed

The studies randomised 3218 participants, of which 3102 were included in the studies' analyses (the latter number does not include any participants from Makkonen 2000, as this study did not report how many of the 40 randomised participants were analysed).

Age and sex

The age of the participants ranged from 1 to 87 years, with four studies only including children and young adults (i.e. up to 18 years) (Cesaro 2013; Gholizadeh 2016; Lucchese 2016a; Lucchese 2016b). Of the 31 studies including adult participants, one had a median age of 29 (Dazzi 2003), two had mean or median ages in their 30s (Linch 1993; Nemunaitis 1995), nine in their 40s (Blazar 2006; Bradstock 2014; Cartee 1995; Chi 1995; Hosseinjani 2017; Jagasia 2012; Spielberger 2004; Vadhan-Raj 2010; van der Lelie 2001), 11 in their 50s (Blijlevens 2013; Brizel 2008; Fink 2011; Freytes 2004; Henke 2011; Katano 1995; Kim 2017; Le 2011; Peterson 2009; Saarilahti 2002; Wu 2009), seven in their 60s (Antoun 2009; Crawford 1999; Makkonen 2000; McAleese 2006; Meropol 2003; Rosen 2006; Su 2006), and one study did not report the age, although the inclusion criteria stated that they must be at least 18 years old (Schneider 1999). In twenty four studies, there was a clear majority of male participants, whilst the male to female ratio was roughly equal in seven studies. In three studies there were more female participants, although two of these exclusively included breast cancer patients (Cartee 1995; Katano 1995), whilst the third included colorectal cancer patients (Peterson 2009).

Cancer type

Fourteen studies enrolled participants with haematological cancers (Blazar 2006; Blijlevens 2013; Bradstock 2014; Fink 2011; Freytes 2004; Gholizadeh 2016; Hosseinjani 2017; Jagasia 2012; Kim 2017; Lucchese 2016a; Lucchese 2016b; Nemunaitis 1995; Spielberger 2004; van der Lelie 2001). Eighteen studies enrolled participants with solid cancers: head and neck: (Brizel 2008; Chi 1995; Henke 2011; Le 2011; Makkonen 2000; McAleese 2006; Saarilahti 2002; Schneider 1999; Su 2006; Wu 2009); colorectal (Antoun 2009; Meropol 2003; Peterson 2009; Rosen 2006); Breast (Cartee 1995; Katano 1995); lung (Crawford 1999); sarcoma (Vadhan-Raj 2010). The remaining three studies enrolled a mixture of participants with solid cancers and particpants with haematological cancers, two of which were 80 to 90% solid (Cesaro 2013; Dazzi 2003), and the other study only 3% solid (Linch 1993).

110 Cancer treatment

In 11 studies, the participants received chemotherapy only (Antoun 2009; Bradstock 2014; Cartee 1995; Chi 1995; Crawford 1999; Gholizadeh 2016; Katano 1995; Meropol 2003; Peterson 2009; Rosen 2006; Vadhan-Raj 2010). Of the 15 studies in which the participants received conditioning therapy prior to stem cell or bone marrow transplantation, five of these involved chemotherapy only (Blijlevens 2013; Dazzi 2003; Fink 2011; Hosseinjani 2017; Kim 2017), and one involved total body irradiation (TBI) only (Lucchese 2016b). In the remaining nine transplant studies, all the participants had chemotherapy, but the proportion of participants also receiving TBI differed: 100% (Lucchese 2016a; Nemunaitis 1995; Spielberger 2004); around 50% (Blazar 2006; Jagasia 2012; van der Lelie 2001); 29% (Linch 1993); 10% or less (Cesaro 2013; Freytes 2004). The remaining nine studies were all on head and neck cancer patients where the participants either had radiotherapy to the head and neck alone (Makkonen 2000; McAleese 2006; Saarilahti 2002; Schneider 1999; Su 2006), or radiotherapy to the head and neck plus chemotherapy (Brizel 2008; Henke 2011; Le 2011; Wu 2009), although in one of those studies only 50% of participants had the chemotherapy (Wu 2009).

Of the 15 transplant studies, four involved allogeneic transplants (Blazar 2006; Jagasia 2012; Lucchese 2016b; Nemunaitis 1995), nine involved autologous transplants (Blijlevens 2013; Cesaro 2013; Dazzi 2003; Fink 2011; Freytes 2004; Hosseinjani 2017; Kim 2017; Lucchese 2016a; Spielberger 2004), with the remaining two involving a mixture (Linch 1993; van der Lelie 2001).

In six studies, all participants received granulocyte-colony stimulating factor (a growth factor) as part of the cancer treatment to prevent neutropenia. Four of these studies were investigating keratinocyte growth factor (Blazar 2006; Bradstock 2014; Spielberger 2004; Vadhan-Raj 2010), and two were investigating granulocyte-macrophage colony- stimulating factor (Cartee 1995; Dazzi 2003). Giving all participants this growth factor would have the potential to lessen the impact of the study intervention.

Characteristics of the interventions and comparisons Keratinocyte growth factor (KGF)

Of the 16 studies investigating KGF, 14 used a placebo comparator (Blazar 2006; Blijlevens 2013; Bradstock 2014; Brizel 2008; Freytes 2004; Henke 2011; Jagasia 2012; Le 2011; Lucchese 2016a; Lucchese 2016b; Meropol 2003; Rosen 2006; Spielberger 2004; Vadhan-Raj 2010), one was KGF plus standard care versus standard care alone (Fink 111 2011), and the remaining study used a chlorhexidine mouthwash comparator (Gholizadeh 2016).

In all studies, KGF was given intravenously. The most common total dosage received was 360 µg/kg in seven studies (Bradstock 2014; Fink 2011; Gholizadeh 2016; Jagasia 2012; Lucchese 2016a; Lucchese 2016b; Spielberger 2004). The dosage varied greatly in the other studies: 120 µg/kg (Rosen 2006); 180 µg/kg (Vadhan-Raj 2010); 600 µg/kg (Brizel 2008); 840 to 960 µg/kg depending on resection type (Henke 2011); 1440 µg/kg (Le 2011). The dosages varied within the remaining studies due to multiple arms receiving different doses: 3 to 240 µg/kg (Meropol 2003); 180 to 360 µg/kg (Blijlevens 2013); 240 to 720 µg/kg (Blazar 2006); 325 to 650 µg/kg (Freytes 2004).

The number of doses received ranged from one (Vadhan-Raj 2010) to 13 (Freytes 2004), but the most common was six (Blijlevens 2013; Bradstock 2014; Fink 2011; Gholizadeh 2016; Lucchese 2016a; Lucchese 2016b; Spielberger 2004).

Reporting of compliance varied too greatly to summarise succinctly but compliance was generally high (see ‘Characteristics of included studies’ tables in Appendix 7).

Granulocyte-macrophage colony-stimulating factor (GM-CSF)

Of the eight studies investigating GM-CSF, four used a placebo comparator (Cartee 1995; Dazzi 2003; Nemunaitis 1995; van der Lelie 2001), two used a no-treatment comparator (Chi 1995; McAleese 2006), one was GM-CSF plus sucralfate versus sucralfate alone (Makkonen 2000), and the remaining study used a sucralfate comparator (Saarilahti 2002).

In three studies, GM-CSF was given by subcutaneous injection (Chi 1995; Makkonen 2000; McAleese 2006). In Makkonen 2000, both arms received sucralfate mouthwash that was swallowed after rinsing. In three studies, GM-CSF was taken as a mouthwash (Cartee 1995; Dazzi 2003; Saarilahti 2002). In Saarilahti 2002, both the GM-CSF and sucralfate comparator mouthwashes were swallowed after rinsing. In one study, GM-CSF was given as an oral gel and swallowed after holding in the mouth (van der Lelie 2001). In the remaining study, GM-CSF was given intravenously (Nemunaitis 1995).

Total dosage varied greatly: 40 µg (Chi 1995); 2100 µg (McAleese 2006); 5250 µg/m² (Nemunaitis 1995). The dosages ranged from 12.6 to 12,600 µg within one study due to multiple arms receiving different doses (Cartee 1995). Another study reported a mean total dosage of 3398 µg, but this total ranged from 300 to 7200 µg depending on the

112 participant's weight and the length of radiotherapy course (Makkonen 2000). In two studies, the dose was 150 µg per day but the total received varied depending on neutrophil recovery (Dazzi 2003), and the length of radiotherapy course (Saarilahti 2002). In the remaining study, the dose was 300 µg per day but the total received varied depending on neutrophil recovery (van der Lelie 2001).

As is obvious from the variation in total dosage, the number of doses received varied greatly both between studies and within studies. Compliance was also reported inconsistently but was generally high (see ‘Characteristics of included studies’ tables in Appendix 7).

Granulocyte-colony stimulating factor (G-CSF)

Of the six studies investigating G-CSF, four used a placebo comparator (Crawford 1999; Linch 1993; Schneider 1999; Su 2006), one used a no-treatment comparator (Katano 1995), and the remaining study compared a type of G-CSF that is given as a single dose (pegfilgrastim) with the standard G-CSF that is given in multiple doses (filgrastim) (Cesaro 2013).

Four studies reported that G-CSF was given by subcutaneous injection (Crawford 1999; Katano 1995; Schneider 1999; Su 2006), whilst one did not specify, but was probably subcutaneous (Cesaro 2013), and the remaining study was intravenous delivery (Linch 1993).

Total dosage varied: 3220 µg/m² (Crawford 1999); 3 µg/kg per day with the total dependent on neutrophil counts and the length of radiotherapy course (Schneider 1999; Su 2006); 2 to 15 µg/kg per day due to multiple arms receiving different dosages with the total was depending on neutrophil recovery (Linch 1993); 125 µg per day with total depending on neutrophil recovery (Katano 1995); 100 µg/kg in the pegfilgrastim arm and at least 45 µg/kg in the filgrastim arm (Cesaro 2013).

The number of doses received varied both between studies and within studies. Compliance was reported as being 100% in one study (Cesaro 2013), whilst one study only reported that the interventions were well tolerated (Schneider 1999), and the remaining four studies did not report on compliance.

Epidermal growth factor (EGF)

113 Two studies investigated an oral spray of EGF, both using a placebo comparator (Kim 2017; Wu 2009). Total dosage was unclear in both studies but the daily dose was 50 µg/ml (6 sprays twice daily) in one study (Kim 2017), and 10 to 100 µg per day (due to multiple arms receiving different dosages) in the other study (Wu 2009). The number of doses varied depending on neutrophil recovery and resolution of oral mucositis in Kim 2017, whilst participants in Wu 2009 received the interventions daily for 5 weeks but it was not clear if that meant only on the radiotherapy days (5 days per week) or 7 days per week. Compliance was reported as a median of 93% and 92% in the EGF and placebo groups respectively in Kim 2017, but compliance was not reported in Wu 2009.

Intestinal trefoil factor (ITF)

One study investigated an oral spray of ITF using a placebo comparator (Peterson 2009). The ITF was not expectorated. The study included two ITF arms with total dosages of 336 mg and 2688 mg. The mode of administration was 3 sprays to the oral mucosa 8 times daily for 14 days. Patient-reported compliance was 97%.

Erythropoietin

One study investigated a mouthwash of erythropoietin using a placebo comparator (Hosseinjani 2017). Neither swallowing nor expectoration was reported. The mouthwash was taken as 15 ml (50 IU/ml) four times daily (daily dosage of 3000 IU) for 14 days or until neutrophil recovery, whichever occurred first. Compliance was reported narratively as being low but no reason was stated.

Transforming growth factor (TGF)

One study investigated TGF-beta(2) using a placebo comparator (Antoun 2009). The dosage was 2 ng of TGF per mg protein mixed with cool boiled water at 0.23 g/mL (100 kcl/100 mL). During each cycle participants received 750 to 1000 mL per day plus any other food desired. The formula was administered for 2 days before, 2 days during, and 3 days after chemotherapy (7 days/cycle), for one to eight cycles. Compliance was poor i.e. nine paraticipants did not eat the formula and were excluded.

Characteristics of the outcomes Primary outcome

For the primary outcome of oral mucositis, we were interested in both the presence/absence of oral mucositis, and also different levels of severity. All 35 studies 114 assessed and reported the incidence of oral mucositis. Twenty-two studies primarily used the WHO 0 to 4 scale, whilst four used the NCI-CTC 0 to 4 scale (Brizel 2008; Dazzi 2003; Freytes 2004; Kim 2017), four used the RTOG 0 to 4 scale (Chi 1995; McAleese 2006; Saarilahti 2002; Wu 2009), one used the CALGB 0 to 4 scale (Cartee 1995), one used an unnamed 0 to 2 scale (Makkonen 2000), one used an unnamed 0 to 3 scale (Su 2006), one used an unnamed 0 to 4 scale (Nemunaitis 1995), and the remaining study did not mention a scale and only reported the incidence of stomatitis (Linch 1993). The different oral mucositis assessment scales are described in Appendix 3.

Twelve studies reported the data in our preferred format which was the maximum oral mucositis score experienced by each participant over the length of the study, allowing us to dichotomise the data into various levels of severity as described in the section 'Methods: Primary outcomes'. Eighteen studies reported a particular level of severity (e.g. grade 3 or above). One study reported the incidence of each oral mucositis grade on multiple assessment days. We were unable to use the data from the remaining four studies for analysis due to unclear or lack of reporting (Linch 1993; Lucchese 2016a; Lucchese 2016b; Makkonen 2000).

The frequency of oral mucositis assessment and the duration for which it was assessed varied greatly across the studies, often depending on whether the participants received radiotherapy, and often depending on the speed of neutrophil recovery, resolution of oral mucositis, or duration of hospitalisation. Four studies did not report the frequency of assessment (Antoun 2009; Cesaro 2013; Linch 1993; Nemunaitis 1995), whilst a further study was unclearly reported (Lucchese 2016b). Twelve studies reported daily assessments, eight reported weekly assessments, with the remainder falling somewhere in between these two frequencies. Where participants had multiple cycles of treatment, we only reported the results for the first cycle if these data were available separately.

Secondary outcomes Interruptions to cancer treatment

Six studies reported data that we were able to use in analyses (Brizel 2008; Henke 2011; Le 2011; Saarilahti 2002; Su 2006; Wu 2009), whilst a further two studies assessed this outcome but either did not report the interruption by treatment arm (Makkonen 2000), or narratively reported that there were no differences, with no numerical data (Schneider 1999).

115 Two studies reported this outcome as the incidence of unscheduled radiotherapy breaks of 5 or more days (Brizel 2008; Henke 2011; Le 2011). Two of those studies also reported on chemotherapy delays/discontinuations (Henke 2011; Le 2011). The remaining studies all reported on the incidence of interruptions to radiotherapy treatment, one of which stated that interruptions were specifically due to oral mucositis (Saarilahti 2002), and another reporting the incidence of 3 or more consecutive days of interruption (Wu 2009).

Oral pain

Four studies reported data that we were able to use in analyses (Dazzi 2003; Freytes 2004; Henke 2011; Le 2011). Two of these studies used a 0 to 4 scale and reported the mean (Henke 2011; Le 2011), whilst the other two studies used a 0 to 10 scale and reported the mean worst score experienced (Dazzi 2003; Freytes 2004).

Of the 11 other studies that reported that oral pain was an outcome of the study, five reported the results as area under the curve (AUC) but, for reasons stated in the section 'Methods: Measures of treatment effect', we did not meta-analyse this data (Blijlevens 2013; Kim 2017; Lucchese 2016a; Rosen 2006; Spielberger 2004). Two studies reported medians, which are not suitable for meta-analysis (Vadhan-Raj 2010; van der Lelie 2001). One study reported the data graphically as a mean over time with no standard deviation (Saarilahti 2002). One study narratively reported that there were no differences, with no numerical data (Wu 2009). The remaining two studies used two different scales: one reported as "no difference" and another reported on a graph with no standard deviation (Makkonen 2000); both reported on a graph over time, with one also reported as AUC (Meropol 2003).

Quality of life

Four studies assessed quality of life using various assessment scales: European Quality Of Life Utility Scale - EQ-5D (Blijlevens 2013); modified Oral Mucositis Daily Questionnaire - OMDQ (Kim 2017); Functional Assessment of Cancer Therapy - FACT (Spielberger 2004); an unnamed 1 to 7 scale (Vadhan-Raj 2010). We did not use the data in our anaylses as they were either reported as AUC (Kim 2017; Spielberger 2004), as a median (Vadhan-Raj 2010), or the mean was reported at one very early timepoint with no standard deviation (Blijlevens 2013).

Normalcy of diet - including use of percutaneous endoscopic gastrostomy (PEG) feeding tubes or total parenteral nutrition (TPN)

116 Fourteen studies reported data that we were able to use in analyses in the form of: incidence of TPN (Blijlevens 2013; Cesaro 2013; Fink 2011; Jagasia 2012; Kim 2017; Spielberger 2004; van der Lelie 2001); incidence of PEG (Brizel 2008; Saarilahti 2002; Su 2006); incidence of TPN, PEG, nasogastric tube or IV hydration (Henke 2011; Le 2011); incidence of "tube feeding" (McAleese 2006); ability to eat using a 1 to 4 scale (Freytes 2004). Only one of these studies explicitly stated that supplemental feeding was due to oral mucositis (Henke 2011).

Two further studies only reported the duration of TPN (Lucchese 2016a; Lucchese 2016b), and another study used 0 to 4 scales to assess difficulty eating and drinking, but reported median scores (Vadhan-Raj 2010).

We combined studies reporting incidence of TPN, PEG, etc, in meta-analyses of 'supplemental feeding'.

Adverse events

This outcome was very poorly reported with some studies reporting numerical data and some reporting narratively. Some studies only reported adverse events if there was a minimum incidence (which varied between studies) or if there was a specified difference in incidence between treatment arms. It was also difficult to determine whether or not many adverse effects were due to the study interventions, or due to the underlying cancer treatment. We presented adverse event data/information only in an additional table.

Number of days in hospital

Two studies reported data that we were able to use in analyses i.e. mean and standard deviation (Blijlevens 2013; Hosseinjani 2017).

Five further studies reported medians (Cesaro 2013; Fink 2011; Linch 1993; Nemunaitis 1995; van der Lelie 2001). One study reported data graphically with no standard deviation or P-value (Crawford 1999). One study listed this as an outcome of the study but did not actually report it (Kim 2017). One study reported the incidence of hospitalisation (Saarilahti 2002).

Number of days of treatment with opioid analgesics

117 Two studies reported data that we were able to use in analyses i.e. mean and standard deviation (Blijlevens 2013; Dazzi 2003; Freytes 2004). Only one study specified that the opioid use was due to oral mucositis (Freytes 2004).

Four further studies reported medians (Fink 2011; Kim 2017; Lucchese 2016a; Spielberger 2004), whilst another study did not state whether the data were means or medians, and there were no standard deviations or P-value (Lucchese 2016b). Three studies reported total dose of opioid analgesic (Henke 2011; Le 2011; Vadhan-Raj 2010), whilst four studies reported the incidence of its use (Hosseinjani 2017; Jagasia 2012; Saarilahti 2002; van der Lelie 2001). One study stated that it assessed the use of opioid analgesics, but did not specify whether this was in terms of duration, quantity or incidence, and did not actually report any data (Wu 2009).

Number of days unable to take medicine orally

No studies reported this outcome.

Excluded studies

We excluded 24 studies from this review for the following reasons.

 Not a randomised controlled trial or unclear (Foncuberta 2001 (137); Gordon 1993 (138); Horsley 2007 (139); Hunter 2009 (140); Iwase 1997 (141); Limaye 2013 (142); Throuvalas 1995 (143); Vitale 2014 (144)).  Stomatitis incidence reported in adverse events table (Kubo 2016 (145); Lee 2016 (146); Nabholtz 2002 (147); Tsurusawa 2016 (148)).  Unclear if mucositis was oral or gastrointestinal (Jones 1996 (149); Legros 1997 (150); Pettengell 1992 (151)).  Study stopped early with very few participants enrolled (Antin 2002 (152); NCT00360971 (153); NCT00626639 (154)).  Oral mucositis not mentioned and unknown if measured (Gebbia 1994 (155); Gladkov 2013 (156)).  Some participants had oral mucositis at baseline (Ryu 2007 (157)).  Cross-over study with no reporting of first-period data (de Koning 2007 (158)).  Results reported by cycle assuming independence (Karthaus 1998 (159)).  Survival/cure was primary outcome with mucositis (unclear if oral or gastrointestinal) as a toxicity (Ifrah 1999 (160)).

118 4.5.2 Risk of bias in included studies

(A risk of bias table is presented for each included study in Appendix 7.)

Allocation (selection bias) Random sequence generation

Nineteen studies described an adequate method of generating a random sequence, so we assessed these as low risk of bias. The remaining 16 studies stated that they were randomised without providing a description of how the random sequence was generated, so we assessed these as unclear risk of bias.

Allocation concealment

Seventeen studies described a process that would have concealed the random sequence from those involved in the study, thus allowing it to be applied as it was generated. We assessed these 17 studies as low risk of bias. The remaining 18 studies did not describe any methods used to conceal the random sequence, so we assessed them as unclear risk of bias.

In total, 16 studies were at low risk of selection bias, meaning that we assessed both of the above domains as low risk of bias. The remaining 19 studies were at unclear risk of bias because one or both of the above domains were rated as unclear.

The majority of studies were carried out in developed countries with strict controls and regulations and we feel that many of them probably had adequate randomisation, and that the unclear ratings for these two domains were probably due to reporting issues rather than actual bias. Therefore, when incorporating risk of bias into our GRADE assessments, we did not downgrade any evidence based on selection bias.

Blinding (performance bias and detection bias) Blinding of participants and personnel (performance bias)

We assessed 28 studies as low risk of bias. Twenty-seven of these studies used a placebo comparator and this ensured that blinding was performed successfully. One further study compared GM-CSF with sucralfate, but the interventions were supplied as identical- looking mouthwashed, the study was described as double-blind, and there was no reason to suspect that participants or personnel were not blinded (Saarilahti 2002).

We assessed seven studies as high risk of bias. Three of these studies used a no-treatment comparator, so blinding was not possible (Chi 1995; Katano 1995; McAleese 2006). Two 119 other studies were similar in that they compared KGF plus best supportive care with best supportive care alone (Fink 2011), and GM-CSF plus sucralfate with sucralfate alone (Makkonen 2000). One study compared intravenous KGF with a chlorhexidine mouthwash (Gholizadeh 2016). The remaining study compared two types of G-CSF, but the dosing schedule was very different, ensuring that blinding was not possible (Cesaro 2013).

Blinding of outcome assessment (detection bias)

We assessed 29 studies as low risk of bias. We assessed four studies as unclear risk of bias because blinding of outcome assessment was not mentioned, but we judged that it would have been possible to achieve (Cesaro 2013; Chi 1995; Katano 1995; Makkonen 2000). We assessed the remaining two studies as high risk of bias because they either stated that there was no blinding of outcome assessors (Fink 2011), or it was implied by the description "single blind" (Linch 1993).

Incomplete outcome data (attrition bias)

Attrition was generally very low and we assessed 31 studies as low risk of bias. We assessed two studies as unclear risk of bias because one did not report how many of the randomised participants were included in the analyses (Makkonen 2000), and the other did not report the attrition by treatment arm but there was potential for bias if the drop-outs were mostly from one arm and had developed the outcome of severe oral mucositis (Cartee 1995). We assessed two studies as high risk of bias because one had very high attrition (Antoun 2009), and the other had 19% attrition in one arm compared to none in the other arm (Fink 2011).

Selective reporting (reporting bias)

It is important to note that we were perhaps quite lenient when rating bias under this domain. We tended to focus on the primary outcome because the vast majority of the data were for this outcome. Also, many studies only reported a particular level of oral mucositis severity, for example grade 2 to 4 (ulcerative/moderate to severe), when they could have reported more usable data by reporting the maximum grade experienced per patient, allowing us to dichotomise this into all severities. Some readers may consider this to be bias but we have reported all this information transparently in the 'Characteristics of included studies' tables (see Appendix 7), thus allowing the reader to decide if they would judge the risk of bias differently. Furthermore, many secondary outcomes were reported poorly or in a way that was not amenable to meta-analysis, which in most cases was a reporting issue rather than a bias issue. This highlighted the problem with the current

120 Cochrane risk of bias tool in that meta-analyses are being biased due to missing information, but this is not being accounted for in the meta-analysis. It does not seem appropriate to rate a study at high risk of bias due to a secondary outcome when it is contributing data to the meta-analysis for the primary outcome, and it is the meta-analysis for the secondary outcome that is affected by bias. Again, all this information was clearly reported in the 'Characteristics of included studies' tables.

We assessed 32 studies as low risk of bias. We assessed the remaining three studies as high risk of bias, two because there were no usable data for the primary outcome (Linch 1993; Makkonen 2000), and one because several outcomes were assessed but not reported (Wu 2009).

Other potential sources of bias

We did not consider there to be any issues arising from other potential sources of bias in any of the studies and we therefore assessed them all as low risk of bias.

Overall risk of bias

 Thirteen studies (37%) were at low overall risk of bias (Blijlevens 2013; Dazzi 2003; Freytes 2004; Henke 2011; Hosseinjani 2017; Kim 2017; Le 2011; Lucchese 2016a; Lucchese 2016b; Saarilahti 2002; Schneider 1999; Su 2006; Vadhan-Raj 2010).  Twelve studies (34%) were at unclear overall risk of bias (Blazar 2006; Bradstock 2014; Brizel 2008; Cartee 1995; Crawford 1999; Jagasia 2012; Meropol 2003; Nemunaitis 1995; Peterson 2009; Rosen 2006; Spielberger 2004; van der Lelie 2001).  Ten studies (29%) were at high overall risk of bias (Antoun 2009; Cesaro 2013; Chi 1995; Fink 2011; Gholizadeh 2016; Katano 1995; Linch 1993; Makkonen 2000; McAleese 2006; Wu 2009).

Risk of bias can be viewed graphically in Figure 6.

121

122

Figure 6: Risk of bias summary: review authors’ judgements about each risk of bias item for each included study

4.5.3 Effects of interventions

Keratinocyte growth factor (KGF) versus placebo Oral mucositis Adults receiving bone marrow/stem cell transplantation after conditioning therapy for haematological cancers

There was insufficient evidence from four studies, one at low (Blijlevens 2013), two at unclear (Blazar 2006; Spielberger 2004), and one at high risk of bias (Fink 2011), to determine whether or not KGF reduces the risk of any level of oral mucositis: RR 0.96, 95% CI 0.88 to 1.05; 655 participants (Analysis 1.1).

Six studies, two at low (Blijlevens 2013; Freytes 2004), three at unclear (Blazar 2006; Jagasia 2012; Spielberger 2004), and one at high risk of bias (Fink 2011), showed a

123 reduction in the risk of moderate to severe oral mucositis in favour of KGF: RR 0.89, 95% CI 0.80 to 0.99; 852 participants; moderate quality evidence (Analysis 1.2).

The same six studies showed a possible reduction in the risk of severe oral mucositis in favour of KGF, but there was also some possibility of harm: RR 0.85, 95% CI 0.65 to 1.11; 852 participants; moderate quality evidence (Analysis 1.3).

Adults receiving radiotherapy to the head and neck with cisplatin/fluorouracil (5FU)

Two studies, both at low risk of bias (Henke 2011; Le 2011), showed a reduction in the risk of any level of oral mucositis in favour of KGF: RR 0.95, 95% CI 0.90 to 1.00; P = 0.04; 374 participants (Analysis 1.1).

Three studies, two at low (Henke 2011; Le 2011), and one at unclear risk of bias (Brizel 2008), showed a reduction in the risk of moderate to severe oral mucositis in favour of KGF: RR 0.91, 95% CI 0.83 to 1.00; P = 0.04; 471 participants; moderate quality evidence (Analysis 1.2).

The same three studies showed a reduction in the risk of severe oral mucositis in favour of KGF: RR 0.79, 95% CI 0.69 to 0.90; 471 participants; high quality evidence (Analysis 1.3).

Adults receiving chemotherapy alone for mixed cancers

Two studies, both at unclear risk of bias (Bradstock 2014; Rosen 2006), showed a reduction in the risk of any level of oral mucositis in favour of KGF: RR 0.71, 95% CI 0.60 to 0.85; 215 participants (Analysis 1.1).

Four studies, one at low (Vadhan-Raj 2010), and three at unclear risk of bias (Bradstock 2014; Meropol 2003; Rosen 2006), showed a reduction in the risk of moderate to severe oral mucositis in favour of KGF: RR 0.56, 95% CI 0.45 to 0.70; 344 participants; high quality evidence (Analysis 1.2).

Three studies, one at low (Vadhan-Raj 2010), and two at unclear risk of bias (Bradstock 2014; Rosen 2006), showed a reduction in the risk of severe oral mucositis in favour of KGF: RR 0.30, 95% CI 0.14 to 0.65; 263 participants; moderate quality evidence (Analysis 1.3).

124

Analysis 1.1: KGF versus placebo; Outcome: Oral mucositis (any) 125

Analysis 1.2: KGF versus placebo; Outcome: Oral mucositis (moderate to severe) 126

Analysis 1.3: KGF versus placebo; Outcome: Oral mucositis (severe) 127 Interruptions to cancer treatment Adults receiving radiotherapy to the head and neck with cisplatin/fluorouracil (5FU)

There was insufficient evidence from three studies, two at low (Henke 2011; Le 2011), and one at unclear risk of bias (Brizel 2008), to determine whether or not KGF reduced the risk of having unscheduled radiotherapy breaks of 5 or more days: RR 1.01, 95% CI 0.65 to 1.59; 473 participants (Appendix 8: Analysis 1.4).

There was insufficient evidence, from the same two studies at low risk of bias, to determine whether or not KGF reduced the risk of having chemotherapy delays/discontinuations: RR 0.96, 95% CI 0.62 to 1.47; 374 participants (Appendix 8: Analysis 1.5).

Oral pain Adults receiving bone marrow/stem cell transplantation after conditioning therapy for haematological cancers

There was insufficient evidence, from one study at low risk of bias (Freytes 2004), to determine whether or not KGF reduced the mean worst pain experienced on a 0 (no pain) to 10 (worst pain) scale: MD -0.85, 95% CI -3.00 to 1.30; 42 participants (Appendix 8: Analysis 1.6).

Adults receiving radiotherapy to the head and neck with cisplatin

There was some evidence, from two studies at low risk of bias (Henke 2011; Le 2011), that KGF might lead to a reduction in the mean pain score on a 0 (no pain) to 4 (worst pain) scale: MD -0.12, 95% CI -0.27 to 0.02; 374 participants (Appendix 8: Analysis 1.6).

Quality of life

No studies assessed this outcome.

Normalcy of diet Adults receiving bone marrow/stem cell transplantation after conditioning therapy for haematological cancers

There was insufficient evidence from four studies, one at low (Blijlevens 2013), two at unclear (Jagasia 2012; Spielberger 2004), and one at high risk of bias (Fink 2011), to determine whether or not KGF reduced the risk of total parenteral nutrition: RR 0.89, 95% CI 0.58 to 1.34; 714 participants (Appendix 8: Analysis 1.7).

128 There was further insufficient evidence, from one study at low risk of bias (Freytes 2004), to determine whether or not KGF reduced the mean worst ability to eat score on a 1 (normal) to 4 (no solids or liquids) scale: MD -0.50, 95% CI -1.21 to 0.21; 42 participants (Appendix 8: Analysis 1.8).

Adults receiving radiotherapy to the head and neck with cisplatin/fluorouracil (5FU)

There was insufficient evidence from three studies, two at low (Henke 2011; Le 2011), and one at unclear risk of bias (Brizel 2008), to determine whether or not KGF reduced the risk of receiving supplemental nutrition (total parenteral nutrition, percutaneous endoscopic gastrostomy, nasogastric tube or IV hydration): RR 1.03, 95% CI 0.77 to 1.37; 473 participants (Appendix 8: Analysis 1.7).

Adverse events

This outcome was difficult to summarise due to poor and inconsistent reporting, and we did not meta-analyse any data. However, there did not appear to be any serious concerns regarding adverse effects of KGF. We tabulated relevant information in Appendix 9: Table 1.

Number of days in hospital Adults receiving bone marrow/stem cell transplantation after conditioning therapy for haematological cancers

There was insufficient evidence, from one study at low risk of bias (Blijlevens 2013), to determine whether or not KGF reduced the mean number of days in hospital: MD 0.00, 95% CI -1.64 to 1.64; 281 participants (Appendix 8: Analysis 1.9).

Number of days of treatment with opioid analgesics Adults receiving bone marrow/stem cell transplantation after conditioning therapy for haematological cancers

There was some imprecise evidence, from two studies at low risk of bias (Blijlevens 2013; Freytes 2004), that KGF might lead to a reduction in the mean number of days of treatment with opioid analgesics: MD -1.41, 95% CI -3.33 to 0.51; 323 participants (Appendix 8: Analysis 1.10). The average effect was around 1.5 days reduction, but the confidence interval was compatible with both a reduction of almost 3.5 days and an increase of half a day.

Number of days unable to take medicine orally 129 No studies assessed this outcome.

Keratinocyte growth factor (KGF) dose comparisons

There was some inconsistent evidence from which no conclusions could be drawn regarding different dosages of KGF (Appendix 8: Analysis 2.1; Analysis 2.2; Analysis 2.3; Analysis 2.4; Analysis 2.5; Analysis 2.6; Analysis 2.7; Analysis 2.8).

Keratinocyte growth factor (KGF) versus chlorhexidine

One study, at high risk of bias and analysing 90 children receiving mixed chemotherapy alone for acute lymphoblastic leukaemia (Gholizadeh 2016), compared KGF by IV infusion with chlorhexidine mouthwash. There was weak evidence (due to risk of bias and low sample size) that KGF performed better than chlorhexidine in reducing the risk of any level of oral mucositis (RR 0.67, 95% CI 0.54 to 0.85; Analysis 3.1), moderate to severe oral mucositis (RR 0.12, 95% CI 0.05 to 0.28; Analysis 3.2), and severe oral mucositis (RR 0.01, 95% CI 0.00 to 0.19; Analysis 3.3).

130

Analysis 3.1: KGF versus chlorhexidine; Outcome: Oral mucositis (any)

Analysis 3.2: KGF versus chlorhexidine; Outcome: Oral mucositis (moderate to severe)

131

Analysis 3.3: KGF versus chlorhexidine; Outcome: Oral mucositis (severe)

132 Granulocyte-macrophage colony-stimulating factor (GM-CSF) versus placebo/no treatment Oral mucositis Adults receiving bone marrow/stem cell transplantation after conditioning therapy for mixed cancers

There was some evidence, from one study at low risk of bias (Dazzi 2003), that GM-CSF might lead to a reduction in the risk of any level of oral mucositis: RR 0.91, 95% CI 0.80 to 1.04; 90 participants (Analysis 4.1).

There was insufficient evidence, from one study at unclear risk of bias (Nemunaitis 1995), to determine whether or not GM-CSF reduced the risk of moderate to severe oral mucositis: (RR 0.94, 95% CI 0.79 to 1.13; 109 participants; very low quality evidence (Analysis 4.2).

There was insufficient evidence from three studies, one at low (Dazzi 2003), and two at unclear risk of bias (Nemunaitis 1995; van der Lelie 2001), to determine whether or not GM-CSF reduced the risk of severe oral mucositis: RR 0.74, 95% CI 0.33 to 1.67; 235 participants; low quality evidence (Analysis 4.3).

Adults receiving radiotherapy to the head and neck

There was insufficient evidence, from one study at high risk of bias (McAleese 2006), to determine whether or not GM-CSF reduced the risk of any level of oral mucositis (RR 1.01, 95% CI 0.82 to 1.23; 29 participants; Analysis 4.1), moderate to severe oral mucositis (RR 0.72, 95% CI 0.49 to 1.06; 29 participants; very low quality evidence; Analysis 4.2), or severe oral mucositis (RR 0.31, 95% CI 0.01 to 7.09; 29 participants; very low quality evidence; Analysis 4.3).

Adults receiving chemotherapy alone for mixed cancers

There was insufficient evidence from two studies, one at unclear (Cartee 1995), and one at high risk of bias (Chi 1995), to determine whether or not GM-CSF reduced the risk of severe oral mucositis: RR 0.59, 95% CI 0.05 to 7.11; 65 participants; very low quality evidence (Analysis 4.3).

133

Analysis 4.1: GM-CSF versus control; Outcome: Oral mucositis (any)

134

Analysis 4.2: GM-CSF versus control; Outcome: Oral mucositis (moderate to severe)

135

Analysis 4.3: GM-CSF versus control; Outcome: Oral mucositis (severe) 136 Interruptions to cancer treatment

No studies assessed this outcome.

Oral pain Adults receiving bone marrow/stem cell transplantation after conditioning therapy for mixed cancers

There was insufficient evidence, from one study at low risk of bias (Dazzi 2003), to determine whether or not GM-CSF reduced the mean pain score on a 0 (no pain) to 10 (worst pain) scale: MD 0.60, 95% CI -0.85 to 2.05; 90 participants (Appendix 8: Analysis 4.4).

Quality of life

No studies assessed this outcome.

Normalcy of diet Adults receiving bone marrow/stem cell transplantation after conditioning therapy for haematological cancers

There was insufficient evidence, from one study at unclear risk of bias (van der Lelie 2001), to determine whether or not GM-CSF reduced the risk of total parenteral nutrition: RR 1.10, 95% CI 0.63 to 1.91; 36 participants (Appendix 8: Analysis 4.5).

Adults receiving radiotherapy to the head and neck

There was insufficient evidence, from one study at high risk of bias (McAleese 2006), to determine whether or not GM-CSF reduced the risk of tube feeding: RR 0.31, 95% CI 0.01 to 7.09; 29 participants (Appendix 8: Analysis 4.5).

Adverse events

This outcome was difficult to summarise due to poor and inconsistent reporting, and we did not meta-analyse any data. However, there did not appear to be any serious concerns regarding adverse effects of GM-CSF. We tabulated relevant information in Appendix 9: Table 2.

Number of days in hospital

No studies assessed this outcome.

137 Number of days of treatment with opioid analgesics Adults receiving bone marrow/stem cell transplantation after conditioning therapy for mixed cancers

There was some evidence, from one study at low risk of bias (Dazzi 2003), that GM-CSF might lead to a reduction in the mean number of days of treatment with opioid analgesics: MD -1.10, 95% CI -1.91 to -0.29; 90 participants (Appendix 8: Analysis 4.6).

Number of days unable to take medicine orally

No studies assessed this outcome.

Granulocyte-macrophage colony-stimulating factor (GM-CSF) dose comparison

There was some very weak evidence, from one study at unclear risk of bias and analysing 36 adults receiving mixed chemotherapy alone for breast cancer (Cartee 1995), that a higher dose of GM-CSF (range 1260 to 12,600 µg) reduced the risk of severe oral mucositis when compared to a lower dose (range 12.6 to 126 µg): RR 2.75, 95% CI 1.07 to 7.04; 36 participants (Appendix 8: Analysis 5.1).

Granulocyte-macrophage colony-stimulating factor (GM-CSF) versus sucralfate

One study, at low risk of bias and analysing 40 adults receiving radiotherapy to the head and neck (Saarilahti 2002), compared GM-CSF with sucralfate, both as a mouthwash. There was insufficient evidence to determine whether GM-CSF or sucralfate performed better in reducing the risk of moderate to severe oral mucositis (RR 0.96, 95% CI 0.80 to 1.14; Analysis 6.1), severe oral mucositis (RR 0.54, 95% CI 0.24 to 1.21; Analysis 6.2), interruptions to cancer treatment (RR 0.13, 95% CI 0.01 to 2.36; Appendix 8: Analysis 6.3), or percutaneous endoscopic gastrostomy (RR 0.18, 95% CI 0.01 to 3.56; Appendix 8: Analysis 6.4).

138

Analysis 6.1: GM-CSF versus sucralfate; Outcome: Oral mucositis (moderate to severe)

Analysis 6.2: GM-CSF versus sucralfate; Outcome: Oral mucositis (severe)

139

Granulocyte-colony stimulating factor (G-CSF) versus placebo/no treatment Oral mucositis Adults receiving radiotherapy to the head and neck

There was insufficient evidence, from two studies at low risk of bias (Schneider 1999; Su 2006), to determine whether or not G-CSF reduced the risk of any level of oral mucositis: RR 1.02, 95% CI 0.86 to 1.22; 54 participants (Analysis 7.1).

The same two studies showed weak evidence (due to a wide confidence interval and low sample size) of a reduction in the risk of severe oral mucositis in favour of G-CSF: RR 0.37, 95% CI 0.15 to 0.87; 54 participants; low quality evidence (Analysis 7.3).

Adults receiving chemotherapy alone for mixed cancers

One study on lung cancer, at unclear risk of bias (Crawford 1999), showed a reduction in the risk of any level of oral mucositis in favour of G-CSF: RR 0.59, 95% CI 0.40 to 0.87; 195 participants (Analysis 7.1).

One study on breast cancer, at high risk of bias (Katano 1995), showed very weak evidence (due to risk of bias, very low sample size and a wide confidence interval) of a reduction in the risk of moderate to severe oral mucositis in favour of G-CSF: RR 0.33, 95% CI 0.12 to 0.95; 14 participants; very low quality evidence (Analysis 7.2).

Adults receiving bone marrow/stem cell transplantation after conditioning therapy for haematological cancers

One study, at high risk of bias and analysing 121 participants (Linch 1993), did not provide any details of how oral mucositis was measured, so it was not clear what severity the information refers to. There were no numerical results reported, only the statement: "There was no difference in the frequency of stomatitis (defined as a sore, infected or ulcerated mouth, lips or pharynx), the incidence being between 29 and 33% in all groups".

140

Analysis 7.1: G-CSF versus control; Outcome: Oral mucositis (any)

141

Analysis 7.2: G-CSF versus control; Outcome: Oral mucositis (moderate to severe)

Analysis 7.3: G-CSF versus control; Outcome: Oral mucositis (severe) 142 Interruptions to cancer treatment Adults receiving radiotherapy to the head and neck

There was insufficient evidence, from one study at low risk of bias (Su 2006), to determine whether or not G-CSF reduced the risk of radiotherapy interruptions: RR 0.22, 95% CI 0.01 to 4.31; 40 participants (Appendix 8: Analysis 7.4).

Oral pain

No studies assessed this outcome.

Quality of life

No studies assessed this outcome.

Normalcy of diet Adults receiving radiotherapy to the head and neck

There was insufficient evidence, from one study at low risk of bias (Su 2006), to determine whether or not G-CSF reduced the risk of percutaneous endoscopic gastrostomy: RR 0.16, 95% CI 0.01 to 2.86; 40 participants (Appendix 8: Analysis 7.5).

Adverse events

This outcome was difficult to summarise due to poor and inconsistent reporting, and we did not meta-analyse any data. However, there did not appear to be any serious concerns regarding adverse effects of G-CSF. We tabulated relevant information in Appendix 9: Table 3.

Number of days in hospital

No studies assessed this outcome.

Number of days of treatment with opioid analgesics

No studies assessed this outcome.

Number of days unable to take medicine orally

No studies assessed this outcome.

G-CSF (pegfilgrastim) versus G-CSF (filgrastim) 143 There was insufficient evidence, from one study at high risk of bias and analysing 61 children receiving bone marrow/stem cell transplantation after conditioning therapy for mixed cancers (Cesaro 2013), to determine whether pegfilgrastim or filgrastim performed better in reducing the risk of any level of oral mucositis (RR 1.02, 95% CI 0.82 to 1.27; Appendix 8: Analysis 8.1), moderate to severe oral mucositis (RR 0.78, 95% CI 0.55 to 1.11; Appendix 8: Analysis 8.2), or total parenteral nutrition (RR 1.00, 95% CI 0.94 to 1.06; Appendix 8: Analysis 8.3).

Epidermal growth factor (EGF) versus placebo Oral mucositis Adults receiving bone marrow/stem cell transplantation after conditioning therapy for haematological cancers

There was insufficient evidence, from one study at low risk of bias and analysing 136 participants (Kim 2017), to determine whether or not EGF reduced the risk of moderate to severe oral mucositis (RR 1.06, 95% CI 0.78 to 1.43; Analysis 9.1), or severe oral mucositis (RR 1.03, 95% CI 0.59 to 1.80; Analysis 9.2).

Adults receiving radiotherapy to the head and neck with/without cisplatin

One study, at high risk of bias (Wu 2009), showed weak evidence (due to risk of bias and low sample size) of a reduction in the risk of moderate to severe oral mucositis in favour of EGF: RR 0.67, 95% CI 0.45 to 0.99; 103 participants (Analysis 9.1).

144

Analysis 9.1: EGF versus placebo; Outcome: Oral mucositis (moderate to severe)

145

Analysis 9.2: EGF versus placebo; Outcome: Oral mucositis (severe)

146 Interruptions to cancer treatment Adults receiving radiotherapy to the head and neck with/without cisplatin

There was insufficient evidence, from one study at high risk of bias (Wu 2009), to determine whether or not EGF reduced the risk of having radiotherapy breaks longer than 2 days: RR 4.38, 95% CI 0.25 to 75.44; 113 participants (Appendix 8: Analysis 9.3).

Oral pain

No studies assessed this outcome.

Quality of life

No studies assessed this outcome.

Normalcy of diet Adults receiving bone marrow/stem cell transplantation after conditioning therapy for haematological cancers

There was insufficient evidence, from one study at low risk of bias (Kim 2017), to determine whether or not EGF reduced the risk of total parenteral nutrition: RR 1.03, 95% CI 0.55 to 1.94; 136 participants (Appendix 8: Analysis 9.4).

Adverse events

There did not appear to be any serious concerns regarding adverse effects of EGF. We tabulated relevant information in Appendix 9: Table 4.

Number of days in hospital

No studies assessed this outcome.

Number of days of treatment with opioid analgesics

No studies assessed this outcome.

Number of days unable to take medicine orally

No studies assessed this outcome.

Intestinal trefoil factor (ITF) versus placebo Oral mucositis 147 Adults receiving chemotherapy alone for colorectal cancer

One study, at unclear risk of bias and analysing 99 participants (Peterson 2009), showed weak evidence (due to low sample size) of a reduction in the risk of any level of oral mucositis (RR 0.52, 95% CI 0.35 to 0.79; Analysis 10.1), and moderate to severe oral mucositis (RR 0.22, 95% CI 0.10 to 0.48; Analysis 10.2), both in favour of ITF.

There was insufficient evidence, from the same study, to determine whether or not EGF reduced the risk of severe oral mucositis: RR 1.52, 95% CI 0.06 to 36.39 (Analysis 10.3).

148

Analysis 10.1: ITF versus placebo; Outcome: Oral mucositis (any)

Analysis 10.2: ITF versus placebo; Outcome: Oral mucositis (moderate to severe) 149

Analysis 10.3: ITF versus placebo; Outcome: Oral mucositis (severe)

150 Interruptions to cancer treatment

No studies assessed this outcome.

Oral pain

No studies assessed this outcome.

Quality of life

No studies assessed this outcome.

Normalcy of diet

No studies assessed this outcome.

Adverse events

There did not appear to be any serious concerns regarding adverse effects of ITF. We tabulated relevant information in Appendix 9: Table 5.

Number of days in hospital

No studies assessed this outcome.

Number of days of treatment with opioid analgesics

No studies assessed this outcome.

Number of days unable to take medicine orally

No studies assessed this outcome.

Intestinal trefoil factor (ITF) dose comparison

There was insufficient evidence, from one study at unclear risk of bias and analysing 66 adults receiving chemotherapy alone for colorectal cancer (Peterson 2009), to determine whether a lower dose (336 mg) or a higher dose (2688 mg) performed better in reducing the risk of oral mucositis of any severity (Appendix 8: Analysis 11.1; Analysis 11.2; Analysis 11.3).

Erythropoietin versus placebo 151 Oral mucositis Adults receiving bone marrow/stem cell transplantation after conditioning therapy for haematological cancers

One study, at low risk of bias and analysing 80 participants (Hosseinjani 2017), showed weak evidence (due to low sample size) of a reduction in the risk of any level of oral mucositis (RR 0.35, 95% CI 0.21 to 0.60; Analysis 12.1), and moderate to severe oral mucositis (RR 0.43, 95% CI 0.24 to 0.79; Analysis 12.2), both in favour of erythropoietin.

The same study showed weak evidence (due to low sample size and a wide confidence interval) that erythropoietin might reduce the risk of severe oral mucositis, but there was also some possibility of harm: RR 0.40, 95% CI 0.14 to 1.17 (Analysis 12.3).

152

Analysis 12.1: Erythropoietin versus placebo; Outcome: Oral mucositis (any)

Analysis 12.2: Erythropoietin versus placebo; Outcome: Oral mucositis (moderate to severe) 153

Analysis 12.3: Erythropoietin versus placebo; Outcome: Oral mucositis (severe)

154 Interruptions to cancer treatment

No studies assessed this outcome.

Oral pain

No studies assessed this outcome.

Quality of life

No studies assessed this outcome.

Normalcy of diet

No studies assessed this outcome.

Adverse events

No studies assessed this outcome.

Number of days in hospital Adults receiving bone marrow/stem cell transplantation after conditioning therapy for haematological cancers

There was insufficient evidence, from one study at low risk of bias (Hosseinjani 2017), to determine whether or not erythropoietin reduced the mean number of days in hospital: MD -2.95, 95% CI -7.73 to 1.83; 80 participants (Appendix 8: Analysis 12.4).

Number of days of treatment with opioid analgesics

No studies assessed this outcome.

Number of days unable to take medicine orally

No studies assessed this outcome.

Transforming growth factor (TGF) versus placebo Oral mucositis Adults receiving chemotherapy alone for colorectal cancer

155 There was insufficient evidence, from one study at high risk of bias and analysing 13 participants (Antoun 2009), to determine whether or not TGF reduced the risk of any level of oral mucositis: RR 0.10, 95% CI 0.01 to 1.71 (no forest plot presented).

Interruptions to cancer treatment

No studies assessed this outcome.

Oral pain

No studies assessed this outcome.

Quality of life

No studies assessed this outcome.

Normalcy of diet

No studies assessed this outcome.

Adverse events

No studies assessed this outcome.

Number of days in hospital

No studies assessed this outcome.

Number of days of treatment with opioid analgesics

No studies assessed this outcome.

Number of days unable to take medicine orally

No studies assessed this outcome.

156 4.6 Summary of findings tables

1 KGF compared to placebo for preventing oral mucositis in adults with cancer receiving treatment

KGF compared to placebo for preventing oral mucositis in adults with cancer receiving treatment Patient or population: adults** receiving treatment for cancer (see subgroup for treatment type) Setting: hospital Intervention: KGF Comparison: placebo Outcomes Anticipated absolute effects* Relative № of Quality of the Comments (95% CI) effect participants evidence Risk with Risk with (95% CI) (studies) (GRADE) placebo KGF Oral mucositis BMT/SCT after conditioning for RR 0.89 852 ⊕⊕⊕⊝ There is probably a benefit for KGF in this (moderate + haematological cancers (0.80 to (6 study) MODERATE 1 population severe) 848 per 1,000 755 per 1,000 0.99) (678 to 839) NNTB = 11 (95% CI 6 to 112) RT to head & neck with RR 0.91 471 ⊕⊕⊕⊝ There is probably a benefit for KGF in this cisplatin/5FU (0.83 to (3 study) MODERATE 1 population 932 per 1,000 848 per 1,000 1.00) (773 to 932) NNTB = 12 (95% CI 7 to ∞) CT alone for mixed cancers RR 0.56 344 ⊕⊕⊕⊕ It is very likely that there is a benefit for KGF 631 per 1,000 353 per 1,000 (0.45 to (4 study) HIGH in this population (284 to 441) 0.70) NNTB = 4 (95% CI 3 to 6) Oral mucositis BMT/SCT after conditioning for RR 0.85 852 ⊕⊕⊕⊝ There might be a benefit for KGF in this (severe) haematological cancers (0.65 to (6 study) population, but there is also some possibility

157 677 per 1,000 575 per 1,000 1.11) MODERATE 2 of harm (440 to 751) NNTB = 10 (95% CI 5 NNTB to 14 NNTH) RT to head & neck with RR 0.79 471 ⊕⊕⊕⊕ It is very likely that there is a benefit for KGF cisplatin/5FU (0.69 to (3 study) HIGH in this population 700 per 1,000 553 per 1,000 0.90) (483 to 630) NNTB = 7 (95% CI 5 to 15) CT alone for mixed cancers RR 0.30 263 ⊕⊕⊕⊝ There is probably a benefit for KGF in this 3 154 per 1,000 46 per 1,000 (0.14 to (3 study) MODERATE population (22 to 100) 0.65) NNTB = 10 (95% CI 8 to 19) Adverse events that were attributed to the study drugs rather than the cancer therapy were typically oral-related or skin-related. Adverse events Events were mostly mild to moderate with very few incidences of serious events. However, reporting was poor and inconsistent, meaning that it was not appropriate to meta-analyse data *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). **Only one study in the subgroup 'BMT/SCT after conditioning for haematological cancers' included some children (but the median age of participants was 46 years). ***The number of people that would need to receive KGF in order to prevent one additional person from developing the outcome. Calculated as 1 divided by the absolute risk reduction (which is the control arm event rate minus the experimental arm event rate). NNTH means the number of people that would need to receive KGF to cause one additional person to develop the outcome. All decimal places have been rounded up to the nearest whole number (i.e. 6.1 = 7).

∞: infinity; 5FU: fluorouracil; BMT: bone marrow transplantation; CI: Confidence interval; CT: chemotherapy; KGF: keratinocyte growth factor; NNTB: number needed to treat to benefit***; NNTH: number needed to treat to harm; RR: Risk ratio; RT: radiotherapy; SCT: stem cell transplantation GRADE Working Group grades of evidence High quality: We are very confident that the true effect lies close to that of the estimate of the effect

158 Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

Footnotes

1 Downgraded by one level for inconsistency (substantial heterogeneity: I² = 50% to 90%, P < 0.1)

2 Downgraded by one level for inconsistency (substantial heterogeneity: I² = 50% to 90%, P < 0.1); we did not downgrade for imprecision because, despite the confidence interval including a small chance of harm, it is a fairly narrow interval and a rating of 'low quality' would seem an overly harsh rating for this body of evidence

3 Downgraded by one level for imprecision (wide confidence interval, small sample size and low event rate)

159 2 GM-CSF compared to placebo/no treatment for preventing oral mucositis in adults with cancer receiving treatment

GM-CSF compared to placebo/no treatment for preventing oral mucositis in adults with cancer receiving treatment Patient or population: adults** receiving treatment for cancer (see subgroup for treatment type) Setting: hospital Intervention: GM-CSF Comparison: placebo/no treatment Outcomes Anticipated absolute effects* (95% Relative № of Quality of the Comments CI) effect participants evidence Risk with placebo/no Risk with (95% CI) (studies) (GRADE) treatment GM-CSF Oral mucositis BMT/SCT after conditioning for RR 0.94 109 ⊕⊝⊝⊝ There is insufficient evidence to determine (moderate + haematological cancers (0.79 to (1 study) VERY LOW 1 a benefit for GM-CSF in this population severe) 839 per 1,000 789 per 1,000 1.13) (663 to 948) NNTB = 20 (95% CI 6 NNTB to 10 NNTH) RT to head & neck RR 0.72 29 ⊕⊝⊝⊝ There is insufficient evidence to determine 2 929 per 1,000 669 per 1,000 (0.49 to (1 study) VERY LOW a benefit for GM-CSF in this population (455 to 984) 1.06) NNTB = 4 (95% CI 3 NNTB to 14 NNTH) Oral mucositis BMT/SCT after conditioning for RR 0.74 235 ⊕⊕⊝⊝ There is insufficient evidence to determine (severe) mixed cancers (0.33 to (3 study) LOW 3 a benefit for GM-CSF in this population 347 per 1,000 257 per 1,000 1.67) (115 to 580) NNTB = 12 (95% CI 5 NNTB to 5 NNTH) RT to head & neck RR 0.31 29 ⊕⊝⊝⊝ There is insufficient evidence to determine 4 71 per 1,000 22 per 1,000 (0.01 to (1 study) VERY LOW a benefit for GM-CSF in this population (1 to 506) 7.09) NNTB = 21 (95% CI 15 NNTB to 3

160 NNTH) CT alone for mixed cancers RR 0.59 65 ⊕⊝⊝⊝ There is insufficient evidence to determine 5 500 per 1,000 295 per 1,000 (0.05 to (2 study) VERY LOW a benefit for GM-CSF in this population (25 to 1,000) 7.11) NNTB = 5 (95% CI 3 NNTB to 2 NNTH) Adverse events that were attributed to the study drugs rather than the cancer therapy were typically bone pain, nausea, fever and Adverse events headache. Events were not reported as being serious. Some studies did not report adverse events and one even reported that there were none. However, reporting was poor and inconsistent, meaning that it was not appropriate to meta-analyse data *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). **There were no studies conducted on children. ***The number of people that would need to receive GM-CSF in order to prevent one additional person from developing the outcome. Calculated as 1 divided by the absolute risk reduction (which is the control arm event rate minus the experimental arm event rate). NNTH means the number of people that would need to receive GM-CSF to cause one additional person to develop the outcome. All decimal places have been rounded up to the nearest whole number (i.e. 6.1 = 7).

BMT: bone marrow transplantation; CI: Confidence interval; CT: chemotherapy; GM-CSF: granulocyte-macrophage colony-stimulating factor; NNTB: number needed to treat to benefit***; NNTH: number needed to treat to harm; RR: Risk ratio; RT: radiotherapy; SCT: stem cell transplantation GRADE Working Group grades of evidence High quality: We are very confident that the true effect lies close to that of the estimate of the effect Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

Footnotes

161 1 Downgraded by two levels for imprecision (single study with a small sample size and the confidence interval includes a possible increase in risk that is of a similar magnitude to the possible reduction in risk); downgraded one further level for indirectness (single study so not widely generalisable)

2 Downgraded by two levels for imprecision (wide confidence interval and very small sample size); downgraded by one further level for high risk of performance bias; downgraded by one further level for indirectness (single study so not widely generalisable)

3 Downgraded by two levels for imprecision (small sample size and the confidence interval includes a possible increase in risk that is of a similar magnitude to the possible reduction in risk); downgraded by one further level for inconsistency (substantial heterogeneity: I² = 50% to 90%, P < 0.1)

4 Downgraded by two levels for imprecision (extremely wide confidence interval incorporating both very large increase and reduction in risk, very small sample size and very low event rate); downgraded by one further level for high risk of performance bias; downgraded by one further level for indirectness (single study so not widely generalisable)

5 Downgraded by two levels for imprecision (extremely wide confidence interval incorporating both very large increase and reduction in risk and very small sample size); downgraded by one further level for high risk of performance bias; downgraded by one further level for inconsistency (substantial heterogeneity: I² = 50% to 90%, P < 0.1)

162 3 G-CSF compared to placebo/no treatment for preventing oral mucositis in adults with cancer receiving treatment

G-CSF compared to placebo/no treatment for preventing oral mucositis in patients with cancer receiving treatment Patient or population: adults** receiving treatment for cancer (see subgroup for treatment type) Setting: hospital Intervention: G-CSF Comparison: placebo/no treatment Outcomes Anticipated absolute effects* Relative № of Quality of the Comments (95% CI) effect participants evidence Risk with placebo/no Risk with (95% CI) (studies) (GRADE) treatment G-CSF Oral mucositis CT alone for breast cancer RR 0.33 14 ⊕⊝⊝⊝ There is very weak evidence that there might 1 (moderate + 1,000 per 1,000 330 per (0.12 to (1 study) VERY LOW be a benefit for G-CSF in this population severe) 1,000 0.95) (120 to NNTB = 2 (95% CI 2 to 20) 950) Oral mucositis RT to head & neck RR 0.37 54 ⊕⊕⊝⊝ There is weak evidence that there might be a 2 (severe) 519 per 1,000 192 per (0.15 to (2 study) LOW benefit for G-CSF in this population 1,000 0.87) (78 to 451) NNTB = 3 (95% CI 3 to 15) There was limited evidence of adverse events for G-CSF. Two of the six studies did not report adverse events. There were low rates Adverse events of mild to moderate events, the most common of which appeared to be bone pain. However, reporting was poor and inconsistent, meaning that it was not appropriate to meta-analyse data *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). **There were no studies conducted on children.

163 ***The number of people that would need to receive G-CSF in order to prevent one additional person from developing the outcome. Calculated as 1 divided by the absolute risk reduction (which is the control arm event rate minus the experimental arm event rate). NNTH means the number of people that would need to receive G-CSF to cause one additional person to develop the outcome. All decimal places have been rounded up to the nearest whole number (i.e. 6.1 = 7).

CI: Confidence interval; CT: chemotherapy; G-CSF: granulocyte-colony stimulating factor; NNTB: number needed to treat to benefit***; NNTH: number needed to treat to harm; RR: Risk ratio; RT: radiotherapy GRADE Working Group grades of evidence High quality: We are very confident that the true effect lies close to that of the estimate of the effect Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

Footnotes

1 Downgraded by two levels for imprecision (wide confidence interval and very small sample size); downgraded by one further level for high risk of performance bias; downgraded by one further level for indirectness (single study so not widely generalisable)

2 Downgraded by two levels for imprecision (wide confidence interval and very small sample size)

164 4.7 Discussion

Summary of main results

Thirty-five studies met our eligibility criteria and were included in this review. We used GRADE methodology to assess the quality of the body of evidence for each of the main comparisons and for the primary outcome of incidence and severity of oral mucositis (45). Most of the evidence we found was for keratinocyte growth factor (KGF: Summary of findings table 1), granulocyte-macrophage colony-stimulating factor (GM-CSF: Summary of findings table 2), and granulocyte-colony stimulating factor (G-CSF: Summary of findings table 3).

Our main findings were as follows.

KGF Moderate to severe oral mucositis

 Adults receiving bone marrow/stem cell transplantation after conditioning therapy for haematological cancer: probably a reduction in risk (11% and ranging from 20% to 1%).  Adults receiving radiotherapy to the head and neck with cisplatin/fluorouracil (5FU): probably a reduction in risk (9% and ranging from 17% to no reduction).  Adults receiving chemotherapy alone for mixed cancers: very likely a reduction in risk (44% and ranging from 55% to 30%).

Severe oral mucositis

 Adults receiving bone marrow/stem cell transplantation after conditioning therapy for haematological cancer: might be a reduction in risk, but some possibility of harm (15% reduction and ranging from 35% reduction to 11% increase).  Adults receiving radiotherapy to the head and neck with cisplatin/fluorouracil (5FU): very likely a reduction in risk (21% and ranging from 31% to 10%).  Adults receiving chemotherapy alone for mixed cancers: probably a reduction in risk (60% and ranging from 86% to 35%).

GM-CSF Moderate to severe oral mucositis

165  Adults receiving bone marrow/stem cell transplantation after conditioning therapy for haematological cancer: insufficient evidence of a benefit.  Adults receiving radiotherapy to the head and neck: insufficient evidence of a benefit.

Severe oral mucositis

 Adults receiving bone marrow/stem cell transplantation after conditioning therapy for mixed cancers: insufficient evidence of a benefit.  Adults receiving radiotherapy to the head and neck: insufficient evidence of a benefit.  Adults receiving chemotherapy alone for mixed cancers: insufficient evidence of a benefit.

G-CSF Moderate to severe oral mucositis

 Adults receiving chemotherapy alone for breast cancer: very weak evidence of a possible reduction in risk (67% and ranging from 88% to 5%).

Severe oral mucositis

 Adults receiving radiotherapy to the head and neck: weak evidence of a possible reduction in risk (63% and ranging from 85% to 13%).

The remaining evidence for the primary outcome was from single-study comparisons.

 Epidermal growth factor might reduce the risk of moderate to severe oral mucositis in adults receiving radiotherapy to the head and neck with or without cisplatin, but there was insufficient evidence of a reduction in the risk of either moderate to severe, or severe oral mucositis in adults receiving bone marrow/stem cell transplantation after conditioning therapy for haematological cancer.  Intestinal trefoil factor might reduce the risk of moderate to severe oral mucositis in adults receiving chemotherapy alone for colorectal cancer.  Erythropoietin might reduce the risk of moderate to severe oral mucositis in adults receiving bone marrow/stem cell transplantation after conditioning therapy for haematological cancer.

166 There was mostly insufficient evidence of a benefit regarding the secondary outcomes of this review. The interventions investigated all appear to be relatively safe, with only mild to moderate adverse effects reported.

Overall completeness and applicability of evidence

The evidence we have presented in this review allows for some conclusions to be made regarding the effects of KGF for preventing oral mucositis in adults receiving certain types of cancer treatment. However, the evidence is lacking for other cytokines and growth factors, and for children. It is unfortunate that the two studies we found on KGF versus placebo in children were unclear in their reporting and we were unable to present any data. All studies reported on our primary outcome, but the evidence for the secondary outcomes of this review is lacking.

The evidence for KGF should have reasonable external validity as most of the adult population were covered in terms of the types of treatment people have for different types of cancer. The studies were also carried out all over the world and often involved multiple sites. One limitation, however, may be the fact that most studies were done in developed countries, so may be less generalisable to people in developing countries.

Numerous studies reported on some of our secondary outcomes but did not report the data in a suitable format for inclusion in our meta-analyses e.g. as median with or without range, area under the curve, or as mean (or a graph) but with no standard deviation/standard error/P-value. In such cases, the meta-analysis is biased by missing information. However, the Cochrane risk of bias tool and meta-analyses do not currently address this issue adequately. The study may be assessed at high risk of selective outcome reporting, but if the study is not included in the meta-analysis due to having no data, then this is not reflected or accounted for. This highlights the need for standardisation in both 'what to measure' and 'how to measure it' in clinical trials in this area of research. Otherwise there will continue to be research waste, with data that are not able to be pooled in data syntheses. There are initiatives such as COMET (Core Outcome Measures in Effectiveness Trials) and COSMIN (COnsensus-based Standards for the selection of health Measurement INstruments) that can help with these issues, and future research in these areas would be beneficial.

During the systematic review process, we developed further concerns regarding the usefulness of the secondary outcomes because it was not clear whether or not they were due to oral mucositis. Hospitalisation, the use of supplemental nutrition or opioid 167 analgesics, and interruptions to cancer treatment could all occur due to reasons other than oral mucositis. Furthermore, it was not always clear if adverse effects were due to the interventions given to prevent oral mucositis. These issues could be improved by clearer and more explicit reporting.

Quality of the evidence

We included 35 randomised controlled trials (RCTs) analysing 3102 participants. Despite this large volume of research, we were not able to make robust conclusions about the effects of most cytokines and growth factors. The strongest body of evidence, both in terms of volume and quality, was for the different populations receiving KGF. We assessed the bodies of evidence for KGF both in preventing moderate to severe oral mucositis in adults receiving chemotherapy alone for mixed cancers, and in preventing severe oral mucositis in adults receiving radiotherapy to the head and neck with or without cisplatin or fluorouracil, as high quality. The remaining evidence for KGF was moderate quality, with the main reason for downgrading being due to inconsistency in the individual study results (heterogeneity). There were no concerns over risk of bias in the studies as they are often large multicentre trials which are carried out well, mostly using placebos for blinding purposes, and with very low attrition.

The evidence for GM-CSF and G-CSF was much weaker, and consequently was rated as being low or very low quality. The reasons for downgrading were mostly due to imprecision because the volume of evidence was much lower and the studies often recruited very few participants, leading to very wide confidence intevals that frequently included both the possibility of benefit and harm. Further reasons were risk of performance bias due to lack of blinding in some of these studies, inconsistency, and also because some of the evidence was from single studies. When a body of evidence was from a single study, we automatically downgraded a level. The reasoning behind this was because often, when using GRADE methodology, bodies of evidence are downgraded for inconsistency due to different effect estimates in the individual studies. This inconsistency is not possible for a single-study body of evidence and therefore not downgrading would falsely inflate the rating of quality, whilst at the same time the larger body of evidence is unfairly penalised, in comparison, due to having more studies. In such cases, we downdgraded the single- study evidence due to indirectness as it may only be generalisable to the particular population who took part in the study.

168 The remaining evidence for other interventions was from single-study comparisons and therefore was all considered to be of low to very low quality, mainly for indirectness (as described above) and imprecision.

Potential biases in the review process

Although systematic review methodology is designed to minimise biases in the process, decisions are often made out or necessity or for practical reasons, and this can introduce some potential bias.

Once we began to assess the literature identified by the searches, we became concerned that we may have missed some relevant studies because we had not included search terms for other conditions for which cytokines and growth factors have been used to manage (e.g. diahhroea, graft-versus-host disease, and neutropenia). In order to assess the extent of this potential problem, a rough scoping search was run including the search terms. The yield was very high and a single review author assessed a sample of 500 records, but no further studies were identified. Therefore, we decided not to amend the search by adding the new search terms. We acknowledge the possibility that we have missed some studies that have measured and reported on oral mucositis but not mentioned it in the abstract. This could introduce bias if there are relevant data missing from the review.

There were some studies that had multiple treatment arms with different doses of the cytokine or growth factor. In all instances, we combined the arms to make a pairwise comparison against the control arm thus losing some possible subtleties of the data and potentially biasing the results.

Agreements and disagreements with other studies or reviews

The Mucositis Study Group (MSG) of the Multinational Association of Supportive Care in Cancer/International Society of Oral Oncology (MASCC/ISOO) is the leading international group in this area of research. In 2013, they published a series of systematic reviews on the different interventions for managing oral mucositis, including one on

169 cytokines and growth factors (22). These reviews feed into the 'MASCC/ISOO Clinical Practice Guidelines for the Management of Mucositis Secondary to Cancer Therapy’ (65). The MASCC/ISOO systematic review is not limited to RCTs. The current guidance from this group is as follows:

 Recommendations in favour of an intervention (i.e. strong evidence supporting effectiveness): The panel recommends that recombinant human keratinocyte growth factor-1 (KGF-1/palifermin) be used to prevent oral mucositis (at a dose of 60 µg/kg per day for 3 days prior to conditioning treatment and for 3 days after transplant) in patients receiving high-dose chemotherapy and total body irradiation, followed by autologous stem cell transplantation, for a hematological malignancy (level II evidence).  Suggestions against an intervention (i.e. weaker evidence indicating lack of effectiveness): The panel suggests that granulocyte-macrophage-colony-stimulating factor mouthwash not be used to prevent oral mucositis in patients receiving high- dose chemotherapy, for autologous or allogeneic stem cell transplantation (level II evidence).

For our meta-analyses for KGF in the above mentioned population, we combined studies of all types of KGF, both with autologous and allogeneic transplants, and with total body irradiation (TBI), without TBI or a mixture of TBI/no TBI. The MASCC/ISOO systematic review separated all of these factors. However, looking at the individual studies in our meta-analyses, the first recommendation appears to be a valid one. Furthermore, the MASCC/ISOO systematic review states "Evidence on the efficacy of palifermin in autologous HSCT without TBI conditioning is conflicting...and these rather small studies did not allow a guideline. In addition, no guideline could be provided for the use of palifermin in the setting of allogeneic HSCT with or without TBI". Despite our meta- analyses including some further RCTs not included in the other review, these statements also appear to be valid.

The suggestion against GM-CSF mouthwash is also a valid one as, although we did not separate studies by mode of administration, it is clear that the two mouthwash studies in our analysis (Analysis 4.3) have conflicting results. However, based on one study on GM- CSF given intravenously in this population (Nemunaitis 1995), there is promising evidence of a benefit, but the MASCC/ISOO systematic review considered this evidence alongside other studies that we did not include, and concluded that there was no guideline possible.

170 Our results are not in agreement with the following statements from the MASCC/ISOO systematic review regarding other populations receiving KGF:

 "No guideline could be provided for the use of palifermin in the setting of CT for solid and hematological tumors...due to insufficient evidence"  "In addition, no guideline could be provided for the use of palifermin in H&N RT due to insufficient evidence"

We present moderate to high quality evidence of a benefit for KGF in both of these populations, possibly warranting new guideline statements in their next update. This evidence would equate to level I evidence in the grading system used in the guidelines ("evidence obtained from meta-analysis of multiple, well-designed, controlled studies").

4.8 Authors' conclusions

Implications for practice

We are confident that keratinocyte growth factor (KGF) reduces the risk of developing moderate to severe and severe oral mucositis in adults who are receiving: a) radiotherapy to the head and neck with cisplatin or fluorouracil; or b) chemotherapy alone for mixed cancers (both solid and haematological). We are less confident about a benefit for KGF in adults receiving bone marrow/stem cell transplant after conditioning therapy for haematological cancers because of multiple factors involved in that population, such as whether or not they received total body irradiation (TBI) and whether the transplant was autologous (the patients' own cells) or allogeneic (cells from a donor). KGF appears to be a relatively safe intervention.

Implications for research

Despite a large volume of research, once studies are categorised by cancer treatment type/population, there is very little we can conclude regarding the effects of most cytokines and growth factors. It is clear that much more research is needed in this area, especially as many of the interventions have shown promise in some populations, yet we have not been

171 able to make robust conclusions due to the limited volume/low sample sizes. More randomised controlled trials (RCTs) of KGF are needed in the population receiving bone marrow/stem cell transplant after conditioning therapy so that in future updates we may be able to include separate subgroups to account for differing factors such as TBI/no TBI and autologous/allogeneic transplant. Further large confirmatory RCTs of KGF would be beneficial in the other two populations: a) radiotherapy to the head and neck with chemotherapy (and possibly without chemotherapy); and b) chemotherapy alone for mixed cancers.

More research is needed on all other cytokines and growth factors in the various populations, including in children. Placebo controls should be used in the first instance to establish whether or not they are effective, and only then should head-to-head comparisons of active interventions be made.

Future RCTs should be adequately powered to detect a difference if one actually exists and they should be reported according to The CONSORT Statement (Consolidated Standards of Reporting Trials). They should measure and report in full all the outcomes listed in this review, most of which are recommended in the core outcome set produced by Bellm et al (71). For our primary outcome of oral mucositis incidence, we urge trialists to use a measurement tool such as the WHO or NCI-NCT scale (Appendix 3), to allow us to combine the data with those already included in this review. Reporting the maximum grade of oral mucositis experienced per participant would allow us to assess the incidence of different severities, thus maximising the usefulness of the data. It would also be useful if oral pain was measured on a 0 to 10 scale and reported as an overall mean and mean maximum score experienced per participant. Numbers included in any analysis should always be reported and any continuous data should be reported as means and standard deviations. Furthermore, measurement of outcomes should be taken with appropriate frequency so as to avoid any problems with ascertainment bias.

172 5 PHARMACOLOGICAL INTERVENTIONS FOR PREVENTING DRY MOUTH AND SALIVARY GLAND DYSFUNCTION FOLLOWING RADIOTHERAPY

5.1 Abstract

Background

Salivary gland dysfunction is an 'umbrella' term for the presence of either xerostomia (subjective sensation of dryness), or salivary gland hypofunction (reduction in saliva production). It is a predictable side effect of radiotherapy to the head and neck region, and is associated with a significant impairment of quality of life. A wide range of pharmacological interventions, with varying mechanisms of action, have been used for the prevention of radiation-induced salivary gland dysfunction.

Objectives

To assess the effects of pharmacological interventions for the prevention of radiation- induced salivary gland dysfunction.

Search methods

Cochrane Oral Health's Information Specialist searched the following databases: Cochrane Oral Health's Trials Register (to 14 September 2016); the Cochrane Central Register of Controlled Trials (CENTRAL; 2016, Issue 8) in the Cochrane Library (searched 14 September 2016); MEDLINE Ovid (1946 to 14 September 2016); Embase Ovid (1980 to 14 September 2016); CINAHL EBSCO (Cumulative Index to Nursing and Allied Health Literature; 1937 to 14 September 2016); LILACS BIREME Virtual Health Library (Latin American and Caribbean Health Science Information database; 1982 to 14 September 2016); Zetoc Conference Proceedings (1993 to 14 September 2016); and OpenGrey (1997 to 14 September 2016). We searched the US National Institutes of Health Ongoing Trials Register and the World Health Organization International Clinical Trials Registry Platform for ongoing trials. No restrictions were placed on the language or date of publication when searching the electronic databases.

Selection criteria

We included randomised controlled trials, irrespective of their language of publication or publication status. Trials included participants of all ages, ethnic origin and gender,

173 scheduled to receive radiotherapy on its own or in addition to chemotherapy to the head and neck region. Participants could be outpatients or inpatients. We included trials comparing any pharmacological agent regimen, prescribed prophylactically for salivary gland dysfunction prior to or during radiotherapy, with placebo, no intervention or an alternative pharmacological intervention. Comparisons of radiation techniques were excluded.

Data collection and analysis

We used standard methodological procedures expected by Cochrane.

Main results

We included 39 studies that randomised 3520 participants; the number of participants analysed varied by outcome and time point. The studies were ordered into 14 separate comparisons with meta-analysis only being possible in three of those.

We found low-quality evidence to show that amifostine, when compared to a placebo or no treatment control, might reduce the risk of moderate to severe xerostomia (grade 2 or higher on a 0 to 4 scale) at the end of radiotherapy (risk ratio (RR) 0.35, 95% confidence interval (CI) 0.19 to 0.67; P = 0.001, 3 studies, 119 participants), and up to three months after radiotherapy (RR 0.66, 95% CI 0.48 to 0.92; P = 0.01, 5 studies, 687 participants), but there is insufficient evidence that the effect is sustained up to 12 months after radiotherapy (RR 0.70, 95% CI 0.40 to 1.23; P = 0.21, 7 studies, 682 participants).

We found very low-quality evidence that amifostine increased unstimulated salivary flow rate up to 12 months after radiotherapy, both in terms of mg of saliva per 5 minutes (mean difference (MD) 0.32, 95% CI 0.09 to 0.55; P = 0.006, 1 study, 27 participants), and incidence of producing greater than 0.1 g of saliva over 5 minutes (RR 1.45, 95% CI 1.13 to 1.86; P = 0.004, 1 study, 175 participants). However, there was insufficient evidence to show a difference when looking at stimulated salivary flow rates.

There was insufficient (very low-quality) evidence to show that amifostine compromised the effects of cancer treatment when looking at survival measures. There was some very low-quality evidence of a small benefit for amifostine in terms of quality of life (10-point scale) at 12 months after radiotherapy (MD 0.70, 95% CI 0.20 to 1.20; P = 0.006, 1 study, 180 participants), but insufficient evidence at the end of and up to three months postradiotherapy. A further study showed no evidence of a difference at 6, 12, 18 and 24

174 months postradiotherapy. There was low-quality evidence that amifostine is associated with increases in: vomiting (RR 4.90, 95% CI 2.87 to 8.38; P < 0.00001, 5 studies, 601 participants); hypotension (RR 9.20, 95% CI 2.84 to 29.83; P = 0.0002, 3 studies, 376 participants); nausea (RR 2.60, 95% CI 1.81 to 3.74; P < 0.00001, 4 studies, 556 participants); and allergic response (RR 7.51, 95% CI 1.40 to 40.39; P = 0.02, 3 studies, 524 participants).

We found insufficient evidence (that was of very low quality) to determine whether or not pilocarpine performed better or worse than a placebo or no treatment control for the outcomes: xerostomia, salivary flow rate, survival, and quality of life. There was some low-quality evidence that pilocarpine was associated with an increase in sweating (RR 2.98, 95% CI 1.43 to 6.22; P = 0.004, 5 studies, 389 participants).

We found insufficient evidence to determine whether or not palifermin performed better or worse than placebo for: xerostomia (low quality); survival (moderate quality); and any adverse effects.

There was also insufficient evidence to determine the effects of the following interventions: biperiden plus pilocarpine, Chinese medicines, bethanechol, artificial saliva, selenium, antiseptic mouthrinse, antimicrobial lozenge, polaprezinc, azulene rinse, and Venalot Depot (coumarin plus troxerutin).

Authors' conclusions

There is some low-quality evidence to suggest that amifostine prevents the feeling of dry mouth in people receiving radiotherapy to the head and neck (with or without chemotherapy) in the short- (end of radiotherapy) to medium-term (three months postradiotherapy). However, it is less clear whether or not this effect is sustained to 12 months postradiotherapy. The benefits of amifostine should be weighed against its high cost and side effects. There was insufficient evidence to show that any other intervention is beneficial.

5.2 Background

Description of the condition

175 Xerostomia (dry mouth) has been defined as the "subjective sensation of dryness" (161), whilst salivary gland hypofunction has been defined as "any objectively demonstrable reduction in either whole and/or individual salivary gland flow rates" (162). Xerostomia is usually the result of a decrease in the volume of saliva secreted. Indeed, healthy individuals complain of a dry mouth when their unstimulated whole salivary flow rate falls below 50% of their normal level (32). However, xerostomia may also occur without a reduction in salivary flow (33), possibly resulting from a change in composition of saliva secreted (34). Thus, xerostomia may, or may not be associated with salivary gland hypofunction. Salivary gland dysfunction is an 'umbrella' term for the presence of either xerostomia, or salivary gland hypofunction.

Salivary gland dysfunction is an extremely common side effect of radiotherapy to the head and neck region (163). The total dose for a course of radiotherapy for head and neck cancer is 50 Gy (gray) to 70 Gy (36). However, doses over 52 Gy will cause severe salivary gland dysfunction (33). A major decrease in saliva flow develops within one week of starting radiotherapy, and continues to deteriorate throughout treatment, culminating in permanent salivary gland dysfunction (36). Indeed, even a dose of 20 Gy is enough to permanently damage salivary flow if it is given as a single dose (33). Salivary gland hypofunction is associated with a variety of oral problems in this group of people (e.g. oral discomfort, taste disturbance, difficulty chewing, difficulty swallowing, speech problems, dental caries, oral candidiasis, and other oral infections). Certainly salivary gland dysfunction is associated with a significant impairment of quality of life in this group of patients.

Description of the intervention

The literature discusses a wide range of pharmacological interventions for preventing radiation-induced salivary gland dysfunction. Examples of these include.

Parasympathomimetic drugs (choline esters, cholinesterase inhibitors)

Parasympathomimetic drugs stimulate salivary secretion by stimulating the parasympathetic nervous system. The parasympathetic nervous system increases bodily secretions such as tears, gastric juices, mucus and saliva to defend the body and help digestion. The most widely used parasympathomimetic drug in this clinical situation is pilocarpine hydrochloride (a choline ester) and has been licensed in many countries for the treatment of radiation-induced salivary gland dysfunction (164). Other indirectly acting parasympathomimetics, for example bethanecol, are much more widely used in other contexts, but have also been used 'off-licence' to treat this condition (165). 176 Parasympatholytic drugs

Parasympatholytic drugs have the opposite effect to parasympathomimetic drugs, their action is anticholinergic, i.e. they inhibit the secretion of saliva. Results from animal tests (166) and a study by Rode et al (167, 168) suggest that the inhibition of salivary secretion during radiotherapy might actually protect later damage of the salivary glands and improve salivation following the treatment.

Cytoprotective agents

Cytoprotective agents can be administered before, with, or after cancer therapy to reduce or prevent damage or toxicity to the normal cells and tissues without compromising therapeutic efficacy. Amifostine is a cytoprotective agent and has been shown to accumulate in the salivary glands (169); there are reports that this might lead to a reduction in parotid parenchymal damage due to radiotherapy (170), and reduce the incidence of radiation-induced xerostomia (171).

Why it is important to do this review

Salivary gland dysfunction is a significant and mostly permanent side effect of radiotherapy to the head and neck region that has numerous knock-on effects, negatively affecting quality of life. Unfortunately, the evidence for prevention using pharmacological agents is weak and some guideline statements do not currently recommend any (39). Although there is a recently published Cochrane Review looking at parasympathomimetic drugs for treating radiation-induced salivary gland dysfunction (172), other drugs with different modes of action have the potential to be effective in this situation, and a broader review of prophylactic measures was needed.

5.3 Objectives

To assess the effects of pharmacological interventions for the prevention of radiation- induced salivary gland dysfunction.

5.4 Methods

Criteria for considering studies for this review Types of studies

177 We included randomised controlled trials of parallel design. Trials were included irrespective of language of publication or publication status.

Types of participants

We included participants of all ages, ethnic origin and gender scheduled to receive radiotherapy on its own or in addition to chemotherapy to the head and neck region. Participants could be outpatients or inpatients.

Types of interventions Active agents

Any pharmacological agent prescribed prophylactically for salivary gland dysfunction prior to or during radiotherapy, by any route, any dose, and for any length of time. Radiation techniques were excluded.

Control groups

No preventative intervention, placebo, or another pharmacological preventative measure for salivary gland dysfunction.

Types of outcome measures

As radiotherapy-induced salivary gland dysfunction is considered to be permanent, we were interested in long-term treatment effects and only collected data starting from the end of radiotherapy, except in the case of adverse effects.

Primary outcomes

The primary outcome measure for the review is salivary gland dysfunction as indicated by either:

 xerostomia, i.e. the subjective sensation of dryness of the mouth. It was anticipated that different investigators would use different scales to assess xerostomia, e.g. visual analogue scales, verbal rating scales;  salivary flow rates (stimulated or unstimulated).

Secondary outcomes

The secondary outcome measures of the review are:

178  adverse effects, e.g. sweating, lacrimation (excess tears, crying), rhinorrhoea (watery discharge from the nose), diarrhoea, nausea;  survival data (overall, disease-free, progression-free, locoregional control);  other oral signs/symptoms, e.g. oral discomfort/pain, dysgeusia (taste disturbance), dysmasesia (difficulty in chewing), dysphagia (difficulty in swallowing), dysphonia (difficulty in speaking);  quality of life, e.g. ability to sleep, work, speak;  patient satisfaction;  cost data.

Search methods for identification of studies Electronic searches

Cochrane Oral Health's Information Specialist conducted systematic searches in the following databases for randomised controlled trials and controlled clinical trials. The search strategies are presented in full in Appendix 10. There were no publication year or publication status restrictions:

 Cochrane Oral Health's Trials Register (searched 14 September 2016);  Cochrane Central Register of Controlled Trials (CENTRAL; 2016, Issue 8) in the Cochrane Library (searched 14 September 2016);  MEDLINE Ovid (1946 to 14 September 2016);  Embase Ovid (1980 to 14 September 2016);  CINAHL EBSCO (Cumulative Index to Nursing and Allied Health Literature; 1937 to 14 September 2016);  LILACS BIREME Virtual Health Library (Latin American and Caribbean Health Science Information database; 1982 to 14 September 2016);  Zetoc Conference Proceedings (1993 to 14 September 2016);  OpenGrey (1997 to 14 September 2016).

Subject strategies were modelled on the search strategy designed for MEDLINE Ovid. Where appropriate, they were combined with subject strategy adaptations of the highly sensitive search strategy designed by Cochrane for identifying randomised controlled trials and controlled clinical trials as described in the Cochrane Handbook for Systematic Reviews of Interventions Chapter 6 (101). The Embase subject search was linked to an adapted version of the Cochrane Crowd Project filter for identifying randomised controlled

179 trials in Embase Ovid (see www.cochranelibrary.com/help/central-creation-details.html for information).

Language

The search attempted to identify all relevant studies irrespective of language. Articles in Chinese were translated and included in the review (173-176). An article in Spanish was translated and subsequently excluded (177).

Searching other resources Ongoing studies

We searched the following trial registries for ongoing studies (see Appendix 10):

 US National Institutes of Health Ongoing Trials Register (http://clinicaltrials.gov; searched 14 September 2016);  World Health Organization International Clinical Trials Registry Platform (http://apps.who.int/trialsearch/default.aspx; searched 14 September 2016).

Reference list searching

The reference lists of review articles and standard clinical oncology textbooks were checked for additional studies. The reference lists of included studies were also checked for additional studies.

Handsearching

Only handsearching done as part of the Cochrane Worldwide Handsearching Programme and uploaded to CENTRAL was included.

Unpublished studies

Requests for information about unpublished studies/studies published in the 'grey literature' were sent to relevant pharmaceutical companies, relevant investigators, editors of radiotherapy journals, and relevant professional organisations.

Data collection and analysis Selection of studies

180 The titles and abstracts of all records identified by the search strategy were scanned independently and in duplicate by two review authors. For both studies that appeared to meet the inclusion criteria, and studies that contained insufficient information in the title and abstract to determine eligibility, we obtained the full-text report and two review authors independently assessed them to establish whether they met the inclusion criteria. Studies excluded at this or subsequent stages were entered in the table of excluded studies with the reasons for exclusion recorded (see Appendix 11). All disagreements were resolved by discussion.

Data extraction and management

Two review authors independently and in duplicate extracted data using specially designed data extraction forms. The data extraction forms were piloted on several papers and modified as required before use. The data extracted included.

 Citation details: including year of publication, country of origin, setting and source of funding.  Details of participants: including demographic characteristics, cancer details (type, stage, location), radiation therapy and criteria for inclusion.  Details of intervention: including type, duration and method of administration.  Details of outcomes reported: including method of assessment (if measurement scales were used, details of whether the scale was validated were recorded).  Sample size calculation and trial registration.

Authors were contacted where possible for clarification and missing information.

Assessment of risk of bias in included studies

Two review authors independently assessed the risk of bias of each included study using the Cochrane domain-based, two-part tool as described in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (42). We contacted study authors for clarification or missing information where necessary and feasible. Disagreements were resolved through discussion, consulting a third review author to achieve consensus when necessary.

We completed a 'Risk of bias' table for each included study (see Appendix 11). For each domain of risk of bias, we described what was reported to have happened in the study. This

181 information provided the rationale for our judgement of whether that domain was at low, high, or unclear risk of bias.

We assessed the following domains:

 sequence generation (selection bias);  allocation concealment (selection bias);  blinding of participants and personnel (performance bias);  blinding of outcome assessment (detection bias);  incomplete outcome data (attrition bias);  selective outcome reporting (reporting bias);  other bias.

We categorised the overall risk of bias of individual studies as being at low, high, or unclear risk of bias according to the following criteria:

We also presented the 'Risk of bias' summary graphically.

Measures of treatment effect

For continuous outcomes (e.g. xerostomia on a visual analogue scale) where studies use the same scale, we used the mean values and standard deviations (SDs) reported in the studies in order to express the estimate of effect as mean difference (MD) with 95% confidence interval (CI). Where different scales were used to measure the same outcome, we expressed the treatment effect as standardised mean difference (SMD) with 95% CI.

For dichotomous outcomes, the estimate of effect of an intervention is expressed as risk ratios (RR) together with 95% CIs.

Unit of analysis issues

The participant is the unit of analysis.

182 Dealing with missing data

We contacted the author(s) of included studies, where feasible, to identify missing data and details of any other outcomes that may have been measured but not reported. We would have used the methods described in Section 7.7.3 of the Cochrane Handbook for Systematic Reviews of Interventions to estimate missing SDs (42) if appropriate. We did not use any other statistical methods or perform any further imputation to account for missing data.

Assessment of heterogeneity

Before any pooling of data was conducted, for comparisons with two or more studies, clinical heterogeneity was assessed by examining the types of participants (e.g. cancer types), interventions (e.g. control group used, dose and mode of administration), and outcomes (e.g. stimulated salivary flow rates or quality of life questionnaires). Statistical heterogeneity was also assessed using a Chi2 test, where a P value < 0.1 indicated statistically significant heterogeneity. We also quantified the heterogeneity using the I2 statistic.

Assessment of reporting biases

We would have assessed publication bias for comparisons where at least 10 studies were included in a meta-analysis. We would have used the recommendations on testing for funnel plot asymmetry (72), as described in Section 10.4 of the Cochrane Handbook for Systematic Reviews of Interventions (42).

Data synthesis

Meta-analyses were only undertaken where there were studies of similar comparisons reporting the same outcomes. We combined MDs for continuous data, and RRs for dichotomous data. Our general approach was to use a random-effects model. Our preference for the more conservative random-effects model is because statistical assessments can miss potentially important between-study heterogeneity in small samples (178).

We presented data not suitable for meta-analysis in additional tables.

Subgroup analysis and investigation of heterogeneity

183 Where possible, subgroup analyses would have been performed according to cancer type and treatment plans for cancer, and age of participants (i.e. children under the age of 18 years).

Sensitivity analysis

Sensitivity analysis was to be undertaken on the primary outcomes by excluding studies at unclear and high risk of bias from the analyses and also excluding unpublished literature.

If any meta-analyses had included studies with a large variation in sample size (for example several small studies and a single very large study), we would have undertaken a sensitivity analysis comparing the effect estimates from both random-effects and fixed- effect models. If these were different we would have reported on both analyses as part of the results section, and we would have considered possible interpretation.

Presentation of main results

We produced a 'Summary of findings' table for each comparison that included more than one study. We included data on: xerostomia, salivary flow rate, survival, quality of life and adverse events. We used GRADE methods (45), and the GRADEpro online tool for developing 'Summary of findings' tables (www.guidelinedevelopment.org). We assessed the quality of the body of evidence for each comparison and outcome by considering the overall risk of bias of the included studies, the directness of the evidence, the inconsistency of the results, the precision of the estimates, and the risk of publication bias. We described our level of certainty in the overall findings for each comparison/outcome in terms of high, moderate, low, very low.

5.5 Results

5.5.1 Description of studies

Results of the search

Electronic searches identified a total of 3536 titles and abstracts. A further study was identified by one of the review authors' knowledge of the topic area. After removal of duplicates, 2284 records were identified for screening. Following screening of these titles and abstracts by two review authors, 87 were identified as potentially relevant. Full papers were retrieved and authors of abstracts were written to in order to gain the full papers.

184 Following a second screening of these studies, 46 were excluded for reasons described in the ‘Excluded studies’ section below and in the ‘Characteristics of excluded studies’ table (see Appendix 11). One study was ongoing and a further study was awaiting classification. Therefore, 39 studies met our eligibility criteria and were included in this review. We presented this process as a flow chart in Figure 7.

Figure 7: PRISMA flow diagram

Included studies 185 Thirty-nine studies were included in this review (see ‘Characteristics of included studies’ tables in Appendix 11):  Abacioglu 1997 (179)  Jellema 2006 (193)  Antonadou 2002 (180)  Jham 2007 (194)  Bardet 2011 (181)  Lajtman 2000 (195)  Brizel 2000 (171)  Lanzós 2010 (196)  Brizel 2008 (106)  Le 2011 (120)  Buentzel 2006 (182)  Lin 2014 (197)  Büntzel 1998 (183)  Lozada-Nur 1998 (198)  Büntzel 2010 (184)  Patni 2004 (199)  Burlage 2008 (185)  Peng 2006 (200)  Duncan 2005 (186)  Pimentel 2014 (201)  Fisher 2003 (187)  Reshma 2012 (202)  Gornitsky 2004 (188)  Rode 1999 (168)  Grötz 2001 (189)  Sangthawan 2001 (203)  Haddad 2002 (190)  Vacha 2003 (204)  Haddad 2009 (191)  Valdez 1993 (205)  Han 2010 (173)  Veerasarn 2006 (206)  He 2004 (174)  Wang 1998 (176)  Henke 2011 (115)  Warde 2002 (207)  Hu 2005 (175)  Watanabe 2010 (208)  Jaguar 2015 (192)

Characteristics of trial setting, publication status and funding

Five trials were multinational (Brizel 2000; Brizel 2008; Buentzel 2006; Henke 2011; Le 2011), six were conducted in China (Han 2010; He 2004; Hu 2005; Lin 2014; Peng 2006; Wang 1998), four were conducted in Germany (Büntzel 1998; Büntzel 2010; Grötz 2001; Vacha 2003), four in the USA (Fisher 2003; Haddad 2009; Lozada-Nur 1998; Valdez 1993), three in Canada (Duncan 2005; Gornitsky 2004; Warde 2002), three in Brazil (Jaguar 2015; Jham 2007; Pimentel 2014), two in Thailand (Sangthawan 2001; Veerasarn 2006), two in India (Patni 2004; Reshma 2012), two in the Netherlands (Burlage 2008; Jellema 2006), one in Croatia (Lajtman 2000), one in Turkey (Abacioglu 1997), one in Greece (Antonadou 2002), one in Iran (Haddad 2002), one in France (Bardet 2011), one in Spain (Lanzós 2010), one in Japan (Watanabe 2010), and one in Slovenia (Rode 1999).

186 All trials had a parallel-group design. Ten trials had more than one published paper, with Büntzel 1998 publishing seven papers relating to the one trial. Abacioglu 1997 was an unpublished dissertation and data were gained from the authors of two trials following publication of their results as conference abstracts (Lozada-Nur 1998; Patni 2004). Eighteen of the trials received external funding, six trials received internal or no funding and the funding source was not stated in 15 trials.

One trial was ongoing (NCT02430298) and would be considered for future updates (see Appendix 11).

Characteristics of the participants

All of the trials recruited adults scheduled to receive radiotherapy to the salivary glands for cancer. The majority of participants were male. The type of cancer was head and neck at different sites in 36 trials and nasopharyngeal in 3 trials (Han 2010; He 2004; Lozada-Nur 1998).

Ten of the trials explicitly stated that chemotherapy was given as part of the treatment regimen in addition to radiotherapy for all patients (Antonadou 2002; Brizel 2008; Buentzel 2006; Büntzel 1998; Han 2010; Henke 2011; Le 2011; Peng 2006; Vacha 2003; Watanabe 2010). Chemotherapy was given to some patients of the following four trials: Bardet 2011; Gornitsky 2004; Haddad 2009; Lozada-Nur 1998. The other trials either undertook no chemotherapy, or were unclear about whether any chemotherapy was given.

Four studies explicitly referred to neck dissection but varied in the clarity of reporting: two clearly reported the proportions in each group that had their submandibular glands removed (Burlage 2008; Vacha 2003); one only reported the proportion that had neck dissection in each group, but did not refer to salivary gland removal (Haddad 2009); and one only stated that participants were stratified by submandibular gland removal, but numbers of participants affected were not reported (Jellema 2006).

The 39 included studies randomised 3520 participants, ranging from 10 to 291.

The percentage of participants lost to follow-up ranged from 0% to 38%.

Characteristics of the intervention

All of the trials provided a detailed description of the intervention including the dose and method of administration for the test and control groups. Twenty-one trials included a placebo control group and 14 a 'no intervention' control group, and the remaining four 187 trials made head-to-head comparisons (Bardet 2011; Jellema 2006; Jham 2007; Watanabe 2010).

 Pilocarpine hydrochloride was assessed in 12 trials at various dosages: Abacioglu 1997; Burlage 2008; Fisher 2003; Gornitsky 2004; Haddad 2002; Lajtman 2000; Lozada-Nur 1998; Pimentel 2014; Rode 1999; Sangthawan 2001; Valdez 1993; Warde 2002.  Biperiden plus pilocarpine was assessed in one trial: Rode 1999.  Amifostine was assessed in 12 trials at various dosages: Antonadou 2002; Bardet 2011; Brizel 2000; Buentzel 2006; Büntzel 1998; Haddad 2009; He 2004; Jellema 2006; Patni 2004; Peng 2006; Vacha 2003; Veerasarn 2006.  Chinese medicine was assessed in five trials, all comparing different herbs: Han 2010; Hu 2005; Lin 2014; Reshma 2012; Wang 1998.  Palifermin was assessed in three trials: Brizel 2008; Henke 2011; Le 2011.  Bethanechol was assessed in two trials: Jaguar 2015; Jham 2007.  Artificial saliva was assessed in one trial: Jham 2007.  Selenium was assessed in one trial: Büntzel 2010.  Antiseptic mouthrinse was assessed in one trial: Lanzós 2010.  Antimicrobial lozenge was assessed in one trial: Duncan 2005.  Polaprezinc was assessed in one trial: Watanabe 2010.  Azulene oral rinse assessed in one trial: Watanabe 2010.  Venalot Depot (coumarin/ troxerutin) was assessed in one trial: Grötz 2001.

The length of follow-up ranged from day 28 of the radiotherapy (RT) to 34 months: day 28/29 of RT (Pimentel 2014; Reshma 2012), end of RT (Abacioglu 1997; Hu 2005; Wang 1998), four weeks from start of RT (Lanzós 2010), four weeks after RT (Lin 2014; Grötz 2001), five weeks after RT (Gornitsky 2004), six weeks after RT (Büntzel 2010; Vacha 2003), seven weeks after RT (He 2004), two months after RT (Jham 2007), three months after RT (Brizel 2008; Han 2010; Jaguar 2015; Lozada-Nur 1998), six months after RT (Duncan 2005; Fisher 2003; Haddad 2002; Sangthawan 2001; Warde 2002), 12 months after RT (Bardet 2011; Buentzel 2006; Büntzel 1998; Burlage 2008; Lajtman 2000; Rode 1999; Valdez 1993), 18 months after RT (Antonadou 2002), 24 months after RT (Brizel 2000; Henke 2011; Jellema 2006; Le 2011; Patni 2004; Veerasarn 2006), and 34 months after RT (Haddad 2009). Duration of follow-up/timing of assessment was unclear in two studies (Peng 2006; Watanabe 2010).

Characteristics of outcome measures

188 The trials used a variety of assessment measures for salivary gland dysfunction. Ten trials included a subjective measure of salivary gland dysfunction, i.e. the patient was involved in the assessment through visual analogue scales (VAS) (Gornitsky 2004; Haddad 2002; Sangthawan 2001; Wang 1998), linear analogue scale (LASA) (Warde 2002), and modified patient questionnaires (Abacioglu 1997; He 2004; Jellema 2006; Lajtman 2000; Veerasarn 2006). One study reported 'acute' or 'chronic' dry mouth only (Peng 2006). Fifteen trials reported a clinical assessment of salivary gland dysfunction using various scales: RTOG/EORTC (Radiation Therapy Oncology Group/European Organisation for Research and Treatment of Cancer) scoring (Antonadou 2002; Brizel 2000; Buentzel 2006; Fisher 2003; Grötz 2001; He 2004; Jellema 2006; Patni 2004; Veerasarn 2006), NCI CTCAE (National Cancer Institute Common Terminology Criteria for Adverse Events) (Brizel 2008; Henke 2011; Le 2011), WHO (World Health Organization) grading/classification (Büntzel 1998), and Late Effects Normal Tissue Task Force (LENT)-Subjective, Objective, Management, Analytic (SOMA) scales (Burlage 2008; Haddad 2002). Unstimulated or stimulated whole saliva secretion data or both were collected in ten trials (Abacioglu 1997; Brizel 2000; Buentzel 2006; Gornitsky 2004; He 2004; Lajtman 2000; Rode 1999; Valdez 1993; Veerasarn 2006; Wang 1998), and salivary gland scintigraphy was used in five trials (Fisher 2003; Grötz 2001; Lozada-Nur 1998; Patni 2004; Veerasarn 2006).

Secondary outcomes were sporadically reported, using various scales. However, the majority of studies reported adverse events.

Excluded studies

Of the 87 trials that were identified as potentially eligible, 46 were excluded (see Appendix 11), with the main reason being the publication of an abstract only (17 publications), with insufficient information to allow thorough assessment.

Other reasons for exclusion were: not a randomised controlled trial or unclear if a randomised controlled trial; prevention of salivary gland dysfunction not the aim of study/not reported; radioactive iodine used rather than radiotherapy; study did not include head and neck cancer patients; the intervention was not a pharmacological agent.

5.5.2 Risk of bias in included studies

(A risk of bias table is presented for each included study in Appendix 11.) Allocation (selection bias) Random sequence generation 189 Twenty of the included studies described an adequate method of random sequence generation and were assessed as at low risk of bias for this domain (Abacioglu 1997; Brizel 2000; Buentzel 2006; Burlage 2008; Gornitsky 2004; Haddad 2002; Haddad 2009; Henke 2011; Jaguar 2015; Jellema 2006; Jham 2007; Lanzós 2010; Le 2011; Lin 2014; Lozada- Nur 1998; Pimentel 2014; Rode 1999; Sangthawan 2001; Valdez 1993; Veerasarn 2006). The remaining 19 studies stated that allocation was random but did not describe their methods and were therefore assessed as being at unclear risk of bias for this domain.

Allocation concealment

Allocation concealment was clearly described in 16 of the included studies and they were assessed as being at low risk of bias for this domain (Abacioglu 1997; Brizel 2000; Buentzel 2006; Burlage 2008; Gornitsky 2004; Haddad 2002; Haddad 2009; Henke 2011; Lanzós 2010; Le 2011; Lozada-Nur 1998; Pimentel 2014; Rode 1999; Sangthawan 2001; Valdez 1993; Veerasarn 2006). The remaining 23 did not describe any methods used to conceal the random sequence, and so were assessed as being at unclear risk of bias.

Blinding (performance bias and detection bias) Blinding of participants and personnel (performance bias)

Twenty-one studies were placebo-controlled and double-blind, and were assessed at low risk of performance bias. In the remaining 18 studies, blinding of the patients and their caregivers to the allocated treatment was not possible because the active and control treatments were administered differently, the control group had no intervention at all, or the personnel administering or patients were not blinded to the intervention (Abacioglu 1997; Antonadou 2002; Bardet 2011; Brizel 2000; Büntzel 1998; Büntzel 2010; Haddad 2009; Han 2010; He 2004; Hu 2005; Jellema 2006; Jham 2007; Patni 2004; Peng 2006; Rode 1999; Vacha 2003; Veerasarn 2006; Watanabe 2010).

Blinding of outcome assessment (detection bias)

Twenty-one studies were assessing the effect of the intervention versus a placebo where the assessor was also blinded and these were assessed as at low risk of bias. A further study which was not placebo-controlled was assessed at low risk of bias because the outcome assessment for salivary gland dysfunction was objective (Rode 1999). The remaining 17 studies were assessed as being at high risk of detection bias, as the assessor was not blinded, the intervention was assessed against no intervention, the administration of the drug was different in the intervention and control groups or the assessment of xerostomia was subjective (Abacioglu 1997; Antonadou 2002; Bardet 2011; Brizel 2000; 190 Büntzel 1998; Büntzel 2010; Haddad 2009; Han 2010; He 2004; Hu 2005; Jellema 2006; Jham 2007; Patni 2004; Peng 2006; Vacha 2003; Veerasarn 2006; Watanabe 2010).

Incomplete outcome data (attrition bias)

Twenty-one studies had no or negligible attrition and were assessed as being low risk. Twelve studies were assessed to be at high risk of attrition bias, due to high dropout rates, no reasons given for dropouts or differential attrition between the groups, which could be linked to the intervention (Bardet 2011; Brizel 2008; Burlage 2008; Grötz 2001; Haddad 2002; Jellema 2006; Jham 2007; Lanzós 2010; Pimentel 2014; Vacha 2003; Veerasarn 2006; Warde 2002). For the six remaining studies, there was insufficient information to determine risk of attrition bias (Fisher 2003; Haddad 2009; Lajtman 2000; Lozada-Nur 1998; Peng 2006; Sangthawan 2001).

Selective reporting (reporting bias)

Nineteen of the included studies reported the outcomes specified in the methods section in full, including information about xerostomia and adverse effects (Abacioglu 1997; Antonadou 2002; Bardet 2011; Brizel 2000; Brizel 2008; Buentzel 2006; Büntzel 1998; Haddad 2002; Han 2010; He 2004; Henke 2011; Hu 2005; Jaguar 2015; Jham 2007; Le 2011; Lin 2014; Lozada-Nur 1998; Veerasarn 2006; Warde 2002). One study was assessed to be at unclear risk of reporting bias (Peng 2006). The remaining 19 studies were assessed as at high risk of reporting bias as they did not report on adverse effects or xerostomia, did not report on all outcomes, only significant data were reported or data on individuals were not reported, and grouped data did not have the standard deviations.

Other potential sources of bias

We did not consider there to be any other issues arising from other potential sources in any of the studies and we therefore assessed them all as being at low risk of bias for this domain.

Overall risk of bias

Overall, three of the included studies (8%) were assessed as at low risk of bias for all domains (Buentzel 2006; Henke 2011; Le 2011), and four studies (10%) were assessed as being at unclear risk of bias for at least one domain (Brizel 2008; Jaguar 2015; Lin 2014; Lozada-Nur 1998). The remaining 32 studies (82%) were at high risk of bias for at least one domain. Risk of bias was presented graphically in Figure 8.

191

192

Figure 8: Risk of bias summary: review authors’ judgements about each risk of bias item for each included study

5.5.3 Effects of interventions

1. Pilocarpine versus no treatment/placebo

Pilocarpine hydrochloride, at various dosages, was assessed in 12 trials: Abacioglu 1997; Burlage 2008; Fisher 2003; Gornitsky 2004; Haddad 2002; Lajtman 2000; Lozada-Nur 1998; Pimentel 2014; Rode 1999; Sangthawan 2001; Valdez 1993; Warde 2002. Over 900 participants were randomised to either pilocarpine or no treatment/placebo; 698 were 193 evaluated (although number varied by outcome/timing of assessment). Eleven of the trials were judged to be at high risk of bias; one was at unclear risk (Lozada-Nur 1998).

Xerostomia

Nine trials evaluated xerostomia, however, the method of assessment varied across studies.

Seven trials presented continuous data on xerostomia obtained by simple VAS or a composite based on a number of xerostomia-focused questions (Abacioglu 1997; Burlage 2008; Gornitsky 2004; Haddad 2002; Lozada-Nur 1998; Sangthawan 2001; Warde 2002). The trial by Burlage 2008 was unable to be included in any statistical pooling as data were presented by Gy dose, but the number receiving each dose was unclear. There was no evidence of a difference between treatment groups at end of radiotherapy (standardised mean difference (SMD) 0.20, 95% confidence interval (CI) -0.16 to 0.56; P = 0.27; 122 participants), up to three months postradiotherapy (SMD 0.02, 95% CI -0.33 to 0.37; P = 0.92; 125 participants), or up to six months postradiotherapy (SMD -0.35, 95% CI -1.04 to 0.33; P = 0.31; 126 participants) (Analysis 1.1). There was substantial statistical heterogeneity present for the six-month data (I2 = 68%, P = 0.08).

One trial used the LENT-SOMA scale to provide an objective assessment of xerostomia (Haddad 2002). This single trial showed a statistically significant difference in favour of pilocarpine (mean difference (MD) -0.40, 95% CI -0.69 to -0.11; P = 0.006; 39 participants) at six months postradiotherapy (Analysis 1.2).

Two trials presented binary data on the number of participants with/without xerostomia (Lajtman 2000; Pimentel 2014). There was no evidence of a difference between treatment groups at the end of radiotherapy (risk ratio (RR) 0.60, 95% CI 0.18 to 2.02; P = 0.41; 11 participants) or at three months postradiotherapy (RR 1.00, 95% CI 0.92 to 1.08; P = 1.00; 48 participants) (Analysis 1.3).

194

Analysis 1.1: Pilocarpine versus control; Outcome: Xerostomia (various scales) 195

Analysis 1.2: Pilocarpine versus control; Outcome: Xerostomia (LENT-SOMA scale)

Analysis 1.3: Pilocarpine versus control; Outcome: Xerostomia (yes/no)

196 Salivary flow rates

Eight studies presented continuous data for either stimulated or unstimulated salivary flow (e.g. ml/min or g) (Abacioglu 1997; Burlage 2008; Gornitsky 2004; Lajtman 2000; Lozada-Nur 1998; Pimentel 2014; Rode 1999; Valdez 1993). The studies by Burlage 2008; Lajtman 2000; Pimentel 2014 and Rode 1999 were unable to be included in any statistical pooling due to insufficient reporting of data. There was no evidence of a difference between treatment groups for unstimulated or stimulated flow rates at any time point (Analysis 1.4; Analysis 1.5).

One study presented binary data on whether stimulated or unstimulated salivary flow was > 0 g (Fisher 2003). There was insufficient evidence to determine whether pilocarpine was beneficial for this outcome at any time point (Analysis 1.6; Analysis 1.7).

197

Analysis 1.4: Pilocarpine versus control; Outcome: Salivary flow rate (unstimulated)

198

Analysis 1.5: Pilocarpine versus control; Outcome: Salivary flow rate (stimulated) 199

Analysis 1.6: Pilocarpine versus control; Outcome: Salivary flow rate > 0 g (unstimulated)

200

Analysis 1.7: Pilocarpine versus control; Outcome: Salivary flow rate > 0 g (stimulated)

201 Survival

Only one trial reported on overall survival within the trial period (six months) (Haddad 2002). There was no evidence of a difference between treatment groups (RR 1.07, 95% CI 0.80 to 1.43; P = 0.66; 60 participants) (Appendix 13: Analysis 1.8).

Quality of life and other oral related symptoms

There was insufficient evidence to determine whether or not pilocarpine improved quality of life measurements for global quality of life, quality of life (HNRQ), oral discomfort, eating difficulties and sleeping problems at the end of radiotherapy (Appendix 13: Analysis 1.9; Appendix 12: Table 1). One trial (Gornitsky 2004) found an increased risk in speech difficulties at the end of radiotherapy in the pilocarpine group (MD 20.20, 95% CI 1.93 to 38.47; P = 0.03; 34 participants) when assessed using a VAS scale (0 to 100 mm) (Appendix 12: Table 1).

There was insufficient evidence from one study (Gornitsky 2004) to determine whether or not pilocarpine improved oral mucosal pain at the end of radiotherapy (MD -14.80, 95% CI -37.07 to 7.47; P = 0.19; 36 participants) (Appendix 12: Table 1).

Side effects

No evidence of a difference was found between treatment groups for any reported adverse event, apart from for sweating where data from five studies showed an increased risk associated with pilocarpine (RR 2.98, 95% CI 1.43 to 6.22; P = 0.004; 389 participants) (Appendix 12: Table 1). There was no observed statistical heterogeneity (I2 = 0%, P = 0.52).

Cost

None of the included studies evaluating the effectiveness of pilocarpine reported cost data.

2. Biperiden plus pilocarpine versus no treatment

One trial, assessed at high risk of bias, compared biperiden and pilocarpine with no treatment (Rode 1999).

Xerostomia 202 No xerostomia data related to the effectiveness of biperiden and pilocarpine were reported.

Salivary flow rates

There was insufficient evidence, from a single trial of 60 participants (Rode 1999), to determine whether or not biperiden and pilocarpine reduced the unstimulated salivary flow rate between patients at the end of radiotherapy (Analysis 2.1; Analysis 2.2).

203

Analysis 2.1: Biperiden plus pilocarpine versus no treatment; Outcome: Salivary flow rate (unstimulated)

204

Analysis 2.2: Biperiden plus pilocarpine versus no treatment; Outcome: Salivary flow rate > 0 g (unstimulated) 205 Survival

Not reported.

Quality of life and other oral related symptoms

There was insufficient evidence to determine whether or not biperiden and pilocarpine reduced the risk of WHO grade 3+ dysphagia up to one year after radiotherapy (Appendix 12: Table 2). No further data on quality of life or other oral related symptoms were reported.

Side effects

No data were reported on side effects.

Cost

No cost data related to the effectiveness of biperiden and pilocarpine were reported.

3. Amifostine versus no treatment/placebo

Eleven trials, one at low risk of bias (Buentzel 2006) and ten at high risk of bias (Antonadou 2002; Brizel 2000; Büntzel 1998; Haddad 2009; He 2004; Jellema 2006; Patni 2004; Peng 2006; Vacha 2003; Veerasarn 2006), randomised 1036 participants (887 analysed, although the number varied by outcome/timing of assessment) to amifostine or no treatment group/placebo. The trial by Jellema 2006 had three comparison groups: two different doses of amifostine and a 'no treatment' group. For the purpose of this comparison, the two amifostine groups were combined.

Xerostomia

Three studies (Büntzel 1998; He 2004; Veerasarn 2006) were combined in a meta-analysis showing that amifostine reduced the risk of developing grade ≥ 2 xerostomia (on a 0 to 4 scale) at the end of radiotherapy (RR 0.35, 95% CI 0.19 to 0.67; P = 0.001; 119 participants) (Analysis 3.1).

Up to and including three months postradiotherapy, a smaller effect was observed in favour of amifostine in a meta-analysis of five studies (Antonadou 2002; Brizel 2000; Buentzel 2006; Patni 2004; Veerasarn 2006) (RR 0.66, 95% CI 0.48 to 0.92; P = 0.01; 687 206 participants) (Analysis 3.1). However, there was substantial heterogeneity present (I2 = 63%).

At 12 months postradiotherapy, there was insufficient evidence of a difference in the risk of grade ≥ 2 xerostomia (RR 0.70, 95% CI 0.40 to 1.23; P = 0.21; 682 participants analysed) (Antonadou 2002; Brizel 2000; Buentzel 2006; Büntzel 1998; Jellema 2006; Patni 2004; Veerasarn 2006) (Analysis 3.1). There was considerable heterogeneity present (I2 = 83%).

Three further studies had no usable data: one failed to report the data by study group and reported that "For the end point xerostomia we are not able to demonstrate that amifostine had a positive effect, and there was no difference detected between the arms in terms of xerostomia, with 41% of patients reporting xerostomia of grade ≥ 2" (Haddad 2009); one failed to report the timing of assessment (Peng 2006); and one only reported xerostomia during radiotherapy (i.e. not at any of the time points we were interested in) (Vacha 2003).

207

Analysis 3.1: Amifostine versus control; Outcome: Xerostomia (0 to 4 scale - grade 2 or above) 208 Salivary flow rates Unstimulated

There was inconsistent evidence regarding the effect of amifostine of unstimulated salivary flow rate. A greater salivary flow rate for those receiving amifostine was shown at the end of radiotherapy (MD 0.34, 95% CI 0.07 to 0.61; P = 0.01; 83 participants) (Analysis 3.2).

There was insufficient evidence of a difference from one study (Veerasarn 2006) up to and including three months postradiotherapy (MD 0.13, 95% CI -0.90 to 1.16; P = 0.8; 41 participants), but the same study showed a slight benefit in favour of amifostine at 12 months postradiotherapy (MD 0.32, 95% CI 0.09 to 0.55; P = 0.006; 27 participants) (Analysis 3.2). A further study (Brizel 2000) showed a benefit at 12 months postradiotherapy in favour of amifostine when looking at incidence of producing > 0.1 g of saliva over 5 minutes (RR 1.45, 95% CI 1.13 to 1.86; P = 0.004; 175 participants) (Analysis 3.4).

Haddad 2009 failed to report the salivary flow data by study group and simply reported that "No difference was observed between the 2 treatment arms."

Stimulated

There was insufficient evidence of a difference from one study (Veerasarn 2006) at the end of radiotherapy (MD -0.09, 95% CI -1.48 to 1.30; P = 0.90; 47 participants), up to and including three months postradiotherapy (MD 0.38, 95% CI -1.43 to 2.19; P = 0.68; 41 participants), or 12 months postradiotherapy (MD 0.82, 95% CI -0.47 to 2.11; P = 0.21; 27 participants) (Analysis 3.3). There was also insufficient evidence of a difference from one study (Brizel 2000), analysing 173 participants, when looking at incidence of producing > 0.1 g of saliva over 5 minutes at 12 months postradiotherapy (RR 1.12, 95% CI 0.89 to 1.41; P = 0.32) (Analysis 3.5).

Haddad 2009 failed to report the salivary flow data by study group and simply reported that "No difference was observed between the 2 treatment arms."

209

Analysis 3.2: Amifostine versus control; Outcome: Salivary flow rate (unstimulated)

210

Analysis 3.3: Amifostine versus control; Outcome: Salivary flow rate (stimulated)

Analysis 3.4: Amifostine versus control; Outcome: Salivary flow rate > 0.1 g in 5 mins (unstimulated) 211

Analysis 3.5: Amifostine versus control; Outcome: Salivary flow rate > 0.1 g in 5 mins (stimulated)

212 Survival

There was insufficient evidence to determine whether or not amifostine reduces overall survival, progression-free survival, disease-free survival or locoregional tumour control up to 24 months postradiotherapy.

Overall survival

There was insufficient evidence from a meta-analysis of two studies (Brizel 2000; Buentzel 2006) to determine whether or not amifostine reduces overall survival at 12 to 24 months postradiotherapy (hazard ratio (HR) 1.18, 95% CI 0.85 to 1.66; P = 0.33; 271 participants) (Appendix 13: Analysis 3.6). Two further studies found no difference in overall survival at two years or more postradiotherapy (Haddad 2009; Jellema 2006) (Appendix 13: Analysis 3.7).

Progression-free survival

There was insufficient evidence from a meta-analysis of two studies (Brizel 2000; Buentzel 2006) to determine whether or not amifostine reduces progression-free survival at 12 to 24 months postradiotherapy (HR 0.94, 95% CI 0.70 to 1.27; P = 0.70; 247 participants) (Appendix 13: Analysis 3.8). A further study (Antonadou 2002) found no difference at 18 months postradiotherapy (RR 1.11, 95% CI 0.81 to 1.51; P = 0.52; 45 participants) (Appendix 13: Analysis 3.9). This was supported by Haddad 2009 who reported "no differences noted" (Appendix 13: Analysis 3.10).

Locoregional tumour control

There was insufficient evidence from a meta-analysis of two studies (Brizel 2000; Buentzel 2006) to determine whether or not amifostine reduced locoregional tumour control at 12 to 24 months postradiotherapy (HR 0.90, 95% CI 0.74 to 1.11; P = 0.33; 279 participants) (Appendix 13: Analysis 3.11). Three further studies reported narrative evidence to support this result (Haddad 2009; Jellema 2006; Patni 2004) (Appendix 13: Analysis 3.12).

Disease-free survival

There was insufficient evidence from one study (Patni 2004) to determine whether or not amifostine reduces disease-free survival at 24 months postradiotherapy (RR 0.94, 95% CI 0.73 to 1.21; P = 0.64; 170 participants) (Appendix 13: Analysis 3.13). Two studies

213 reported narrative evidence to support this result (Patni 2004; Veerasarn 2006) (Appendix 13: Analysis 3.14).

Quality of life and other oral related symptoms

There was insufficient evidence of a difference in quality of life from one study (Brizel 2000), both at the end of radiotherapy (MD 0.38, 95% CI -0.07 to 0.83; P = 0.1; 298 participants), and up to and including three months postradiotherapy (MD 0.52, 95% CI - 0.02 to 1.06; P = 0.06; 233 participants). The same study showed a benefit in favour of amifostine at 12 months postradiotherapy (MD 0.70, 95% CI 0.20 to 1.20; P = 0.006; 180 participants) (Appendix 13: Analysis 3.15). A further study reported no differences in quality of life but did not present data (Jellema 2006) (Appendix 12: Table 3).

Two of the 11 studies presented data on dysphagia (Antonadou 2002; Büntzel 1998). There was insufficient evidence of a difference in the risk of developing grade ≥ 3 dysphagia (on a 0 to 4 scale) at the end of radiotherapy (RR 0.50, 95% CI 0.17 to 1.48; P = 0.21; 73 participants) and up to and including three months postradiotherapy (four weeks after) (RR 0.70, 95% CI 0.13 to 3.78; P = 0.68; 45 participants) (Appendix 12: Table 3). In Antonadou 2002, no participants had grade 3 or above dysphagia by eight weeks after radiotherapy.

One study presented data on dysgeusia (Büntzel 1998). The study showed that amifostine reduced the risk of developing grade ≥ 2 dysgeusia (on a 0 to 4 scale) at the end of radiotherapy (RR 0.24, 95% CI 0.10 to 0.61; P = 0.003; 28 participants) (Appendix 12: Table 3).

Side effects

Adverse events were reported inconsistently across the 11 included trials. There was a higher risk of vomiting in the amifostine group (RR 4.90, 95% CI 2.87 to 8.38; P < 0.00001; five studies; 601 participants) (Antonadou 2002; Brizel 2000; Buentzel 2006; He 2004; Jellema 2006). Two further studies reported high rates of vomiting in the amifostine group but did not mention vomiting in the control group (Peng 2006; Veerasarn 2006). The risk of hypotension was higher in the amifostine group (RR 9.20, 95% CI 2.84 to 29.83; P = 0.0002; three studies; 376 participants) (Antonadou 2002; Brizel 2000; Büntzel 1998). Another study reported hypotension only in the amifostine group (Veerasarn 2006). The risk of nausea was higher in the amifostine group (RR 2.60, 95% CI 1.81 to 3.74; P < 0.00001; four studies; 556 participants) (Brizel 2000; Buentzel 2006; He 2004; Jellema

214 2006). Two further studies reported high rates of nausea in the amifostine group but did not mention nausea in the control group (Peng 2006; Veerasarn 2006). The risk of allergic response was higher in the amifostine group (RR 7.51, 95% CI 1.40 to 40.39; P = 0.02; three studies; 524 participants) (Brizel 2000; Buentzel 2006; Jellema 2006). There was insufficient evidence of a difference in asthenia (weakness/lack of energy) from one study (RR 2.91, 95% CI 0.31 to 27.24; P = 0.35; 130 participants) (Buentzel 2006). Other side effects (alopecia, skin toxicity, hot flush, drowsiness, sneezing, hiccupping, dizziness and fatigue) were reported either narratively or only for the amifostine group (Appendix 12: Table 3).

Cost

One study (Büntzel 1998), analysing 28 participants, reported economic cost data in a separate paper (Bennett 2001). In 2001, the mean per patient supportive care costs were lower in the amifostine group (USD 4401) than the control group (USD 5873) (P = 0.02) (Appendix 12: Table 3).

4. Amifostine (comparison of dosages)

One trial, at high risk of bias, compared two different amifostine regimens of 200 mg/m² either five or three times a week (Jellema 2006) (a third 'no treatment' group was not considered in this comparison).

Xerostomia

There was insufficient evidence to determine whether or not different amifostine dosages reduced the risk of developing grade ≥ 2 xerostomia (on a 0 to 4 scale) at 12 months postradiotherapy (RR 0.94, 95% CI 0.58 to 1.53; P = 0.80; 49 participants) (Analysis 4.1).

215

Analysis 4.1: Amifostine (comparison of doses); Outcome: Xerostomia (0 to 4 scale - grade 2 or above)

216 Salivary flow rates

No salivary flow rate data were reported.

Survival

There was insufficient evidence reported on overall survival or locoregional tumour control (Appendix 13: Analysis 4.2; Analysis 4.3).

Quality of life and other oral related symptoms

The paper reported "no significant differences between the three treatment arms" in quality of life assessed at the end of radiotherapy and 6, 12, 18 and 24 months after radiotherapy (Appendix 12: Table 4).

Side effects

There was insufficient evidence of a difference in nausea (RR 0.64, 95% CI 0.33 to 1.25; P = 0.19; 60 participants), vomiting (RR 0.25, 95% CI 0.06 to 1.08; P = 0.06; 60 participants), or allergic response (RR 1.00, 95% CI 0.15 to 6.64; P = 1; 60 participants) (Appendix 12: Table 4).

Cost

No cost data related to the effectiveness of different doses of amifostine were reported.

5. Amifostine (intravenous versus subcutaneous)

One study, at high risk of bias, compared intravenous and subcutaneous delivery of amifostine (Bardet 2011).

Xerostomia

There was insufficient evidence to determine whether or not different methods of amifostine delivery reduced the risk of developing grade ≥ 2 xerostomia (on a 0 to 4 scale) up to and including three months postradiotherapy (RR 1.03, 95% CI 0.76 to 1.40; P = 0.86; 263 participants). There was a benefit in favour of amifostine at 12 months postradiotherapy (RR 0.61, 95% CI 0.42 to 0.88; P = 0.008; 127 participants) (Analysis 5.1). 217

Analysis 5.1: Amifostine (intravenous versus subcutaneous); Outcome: Xerostomia (0 to 4 scale - grade 2 or above)

218 Salivary flow rates

No salivary flow rate data, related to the effectiveness of different routes of administration of amifostine, were reported.

Survival

There was insufficient evidence of a difference in overall survival (HR 1.36, 95% CI 0.89 to 2.10; P = 0.16; Appendix 13: Analysis 5.2) or locoregional tumour control (HR 1.34, 95% CI 0.76 to 2.36; P = 0.32; Appendix 13: Analysis 5.3), both at 48 months postradiotherapy.

Quality of life and other oral related symptoms

No data on either quality of life or other oral related symptoms were reported.

Side effects

The single trial comparing intravenous and subcutaneous delivery of amifostine reported increased incidence of hypotension for intravenous delivery. Skin rash and local pain at injection site were worse for subcutaneous delivery (Appendix 12: Table 5). Results were inconclusive with regard to nausea/vomiting, fever, and asthenia (weakness/lack of energy).

Cost

No cost data related to the effectiveness of different routes of administration of amifostine were reported.

6. Chinese medicine versus no treatment/placebo

Five studies compared some form of Chinese medicine with no treatment/placebo. Four studies were assessed as being at high risk of bias (Han 2010; Hu 2005; Reshma 2012; Wang 1998); one study was assessed as being at unclear risk of bias (Lin 2014).

Xerostomia

Hu 2005 found that patients who received Shenqi Fanghon recipe had a reduced risk of xerostomia at the end of radiotherapy compared to those in the no treatment control group 219 (RR 0.39, 95% CI 0.28 to 0.55; P < 0.00001; 140 participants) (Analysis 6.1). The paper was translated from Chinese, the methods were unclear.

Lin 2014, a trial of 71 participants, however, found no evidence of a difference for Tianwang Buxin Mini-pills when compared with placebo when xerostomia was evaluated using both dichotomous data (Analysis 6.1) or continuous data (Analysis 6.2). Similarly, Han 2010, a trial of 95 participants, found no evidence of a difference for Jinlong capsules when compared with no intervention (Analysis 6.1).

Wang 1998, a trial of 50 participants, found a difference in favour of Chinese medicine in an assessment of xerostomia (VAS) against a no treatment group at the end of radiotherapy (P < 0.05). The results were graphically represented and the standard deviations were not available from the paper.

220

Analysis 6.1: Chinese medicine versus control; Outcome: Xerostomia (yes/no) 221

Analysis 6.2: Chinese medicine versus control; Outcome: Xerostomia

222 Salivary flow rates

One study showed an increase in stimulated salivary flow rate in favour of Chinese medicine when compared with no treatment at the end of radiotherapy (MD 0.09, 95% CI 0.03 to 0.15; P = 0.001; 50 participants) (Analysis 6.3). The paper was translated from Chinese, with the standard deviations being estimated (Wang 1998). Reshma 2012 mentioned salivary status but provided no data.

223

Analysis 6.3: Chinese medicine versus control; Outcome: Salivary flow rate (stimulated)

224 Survival

Hu 2005 evaluated overall survival (12 months postradiotherapy) but there was insufficient evidence to determine any effect (Appendix 13: Analysis 6.4).

Quality of life and other oral related symptoms

Lin 2014 evaluated quality of life (at end of intervention and up to and including three months postradiotherapy) but there was insufficient evidence to determine any effect (Appendix 13: Analysis 6.5). The same study showed insufficient evidence of a difference for other oral related symptoms (both at end of radiotherapy and one month after) (Appendix 12: Table 6).

Side effects

There was insufficient evidence of a difference for any side effects (at end of intervention and one month after) (Appendix 12: Table 6).

Cost

No cost data related to the effectiveness of Chinese medicine were reported.

7. Palifermin versus placebo

Three trials, two at low (Henke 2011; Le 2011) and one at unclear risk of bias (Brizel 2008), evaluated palifermin versus placebo.

Xerostomia

In a meta-analysis of all three trials, there was insufficient evidence of a difference in the incidence of > grade 2 xerostomia up to three months postradiotherapy (RR 0.97, 95% CI 0.77 to 1.22; P = 0.78; 471 participants). There was considerable heterogeneity present (I2 = 76%, P = 0.02) (Analysis 7.1). It should be noted that a large proportion of participants in Henke 2011 did not have assessments for xerostomia, but the intention-to-treat (ITT) rules stated that such participants would be assumed to have the outcome, and this may have had a substantial effect on the meta-analysis result.

Xerostomia was measured up to 12 months in two studies but no data were reported (Henke 2011; Le 2011). 225

Analysis 7.1: Palifermin versus placebo; Outcome: Xerostomia (0 to 4 scale - grade 2 or above)

226 Salivary flow rates

None of the trials evaluating palifermin provided data on salivary flow rates.

Survival

All three trials reported data on overall and progression-free survival at 42 to 72 months from baseline. There was insufficient evidence of a difference in both overall survival (HR 1.00, 95% CI 0.72 to 1.39; P = 0.99; Appendix 13: Analysis 7.2) and progression-free survival (HR 1.06, 95% CI 0.80 to 1.42; P = 0.67; Appendix 13: Analysis 7.3).

Quality of life and other oral related symptoms

All three trials provided data for a meta-analysis of dysphagia at three months postradiotherapy, with insufficient evidence of a difference (RR 1.32, 95% CI 0.55 to 3.13; P = 0.54) (Appendix 12: Table 7). There was also insufficient evidence of a difference in mouth and throat soreness at three months postradiotherapy in a meta-analysis of two trials (MD -0.12, 95% CI -0.27 to 0.02; P = 0.10) (Appendix 12: Table 7).

Side effects

All three trials provided information on possible adverse events, sometimes reporting the same adverse event (Appendix 12: Table 7). There was no evidence of patients in either group experiencing more or less of these adverse events.

Cost

No cost data related to the effectiveness of palifermin were reported.

8. Bethanechol versus placebo

One study, at unclear risk of bias, compared bethanechol with placebo (Jaguar 2015).

Xerostomia

Bethanechol reduced the risk of developing grade ≥ 2 xerostomia (on a 0 to 3 scale) at the end of radiotherapy (RR 0.43, 95% CI 0.28 to 0.66; P = 0.0001; 84 participants). However, there was insufficient evidence of a difference up to and including three months postradiotherapy (RR 0.81, 95% CI 0.65 to 1.01; P = 0.06; 84 participants) (Analysis 8.1).

227

Analysis 8.1: Bethanechol versus placebo; Outcome: Xerostomia (0 to 3 scale - grade 2 or above)

228 Salivary flow rates

Bethanechol increased unstimulated saliva flow (ml/min) at two months postradiotherapy (MD 0.19, 95% CI 0.06 to 0.32; P = 0.004; 97 participants) (Analysis 8.2).

There was insufficient evidence of a difference in stimulated saliva flow (ml/min) at two months postradiotherapy (MD 0.15, 95% CI -0.03 to 0.33; P = 0.11; 97 participants; Analysis 8.3).

229

Analysis 8.2: Bethanechol versus placebo; Outcome: Salivary flow rate in ml/min (unstimulated)

Analysis 8.3: Bethanechol versus placebo; Outcome: Salivary flow rate in ml/min (stimulated) 230 Survival

No survival data were reported.

Quality of life and other oral related symptoms

No data on either quality of life or other oral related symptoms were reported.

Side effects

The study reported narratively that there were no statistical differences between the groups in bethanechol-related toxicity and that "no patient experienced severe (grade 3) toxicity and no one dropped out of the study due to adverse effects" (Appendix 12: Table 8).

Cost

No cost data related to the effectiveness of bethanecol versus placebo were reported.

9. Bethanechol versus artificial saliva

One study, at high risk of bias, compared bethanechol with artificial saliva (Jham 2007).

Xerostomia

There was insufficient evidence of a difference in having a dry mouth (yes/no) either at the end of radiotherapy (RR 0.63, 95% CI 0.30 to 1.29; P = 0.2; 36 participants) or at 8 to 40 weeks postradiotherapy (RR 0.56, 95% CI 0.30 to 1.05; P = 0.07; 30 participants) (Analysis 9.1).

231

Analysis 9.1: Bethanechol versus artificial saliva; Outcome: Xerostomia (yes/no)

232 Salivary flow rates

Bethanechol increased unstimulated saliva flow (ml/min) at the end of radiotherapy (MD 0.12, 95% CI 0.01 to 0.23; P = 0.03; 36 participants), but there was insufficient evidence of a difference at 8 to 40 weeks postradiotherapy (MD 0.07, 95% CI -0.02 to 0.16; P = 0.13; 33 participants) (Analysis 9.2).

There was insufficient evidence of a difference in stimulated saliva flow (ml/min) at the end of radiotherapy (MD 0.13, 95% CI -0.03 to 0.29; P = 0.12; 32 participants), but there was a benefit in favour of bethanechol at 8 to 40 weeks postradiotherapy (MD 0.21, 95% CI 0.01 to 0.41; P = 0.04; 29 participants) (Analysis 9.3).

233

Analysis 9.2: Bethanechol versus artificial saliva; Outcome: Salivary flow rate in ml/min (unstimulated)

234

Analysis 9.3: Bethanechol versus artificial saliva; Outcome: Salivary flow rate in ml/min (stimulated)

235 Survival

There was insufficient evidence of a difference in overall survival at 40 weeks postradiotherapy (RR 1.59, 95% CI 0.43 to 5.84; P = 0.48; 43 participants; Appendix 13: Analysis 9.4).

Quality of life and other oral related symptoms

No data on either quality of life or other oral related symptoms were reported.

Side effects

There were low rates of adverse effects (watering eyes, nervousness, frequent urination, sweating, warm face, cramps, diarrhoea, nausea) with insufficient evidence of any differences (Appendix 12: Table 9).

Cost

No cost data related to the effectiveness of bethanechol versus artifical saliva were reported.

10. Selenium versus no intervention

Selenium was compared to no intervention in one trial assessed as at high risk of bias (Büntzel 2010).

Xerostomia

We were unable to use any of the data as bar charts of mean scores (baseline to postradiotherapy) were presented with no standard deviations. The results reported in the text indicated that there was no evidence that selenium reduced xerostomia.

Salivary flow rates

No data on salivary flow rates were reported.

Survival

No survival data were reported.

236 Quality of life and other oral related symptoms

We were unable to use any of the data for loss of taste or dysphagia as bar charts of mean scores were presented with no standard deviations, but some information was reported in the text (Appendix 12: Table 10). There was insufficient evidence of any differences.

Side effects

The text around adverse events implied that there was no evidence that selenium caused a higher number of events. We were unable to analyse the data as they were clustered (Appendix 12: Table 10).

Cost

No cost data related to the effectiveness of selenium were reported.

11. Antiseptic mouthrinse versus placebo

Antiseptic mouthrinse was assessed in one small trial at high risk of bias (Lanzós 2010).

Xerostomia

No data on xerostomia were reported.

Salivary flow rates

This outcome was only reported during radiotherapy (i.e. not at any of the time points we were interested in).

Survival

No survival data were reported.

Quality of life and other oral related symptoms

No quality of life data were reported, however, data on hyposialosis (drooling) were reported; there was insufficient evidence to determine whether or not antiseptic mouthrinse reduced or increased drooling in patients after four weeks (Appendix 12: Table 11).

Side effects

237 The study reported "no relevant adverse events were reported in any group" (Appendix 12: Table 11).

Cost

No cost data related to the effectiveness of antiseptic mouthrinses were reported.

12. Antimicrobial lozenge versus placebo

Antimicrobial lozenge was assessed in one trial assessed as at high risk of bias (Duncan 2005).

Xerostomia

There was insufficient evidence that the antimicrobial lozenges reduced xerostomia up to and including three months postradiotherapy (RR 1.16, 95% CI 0.97 to 1.40; P = 0.11; 133 participants) (Analysis 12.1).

238

Analysis 12.1: Antimicrobial lozenge versus placebo; Outcome: Xerostomia (QoL response for dryness: yes/no)

239 Salivary flow rates

No data on salivary flow rates were reported.

Survival

No survival data were reported.

Quality of life and other oral related symptoms

There was insufficient evidence of a difference between the groups for global quality of life (Appendix 13: Analysis 12.2), mouth pain, sore/burning mouth, or throat pain (Appendix 12: Table 12).

Side effects

There was weak evidence that the antimicrobial lozenge may cause nausea, but insufficient evidence of a difference in dryness in the mouth, diarrhoea, or constipation (Appendix 12: Table 12).

Cost

No cost data related to the effectiveness of antimicrobial lozenges were reported.

13. Polaprezinc versus azulene oral rinse

One study at high risk of bias compared polaprezinc with azulene oral rinse (Watanabe 2010).

Xerostomia

There was some weak evidence that polaprezinc reduced severe xerostomia at the end of radiotherapy when compared with azulene rinse (RR 0.17, 95% CI 0.04 to 0.65; P = 0.009; 31 participants) (Analysis 13.1).

240

Analysis 13.1: Polaprezinc versus azulene oral rinse; Outcome: Xerostomia (0 to 3 scale - grade 2 or above)

241 Salivary flow rates

No data on salivary flow rates were reported.

Survival

The study reported tumour response by RECIST (Response Evaluation Criteria In Solid Tumors) criteria for a specific group of patients only.

Quality of life and other oral related symptoms

There was some weak evidence that polaprezinc reduced severe oral pain and severe taste disturbance, however there was no evidence that polaprezinc helped patients to eat more when compared with azulene oral rinse (Appendix 12: Table 13).

Side effects

No data on adverse events were reported.

Cost

No cost data related to the effectiveness of polaprezinc or azulene were reported.

14. Venalot Depot (coumarin/troxerutin) versus placebo

One small trial, assessed as at high risk of bias, compared Venalot Depot with placebo (Grötz 2001).

Xerostomia

RTOG scores were reported, however, these scores were a composite of radiation side effects on different sites, so could not be used for assessing xerostomia.

Salivary flow rates

The sialometric data showed the reading dropped to an unmeasurable level in both groups so this was abandoned as a primary marker of efficacy and the protocol changed.

Survival

242 We were unable to use the data on locoregional control.

Quality of life and other oral related symptoms

No data on either quality of life or other oral related symptoms were reported.

Side effects

The study reported that "no adverse events could be attributed to the experimental medication" (Appendix 12: Table 14).

Cost

No cost data related to the effectiveness of Venalot Depot were reported.

Sensitivity analysis Risk of bias

There were too few studies at low risk of bias to carry out sensitivity analyses based on this factor.

Publication status

Four of the 39 studies were unpublished; Abacioglu 1997 was a dissertation, Lozada-Nur 1998, Patni 2004 and Veerasarn 2006 were conference abstracts. The authors of these papers were contacted and their data were provided. Sensitivity analysis was undertaken to assess whether the inclusion of unpublished information had an effect on the results of the review.

Pilocarpine

Abacioglu 1997 and Lozada-Nur 1998 assessed pilocarpine. Removing these unpublished trials from the results did not alter the findings of the review i.e. xerostomia and salivary flow rates at the end of radiotherapy and three months postradiotherapy were still not significant.

Amifostine

Patni 2004 and Veerasarn 2006 assessed amifostine. 243 Removing Veerasarn 2006 from the analysis of xerostomia at the end of radiotherapy increased the effect estimate from RR 0.35 (95% CI 0.19 to 0.67; P = 0.001; 119 participants) to RR 0.22 (95% CI 0.09 to 0.53; P = 0.0008; 60 participants).

Removing Patni 2004 and Veerasarn 2006 from the analysis of xerostomia at three months postradiotherapy did not change the result but increased the uncertainty around the effect estimate from RR 0.66 (95% CI 0.48 to 0.92; P = 0.01; 687 participants) to RR 0.66 (95% CI 0.40 to 1.09; P = 0.1; 473 participants), thus including the possibility of harm associated with amifostine.

Removing Patni 2004 and Veerasarn 2006 from the analysis of xerostomia at 12 months postradiotherapy did not affect the result i.e. from RR 0.70 (95% CI 0.40 to 1.23; P = 0.21; 682 participants analysed) to RR 0.64 (95% CI 0.38 to 1.08; P = 0.09; 479 participants analysed).

244 5.6 Summary of findings tables

1 Pilocarpine compared to no treatment/placebo for preventing salivary gland dysfunction following radiotherapy

Pilocarpine compared to no treatment/placebo for preventing salivary gland dysfunction following radiotherapy Patient or population: patients receiving radiotherapy on its own or in addition to chemotherapy to the head and neck region Intervention: pilocarpine Comparison: no treatment/placebo Outcomes Anticipated absolute effects* (95% Relative Number of Quality of Comments CI) effect participants the Risk with no Risk with (95% (studies) evidence treatment/placebo pilocarpine CI) (GRADE) Xerostomia - Up to and - SMD 0.35 - 126 ⊕⊝⊝⊝ Insufficient evidence of a difference at this including 6 months lower (2 RCTs) VERY time point and also at the end of RT and 3 postRT (1.04 lower to LOW1 months postRT 0.33 higher) Studies used different 1 of the 2 studies in this assessment showed ways of measuring the inconsistent results when using an alternative outcome and therefore we way of measuring this outcome at the 6- combined the studies month time point. 2 further studies showed using SMD insufficient evidence of a difference, 1 at the end of RT and the other at 3 months postRT Salivary flow rate - MD 0.06 lower - 24 ⊕⊝⊝⊝ Insufficient evidence of a difference at this (unstimulated) - Up to and (0.23 lower to (1 RCT) VERY time point and also at the end of RT including 3 months 0.11 higher) LOW2 postRT Same results for stimulated salivary flow rates at end of RT, and 3, 6 and 12 months Studies used different

245 ways of measuring the postRT outcome and therefore we combined the studies Same results for a further study at the end of using SMD RT and 3 months postRT looking at whether or not stimulated and unstimulated salivary flow was > 0 g Overall survival - Up to 724 per 1000 775 per 1000 RR 1.07 60 ⊕⊝⊝⊝ Insufficient evidence of a difference and including 6 months (579 to 1000) (0.80 to (1 RCT) VERY postRT 1.43) LOW3 Quality of life - Up to and Control group mean MD 0.20 higher - 90 ⊕⊝⊝⊝ Insufficient evidence of a difference at this including 6 months was 5.3 (0.19 lower to (1 RCT) VERY time point and also at the end of RT and 3 postRT 0.59 higher) LOW3 months postRT

McMaster University Head and Neck Questionnaire (HNRQ). Score 1-7, lower score = poorer quality of life Insufficient evidence of a difference between groups for any reported adverse event, apart from for sweating where data Adverse effects from 5 studies showed an increased risk associated with pilocarpine (RR 2.98, 95% CI 1.43 to 6.22; P = 0.004; I2 = 0%; 389 participants; ⊕⊕⊝⊝ LOW4) *The risk in the intervention group (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI) CI: confidence interval; MD: mean difference; RCT: randomised controlled trial; RR: risk ratio; SMD: standardised mean difference; RT: radiotherapy GRADE Working Group grades of evidence High quality: we are very confident that the true effect lies close to that of the estimate of the effect Moderate quality: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a

246 possibility that it is substantially different Low quality: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect Very low quality: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect

Footnotes

1Downgraded by 1 level for risk of bias, 1 level for imprecision (small sample size and 95% CIs include both possibility of benefit and harm), and 1 level for inconsistency (I2 = 68%). 2Downgraded by 1 level for risk of bias, and 2 levels for imprecision (single study with 12 participants per group and 95% CIs include both possibility of benefit and harm). 3Downgraded by 1 level for risk of bias, and 2 levels for imprecision (single study and 95% CIs include both possibility of benefit and harm). 4Downgraded by 1 level for risk of bias, and 1 level for imprecision (very wide 95% CIs).

247 2 Amifostine compared to no treatment/placebo for preventing salivary gland dysfunction following radiotherapy

Amifostine compared to no treatment/placebo for preventing salivary gland dysfunction following radiotherapy Patient or population: patients receiving radiotherapy on its own or in addition to chemotherapy to the head and neck region Intervention: amifostine Comparison: no treatment/placebo Outcomes Anticipated absolute effects* Relative Number of Quality of Comments (95% CI) effect participants the Risk with no Risk with (95% (studies) evidence treatment/placebo amifostine CI) (GRADE) Xerostomia (0-4 418 per 1000 292 per 1000 RR 0.70 682 ⊕⊕⊝⊝ Insufficient evidence of a difference at this time scale - grade 2 or (167 to 514) (0.40 to (7 studies) LOW1 point. However, both at the end of RT (RR 0.35, 95% above) - 12 months 1.23) CI 0.19 to 0.67; 3 studies, 119 participants) and up to postRT 3 months postRT (RR 0.66, 95% CI 0.48 to 0.92; 5 studies, 687 participants), amifostine reduced the risk of developing grade ≥ 2 xerostomia Salivary flow rate Control group mean MD 0.32 - 27 ⊕⊝⊝⊝ Amifostine led to increased unstimulated saliva flow (mg/5 min) was 0.16 higher (1 study) VERY both at 12 months postRT and at the end of RT, but (unstimulated) - 12 (0.09 higher LOW2 there was insufficient evidence of a difference at 3 months postRT to 0.55 months postRT. This evidence was supported by a higher) further study showing a benefit for amifostine at 12 months postRT when looking at incidence of producing > 0.1 g of saliva over 5 minutes (RR 1.45, 95% CI 1.13 to 1.86; 175 participants). A further study narratively reported no difference

Insufficient evidence of a difference in stimulated saliva flow at any time point

248 Overall survival at 450 per 1000** 531 per 1000 HR 1.18 271 ⊕⊝⊝⊝ Insufficient evidence to determine whether or not 12 to 24 months (383 to 747) (0.85 to (2 studies) VERY amifostine reduces overall survival, progression-free postRT 1.66) LOW3 survival, disease-free survival or locoregional tumour control up to 24 months postRT Quality of life Control group mean MD 0.7 - 180 ⊕⊝⊝⊝ Amifostine led to a small improvement in quality of (Patient Benefit was 6.66 higher (1 study) VERY life at 12 months postRT, but there was insufficient Questionnaire) - 12 (0.2 higher LOW2 evidence of a difference at the end of RT and 3 months postRT to 1.2 months postRT higher) 8 items each on a 10- A further study narratively reported no difference at point scale where end of RT and 6, 12, 18, and 24 months postRT higher = better QoL  Data from 5 studies showed an increased risk of vomiting associated with amifostine (RR 4.90, 95% CI 2.87 to 8.38; 601 participants; ⊕⊕⊝⊝ LOW4)  Data from 3 studies showed an increased risk of hypotension associated with amifostine (RR 9.20, 95% CI 2.84 to 29.83; 376 participants; ⊕⊕⊝⊝ LOW4)  Data from 4 studies showed an increased risk of nausea associated with amifostine (RR 2.60, 95% CI 1.81 to 3.74; Adverse effects 556 participants; ⊕⊕⊝⊝ LOW4)  Data from 3 studies showed an increased risk of allergic response associated with amifostine (RR 7.51, 95% CI 1.40 to 40.39; 524 participants; ⊕⊕⊝⊝ LOW4)

There was insufficient evidence of a difference between groups for any other adverse events *The risk in the intervention group (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI) **2014 5-year overall survival rate of patients with head and neck squamous cell carcinoma (www.who.int/selection_medicines/committees/expert/20/applications/HeadNeck.pdf) CI: confidence interval; HR: hazard ratio; MD: mean difference; QoL: quality of life; RR: risk ratio; RT: radiotherapy

Footnotes

249 1Downgraded by 1 level for risk of bias, and 1 level for inconsistency (I2 = 83%). 2Downgraded by 1 level for risk of bias, and 2 levels for imprecision (single study and small sample size). 3Downgraded by 1 level for risk of bias, and 2 levels for imprecision (small sample size and 95% CIs include both possibility of benefit and harm). 4Downgraded by 1 level for risk of bias, and 1 level for imprecision (very wide 95% CIs).

250 3 Palifermin compared to placebo for preventing salivary gland dysfunction following radiotherapy

Palifermin compared to placebo for preventing salivary gland dysfunction following radiotherapy Patient or population: patients receiving radiotherapy on its own or in addition to chemotherapy to the head and neck region Intervention: palifermin Comparison: placebo Outcomes Anticipated absolute Relative Number of Quality of the Comments effects* (95% CI) effect participants evidence Risk with Risk with (95% CI) (studies) (GRADE) placebo palifermin Xerostomia (0-4 scale - grade 727 per RR 0.97 471 ⊕⊕⊝⊝ Insufficient evidence of a difference at this 705 per 1000 2 or above) - Up to and 1000 (0.77 to (3 studies) LOW1 time point (560 to 887) including 3 months postRT 1.22) Overall survival at 42 to 72 450 per HR 1.00 (3 studies) ⊕⊕⊕⊝ Insufficient evidence to determine whether months from baseline 1000** 450 per 1000 (0.72 to MODERATE2 or not amifostine reduces overall survival (324 to 626) 1.39) and progression-free survival up to 72 months Adverse effects There was insufficient evidence of patients in either group experiencing more or less adverse events *The risk in the intervention group (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI) **2014 5-year overall survival rate of patients with head and neck squamous cell carcinoma (www.who.int/selection_medicines/committees/expert/20/applications/HeadNeck.pdf) CI: confidence interval; HR: hazard ratio; RR: risk ratio; RT: radiotherapy

Footnotes

1Downgraded by 1 level for imprecision (95% CIs include both possibility of benefit and harm), and 1 level for inconsistency (I2 = 76%). 2Downgraded by 1 level for imprecision (95% CIs include both possibility of benefit and harm).

251 5.7 Discussion

Summary of main results

A total of 39 trials were included in this review. We assessed the quality of the body of evidence for each outcome within a comparison (providing there was more than one study) using GRADE methodology (45).

Pilocarpine (Summary of findings table 1), compared with no treatment/placebo, was evaluated in 12 trials. There was no evidence of a difference in xerostomia between treatment groups at end of radiotherapy, three or six months. Similarly, there was also no evidence of a difference between treatment groups for salivary flow rates (stimulated or unstimulated) at any time point. There was insufficient evidence to determine the benefit of pilocarpine with regard to improving quality of life or increasing survival. There was no difference in reported adverse events, apart from sweating, where data from five studies showed an increased risk with pilocarpine. The body of evidence for each outcome was rated as very low quality, except for the adverse event of sweating which was low quality.

Amifostine (Summary of findings table 2), compared with no treatment/placebo, was evaluated in 11 studies. There is some (low-quality) evidence that amifostine reduced the risk of developing grade ≥ 2 xerostomia (0 to 4 scale) at end of radiotherapy and, to a lesser extent, up to and including three months postradiotherapy. At 12 months postradiotherapy, there was insufficient evidence of a difference in the risk of grade ≥ 2 xerostomia. There was inconsistent (very low-quality) evidence regarding the effect of amifostine on salivary flow rate. There was insufficient (very low-quality) evidence to determine whether or not amifostine reduced overall survival, progression-free survival, disease-free survival or locoregional tumour control. Similarly, there was insufficient (very low-quality) evidence to determine the benefit of amifostine in terms of quality of life. In general, adverse effects were poorly reported but there was (low-quality) evidence that amifostine was associated with an increased risk of vomiting, hypotension, nausea and allergic response.

Palifermin (Summary of findings table 3), compared with placebo, was evaluated in three trials. There is insufficient (low-quality) evidence to determine whether or not palifermin reduced the incidence of grade ≥ 2 xerostomia (0 to 4 scale) up to three months postradiotherapy. There was insufficient (moderate-quality) evidence to determine the effect of palifermin on overall or progression-free survival. There was no evidence of a difference in reported adverse effects.

252 All evidence from any remaining comparisons did not undergo formal GRADE assessment but was considered to be of very low quality.

Five trials (four at high risk of bias and one at unclear risk of bias) evaluated different forms of Chinese medicine. There is some evidence to suggest a benefit from Shenqi Fanghon recipe and an unspecified Chinese medicine at reducing xerostomia. Similarly, the unspecified Chinese medicine improved salivary flow rates. However, these findings were from single studies at high risk of bias. There was insufficient evidence to determine if any of the Chinese medicines had any effect on quality of life and survival.

Other interventions evaluated, for which there is currently insufficient evidence to draw conclusions were:

 amifostine - comparison of doses (single trial, at high risk of bias);  amifostine - different routes of administration (single trial, at high risk of bias);  biperiden (single trial, at high risk of bias);  bethanechol (single trial, at unclear risk of bias);  bethanecol versus artificial saliva (single trial, at high risk of bias);  selenium (single trial, at high risk of bias);  antiseptic mouthrinse verus placebo (single trial, at high risk of bias);  antimicrobial lozenge versus placebo (single trial, at high risk of bias);  polaprezinc versus azulene oral rinse (single trial, at high risk of bias);  Venalot Depot versus placebo (single trial, at high risk of bias).

Overall completeness and applicability of evidence

Although we found 39 eligible studies that covered a wide range of interventions, the evidence found is not sufficient to highlight much promise in terms of effective preventative treatments for salivary gland dysfunction. This is because, despite there being a reasonable number of studies for three of the interventions (amifostine: 12; pilocarpine: 12; palifermin: 3), there was inconsistency in the way outcomes were reported and in the timing of outcome measurement. Most comparisons included only a single small study, the large majority being at high risk of bias. The most complete body of evidence was for amifostine and the outcome of incidence of moderate to severe xerostomia. Guideline statements point out the lack of an established pharmacological prophylaxis for salivary gland dysfunction and highlight the potential of radiotherapeutic techniques/precautions in reducing damage to the salivary glands (for example parotid-sparing plans) (39). Therefore, for completeness, it may be sensible to also carry out a Cochrane Review of 253 non-pharmacological interventions, although we are not aware of many randomised controlled trials.

As mentioned, a wide range of interventions were assessed, but the studies were also conducted in both middle-income and high-income countries with no exclusion criteria in terms of the population included. Unfortunately, many studies did not include an objective measure of saliva flow to go with the more subjective measure of xerostomia. Furthermore, xerostomia was often measured differently between comparisons making it difficult to get an overall picture of the comparative effectiveness of the different interventions.

It was interesting that two of the three interventions for which we were able to carry out meta-analyses (amifostine and palifermin) have shown promise in a Cochrane Review on the prevention of oral mucositis, another major side effect of cancer treatment (14). As the evidence for amifostine in the prevention of salivary gland dysfunction is promising, this could be beneficial for patients as they may require fewer medications. However, amifostine is not currently recommended in clinical practice guidelines due to high costs and its side effects (39). Furthermore, the evidence for its long-term benefit is weak. There was insufficient evidence to support the use of palifermin in this review, but it is possible that the intention-to-treat rules in one of the three studies in the xerostomia meta-analysis may have influenced the result (115). Therefore, further studies assessing palifermin may be of interest. Worthington 2011 reported that there was no evidence that pilocarpine prevents oral mucositis (14).

Quality of the evidence

We included 39 studies that randomised 3520 participants; the number of participants analysed varied by outcome and time point.

Pilocarpine

We have very little confidence in the effect estimates for the outcomes of xerostomia, salivary flow rate, survival and quality of life (none of which showed a difference), mainly due to concerns regarding the risk of bias of the studies and imprecision of the results, but also due to inconsistency in the case of xerostomia. Further studies are likely to change the results. We had a little more confidence (although still limited) in the effect estimate for the adverse effect of sweating, which occurred more frequently in those receiving pilocarpine. Again it was risk of bias and imprecision which limited our confidence, and

254 further studies would probably change the effect estimate. For more details see Summary of findings table 1.

Amifostine

Our confidence in the effect estimates for xerostomia was limited by concerns regarding the risk of bias and inconsistency in the results of the individual studies. New studies could change the results. We had a similar level of confidence in the results for the adverse effects of vomiting, hypotension, nausea and allergic response, which were all more frequent in those receiving amifostine. Risk of bias and imprecision were the factors affecting our confidence. We had very little confidence in the effect estimates for salivary flow rate, survival and quality of life due to risk of bias and imprecision. For more details see Summary of findings table 2.

Palifermin

Our confidence in the effect estimate for xerostomia was limited by concerns regarding imprecision and inconsistency in the results of the individual studies. We were moderately confident that palifermin did not compromise survival. For more details see Summary of findings table 3.

We did not formally assess the quality of the evidence for all other comparisons in this review, but it is all considered to be very low quality due to single small studies that are mostly at high risk of bias. Further studies would very likely change the effect estimates for all outcomes and time points within these comparisons.

Potential biases in the review process

Standard Cochrane methods were followed to avoid biases in the review process. However, we acknowledge that the decision to exclude data measured and reported during radiotherapy may be considered by some readers to be an arbitrary one. These data are potentially of interest and their exclusion may be thought of as a bias.

Selection bias in the review process may have occured where studies were excluded due to not reporting our primary outcomes. In future updates we will write to the authors of such studies to ask if they measured the outcomes. If any other studies are identified in the search updates with the same issue, we will write to the authors prior to deciding on eligibility. Indeed obtaining data from trial authors is also essential to reduce outcome reporting bias and we will attempt to encourage authors to reply to emails requesting 255 further information in future updates in various ways. This will include providing copies of the previous version of the review and explaining the benefits of having their research included.

Agreements and disagreements with other studies or reviews

Although there are some systematic reviews on the treatment of salivary gland dysfunction caused by radiotherapy (172, 209), we are not aware of any high quality systematic reviews on prevention.

5.8 Authors' conclusions

Implications for practice

Implications for research

Further well conducted, well reported and adequately powered randomised controlled trials are needed to add to the evidence base for the interventions assessed in the single-study comparisons of this systematic review. Amifostine should be assessed with longer term follow-up to establish whether the promising shorter term effects are sustained. Palifermin should also be studied further and with longer follow-up.

Trialists should endeavour to use similar scales to measure xerostomia i.e. one that can be dichotomised to report the incidence of moderate to severe or severe xerostomia or both. Buglione et al recommend several established standardised scales such as NCI CTCAE (National Cancer Institute Common Terminology Criteria for Adverse Events), RTOG (Radiation Therapy Oncology Group), and LENT-SOMA (Late Effects Normal Tissue Task Force - Subjective, Objective, Management, Analytic scale) (39). This should be reported alongside a more objective measure such as salivary flow rate. Adverse effects

256 should also be clearly reported and quality of life would be a useful patient-important outcome.

257 6 THE QUALITY OF CLINICAL GUIDELINES ON MOUTH CARE FOR CANCER PATIENTS

6.1 Abstract

Background: Oral side effects of cancer treatment can be harmful to patients and affect their chances of survival and quality of life. High quality evidence-based clinical practice guidelines are needed in order to help clinicians to manage these patients. Aims: To assess the quality of guidelines on the management of oral complications caused by cancer treatment. Methods: Electronic searching of MEDLINE Ovid was combined with searching common guideline websites to identify relevant guidelines. The AGREE II instrument was used to appraise the guidelines, which were also assessed to determine whether or not their recommendations were based on high quality systematic review evidence. Results: There were 20 guidelines that met the eligibility criteria and were appraised as described above. The quality of the guidelines was generally found to be poor, such that only 5 would be recommended, based on the quality of the methods, reporting, or both. Only 4 of the 37 sets of recommendations, made across the 20 guidelines, were judged to be valid and based on rigorous systematic review evidence. Conclusions: The current state of guidelines on mouth care for cancer patients is very worrying. Guideline developers and systematic review producers should collaborate more to produce fewer guidelines and ensure that they are reliable.

6.2 Background

Treatments for cancer can lead to oral complications such as mucositis, candidiasis, salivary gland dysfunction, herpes simplex virus, and dental caries. These problems can lead to a reduction in quality of life, with oral mucositis potentially causing disruption to cancer treatment in the form of missed or reduced doses. Evidence-based strategies for the management of these sequelae are needed. Unfortunately, clinical practice may vary internationally, nationally, regionally, or even by individual doctor in the same location (49). Indeed it has previously been shown that oral care for children receiving cancer treatment in the UK varies between national specialist centres (15).

The Institute of Medicine’s 2011 definition of clinical practice guidelines describes them as “statements that include recommendations intended to optimize patient care. They are informed by a systematic review of evidence and an assessment of the benefits and harms 258 of alternative care options” (47). Theoretically, clinical guidelines based on rigorous systematic reviews should reduce this kind of inappropriate variation, both by recommending effective practices and discouraging those that have not been proven to be effective (49).

There are many examples of the potential for clinical guidelines to change practice. Previous Cochrane systematic reviews on mouth care for cancer patients have shown a lack of evidence to support the use of prophylactic chlorhexidine for the prevention of oral mucositis and oral candidiasis, and of prophylactic nystatin for the prevention of oral candidiasis (14, 50). These reviews informed the Children’s Cancer and Leukaemia Group and Paediatric Oncology Nurses Forum (CCLG-PONF) Mouth Care Group UK national guidelines on mouth care for children, teenagers and young adults receiving cancer treatment (16). A subsequent audit carried out to monitor the uptake of national guidelines in the UK found that these guidelines had resulted in significant changes in the prescribing of these two interventions, leading to a 40% decrease and a 34% decrease in the routine use of prophylactic nystatin and chlorhexidine respectively (51). This example represents an ideal flow of evidence through the healthcare system

As with primary studies and systematic reviews, guidelines should not be followed blindly without careful critical appraisal to discover whether or not they have been conducted in a rigorous and transparent manner. One such tool for this purpose is the Appraisal of Guidelines for Research and Evaluation (AGREE) instrument, which is a checklist that can be used both by guideline developers to guide their work, and by end users to evaluate their methodological quality. The AGREE instrument has become the internationally accepted standard (56).

6.3 Objective

To assess the quality of guidelines on the management of oral complications caused by cancer treatment.

6.4 Methods

Types of studies Guidelines were included if they met the following criteria:

259  make “recommendations intended to optimize patient care and assist physicians and/or other health care practitioners and patients to make decisions about appropriate health care for specific clinical circumstances” (210);  published in 2005 or after;  published in English.

Guidelines were excluded if:  they were unamended reproductions of other published guidelines;  they had been superseded by a more recent version conducted by the same guideline group.

Guideline topic Guidelines on adults (16 years or older) receiving treatment for cancer and who have either developed or are at risk of developing the following conditions:  oral mucositis;  oral candidiasis;  salivary gland dysfunction (encompassing xerostomia and salivary gland hypofunction);  herpes simplex virus;  dental caries.

Data collection An electronic search of MEDLINE Ovid was carried out on 11 May 2017 using the following search strategy, limited to publication years 2005 to 2017: 1. exp NEOPLASMS/ 2. exp RADIOTHERAPY/ 3. exp Antineoplastic agents/ 4. Anti-neoplastic combined chemotherapy protocols/ 5. Bone Marrow Transplantation/ 6. Hematopoietic Stem Cell Transplantation/ 7. (neoplasm$ or cancer$ or leukaemi$ or leukemi$ or tumour$ or tumor$ or or malignan$ or neutropeni$ or carcino$ or adenocarcinoma$ or lymphoma$).ti,ab. 8. (radioth$ or radiat$ or irradiat$).ti,ab. 9. ((bone adj marrow adj5 transplant$) or "hematopoietic stem cell transplant$" or "haematopoietic stem cell transplant$" or HSCT).ti,ab. 10. chemo$.ti,ab. 260 11. or/1-10 12. exp STOMATITIS/ 13. Candidiasis, Oral/ 14. (stomatitis or mucositis or (oral adj4 candid$) or (oral and fung$) or mycosis or mycotic or thrush).ti,ab. 15. (oral adj6 mucos$).ti,ab. 16. mIAS.ti,ab. 17. Salivary gland diseases/ 18. Xerostomia/ 19. (salivary gland$ adj4 (dysfunction or hypofunction)).ti,ab. 20. (xerostomia or “dry mouth” or radioxerostomia or hyposalivat* or asialia).ti,ab. 21. Herpesvirus 1, Human/ 22. Stomatitis, herpetic/ 23. (“cold sore$” or coldsore$ or herpetic or herpes).ti,ab. 24. Dental caries/ 25. (caries or carious or (tooth adj5 (decay or deminerali$ or reminerali$)).ti,ab. 26. ((oral or mouth) adj5 care).ti,ab. 27. Or/12-26 28. 11 and 27 29. exp consensus development conference/ 30. exp guideline/ 31. guidelines as topic/ 32. exp practice guideline/ 33. practice guidelines as topic/ 34. health planning guidelines/ 35. (guideline or practice guideline or consensus development conference or consensus development conference, NIH).pt. 36. (standards or guideline or guidelines or guidance$).ti,kf,kw. 37. ((practice or treatment$ or clinical) adj guideline$).ab. 38. (CPG or CPGs).ti. 39. or/29-38 40. 28 and 39 After the removal of duplicates, this search resulted in 244 records for screening.

Eight websites were also searched for guidelines using the following search terms:  cancer and mucositis;

261  cancer and stomatitis;  cancer and mouth care;  cancer and caries.

These websites were searched on 11 May 2017: National Guidelines Clearing House https://www.guideline.gov/ Guidelines International Network http://www.g-i-n.net/library/international- guidelines-library NICE National Library of Guidelines https://www.nice.org.uk/guidance/ Scottish Intercollegiate Guidelines http://www.sign.ac.uk/guidelines/ Network New Zealand Guidelines Group http://www.health.govt.nz/publications?f[0 ]=im_field_publication_type%3A26 Scottish Dental Clinical Effectiveness http://www.sdcep.org.uk/published- Programme guidance/ Canadian Medical Association Infobase https://www.cma.ca/en/Pages/cpg- specialties.aspx Australian Government National Health https://www.nhmrc.gov.au/ and Medical Research Council

Also checked were guidelines that cited Cochrane reviews on the topic, as reported to the National Institute of Health Research in Cochrane Oral Health’s annual reports from 2013 to 2017.

Two assessors examined all the retrieved records independently and in duplicate to determine whether or not they met the inclusion criteria. Full text copies were retrieved for any potentially eligible publications, and also where it was not clear from the title/abstract. The full text copies were also checked against eligibility criteria independently and in duplicate.

Data analysis The AGREE II Instrument was used to appraise each included guideline document (211). Four assessors appraised each included guideline independently as recommended in the AGREE II User’s Manual (211). This involved rating 23 items divided into 6 domains plus a further two overall assessment items, structured as follows:

262 Domain Item Description 1. Scope and 1 The overall objective(s) of the guideline is (are) specifically Purpose described 2 The health question(s) covered by the guideline is (are) specifically described 3 The population (patients, public, etc.) to whom the guideline is meant to apply is specifically described 2. Stakeholder 4 The guideline development group includes individuals from Involvement all relevant professional groups 5 The views and preferences of the target population (patients, public, etc.) have been sought 6 The target users of the guideline are clearly defined 3. Rigour of 7 Systematic methods were used to search for evidence Development 8 The criteria for selecting the evidence are clearly described 9 The strengths and limitations of the body of evidence are clearly described 10 The methods for formulating the recommendations are clearly described 11 The health benefits, side effects, and risks have been considered in formulating the recommendations 12 There is an explicit link between the recommendations and the supporting evidence 13 The guideline has been externally reviewed by experts prior to its publication 14 A procedure for updating the guideline is provided 4. Clarity of 15 The recommendations are specific and unambiguous Presentation 16 The different options for management of the condition or health issue are clearly presented 17 Key recommendations are easily identifiable 5. Applicability 18 The guideline describes facilitators and barriers to its application 19 The guideline provides advice and/or tools on how the recommendations can be put into practice 20 The potential resource implications of applying the recommendations have been considered

263 21 The guideline presents monitoring and/or auditing criteria 6. Editorial 22 The views of the funding body have not influenced the Independence content of the guideline 23 Competing interests of guideline development group members have been recorded and addressed

The four assessors were: Helen Worthington, a biostatistician and Professor of Evidence Based Care; Anne-Marie Glenny, a methodologist and Professor of Health Sciences Research; Anne Littlewood, an Information Specialist; and Philip Riley, a methodologist and Research Fellow. The assessors were chosen due to their expertise and knowledge of systematic review and guideline methodology, and also because they have knowledge of the topic area from previous research. The research project was part of Philip Riley’s PhD and the Professors were his supervisors.

Each of these items was rated on a 1 (strongly disagree) to 7 (strongly agree) scale, although a rating of 1 would also be assigned where there was insufficient information available to rate the item, or where reporting in the guideline was very poor.

The overall assessment involved two further items, the first asking the assessor to rate the overall quality of the guideline on a 1 (lowest possible quality) to 7 (highest possible quality) scale, and the second asking the assessor if they would recommend the guideline for use, with three possible answers: yes; yes with modifications; no.

A quality score for each domain was then calculated by summing all individual scores of all assessors within a domain and scaling the total as a percentage of the maximum possible score for that domain, as described in the AGREE II User’s Manual and shown as follows (211):

Obtained score - Minimum possible score

Maximum possible score - Minimum possible score

The maximum score for domain 1 would be 84: 7 (highest possible rating) x 3 (items) x 4 (appraisers). The minimum score would be 12: 1 (lowest possible rating) x 3 x 4.

264 For the overall quality assessment of each guideline, the median of the four assessors’ scores was reported. For question of whether the assessor would recommend the guideline for use, the individual responses were reported. All results were tabulated.

6.5 Results

Twenty guidelines were deemed eligible for inclusion and were critically appraised using the AGREE II instrument. The overall results are presented below in Tables 1 and 2. For a table of the individual item scores of each of the four assessors per included guideline, see Appendix 14.

265 AGREE II DOMAIN Guideline Scope and Stakeholder Rigour of Clarity of Applicability Editorial Overall Would you purpose (%) involvement development presentation (%) independence assessment recommend? (%) (%) (%) (%) (median) Alberta 2017 97 26 49 71 25 75 5 Y+, Y+, Y+, Y+ (212) Bensinger 54 47 17 67 13 54 3 Y+, N, N, N 2008 (213) Buglione 2016 57 38 52 63 2 69 4 Y+, Y+, Y+, Y+ (39, 214) Butterworth 31 14 5 33 0 0 1.5 N, N, N, N 2016 (215) Califano 2015 67 57 31 72 27 85 2.5 Y+, N, N, N (216) Cohen 2016 90 86 78 83 64 98 6 Y, Y, Y, Y+ (217) Davies 2010 64 25 35 63 9 35 3.5 Y+, Y+, N, N (218) De Sanctis 71 36 47 57 10 67 4 Y+, Y+, Y+, N 2016 (219) Edmonds 68 31 27 58 9 42 3 Y+, N, N, N 2012 (220) Fogh 2014 53 21 8 17 8 33 1.5 N, N, N, N (221) Hensley 2009 63 79 74 86 18 65 5 Y, Y, Y+, Y+ (222) Hutchinson 65 4 7 7 0 31 1.5 N, N, N, N 2014 (223) Jensen 2010 82 33 65 65 17 52 4.5 Y, Y+, Y+, N (30)

266 Lacouture 72 25 30 54 9 90 3.5 Y+, Y+, N, N 2011 (224) Lalla 2014 75 33 64 90 21 96 5 Y, Y, Y+, Y+ (65) Mirabile 2016 68 39 44 58 2 63 3 Y+, N, N, N (225) RCS_BSDOH 75 21 11 57 11 0 2.5 N, N, N, N 2012 (226) Sandherr 2015 81 33 16 28 4 42 2.5 Y+, N, N, N (227) Snowden 71 42 28 38 6 10 3 Y+, N, N, N 2011 (228) Tolentino 65 4 5 11 0 0 1.5 N, N, N, N 2011 (229)

Table 1: Quality scores for each included guideline by AGREE II domain Footnotes  The scores for the six AGREE II domains are percentages of the maximum possible score for each domain (see methods for details of calculation)  The overall assessment is the median of the four assessors’ scores on a 1 (lowest possible score) to 7 (highest possible score) scale  For the question ‘would you recommend this guideline for use?’: Y = yes, Y+ = yes with modifications, N = no

267 Domain Median score Range of scores Scope and purpose 68 31 to 97 Stakeholder involvement 33 4 to 86 Rigour of development 30 5 to 78 Clarity of presentation 58 7 to 90 Applicability 9 0 to 64 Editorial independence 52 0 to 98

Table 2: Summary of AGREE II domain scores

The assessors’ ratings presented in Table 1 indicate that guidelines on mouth care for cancer patients were generally of poor quality according to the AGREE II instrument. This was perhaps best illustrated by the fact that only 5 of the 20 guidelines would be recommended by all four assessors (Alberta 2017; Buglione 2016; Cohen 2016; Hensley 2009; Lalla 2014), albeit with modifications. Another way to summarise the quality of the guidelines was to observe the median overall assessment score (on a 1 to 7 scale). Using a rating of 4 as the mid-point on the scale, again only 5 of the 20 guidelines scored in the upper half of the scale i.e. > 4 (Alberta 2017; Cohen 2016; Hensley 2009; Jensen 2010; Lalla 2014). Interestingly, Buglione 2016 dropped out of the top 5 and was replaced by Jensen 2010, with them scoring 4 and 4.5 respectively. This demonstrates the subjectivity inherent in the AGREE II instrument, and that the interpretation of the scores is not straightforward. Indeed the AGREE II User’s Manual states that the instrument was not developed with any definitions of what scores would constitute high or low quality, and the user should interpret using their own judgement (211). For the six domains which were scored as a percentage, a previous study assessing oral mucositis guidelines used a cut-off score of > 60% to indicate high quality (230).

The first domain ‘Scope and Purpose’ generally scored well, with a median score across the 20 guidelines of 68% (range 31% to 97%). Sixteen of the 20 guidelines scored > 60%. However, it is the least that can be expected for guidelines to state what they are about and it would be very worrying if this were not the case. The second domain ‘Stakeholder Involvement’ generally scored poorly, with a median score of 33% (range 4% to 86%). Only two guidelines scored > 60% (Cohen 2016; Hensley 2009). Item 5 within this domain scored particularly badly as most guidelines had not sought the views of patients or the public. The third domain ‘Rigour of Development’ represents the methodology of the guidelines and therefore might reasonably be considered the most critically important 268 domain, but unfortunately scored poorly, with a median score of 30% (range 5% to 78%). Only four guidelines scored > 60% (Cohen 2016; Hensley 2009; Jensen 2010; Lalla 2014). The fourth domain ‘Clarity of Presentation’ scored reasonably well, with a median score of 58% (range 7% to 90%). Nine guidelines scored > 60%, but it was easy to score higher points simply by presenting the recommendations clearly in a box or table. The fifth domain ‘Applicability’ scored extremely poorly, with a median score of 9% (range 0% to 64%), and only one guideline scoring > 60% (Cohen 2016). It is perhaps a lot to expect from some of the guidelines that are not a product of large organisations, such as the American Cancer Society (Cohen 2016), to have the expertise and resources to be able to score highly on the items in this domain. However, it is apparent that even such large organisations are not giving sufficient consideration to these items. The sixth and final domain ‘Editorial Independence’ scored reasonably well but the median score (52%) was brought down by the three studies that scored 0% (Butterworth 2016; RCS_BSDOH 2012; Tolentino 2011). This domain had the widest range of scores (0% to 98%).

As it became apparent when using the AGREE II instrument that it does not ask whether the recommendations are based on appropriate evidence, a post-hoc assessment was carried out to investigate this. However, it should be noted that this was purely a qualitative assessment and that different assessors may arrive at different conclusions. Guidelines were checked to see if the recommendations were based on high quality systematic reviews in an attempt to make the assessment more objective. Conversely, this could again be viewed as being subjective as: a) the assessors’ relied on a brief appraisal of any cited systematic reviews; and b) many guidelines had carried out their own ‘systematic review’ which the assessors normally did not consider to be a true systematic review. Having said this, the assessors were both Editors of Cochrane Oral Health and are experienced appraisers of systematic reviews. The guidelines were not assessed in duplicate. The results are presented below in Table 3.

269 Guideline Population Condition Grade/level of Recommendation(s) Comment recommendation(s) supported by high quality systemic review? Alberta 2017 Head and neck Oral None stated No Cochrane review cited as evidence for use of oral cancer patients mucositis cryotherapy only. However, probably not relevant to this population.

Other recommendations not adequately supported by evidence. Oral None stated No Mentions using nystatin, for which there is insufficient candidiasis evidence of benefit in Cochrane reviews. Salivary None stated No Focuses on pilocarpine for which there is insufficient gland evidence for preventative use in the Cochrane review dysfunction that is part of this thesis. Does not cite Davies 2015 which is Cochrane review on treatment. Herpes None stated No Does not cite any systematic review but mentions using simplex acyclovir, which is described as being effective in virus Cochrane review. Vaguely mentions “Oral antivirals can also be used”. Dental None stated No caries Bensinger All cancer Oral None stated Yes The recommendations are largely informed by the 2008 patients mucositis MASCC guidelines (older version of Lalla 2014), which are based on an internal systematic review process. They are also informed by the Cochrane review, plus other clinical studies and expert opinion. Buglione Head and neck Salivary None stated. Based ? Based on the group’s review of the literature but does 2016 cancer patients gland on level of consensus not appear to be a true systematic review. dysfunction

270 Recommendation against amifostine is partly informed by a systematic review following Cochrane methodology. Recommendation on pilocarpine for treatment does not cite Cochrane review (Davies 2015) or another systematic review. Dental None stated. Based No Based on the group’s review of the literature but does caries on level of consensus not appear to be a true systematic review. Butterworth Head and neck Oral None stated No Cochrane review cited but statements not in line with 2016 cancer patients mucositis (recommendations results of the review. assigned a ‘G’ or ‘R’ but this is unexplained) Oral None stated (as No No evidence is cited yet, although vague, the candidiasis above) statements generally agree with the Cochrane review. Salivary None stated (as No Statement on pilocarpine for treatment is in line with gland above) Cochrane review (Davies 2015). dysfunction Dental None stated (as No caries above) Califano 2015 EGFR-mutant Oral None stated ? Very vague and only cites the NCCN Task Force advanced mucositis guidelines (Bensinger 2008). NSCLC (referred to patients as It is important to note that there is very little evidence receiving stomatitis relating to targeted therapy-induced oral mucositis. EGFR tyrosine to kinase distinguish inhibitors such targeted as gefitinib, therapy- erlotinib, and induced oral

271 afatinib mucositis from CT/RT- induced oral mucositis) Cohen 2016 Head and neck Oral 0 = expert opinion/ No Cochrane review on treatment states there is cancer candidiasis observational study insufficient evidence of benefit for any intervention. survivors (late (excluding case- These guidelines recommend fluconazole/clotrimazole effects) control and based on expert opinion. prospective cohort studies)/clinical practice/literature review/pilot study Salivary 0 = (as above) No Does not cite Cochrane review on treatment (Davies gland 2015). dysfunction Dental IA/0 = RCT of HNC No caries survivors/as above Davies 2010 All cancer Salivary SIGN grading levels ? Partially i.e. the recommendation on amifostine for patients gland C and D prevention is informed by a systematic review; the dysfunction recommendation on pilocarpine for treatment is informed by the Cochrane review. De Sanctis Head and neck Oral None stated. Based Yes Recommendations are based on systematic reviews 2016 cancer patients mucositis on level of consensus along with individual studies. Edmonds Renal cell and Oral None stated ? Partially i.e. several of the recommendations were 2012 hepatocellular mucositis/ based on systematic review evidence. cancer patients stomatitis receiving It is important to note that there is very little evidence sorafenib and relating to targeted therapy-induced oral mucositis.

272 other targeted therapies Fogh 2014 All cancer Oral None stated No This publication aims to provide “a set of general patients mucositis recommendations, for symptom management” but does receiving RT Salivary None stated No not do so. gland dysfunction Hensley 2009 All cancer Oral None stated ? Partially i.e. the Cochrane review and other systematic patients mucositis reviews are cited with regards to amifostine, but the recommendations on palifermin appear to be based on an internal literature review. Salivary None stated No Based on an internal literature review. gland dysfunction Hutchinson Head and neck Salivary Level 1 (cholinergic No No systematic reviews are cited. Only four studies are 2014 cancer patients gland stimulants and cited as evidence for all three categories of dysfunction salivary substitutes); interventions. level 1b (acupuncture) Jensen 2010 All cancer Salivary Levels II to IV Yes An internal systematic review was carried out. The patients gland methods are described in two other papers and appear dysfunction rigorous. Systematic reviews were excluded as part of the eligibility criteria. Lacouture Cancer Oral Levels I to IV ? Partially i.e. The recommendations are largely 2011 patients mucositis informed by the MASCC guidelines (older version of receiving Lalla 2014), which are based on an internal systematic EGFR review process, but supplemented with other individual inhibitors studies. The review process for these guidelines is not described in sufficient detail to conclude that it is adequate.

273 Lalla 2014 All cancer Oral Levels I to IV Yes Internal systematic reviews carried out for each patients mucositis category of interventions. Some of those reviews cite the Cochrane reviews. Mirabile 2016 Head and neck Oral None stated. Based No Based on the group’s review of the literature but does cancer patients mucositis on level of consensus not appear to be a true systematic review.

RCS_BSDOH All cancer Oral NICE grading No Grading applied to each individual piece of supporting 2012 patients mucositis system evidence, rather than the recommendation itself. Inappropriate grading of some evidence, e.g “Intensive Varies from IA to oral hygiene reduces mucositis” graded as IA as taken IIIC from SIGN guidance, despite SIGN recommending this as best practice only (not based on research evidence). Oral NICE grading No Ineffective interventions recommended with grading candidiasis system IA. For example, nystatin recommended with reference to a systematic review (hence IA). However, the Varies from IA to systematic review does not actually provide evidence IIIC to support this recommendation. Salivary NICE grading No Grading applied to each individual piece of supporting gland system evidence, rather than the recommendation itself. dysfunction Some, but not all recommendations taken from SIGN Varies from IA to guidance; unclear justification for selection. IIIC Herpes NICE grading No Research evidence available at the time to make simplex system stronger recommendation. virus IIIC

Sandherr Patients with Herpes Infectious Diseases No Systematic review evidence not included. 2015 solid tumours simplex Society of America - and virus United States Public haematological Health Service

274 malignancies grading system for ranking recommendations Snowden Multiple Oral None stated No Limited data presented; incomplete assessment of 2011 myeloma mucositis available evidence. patients Tolentino Head and neck Oral None stated No Recommendations made based on incomplete 2011 cancer patients mucositis evaluation of research evidence/opinion. Salivary None stated No gland dysfunction Dental None stated No caries (radiation caries)

Table 3: Appraisal of guideline recommendations Footnotes CT = chemotherapy EGFR = Epidermal growth factor receptor HNC = head and neck cancer NSCLC = non-small cell lung cancer RCT = randomised controlled trial RT = radiotherapy SIGN = Scottish Intercollegiate Guidelines Network

275 There were 37 sets of recommendations made across the 20 guidelines (i.e. a guideline may have addressed more than one condition, and there may have been more than one recommendation per condition). The most common condition for which there were recommendations, is oral mucositis (in 14 of the 20 guidelines). This was followed by salivary gland dysfunction (11 guidelines), dental caries (5 guidelines), oral candidiasis (4 guidelines), and herpes simplex virus (3 guidelines). This was in line with the burden and seriousness of the conditions. Of these 37 sets of recommendations, only four were judged to have been adequately supported by high quality systematic review evidence (Bensinger 2008; De Sanctis 2016; Jensen 2010; Lalla 2014). Twenty-seven of the sets of recommendations were not supported by high quality systematic review evidence, according to the judgement of the assessors. In some of these cases, systematic review evidence had been cited but, on further investigation, it had either been inappropriately used or misinterpreted. The remaining six sets of recommendations used high quality systematic review evidence for some of the recommendations but not others. The majority of recommendations were also not assigned a grade/level, which added to the general lack of clarity.

6.6 Discussion

Before discussing the findings of this study, it should be acknowledged that there were some limitations that may affect the interpretation by the reader. Firstly, it is important to stress that this was not an exhaustive examination of the world’s guidelines on mouth care for cancer patients. Some guidelines will have been missed through either limitations of the electronic search process or through individual error when screening the search results. Due to resource restraints, general guidelines on the management of cancer patients were not targeted as the volume would not be manageable. However, the purpose of this study was to present a snapshot of the current and recent state of guidelines in this field, and arguably this aim was achieved. Another potential limitation is that the four assessors for the AGREE II part of the study were all methodologists. It is feasible to suggest that clinicians or other stakeholders may have reached different scores/conclusions when appraising the guidelines. An additional limitation is the post-hoc assessment of the evidence base on which the recommendations were made. Analyses that are not stated a priori may always be considered to be inherently biased as they are likely to be driven by the findings of the research. Furthermore, this part of the study is subjective and was not carried out in duplicate using discussion to reach consensus.

276 Looking first at the AGREE II assessments, it is clear that guidelines in this field are either not carried out to the expected standard, or are not reported adequately, or perhaps both. It was worrying that only five of the 20 guidelines were judged to be worthy of recommendation by the four assessors. However, it became apparent during the process of appraising these guidelines that the assessors had reservations about the appropriateness of using AGREE II to make decisions on the quality of guidelines. It seems that AGREE II might only be currently suited to assessing the quality of the reporting of guidelines. This is because there are no items or domains explicitly assessing whether or not the recommendations are actually appropriate i.e. are they supported by relevant and rigorous evidence? This is a severe limitation of the AGREE II instrument, especially considering the Institute of Medicine’s assertion that trustworthy guidelines should be “based on a systematic review of the existing evidence” (47). Whether or not the recommendations made are valid is surely more important than some of the existing domains when considering the purpose of guidelines. Taking the example of the domain ‘Applicability’, the four items could be considered harsh and going over and beyond what guidelines should fundamentally do. A low score in this domain would not necessarily mean that the guidelines are not able to be applied in clinical practice. However, if their recommendations are not valid then they should certainly not be applied in clinical practice. Of course, this is something that could be easily modified with an amendment of the AGREE II instrument. These findings may warrant discussion with the AGREE team and could form part of further research, or perhaps a PhD project, looking into how to rectify the issues. It could be that the author team alert the AGREE team to the issues and allow them to decide how to proceed, or collaborate with them under their guidance.

The second part of the study, presented in Table 3, involved an assessment of whether or not the recommendations made in the guidelines were based on high quality systematic review evidence i.e. were the recommendations valid? The findings were highly alarming. Of the 37 sets of recommendations that were assessed across the 20 guidelines, only four were judged to have been adequately supported by high quality systematic review evidence. Most guidelines did carry out a literature search in order to identify evidence on which to base their recommendations, but they could not be justifiably described as true systematic reviews. This assessment was in conflict with the AGREE II appraisal as only one of those four guidelines with valid recommendations (Lalla 2014) was included in the list of five guidelines that all four assessors would recommend. It is possible that the use of two Cochrane Oral Health Editors when carrying out this assessment could be perceived as being biased towards Cochrane reviews. It would be highly beneficial to carry out this

277 assessment again using other or additional assessors, perhaps as an international Delphi- type process involving a mixture of international content experts and methodologists.

A previous study carried out almost 10 years ago demonstrates the general lack of improvement in guidelines over this time period (230). Potting et al examined guidelines for managing oral mucositis in adult cancer patients using the original AGREE instrument. The findings were similar to this study as “none of the guidelines were of good overall quality” and “none of the guidelines could be recommended”. The authors believed that the development and introduction of the AGREE instrument may lead to improvements in the quality of future guidelines. This clearly has not been the case judging by the results of this study. Further similarities exist between the studies in that Potting et al noted both the strictness of the AGREE instrument and that it does not assess the validity of the recommendations of guidelines.

The current situation is surely not acceptable in such a highly vulnerable population such as cancer patients, where care must be of the highest quality. Another issue is the large number of guideline-type documents in existence. As already mentioned, the guidelines included in this study is by no means a comprehensive list of all those available. The question is ‘why are there so many?’ as, inevitably, the recommendations differ. Clinicians will surely ask which one should they use and which one is correct, and they cannot be blamed for being confused. It is unfathomable that the scientific community have not come together to eradicate such duplication of effort and conflicting recommendations in guidelines. The aim for the future should be to produce fewer guidelines for each of the oral side effects of cancer treatment and disseminate widely to ensure that all patients are treated optimally and equally.

6.7 Conclusion

Most guidelines on the management of oral side effects of cancer treatment are inadequate for their purpose. The methods are either not rigorous or are poorly reported, or both. Very few guidelines are explicitly based on valid systematic review evidence and in some cases the evidence has been misinterpreted or misused. There are also too many guidelines. Guideline developers and systematic review producers should work together to producer fewer, better and more valid guidelines.

278 7 DISCUSSION

This thesis has involved the gathering of evidence, the extraction of relevant information and data, its appraisal and synthesis, discussions around the results, and the formulation of conclusions. The volume of research into cancer treatment and the side effects of those treatments, particularly oral side effects in the case of this body of work, is huge and shows no sign of abating, and with good cause. However, although the studies included in this thesis address clinically important questions, research resources are not always used to their optimum. Although some clinically useful findings have been identified, much more research is still needed. It is not the purpose of this section to repeat previous statements. Instead, areas of concern with the current evidence base regarding mouth care for patients receiving cancer treatment will be discussed.

Issues that influence the usefulness and interpretation/uptake of the evidence Due to the increasingly large volume of research on prevention of oral mucositis in cancer patients, the original Cochrane review was becoming less viable to maintain and update, and possibly even to use in guidelines and decision making (14). In addition, a major criticism from guideline producers was that the meta-analyses “did not discriminate between different clinical settings” (22). The purpose of systematic reviews is to produce summaries of the evidence that are easily accessible to stakeholders to help inform patient care. If reviews become unwieldly, and lack the structure required by decision makers, the clinical relevance and usefulness of the review may be lost. Systematic reviewers often face dilemmas over lumping (having a broad scope) or splitting (having a narrow scope). The previous versions of the Cochrane review have opted for the lumping approach which can have advantages such as producing more generalisable results and reducing the possibility of chance results (231). It has been highlighted that the decision on lumping or splitting in a systematic review is “an important methodological consideration and impacts the usability of the review” (232). This decision will affect how the review is interpreted by end users (232). Thus, following discussions with the authors of the original Cochrane review, the decision was made to split the review into separate reviews based on interventions or categories of interventions. In order to help facilitate the uptake of the findings of the Cochrane reviews into clinical guidelines, the categories of interventions were based on those used by the leading guideline producers, the Mucositis Study Group (MSG) of the Multinational Association of Supportive Care in Cancer/International Society of Oral Oncology (MASCC/ISOO) (65). It was hoped that this splitting would also

279 facilitate more timely updating of the reviews as new evidence becomes available. Indeed, following the peer review process for the review on cytokines and growth factors, presented in Chapter 4, dialogue has been started between Cochrane Oral Health and the MSG with a view to discussing how to work together in the future to minimise duplication of effort and to ensure that the guidelines are based on the best possible evidence. The MSG guidelines involve a series of systematic reviews that inform their recommendations and suggestions. However, these systematic reviews do not include a meta-analysis of results, and often make inappropriate, naïve indirect comparisons of interventions. It is hoped that Cochrane Oral Health will work more closely with the MSG to provide the evidence during the guideline process, thus negating the need for MSG systematic reviews and reducing duplication of effort. Instead the MSG could provide clinical experts to guide the methodologists at Cochrane in producing clinically relevant systematic reviews. A recent example of such a collaborative approach has occurred between the American Dental Association (ADA) and Cochrane Oral Health. The two organisations worked co- operatively with Cochrane supplying data from their systematic reviews of diagnostic test accuracy to the ADA during their update of guidelines on screening for oral cancer (upcoming publication in JADA October 2017). The ADA have reported to Cochrane that this sharing of information has led to a reduction in the time taken to update their guidelines from 2 years to 3 months. This collaborative approach may enable the improvement of guidelines on mouth care for cancer patients, and if there is no progress made between Cochrane and the MSG, other avenues may be possible, for example with SIGN, NICE, ADA, etc. A collaboration between MacMillan and Cochrane is also currently being explored as another way to disseminate the findings of the Cochrane reviews in the form of information aimed at patients. Furthermore, a podcast is currently being prepared to help with the dissemination of the findings from the Cochrane review on salivary gland dysfunction presented in Chapter 5.

In addition to splitting the reviews according to intervention type, a further splitting within the oral mucositis prevention reviews into patient subgroups was undertaken. Again, this was undertaken to reflect feedback from guideline developers and improve the clinical relevance of the analyses. However, despite one peer reviewer requesting further multiple breakdown of the cytokine and growth factor review results, a balance between lumping and splitting must be reached. There are so many factors involved in the treatment of the cancer patients participating in the included studies that such breakdown by the minutiae could result in an endless stream of single-study analyses. Therefore it was felt that splitting by the main types of cancer treatment, for which there is biological plausibility

280 and reasoning to group patients separately, was the optimum structure for the analyses. Again, working with the guideline producers in the future, it may be necessary to produce various breakdowns of the results to suit their needs and answer any specific guideline questions.

The registration of all clinical trials, access to protocols, and reporting of all results are big issues in healthcare research. It is always possible that unfavourable results have never been published as they are less interesting, or because they do not reflect well on pharmaceutical products, leading to reporting or publication biases. This is, of course, more likely where industry has some link to the trials, and the systematic reviews in this thesis are potentially affected by these issues. However, this can be difficult to deal with in systematic reviews. Firstly, in order to produce a funnel plot to detect publication bias, the Cochrane Handbook states that at least 10 trials are needed (42). There were not enough studies in the meta-analyses carry out this test. Secondly, if nearly all the studies are linked to industry, as was the case in the keratinocyte growth factor (KGF) versus placebo comparison in Chapter 4, subgroup analyses may not be of use to show any influence on the results (as is the recommended method of dealing with industry funding suggested in the Cochrane Handbook). There was some discussion around potential over-leniency regarding the GRADE assessments for the cytokines and growth factors review (Chapter 4). This was owing to the fact that there were three studies awaiting classification. One of those studies would have been in the subgroup of chemotherapy alone for mixed cancers and, with 100 participants, could feasibly alter the effect estimates (NCT00393822). The two other studies would be in the subgroup of transplantation after conditioning for haematological cancers, adding a potential 217 participants to the meta-analysis, and again potentially altering the effect estimates (NCT02313792, (233)). There remains the question of whether or not the evidence should be downgraded for publication bias. If it should be, then would only the evidence for the two mentioned subgroups be downgraded, or should the other subgroup of patients receiving radiotherapy to the head and neck plus chemotherapy be downgraded too, despite there being no knowledge of missing studies? Of the 15 KGF versus placebo studies, 12 of these were funded by Amgen, with one not mentioning funding and two receiving government-only funding. Interestingly, eight of the 12 Amgen-funded studies, plus two of the three studies awaiting classification, were found to be on trial registries. Therefore the knowledge that Amgen, who produce this drug, have a good history of registering their trials may raise some doubt about potential publication bias.

281 The two KGF versus placebo studies that explicitly reported government-only funding, and only included children, reported their data so unclearly that meta-analysis was not possible. Evidence of interventions for preventing oral mucositis in children receiving cancer treatment is extremely scarce, so this was particularly disappointing. Neither study was on a trials registry, but it may have been helpful to have delineated how the outcomes would be assessed and reported. Also, on the trials registry ClinicalTrials.gov, there is a tab for study results, but unfortunately this remains quite underused. If it was made compulsory for trialists to share the raw data, so that systematic reviewers can take what they require for their analyses, this would be a very useful advancement. This would also minimise time lag bias, a form of publication bias, so that people awaiting the results would not have to wait for the results to be published, which may not be published anyway if they are negative/unfavourable.

Ultimately, the decisions made in the reviews in this thesis are transparent and the reader may come to their own conclusion. There is currently an international initiative called AllTrials which aims to address and fix the problem of missing trial results by encouraging prospective registration for all new trials and retrospective registration for all those that were not previously registered (www.alltrials.net).

Another issue that consistently came up when carrying out the systematic reviews was the inconsistency in the outcomes considered and the way in which they were reported. Researchers use the PICO method to help clarify and formulate answerable research questions (population/problem, intervention, comparison, outcome/s) (234). Thus outcome measures are one of the most important factors to consider when assessing the effects of interventions. Indeed it has been stated that “clinical trials are only as credible as their endpoints” (235). Research assessing outcomes that are inappropriate or unimportant to the users of such research will be less likely to be used, and therefore limited healthcare resources would have been wasted, which must be considered unethical. One major problem resulting from inconsistencies in the consideration of outcomes in clinical studies is heterogeneity, which makes it difficult to compare and summarise research, for example when conducting systematic reviews (67). In fact, the most accessed and cited Cochrane reviews all describe serious problems with the reporting of outcomes in the included studies (236). The other major problem is the potential for outcome reporting bias, whereby results for outcomes are selectively reported according to statistical significance (70). A study found that statistically significant outcomes were 2.2 to 4.7 times more likely to be fully reported than non-significant outcomes, and that 40% to 62% of studies had at

282 least one primary outcome that was changed, introduced, or omitted (69). This may give the user of such research a distorted view of the effects of an intervention, leading to them incorrectly considering its effects to be of greater benefit than the alternative choice of intervention (or of the placebo/no treatment) (68). A further study looking at 283 Cochrane reviews found that 55% of the reviews were unable to include the full data for their primary outcome and concluded that this affected the conclusions of a worryingly large number of these reviews (237).

Clarke has stated that, in order to make well-informed decisions on healthcare interventions, we require measurement and reporting of the same outcomes, measured in the same way, and he has suggested that the solution to this and the above problems is to develop disease/condition-specific core outcome sets to be used in the conduct and reporting of clinical trials (68). In an area where there is a known core outcome set, it would be expected that it would be included in any subsequent clinical trials as a minimum, whilst researchers would be free to research additional outcomes (67). Indeed the Health Technology Assessment (HTA) Programme in the UK require those submitting applications to justify any deviation from a core outcome set where one is known. This is an indicator of the growing recognition of the importance of core outcome sets and the COMET (Core Outcome Measures in Effectiveness Trials) Initiative (www.comet- initiative.org) is the leading organisation in the development of core outcome sets and appropriate associated methodology. The COMET initiative maintain a free online database containing details of both completed and ongoing work on core outcome sets, and this is intended to avoid duplication of effort (67).

In the area of oral mucositis, a previous study, published in 2002, has attempted to identify the most important outcomes to measure (71). In order to do so, a panel of 11 people representing different stakeholders was formed. The panel then discussed outcomes at a single meeting until they reached consensus. The resulting list of outcomes from the study were (in descending order of importance): oral pain; need for opioid analgesics; inability to eat soft foods; quality of life and functional status reductions; increase in hospital stay; and inability to take medication orally. The authors also concluded that an objective measurement of oral mucositis, which truly represents the severity of the condition, should also be included as an outcome. The scope of the study is focussed on bone marrow transplant patients, nurses and clinicians, and it is not clear whether important outcomes would differ for other types of cancer. Whilst the authors of the study should be commended for including most relevant stakeholders (dentist, oncologists/haematologists,

283 patients, statistician, health economist, oncology nurses, pharmaceutical industry representatives), the sample size is very small and, because it is not internationally represented, the results may have limited external validity. The authors also acknowledge that they cannot be certain that the views of the panel are representative of their respective stakeholder groups. A further problem could be that the panel met face-to-face resulting in the loss of anonymity in their answers/opinions, and the possibility that certain stakeholders may have had greater influence than others in achieving consensus on the final outcome set (i.e. it is not difficult to imagine that patients, nurses, and possibly others may have acted deferentially towards, for example, the oncologists).

The amount of time since this study was done, and the fact that there are some obvious flaws in the methodology, suggest that up-to-date, more rigorous work should be carried out. This is an area for future work that has been developed into a protocol and will be pursued if resources allow (see the protocol in Appendix 15). It is not clear how this would change the reviews included in this PhD but any new outcomes identified by such a process would have to be included in any future updates, possibly necessitating further data extraction and analyses. On the other hand, a new core outcome set with increased patient input could potentially highlight the lack of importance of one or more of the currently used outcomes. This could lead to removing some of the analyses from the current reviews, which could either simplify them and improve readability, or lead to loss of information that is potentially of interest or importance to one or more stakeholder groups. However, focusing on outcomes that have been identified as important to a range of stakeholders should overall lead to more relevant and valued research, which should in turn feed into guidelines and lead to improved clinical practice and patient outcomes.

For the two oral mucositis prevention reviews in this thesis, it was decided to use the Bellm et al core outcome set, adding the outcomes ‘adverse effects’ and ‘interruptions to cancer treatment’ which were considered important by the Cochrane review team. All studies reported oral mucositis but the reporting of the secondary outcomes was highly variable. Across the 49 studies included in the two reviews, for the important outcome of incidence of cancer treatment interruptions, only seven reported data that could be used in meta-analysis. For the outcome oral pain, arguably the thing that affects patients the most, only six of 49 studies reported data that could be pooled. Another 14 studies reported data but for various reasons (e.g. presented as either area under the curve or median, or no standard deviation, standard error or P-value) it was not analysable. Only four of the 49 studies reported data on quality of life, but none of these data were usable in the analyses

284 for the same reasons as those mentioned above. For such an important outcome, it is surprising that it has not been considered more frequently. One explanation for this may be that there are so many other factors at play in these patients affecting quality of life, that the trial investigators may feel the results will be confounded. There were similar problems for all three systematic reviews when trying to summarise adverse effects i.e. it is difficult to distinguish between effects of the interventions given to prevent oral complications of cancer treatment and any effects caused by the cancer treatment itself, any other interventions being received for the inevitable comorbidities, and the underlying cancer. It did not help that adverse effects were often reported only where there was a, for example, 10% difference in incidence between the groups. Some studies reported incidence of different grades of severity of adverse effects. All these issues made it difficult to summarise and present the information in a satisfactory manner, and it was generally decided to report the data narratively in additional tables. It would be helpful for trialists to more clearly distinguish between the adverse effects definitely due to the interventions under study from other adverse effects, and to report all data, perhaps using online appendices, to allow systematic reviewers to find a way to pool data across studies.

In contrast to the idea of reporting a longer list of outcomes, it could be argued that it would be better to focus on fewer outcomes and report them more thoroughly. For example, a criticism of the Cochrane reviews on oral mucositis prevention has been that they have only ever considered the incidence of oral mucositis, whilst factors such as duration of oral mucositis are also important (213). Furthermore, the most commonly used oral mucositis scales, the WHO and the National Cancer Institute common toxicity criteria (NCI-CTC) scales, already take some of the secondary outcomes into account, such as pain and supplemental feeding (the latter of which will indicate hospitalisation). Another issue with the secondary outcomes, namely interruptions to cancer treatment, ability to eat and drink/use of supplemental feeding, duration of hospitalisation and duration of treatment with opioid analgesics, is that it is not always clear if they are due to oral mucositis or something else. In fact, this is rarely reported. On balance, it could be concluded that trial authors should either standardise the way they report data for oral mucositis trials, or report the data in various ways to allow meta-analysis of a greater number of trials. This would increase the power, precision and external validity of the effect estimates, and would benefit cancer patients internationally.

After moving through trials to systematic reviews, the evidence pathway then progresses to guidelines. However, the current state of guidelines in the field of mouth care for cancer

285 patients has been shown to be of low quality in Chapter 6. It is not clear why this was the case, but it seemed that there was a lack of understanding by the developers of some of the guidelines examined regarding critical appraisal of the literature and the process of guideline development. Many of the guidelines carried out their own ‘systematic reviews’ which were not sufficiently rigorous and often did not appear to involve an appraisal of the methodological aspects of the cited evidence. They often did not cite important systematic reviews with meta-analyses and were therefore trying to make sense of individual, often conflicting, effect estimates. One example that was worrying was that of a guideline on the oral management of oncology patients receiving radiotherapy, chemotherapy with or without bone marrow transplantation (226). The scope was wide, covering all oral complications in patients of all ages and receiving all types of cancer treatment with the exception of targeted therapy. The guidelines are produced by The Royal College of Surgeons of England and therefore are likely to be trusted by the end users. However, all of the four assessors using the AGREE II instrument for appraising the guidelines stated that they would not recommend their use. Furthermore, the assessment of the evidence base for their recommendations revealed multiple flaws. For example, there was misleading grading of evidence and ineffective interventions were recommended. As already stated, these guidelines have been produced by a respected organisation. This may increase the uptake of the recommendations and result in clinicians providing care to their patients that they believe is evidence based but is not.

As with clinical trials and systematic reviews, clinical guidelines must be critically appraised to determine whether or not they are reliable and should be used. In order to do so, an appropriate checklist or tool is required. The AGREE II instrument appears to be the most well-known amongst healthcare researchers and indeed has gained recognition as the internationally accepted standard (56, 57). However, the use of the instrument in the study presented in Chapter 6 identified some weaknesses inherent in this tool. The main weakness is that it has the potential to result in misleading conclusions being drawn regarding the quality of some guidelines. This is because it essentially assesses the quality of reporting in the guidelines, but does not explicitly guide the assessor to make judgements about the supporting evidence for the recommendations made. This could be problematic for those who do not have a working knowledge of the literature in a particular field, and also for those who do not have a good understanding of research methodology or the appraisal of evidence. The assessors carrying out the appraisal in Chapter 6 did have these skills and knowledge and therefore were able to identify many problems with the guidelines and their evidence base. It is possible that assessors without specialist skills

286 could rate these guidelines as being more reliable than they actually are. This is not helped by the often misleading terminology in many of the guidelines, such as ‘evidence based’ and ‘systematic review’, that were used inaccurately. There are further weaknesses with the instrument relating to the domains assessed. This is due to a mixture of a tangible difference in the importance of some domains, and also some being easier or much harder to score highly on. This could again lead to misinterpretation regarding the quality of a guideline. To conclude, the most important aspect of a guideline is surely the evidence underpinning it, yet the AGREE II instrument does not currently assess this. In the study presented in Chapter 6, the appraisals using this instrument were not in agreement with those examining the supporting evidence for the recommendations.

Evidence based practice has come a long way over recent years and has helped lead to improvements in the methodology and reporting of clinical trials, systematic reviews and clinical guidelines. However, what has become apparent throughout this thesis is the need to ‘take stock’ of the current evidence base and the direction it should now take. The justification for ever increasing numbers of systematic reviews and clinical guidelines in this field must be questioned. Are they undertaken to truly improve patient care, or is it a case of academic/self-promotion? It would be beneficial for all stakeholders to perhaps work together on developing a rigorous evidence base from which clinical guidelines can be made. Researchers should be encouraged to check for existing protocols (e.g. on PROSPERO) before embarking on research projects, and to approach others found to be carrying out similar work with a view to collaboration. There needs to be recognition of the different skills required in all stages of the evidence based pathway, with clinicians, methodologists, statisticians, information specialists, and patients working together to ensure that the care provided at the end of this pathway is truly evidence based.

287

8 REFERENCES

1. Al-Dasooqi N, Sonis ST, Bowen JM, Bateman E, Blijlevens N, Gibson RJ, et al. Emerging evidence on the pathobiology of mucositis. Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer. 2013;21(11):3233-41. 2. Scully C, Sonis S, Diz PD. Oral mucositis. Oral diseases. 2006;12(3):229-41. 3. Sonis ST, Elting LS, Keefe D, Peterson DE, Schubert M, Hauer-Jensen M, et al. Perspectives on cancer therapy-induced mucosal injury: pathogenesis, measurement, epidemiology, and consequences for patients. Cancer. 2004;100(9 Suppl):1995-2025. 4. Trotti A, Bellm LA, Epstein JB, Frame D, Fuchs HJ, Gwede CK, et al. Mucositis incidence, severity and associated outcomes in patients with head and neck cancer receiving radiotherapy with or without chemotherapy: a systematic literature review. Radiotherapy and oncology : journal of the European Society for Therapeutic Radiology and Oncology. 2003;66(3):253-62. 5. Miller M, Kearney N. Oral care for patients with cancer: a review of the literature. Cancer nursing. 2001;24(4):241-54. 6. Jensen SB, Peterson DE. Oral mucosal injury caused by cancer therapies: current management and new frontiers in research. Journal of Oral Pathology & Medicine. 2014;43(2):81- 90. 7. Peterson DE, Srivastava R, Lalla RV. Oral mucosal injury in oncology patients: perspectives on maturation of a field. Oral diseases2015. p. 133-41. 8. Sonis ST. Mucositis: The impact, biology and therapeutic opportunities of oral mucositis. Oral oncology. 2009;45(12):1015-20. 9. Barasch A, Peterson DE. Risk factors for ulcerative oral mucositis in cancer patients: unanswered questions. Oral oncology. 2003;39(2):91-100. 10. Elting LS, Cooksley CD, Chambers MS, Garden AS. Risk, outcomes, and costs of radiation-induced oral mucositis among patients with head-and-neck malignancies. International journal of radiation oncology, biology, physics. 2007;68(4):1110-20. 11. Sonis ST, Peterson RL, Edwards LJ, Lucey CA, Wang L, Mason L, et al. Defining mechanisms of action of interleukin-11 on the progression of radiation-induced oral mucositis in hamsters. Oral oncology. 2000;36(4):373-81. 12. National Institute for Health and Care Excellence (NICE). Neutropenic sepsis: prevention and management of neutropenic sepsis in cancer patients (CG151). Issued: September 2012. Available from http://guidance.nice.org.uk/CG151 [accessed 29 January 2014]. 13. Napenas JJ, Brennan MT, Coleman S, Kent ML, Noll J, Frenette G, et al. Molecular methodology to assess the impact of cancer chemotherapy on the oral bacterial flora: a pilot study. Oral surgery, oral medicine, oral pathology, oral radiology, and endodontics. 2010;109(4):554-60. 14. Worthington HV, Clarkson JE, Bryan G, Furness S, Glenny AM, Littlewood A, et al. Interventions for preventing oral mucositis for patients with cancer receiving treatment. The Cochrane database of systematic reviews. 2011(4):CD000978. 15. Glenny AM, Gibson F, Auld E, Coulson S, Clarkson JE, Craig JV, et al. A survey of current practice with regard to oral care for children being treated for cancer. European journal of cancer. 2004;40(8):1217-24. 16. Glenny AM, Gibson F, Auld E, Coulson S, Clarkson JE, Craig JV, et al. The development of evidence-based guidelines on mouth care for children, teenagers and young adults treated for cancer. European journal of cancer. 2010;46(8):1399-412. 17. Bowen JM, Elad S, Hutchins RD, Lalla RV, Mucositis Study Group of the Multinational Association of Supportive Care in Cancer/International Society of Oral O. Methodology for the MASCC/ISOO Mucositis Clinical Practice Guidelines Update. Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer. 2013;21(1):303-8. 18. Elad S, Bowen J, Zadik Y, Lalla RV, Mucositis Study Group of the Multinational Association of Supportive Care in Cancer/International Society of Oral O. Development of the MASCC/ISOO Clinical Practice Guidelines for Mucositis: considerations underlying the process.

288 Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer. 2013;21(1):309-12. 19. McGuire DB, Fulton JS, Park J, Brown CG, Correa ME, Eilers J, et al. Systematic review of basic oral care for the management of oral mucositis in cancer patients. Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer. 2013;21(11):3165-77. 20. Mucositis Study Group Multinational Association of Supportive Care in Cancer/International Society of Oral Oncology (MASCC/ISOO). Evidence-based Clinical Practice Guidelines for Mucositis Secondary to Cancer Therapy, 2013. Available at: http://www.mascc.org/mucositis-guidelines (accessed on 16 May 2014). 21. Lalla RV, Sonis ST, Peterson DE. Management of oral mucositis in patients who have cancer. Dental clinics of North America. 2008;52(1):61-77, viii. 22. Raber-Durlacher JE, von Bultzingslowen I, Logan RM, Bowen J, Al-Azri AR, Everaus H, et al. Systematic review of cytokines and growth factors for the management of oral mucositis in cancer patients. Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer. 2013;21(1):343-55. 23. Nicolatou-Galitis O, Sarri T, Bowen J, Di Palma M, Kouloulias VE, Niscola P, et al. Systematic review of anti-inflammatory agents for the management of oral mucositis in cancer patients. Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer. 2013;21(11):3179-89. 24. Saunders DP, Epstein JB, Elad S, Allemano J, Bossi P, van de Wetering MD, et al. Systematic review of antimicrobials, mucosal coating agents, anesthetics, and analgesics for the management of oral mucositis in cancer patients. Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer. 2013;21(11):3191-207. 25. Migliorati C, Hewson I, Lalla RV, Antunes HS, Estilo CL, Hodgson B, et al. Systematic review of laser and other light therapy for the management of oral mucositis in cancer patients. Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer. 2013;21(1):333-41. 26. Peterson DE, Ohrn K, Bowen J, Fliedner M, Lees J, Loprinzi C, et al. Systematic review of oral cryotherapy for management of oral mucositis caused by cancer therapy. Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer. 2013;21(1):327-32. 27. Nicolatou-Galitis O, Sarri T, Bowen J, Di Palma M, Kouloulias VE, Niscola P, et al. Systematic review of amifostine for the management of oral mucositis in cancer patients. Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer. 2013;21(1):357-64. 28. Yarom N, Ariyawardana A, Hovan A, Barasch A, Jarvis V, Jensen SB, et al. Systematic review of natural agents for the management of oral mucositis in cancer patients. Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer. 2013;21(11):3209-21. 29. Jensen SB, Jarvis V, Zadik Y, Barasch A, Ariyawardana A, Hovan A, et al. Systematic review of miscellaneous agents for the management of oral mucositis in cancer patients. Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer. 2013;21(11):3223-32. 30. Jensen SB, Pedersen AM, Vissink A, Andersen E, Brown CG, Davies AN, et al. A systematic review of salivary gland hypofunction and xerostomia induced by cancer therapies: management strategies and economic impact. Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer. 2010;18(8):1061-79. 31. Sreebny LM. Saliva in health and disease: an appraisal and update. International dental journal. 2000;50(3):140-61. 32. Dawes C. Physiological factors affecting salivary flow rate, oral sugar clearance, and the sensation of dry mouth in man. Journal of dental research. 1987;66 Spec No:648-53. 33. Porter SR, Scully C, Hegarty AM. An update of the etiology and management of xerostomia. Oral surgery, oral medicine, oral pathology, oral radiology, and endodontics. 2004;97(1):28-46. 34. Pankhurst CL, Smith EC, Rogers JO, Dunne SM, Jackson SH, Proctor G. Diagnosis and management of the dry mouth: Part 1. Dental update. 1996;23(2):56-62. 35. Jensen SB, Pedersen AM, Vissink A, Andersen E, Brown CG, Davies AN, et al. A systematic review of salivary gland hypofunction and xerostomia induced by cancer therapies: 289 prevalence, severity and impact on quality of life. Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer. 2010;18(8):1039-60. 36. Shiboski CH, Hodgson TA, Ship JA, Schiodt M. Management of salivary hypofunction during and after radiotherapy. Oral surgery, oral medicine, oral pathology, oral radiology, and endodontics. 2007;103 Suppl:S66 e1-19. 37. Nieuw Amerongen AV, Veerman EC. Current therapies for xerostomia and salivary gland hypofunction associated with cancer therapies. Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer. 2003;11(4):226-31. 38. Jensen SB, Pedersen AM, Reibel J, Nauntofte B. Xerostomia and hypofunction of the salivary glands in cancer therapy. Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer. 2003;11(4):207-25. 39. Buglione M, Cavagnini R, Di Rosario F, Maddalo M, Vassalli L, Grisanti S, et al. Oral toxicity management in head and neck cancer patients treated with chemotherapy and radiation: Xerostomia and trismus (Part 2). Literature review and consensus statement. Critical reviews in oncology/hematology. 2016;102:47-54. 40. Sackett DL, Rosenberg WM, Gray JA, Haynes RB, Richardson WS. Evidence based medicine: what it is and what it isn't. Bmj. 1996;312(7023):71-2. 41. Mulrow CD. Rationale for systematic reviews. Bmj. 1994;309(6954):597-9. 42. Higgins JPT, Green S (editors). Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 [updated March 2011]. The Cochrane Collaboration, 2011. Available from www.cochrane-handbook.org. 43. Akobeng AK. Understanding randomised controlled trials. Archives of disease in childhood. 2005;90(8):840-4. 44. Roberts C, Torgerson D. Randomisation methods in controlled trials. Bmj. 1998;317(7168):1301. 45. Atkins D, Best D, Briss PA, Eccles M, Falck-Ytter Y, Flottorp S, et al. Grading quality of evidence and strength of recommendations. Bmj. 2004;328(7454):1490. 46. Field MJ, Lohr KN, editors. Institute of Medicine Committee to advise the public health service on clinical practice guidelines. Clinical practice guidelines: directions for a new program. Washington, DC: National Academy Press; 1990. 47. IOM (Institute of Medicine). 2011. Clinical Practice Guidelines We Can Trust. Washington, DC: The National Academies Press. 48. Scottish Intercollegiate Guidelines Network. How are guidelines developed? Available from http://www.sign.ac.uk/how-are-guidelines-developed.html [accessed 14 August 2017]. 49. Woolf SH, Grol R, Hutchinson A, Eccles M, Grimshaw J. Clinical guidelines: potential benefits, limitations, and harms of clinical guidelines. Bmj. 1999;318(7182):527-30. 50. Clarkson JE, Worthington HV, Eden OB. Interventions for preventing oral candidiasis for patients with cancer receiving treatment. The Cochrane database of systematic reviews. 2007(1):CD003807. 51. Craig JV, Gibson F, Glenny AM, Children's C, Leukaemia G, Paediatric Oncology Nurses Forum Mouth Care G. Audit to monitor the uptake of national mouth care guidelines for children and young people being treated for cancer. Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer. 2011;19(9):1335-41. 52. Glenny AM, Oliver R, Roberts GJ, Hooper L, Worthington HV. Antibiotics for the prophylaxis of bacterial endocarditis in dentistry. The Cochrane database of systematic reviews. 2013(10):CD003813. 53. National Institute for Health and Care Excellence (NICE). Prophylaxis against infective endocarditis: antimicrobial prophylaxis against infective endocarditis in adults and children undergoing interventional procedures (CG64). Available from https://www.nice.org.uk/guidance/cg64/evidence [accessed 15 August 2017] 54. Thornhill MH, Dayer MJ, Forde JM, Corey GR, Chu VH, Couper DJ, et al. Impact of the NICE guideline recommending cessation of antibiotic prophylaxis for prevention of infective endocarditis: before and after study. Bmj. 2011;342:d2392. 55. Shaneyfelt TM, Centor RM. Reassessment of clinical practice guidelines: go gently into that good night. Jama. 2009;301(8):868-9. 56. Brouwers MC, Kerkvliet K, Spithoff K, Consortium ANS. The AGREE Reporting Checklist: a tool to improve reporting of clinical practice guidelines. Bmj. 2016;352:i1152.

290 57. Scottish Intercollegiate Guidelines Network (SIGN). SIGN 50: a guideline developer’s handbook. Edinburgh: SIGN; 2015. (SIGN publication no. 50). [November 2015]. Available from URL: http://www.sign.ac.uk. 58. Boers-Doets CB, Raber-Durlacher JE, Treister NS, Epstein JB, Arends AB, Wiersma DR, et al. Mammalian target of rapamycin inhibitor-associated stomatitis. Future oncology. 2013;9(12):1883-92. 59. Elting LS, Keefe DM, Sonis ST, Garden AS, Spijkervet FK, Barasch A, et al. Patient- reported measurements of oral mucositis in head and neck cancer patients treated with radiotherapy with or without chemotherapy: demonstration of increased frequency, severity, resistance to palliation, and impact on quality of life. Cancer. 2008;113(10):2704-13. 60. Epstein JB, Emerton S, Kolbinson DA, Le ND, Phillips N, Stevenson-Moore P, et al. Quality of life and oral function following radiotherapy for head and neck cancer. Head & neck. 1999;21(1):1-11. 61. Nonzee NJ, Dandade NA, Patel U, Markossian T, Agulnik M, Argiris A, et al. Evaluating the supportive care costs of severe radiochemotherapy-induced mucositis and pharyngitis : results from a Northwestern University Costs of Cancer Program pilot study with head and neck and nonsmall cell lung cancer patients who received care at a county hospital, a Veterans Administration hospital, or a comprehensive cancer care center. Cancer. 2008;113(6):1446-52. 62. Sonis ST, Oster G, Fuchs H, Bellm L, Bradford WZ, Edelsberg J, et al. Oral mucositis and the clinical and economic outcomes of hematopoietic stem-cell transplantation. Journal of Clinical Oncology. 2001;19(8):2201-5. 63. Lilleby K, Garcia P, Gooley T, McDonnnell P, Taber R, Holmberg L, et al. A prospective, randomized study of cryotherapy during administration of high-dose melphalan to decrease the severity and duration of oral mucositis in patients with multiple myeloma undergoing autologous peripheral blood stem cell transplantation. Bone Marrow Transplantation. 2006;37(11):1031-5. 64. Karagozoglu S, Filiz Ulusoy M. Chemotherapy: the effect of oral cryotherapy on the development of mucositis. Journal of clinical nursing. 2005;14(6):754-65. 65. Lalla RV, Bowen J, Barasch A, Elting L, Epstein J, Keefe DM, et al. MASCC/ISOO clinical practice guidelines for the management of mucositis secondary to cancer therapy. Cancer. 2014;120(10):1453-61. 66. Worthington H, Clarkson J, Weldon J. Priority oral health research identification for clinical decision-making. Evidence-based Dentistry. 2015;16(3):69-71. 67. Williamson PR, Altman DG, Blazeby JM, Clarke M, Devane D, Gargon E, et al. Developing core outcome sets for clinical trials: issues to consider. Trials. 2012;13:132. 68. Clarke M. Standardising outcomes for clinical trials and systematic reviews. Trials. 2007;8:39. 69. Dwan K, Altman DG, Arnaiz JA, Bloom J, Chan AW, Cronin E, et al. Systematic review of the empirical evidence of study publication bias and outcome reporting bias. PloS one. 2008;3(8):e3081. 70. Williamson PR, Gamble C, Altman DG, Hutton JL. Outcome selection bias in meta- analysis. Statistical methods in medical research. 2005;14(5):515-24. 71. Bellm LA, Cunningham G, Durnell L, Eilers J, Epstein JB, Fleming T, et al. Defining clinically meaningful outcomes in the evaluation of new treatments for oral mucositis: oral mucositis patient provider advisory board. Cancer investigation. 2002;20(5-6):793-800. 72. Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. Bmj. 1997;315(7109):629-34. 73. Askarifar M, Lakdizaji S, Ramzi M, Rahmani A, Jabbarzadeh F. The effect of oral cryotherapy on chemotherapy-induced oral mucositis in patients undergoing autologous transplantation of blood stem cells: a clinical trial. Iranian Red Crescent Medical Journal2015. 74. Cascinu S, Fedeli A, Fedeli SL, Catalano G. Oral cooling (cryotherapy), an effective treatment for the prevention of 5-fluorouracil-induced stomatitis. European Journal of Cancer Part B, Oral Oncology. 1994;30B(4):234-6. 75. Gori E, Arpinati M, Bonifazi F, Errico A, Mega A, Alberani F, et al. Cryotherapy in the prevention of oral mucositis in patients receiving low-dose methotrexate following myeloablative allogeneic stem cell transplantation: a prospective randomized study of the Gruppo Italiano Trapianto di Midollo Osseo nurses group. Bone Marrow Transplantation2007. p. 347-52. 76. Heydari A, Sharifi H, Salek R. Effect of oral cryotherapy on combination chemotherapy- induced oral mucositis: a randomized clinical trial. Middle East Journal of Cancer. 2012;3(2/3):55- 64. 291 77. Kakoei S, Ghassemi A, Nakhaei NR. Effect of cryotherapy on oral mucositis in patients with head and neck cancers receiving radiotherapy. International Journal of Radiation Research. 2013;11(2):117-20. 78. Katranci N, Ovayolu N, Ovayolu O, Sevinc A. Evaluation of the effect of cryotherapy in preventing oral mucositis associated with chemotherapy - a randomized controlled trial. European Journal of Oncology Nursing. 2012;16(4):339-44. 79. Mahood DJ, Dose AM, Loprinzi CL, Veeder MH, Athmann LM, Thereau TM, et al. Inhibition of fluorouracil-induced stomatitis by oral cryotherapy. Journal of Clinical Oncology. 1991;9(3):449-52. 80. Rocke LK, Loprinzi CL, Lee JK, Kunselman SJ, Iverson RK, Finck G, et al. A randomized clinical trial of two different durations or oral cryotherapy for prevention of 5-fluorouracil-related stomatitis. Cancer. 1993;72(7):2234-8. 81. Salvador P, Azusano C, Wang L, Howell D. A pilot randomized controlled trial of an oral care intervention to reduce mucositis severity in stem cell transplant patients. Journal of Pain and Symptom Management. 2012;44(1):64-73. 82. Sorensen JB, Skovsgaard T, Bork E, Damstrup L, Ingeberg S. Double-blind, placebo- controlled, randomized study of chlorhexidine prophylaxis for 5-fluorouracil-based chemotherapy- induced oral mucositis with nonblinded randomized comparison to oral cooling (cryotherapy) in gastrointestinal malignancies. Cancer2008. p. 1600-6. 83. Svanberg A, Birgegard G, Ohrn K. Oral cryotherapy reduces mucositis and opioid use after myeloablative therapy--a randomized controlled trial. Supportive Care in Cancer2007. p. 1155-61. 84. Toro J, Schneider D, Alonzo R, Hasan A, Lee S, Gushiken F, et al. A randomized trial of oral cryotherapy, saline solution and Caphosol for the prevention of high-dose melphalan-induced oral mucositis followed by autologous hematopoietic stem cell transplantation. Supportive Care in Cancer. 2013;21(1 Suppl):S138-S. 85. Zhang W, Li LS, Lu YW, Zhen JC, Zhang XL. Intervention research on preventing oral mucositis after using high dose methotrexate chemotherapy in osteosarcoma by gargling with calcium folinic. Chinese Pharmaceutical Journal. 2011;46(14):1126-8. 86. Papadeas E, Naxakis S, Riga M, Kalofonos Ch. Prevention of 5-fluorouracil-related stomatitis by oral cryotherapy: a randomized controlled study. European Journal Oncology Nursing. 2007;11(1):60-5. 87. Sato A, Saisho-Hattori T, Koizumi Y, Minegishi M, Iinuma K, Imaizumi M. Prophylaxis of mucosal toxicity by oral propantheline and cryotherapy in children with malignancies undergoing myeloablative chemo-radiotherapy. The Tohoku Journal of Experimental Medicine. 2006;210(4):315-20. 88. Aisa Y, Mori T, Kudo M, Yashima T, Kondo S, Yokoyama A, et al. Oral cryotherapy for the prevention of high-dose melphalan-induced stomatitis in allogeneic hematopoietic stem cell transplant recipients. Supportive Care in Cancer. 2005;13(4):266-9. 89. Mori T, Yamazaki R, Aisa Y, Nakazato T, Kudo M, Yashima T, et al. Brief oral cryotherapy for the prevention of high-dose melphalan-induced stomatitis in allogeneic hematopoietic stem cell transplant recipients. Supportive Care in Cancer. 2006;14(4):392-5. 90. Ohyama W, Kano Y, Akutsu M, Tsunoda S. Ice ball cryotherapy for chemotherapy- induced mucositis. Gan To Kagaku Ryoho. 1994;21(15):2675-7. 91. Sato A, Kumagai S, Sakaki K, Morikawa H, Song ST, Mori S. Inhibition of 5-fluorouracil- cisplatin-induced stomatitis by oral cryotherapy: use of an ice-bar containing fibrinolysin and deoxyribonuclease comiben (Elase). Gan To Kagaku Ryoho. 1997;24(9):1135-9. 92. Baydar M, Dikilitas M, Sevinc A, Aydogdu I. Prevention of oral mucositis due to 5- fluorouracil treatment with oral cryotherapy. Journal of the National Medical Association. 2005;97(8):1161-4. 93. Castelino F, Devi Elsa S, Jyothi RK. Effectiveness of plain ice cubes versus flavoured ice cubes in preventing oral mucositis associated with injection 5-fluorouracil among cancer patients. International Journal of Nursing Education Scholarship. 2011;3(2):38-40. 94. Nikoletti S, Hyde S, Shaw T, Myers H, Kristjanson LJ. Comparison of plain ice and flavoured ice for preventing oral mucositis associated with the use of 5 fluorouracil. Journal of clinical nursing. 2005;14(6):750-3. 95. de Paula Eduardo F, Bezinelli LM, da Graça Lopes RM, Nascimento Sobrinho JJ, Hamerschlak N, Correa L. Efficacy of cryotherapy associated with laser therapy for decreasing severity of melphalan-induced oral mucositis during hematological stem-cell transplantation: a prospective clinical study. Hematological Oncology2014. 292 96. Juergens C, Weston C, Lewis I, Whelan J, Paulussen M, Oberlin O, et al. Safety assessment of intensive induction with vincristine, ifosfamide, doxorubicin, and etoposide (VIDE) in the treatment of Ewing tumors in the EURO-E.W.I.N.G. 99 clinical trial. Pediatric Blood & Cancer. 2006;47(1):22-9. 97. Wang L, Gu Z, Zhai R, Zhao S, Luo L, Li D, et al. Efficacy of oral cryotherapy on oral mucositis prevention in patients with hematological malignancies undergoing hematopoietic stem cell transplantation: a meta-analysis of randomized controlled trials. PloS one. 2015;10(5):e0128763-e. 98. Lu Y, Zhu X, Depei W. Oral cryotherapy for the prevention of mucositis following myeloablative conditioning and hematopoietic stem cell transplantation. Blood. 2013;122(21):5466-. 99. von Bültzingslöwen I, Brennan MT, Spijkervet FK, Logan R, Stringer A, Raber-Durlacher JE, et al. Growth factors and cytokines in the prevention and treatment of oral and gastrointestinal mucositis. Supportive Care in Cancer. 2006;14(6):519-27. 100. Lyman GH, Dale DC, Wolff DA, Culakova E, Poniewierski MS, Kuderer NM, et al. Acute myeloid leukemia or myelodysplastic syndrome in randomized controlled clinical trials of cancer chemotherapy with granulocyte colony-stimulating factor: a systematic review. Journal of Clinical Oncology. 2010;28(17):2914-24. 101. Lefebvre C, Manheimer E, Glanville J. Chapter 6: Searching for studies. In: Higgins JP, Green S, editor(s). Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 (updated March 2011). The Cochrane Collaboration, 2011. Available from handbookcochraneorg. 102. Antoun S, Boige V, Ducreux M, Paintain M, van't Hof M, Enslen M, et al. Protective effect of an enteral formula containing TGF-beta(2) in the prevention of chemotherapy-induced diarrhoea: A pilot study. e-SPEN, the European e-Journal of Clinical Nutrition and Metabolism. 2009;4(6):e348-e50. 103. Blazar BR, Weisdorf DJ, Defor T, Goldman A, Braun T, Silver S, et al. Phase 1/2 randomized, placebo-control trial of palifermin to prevent graft-versus-host disease (GVHD) after allogeneic hematopoietic stem cell transplantation (HSCT). Blood. 2006;108(9):3216-22. 104. Blijlevens N, de Château M, Krivan G, Rabitsch W, Szomor A, Pytlik R, et al. In a high- dose melphalan setting, palifermin compared with placebo had no effect on oral mucositis or related patient's burden. Bone Marrow Transplantation. 2013;48(7):966-71. 105. Bradstock KF, Link E, Collins M, Di Iulio J, Lewis ID, Schwarer A, et al. A randomized trial of prophylactic palifermin on gastrointestinal toxicity after intensive induction therapy for acute myeloid leukaemia. British Journal of Haematology. 2014;167(5):618-25. 106. Brizel DM, Murphy BA, Rosenthal DI, Pandya KJ, Glück S, Brizel HE, et al. Phase II study of palifermin and concurrent chemoradiation in head and neck squamous cell carcinoma. Journal of Clinical Oncology. 2008;26(15):2489-96. 107. Cartee L, Petros WP, Rosner GL, Gilbert C, Moore S, Affronti ML, et al. Evaluation of GM-CSF mouthwash for prevention of chemotherapy-induced mucositis: a randomized, double- blind, dose-ranging study. Cytokine. 1995;7(5):471-7. 108. Cesaro S, Nesi F, Tridello G, Abate M, Panizzolo IS, Balter R, et al. A randomized, non- inferiority study comparing efficacy and safety of a single dose of pegfilgrastim versus daily filgrastim in pediatric patients after autologous peripheral blood stem cell transplant. PloS one. 2013;8(1):e53252-e. 109. Chi KH, Chen CH, Chan WK, Chow KC, Chen SY, Yen SH, et al. Effect of granulocyte- macrophage colony-stimulating factor on oral mucositis in head and neck cancer patients after cisplatin, fluorouracil and leucovorin chemotherapy. Journal of Clinical Oncology. 1995;13(10):2620-8. 110. Crawford J, Tomita DK, Mazanet R, Glaspy J, Ozer H. Reduction of oral mucositis by filgrastim (r-metHuG-CSF) in patients receiving chemotherapy. Cytokines, Cellular and Molecular Therapy. 1999;5(4):187-93. 111. Dazzi C, Cariello A, Giovanis P, Monti M, Vertogen B, Leoni M, et al. Prophylaxis with GM-CSF mouthwashes does not reduce frequency and duration of severe oral mucositis in patients with solid tumors undergoing high-dose chemotherapy with autologous peripheral blood stem cell transplantation rescue: a double blind, randomized, placebo-controlled study. Annals of Oncology. 2003;14(4):559-63. 112. Fink G, Ihorst G, Burbeck M, Kleber M, Kohlweyer U, Engelhardt M, editors. Prospective randomized trial of Palifermin (Keratinocyte growth factor) versus best supportive care measures (BSC) for the decrease of oral mucositis (OM) in lymphoma patients receiving high-dose BEAM- 293 (BCNU, Etoposide, Ara-C, Melphalan) conditioning and autologous stem-cell transplantation (ASCT). Onkologie; 2011: S. Karger AG. 113. Freytes CO, Ratanatharathorn V, Taylor C, Abboud C, Chesser N, Restrepo A, et al. Phase I/II randomized trial evaluating the safety and clinical effects of repifermin administered to reduce mucositis in patients undergoing autologous hematopoietic stem cell transplantation. Clinical Cancer Research. 2004;10(24):8318-24. 114. Gholizadeh N, Mehdipoor M, Sajadi H, Moosavi MS. Palifermin and chlorhexidine mouthwashes in prevention of chemotherapy-induced mucositis in children with acute lymphocytic leukemia: a randomized controlled trial. Journal of Dentistry (Shiraz, Iran). 2016;17(4):343-7. 115. Henke M, Alfonsi M, Foa P, Giralt J, Bardet E, Cerezo L, et al. Palifermin decreases severe oral mucositis of patients undergoing postoperative radiochemotherapy for head and neck cancer: a randomized, placebo-controlled trial. Journal of Clinical Oncology. 2011;29(20):2815-20. 116. Hosseinjani H, Hadjibabaie M, Gholami K, Javadi M, Radfar M, Jahangard-Rafsanjani Z, et al. The efficacy of erythropoietin mouthwash in prevention of oral mucositis in patients undergoing autologous hematopoietic SCT: a double-blind, randomized, placebo-controlled trial. Hematological Oncology. 2017;35(1):106-12. 117. Jagasia MH, Abonour R, Long GD, Bolwell BJ, Laport GG, Shore TB, et al. Palifermin for the reduction of acute GVHD: a randomized, double-blind, placebo-controlled trial. Bone Marrow Transplantation. 2012;47(10):1350-5. 118. Katano M, Nakamura M, Matsuo T, Iyama A, Hisatsugu T. Effect of granulocyte colony- stimulating factor (G-CSF) on chemotherapy-induced oral mucositis. Surgery Today. 1995;25(3):202-6. 119. Kim JW, Kim MG, Lee HJ, Koh Y, Kwon JH, Kim I, et al. Topical recombinant human epidermal growth factor for oral mucositis induced by intensive chemotherapy with hematopoietic stem cell transplantation: final analysis of a randomized, double-blind, placebo-controlled, phase 2 trial. PloS one. 2017;12(1):e0168854-e. 120. Le QT, Kim HE, Schneider CJ, Muraközy G, Skladowski K, Reinisch S, et al. Palifermin reduces severe mucositis in definitive chemoradiotherapy of locally advanced head and neck cancer: a randomized, placebo-controlled study. Journal of Clinical Oncology. 2011;29(20):2808- 14. 121. Linch DC, Scarffe H, Proctor S, Chopra R, Taylor PR, Morgenstern G, et al. Randomised vehicle-controlled dose-finding study of glycosylated recombinant human granulocyte colony- stimulating factor after bone marrow transplantation. Bone Marrow Transplantation. 1993;11(4):307-11. 122. Lucchese A, Matarese G, Ghislanzoni LH, Gastaldi G, Manuelli M, Gherlone E. Efficacy and effects of palifermin for the treatment of oral mucositis in patients affected by acute lymphoblastic leukemia. Leukemia & Lymphoma. 2016;57(4):820-7. 123. Lucchese A, Matarese G, Manuelli M, Ciuffreda C, Bassani L, Isola G, et al. Reliability and efficacy of palifermin in prevention and management of oral mucositis in patients with acute lymphoblastic leukemia: a randomized, double-blind controlled clinical trial. Minerva Stomatologica. 2016;65(1):43-53. 124. Makkonen TA, Minn H, Jekunen A, Vilja P, Tuominen J, Joeensuu H. Granulocyte Macrophage-colony stimulating factor (GM-CSF) and sucralfate in prevention of radiation-induced mucositis: a prospective randomized study. International journal of radiation oncology, biology, physics. 2000;46(3):525-34. 125. McAleese JJ, Bishop KM, A'Hern R, Henk JM. Randomized phase II study of GM-CSF to reduce mucositis caused by accelerated radiotherapy of laryngeal cancer. The British Journal of Radiology. 2006;79(943):608-13. 126. Meropol NJ, Somer RA, Gutheil J, Pelley RJ, Modiano MR, Rowinsky Ek et al. Randomized phase I trial of recombinant human keratinocyte growth factor plus chemotherapy: potential role as mucosal protectant. Journal of Clinical Oncology. 2003;21(8):1452-8. 127. Nemunaitis J, Rosenfeld CS, Ash R, Freedman MH, Deeg HJ, Appelbaum F, et al. Phase III randomized, double-blind placebo-controlled trial of rhGM-CSF following allogeneic bone marrow transplantation. Bone Marrow Transplantation. 1995;15(6):949-54. 128. Peterson DE, Barker NP, Akhmadullina LI, Rodionova I, Sherman NZ, Davidenko IS, et al. Phase II, randomized, double-blind, placebo-controlled study of recombinant human intestinal trefoil factor oral spray for prevention of oral mucositis in patients with colorectal cancer who are receiving fluorouracil-based chemotherapy. Journal of Clinical Oncology. 2009;27(26):4333-8.

294 129. Rosen LS, Abdi E, Davis ID, Gutheil J, Schnell FM, Zalcberg J, et al. Palifermin reduces the incidence of oral mucositis in patients with metastatic colorectal cancer treated with fluorouracil-based chemotherapy. Journal of Clinical Oncology. 2006;24(33):5194-200. 130. Saarilahti K, Kajanti M, Joensuu T, Kouri M, Joensuu H. Comparison of granulocyte- macrophage colony-stimulating factor and sucralfate mouthwashes in the prevention of radiation- induced mucositis: a double-blind prospective randomized phase III study. International journal of radiation oncology, biology, physics. 2002;54(2):479-85. 131. Schneider SB, Nishimura RD, Zimmerman RP, Tran L, Shiplacoff J, Tormey M, et al. Filgrastim (r-metHuG-CSF) and its potential use in the reduction of radiation-induced oropharyngeal mucositis: an interim look at a randomized, double blind, placebo-controlled trial. Cytokines, Cellular and Molecular Therapy. 1999;5(3):175-80. 132. Spielberger R, Stiff P, Bensinger W, Gentile T, Weisdorf D, Kewalramani T, et al. Palifermin for oral mucositis after intensive therapy for hematologic cancers. The New England Journal of Medicine. 2004;351(25):2590-8. 133. Su YB, Vickers AJ, Zelefsky MJ, Kraus DH, Shaha AR, Shah JP, et al. Double-blind, placebo-controlled, randomized trial of granulocyte-colony stimulating factor during postoperative radiotherapy for squamous head and neck cancer. Cancer Journal. 2006;12(3):182-8. 134. Vadhan-Raj S, Trent J, Patel S, Zhou X, Johnson MM, Araujo D, et al. Single-dose palifermin prevents severe oral mucositis during multicycle chemotherapy in patients with cancer: a randomised trial. Annals of Internal Medicine. 2010;153(6):358-67. 135. van der Lelie H, Thomas BL, van Oers RH, Ek-Post M, Sjamsoedin SA, van Dijk- Overtoom ML, et al. Effect of locally applied GM-CSF on oral mucositis after stem cell transplantation: a prospective placebo-controlled double-blind study. Annals of Hematology. 2001;80(3):150-4. 136. Wu HG, Song SY, Kim YS, Oh YT, Lee CG, Keum KC, et al. Therapeutic effect of recombinant human epidermal growth factor (RhEGF) on mucositis in patients undergoing radiotherapy, with or without chemotherapy, for head and neck cancer: a double-blind placebo- controlled prospective phase 2 multi-institutional clinical trial. Cancer. 2009;115(16):3699-708. 137. Foncuberta MC, Cagnoni PJ, Brandts CH, Mandanas R, Fields K, Derigs HG, et al. Topical transforming growth factor-beta3 in the prevention or alleviation of chemotherapy-induced oral mucositis in patients with lymphomas or solid tumors. Journal of Immunotherapy. 2001;24(4):384-8. 138. Gordon B, Spadinger A, Hodges E, Coccia P, editors. Effect of granulocyte macrophage colony stimulating factor (GMCSF) on oral mucositis after autologous bone marrow transplantation. Proceedings of the 29th Annual Meeting of the American Society of Clinical Oncology; 1993, May 16-18; Orlando, Florida, USA; 1993. 139. Horsley P, Bauer JD, Mazkowiack R, Gardner R, Bashford J. Palifermin improves severe mucositis, swallowing problems, nutrition impact symptoms, and length of stay in patients undergoing hematopoietic stem cell transplantation. Supportive Care in Cancer. 2007;15(1):105-9. 140. Hunter A, Mahendra P, Wilson K, Fields P, Cook G, Peniket A, et al. Treatment of oral mucositis after peripheral blood SCT with ATL-104 mouthwash: results from a randomized, double-blind, placebo-controlled trial. Bone Marrow Transplantation. 2009;43(7):563-9. 141. Iwase M, Yoshiya M, Kakuta S, Nagumo M. Clinical trial of recombinant granulocyte colony-stimulating factor for chemotherapy-induced neutropenia patients with oral cancer. Journal of Oral and Maxillofacial Surgery. 1997;55(8):836-40. 142. Limaye SA, Haddad RI, Cilli F, Sonis ST, Colevas AD, Brennan MT, et al. Phase 1b, multicenter, single blinded, placebo-controlled, sequential dose escalation study to assess the safety and tolerability of topically applied AG013 in subjects with locally advanced head and neck cancer receiving induction chemotherapy. Cancer. 2013;119(24):4268-76. 143. Throuvalas N, Antonadou D, Pulizzi M, Sarris G. Evaluation of the efficacy and safety of GM-CSF in the prophylaxis of mucositis in patients with head and neck cancer treated with RT. European journal of cancer. 1995;31 Suppl 5:S93 (Abstract 431)-S93 (Abstract ). 144. Vitale KM, Violago L, Cofnas P, Bishop J, Jin Z, Bhatia M, et al. Impact of palifermin on incidence of oral mucositis and healthcare utilization in children undergoing autologous hematopoietic stem cell transplantation for malignant diseases. Pediatric Transplantation. 2014;18(2):211-6. 145. Kubo K, Miyazaki Y, Murayama T, Shimazaki R, Usui N, Urabe A, et al. A randomized, double-blind trial of pegfilgrastim versus filgrastim for the management of neutropenia during

295 CHASE(R) chemotherapy for malignant lymphoma. British Journal of Haematology. 2016;174(4):563-70. 146. Lee KH, Kim JY, Lee MH, Han HS, Lim JH, Park KU, et al. A randomized, multicenter, phase II/III study to determine the optimal dose and to evaluate the efficacy and safety of pegteograstim (GCPGC) on chemotherapy-induced neutropenia compared to pegfilgrastim in breast cancer patients: KCSG PC10-09. Supportive Care in Cancer. 2016;24(4):1709-17. 147. Nabholtz JM, Cantin J, Chang J, Guevin R, Patel R, Tkaczuk K, et al. Phase III trial comparing granulocyte colony-stimulating factor to leridistim in the prevention of neutropenic complications in breast cancer patients treated with docetaxel/doxorubicin/cyclophosphamide: results of the BCIRG 004 trial. Clinical Breast Cancer. 2002;3(4):268-75. 148. Tsurusawa M, Watanabe T, Gosho M, Mori T, Mitsui T, Sunami S, et al. Randomized study of granulocyte colony stimulating factor for childhood B-cell non-Hodgkin lymphoma: a report from the Japanese pediatric leukemia/lymphoma study group B-NHL03 study. Leukemia & Lymphoma. 2016;57(7):1657-64. 149. Jones SE, Schottstaedt MW, Duncan LA, Kirby RL, Good RH, Mennel RG, et al. Randomized double-blind prospective trial to evaluate the effects of sargramostim versus placebo in a moderate-dose fluorouracil, doxorubicin, and cyclophosphamide adjuvant chemotherapy program for stage II and III breast cancer. Journal of Clinical Oncology. 1996;14(11):2976-83. 150. Legros M, Fleury J, Bay JO, Choufi B, Basile M, Condat P, et al. rhGM-CSF vs placebo following rhGM-CSF-mobilized PBPC transplantation: a phase III double-blind randomized trial. Bone Marrow Transplantation. 1997;19(3):209-13. 151. Pettengell R, Gurney H, Radford JA, Deakin DP, James R, Wilkinson PM, et al. Granulocyte colony-stimulating factor to prevent dose-limiting neutropenia in non-Hodgkin's lymphoma: a randomized controlled trial. Blood. 1992;80(6):1430-6. 152. Antin JH, Lee SJ, Neuberg D, Alyea E, Soiffer RJ, Sonis S, et al. A phase I/II double- blind, placebo-controlled study of recombinant human interleukin-11 for mucositis and acute GVHD prevention in allogeneic stem cell transplantation. Bone Marrow Transplantation. 2002;29(5):373-7. 153. Rosenthal DI, Radiation Therapy Oncology Group, National Cancer Institute. Palifermin in lessening oral mucositis in patients undergoing radiation therapy and chemotherapy for locally advanced head and neck cancer. ClinicalTrialsgov2006. p. NCT00360971-NCT. 154. NCT00626639. A phase 1/2 study to evaluate safety, pharmacokinetics, pharmacodynamics and preliminary efficacy of weekly doses of palifermin (rHuKGF) for the reduction of oral mucositis in subjects with locally advanced head and neck cancer (HNC) receiving postoperative radiotherapy with concurrent chemotherapy (available from clinicaltrials.gov). 155. Gebbia V, Valenza R, Testa A, Cannata G, Borsellino N, Gebbia N. A prospective randomized trial of thymopentin versus granulocyte-colony stimulating factor with or without thymopentin in the prevention of febrile episodes in cancer patients undergoing highly cytotoxic chemotherapy. Anticancer Research. 1994;14(2B):731-4. 156. Gladkov O, Buchner A, Bias P, Mueller U, Elsaesser R, editors. Chemotherapy-associated treatment burden in breast cancer patients receiving lipegfilgrastim or pegfilgrastim: Secondary efficacy data from a phase III study. Haematologica; 2013: Ferrata Storti Foundation. 157. Ryu JK, Swann S, LeVeque F, Scarantino CW, Johnson D, Chen A, et al. The impact of concurrent granulocyte macrophage-colony stimulating factor on radiation-induced mucositis in head and neck cancer patients: a double-blind placebo-controlled prospective phase III study by Radiation Therapy Oncology Group 9901. International journal of radiation oncology, biology, physics. 2007;67(3):643-50. 158. de Koning BA, Philipsen-Geerling B, Hoijer M, Hahlen K, Buller HA, Pieters R. Protection against chemotherapy induced mucositis by TGF-beta(2) in childhood cancer patients: results from a randomized cross-over study. Paediatric Blood & Cancer. 2007;48(5):532-9. 159. Karthaus M, Rosenthal C, Huebner G, Paul H, Elser C, Hertenstein B, et al. Effect of topical oral G-CSF on oral mucositis: a randomised placebo-controlled trial. Bone Marrow Transplantation. 1998;22(8):781-5. 160. Ifrah N, Witz F, Jouet JP, Francois S, Lamy T, Linassier C, et al. Intensive short term therapy with granulocyte-macrophage-colony stimulating factor support, similar to therapy for acute myeloblastic leukemia, does not improve overall results for adults with acute lymphoblastic leukemia. American Cancer Society. 1999;86(8):1496-505.

296 161. Sreebny LM. Xerostomia: diagnosis, management and clinical complications. In: Edgar Wm OMDM, editor. Saliva and Oral Health. London: British Dental Association; 1996. p. 43-66. 162. Navazesh M, Christensen C, Brightman V. Clinical criteria for the diagnosis of salivary gland hypofunction. Journal of dental research. 1992;71(7):1363-9. 163. Guchelaar HJ, Vermes A, Meerwaldt JH. Radiation-induced xerostomia: pathophysiology, clinical course and supportive treatment. Supportive Care in Cancer. 1997;5(4):281-8. 164. Wiseman LR, Faulds D. Oral pilocarpine: a review of its pharmacological properties and clinical potential in xerostomia. Drugs. 1995;49(1):143-55. 165. Epstein JB, Burchell JL, Emerton S, Le ND, Silverman S Jr. A clinical trial of bethanechol in patients with xerostomia after radiation therapy. A pilot study. Oral Surgery, Oral Medicine, and Oral Pathology. 1994;77(6):610-4. 166. Ahlner BH, Hagelqvist E, Lind MG. Influence on rabbit submandibular gland injury by stimulation or inhibition of gland function during irradiation. Histology and morphometry after 15 gray. Annals of Otology, Rhinology, and Laryngology. 1994;103(2):125-34. 167. Rode M, Smid L, Budihna M, Gassperssic D, Rode M, Soba E. The influence of pilocarpine and biperiden on pH value and calcium, phosphate, and bicarbonate concentrations in saliva during and after radiotherapy for head and neck cancer. Oral surgery, oral medicine, oral pathology, oral radiology, and endodontics. 2001;92(5):509-14. 168. Rode M, Smid L, Budihna M, Soba E, Rode M, Gaspersic D. The effect of pilocarpine and biperiden on salivary secretion during and after radiotherapy in head and neck cancer patients. International journal of radiation oncology, biology, physics. 1999;45(2):373-8. 169. Takahashi I, Nagai T, Miyaishi K, Maehara Y, Niibe H. Clinical study of the radioprotective effects of amifostine (YM-08310, WR-2721) on chronic radiation injury. International journal of radiation oncology, biology, physics. 1986;12(6):935-8. 170. Bohuslavizki KH, Klutmann S, Brenner W, Mester J, Henze E, Clausen M. Salivary gland protection by amifostine in high-dose radioiodine treatment: results of a double-blind placebo- controlled study. Journal of Clinical Oncology. 1998;16(11):3542-9. 171. Brizel DM, Wasserman TH, Henke M, Strnad V, Rudat V, Monnier A, et al. Phase III randomized trial of amifostine as a radioprotector in head and neck cancer. Journal of Clinical Oncology. 2000;18(19):3339-45. 172. Davies AN, Thompson J. Parasympathomimetic drugs for the treatment of salivary gland dysfunction due to radiotherapy. The Cochrane database of systematic reviews. 2015(10):CD003782. 173. Han F, Zhang JD, Shao ZY, Liu FJ. Effects of concurrent chemoradiotherapy combined with Jinlong capsules for patients with advanced nasopharyngeal squamous carcinoma. Chinese Journal of Cancer Prevention and Treatment. 2010;(2):138-9. 174. He X, Hu C, Wu Y. Radioprotective effect of amifostine in nasopharyngeal carcinoma. Fudan University Journal of Medical Sciences. 2004;31(1):53-8. 175. Hu YR, Wu CQ, Liu YJ, Wang Y, Li X, Zhong H, et al. Clinical observation on effect of shenqi fanghou recipe in preventing and treating radiation injury in patients with head and neck tumour. Zhongguo Zhong Xi Yi Jie He Za Zhi. 2005;25(7):623-5. 176. Wang Q, Liu H, Qiao N. Effects of traditional Chinese medicine on salivary glands in the patients with head and neck cancer during radiotherapy. Zhongguo Zhong Xi Yi Jie He Za Zhi. 1998;18(11):662-4. 177. Fuertes Cabero S, Setoain Perego X, Rovirosa Casino A, Mateos Fernández JJ, Fuster Pelfort D, Ferre Jorge J, et al. Usefulness of pilocarpine in the prevention of xerostomia in patients with head and neck cancer treated with radiotherapy. Assessment with gammagraphy and salivary flow. Revista Española de Medicina Nuclear. 2004;23(4):259-66. 178. Kontopantelis E, Reeves D. Performance of statistical methods for meta-analysis when true study effects are non-normally distributed: A simulation study. Statistical methods in medical research. 2012;21(4):409-26. 179. Abacioglu MU. Effect of Pilocarpine for the Prevention of Radiation-Induced Xerostomia [Dissertation]. 180. Antonadou D, Pepelassi M, Synodinou M, Puglisi M, Throuvalas N. Prophylactic use of amifostine to prevent radiochemotherapy-induced mucositis and xerostomia in head-and-neck cancer. International journal of radiation oncology, biology, physics. 2002;52(3):739-47. 181. Bardet E, Martin L, Calais G, Alfonsi M, Feham N E, Tuchais C, et al. Subcutaneous compared with intravenous administration of amifostine in patients with head and neck cancer

297 receiving radiotherapy: final results of the GORTEC2000-02 phase III randomized trial. Journal of Clinical Oncology. 2011;29(2):127-33. 182. Buentzel J, Micke O, Adamietz IA, Monnier A, Glatzel M, de Vries A. Intravenous amifostine during chemoradiotherapy for head-and-neck cancer: a randomized placebo-controlled phase III study. International journal of radiation oncology, biology, physics. 2006;64(3):684-91. 183. Büntzel J, Küttner K, Fröhlich D, Glatzel M. Selective cytoprotection with amifostine in concurrent radiochemotherapy for head and neck cancer. Annals of Oncology. 1998;9(5):505-9. 184. Büntzel J, Riesenbeck D, Glatzel M, Berndt-Skorka R, Riedel T, Mücke R, et al. Limited effects of selenium substitution in the prevention of radiation-associated toxicities. Results of a randomized study in head and neck cancer patients. Anticancer Research. 2010;30(5):1829-32. 185. Burlage FR, Roesink JM, Kampinga HH, Coppes RP, Terhaard C, Langendijk JA, et al. Protection of salivary function by concomitant pilocarpine during radiotherapy: a double-blind, randomized, placebo-controlled study. International journal of radiation oncology, biology, physics. 2008;70(1):14-22. 186. Duncan GG, Epstein JB, Tu D, El Sayed S, Bezjak A, Ottaway J, et al. Quality of life, mucositis, and xerostomia from radiotherapy for head and neck cancers: a report from the NCIC CTG HN2 randomized trial of an antimicrobial lozenge to prevent mucositis. Head & neck. 2005;27(5):421-8. 187. Fisher J, Scott C, Scarantino CW, Leveque FG, White RL, Rotman M, et al. Phase III quality-of-life study results: impact on patients' quality of life to reducing xerostomia after radiotherapy for head-and-neck cancer - RTOG 97-09. International journal of radiation oncology, biology, physics. 2003;56(3):832-6. 188. Gornitsky M, Shenouda G, Sultanem K, Katz H, Hier M, Black M, et al. Double-blind randomized, placebo-controlled study of pilocarpine to salvage salivary gland function during radiotherapy of patients with head and neck cancer. Oral surgery, oral medicine, oral pathology, oral radiology, and endodontics. 2004;98(1):45-52. 189. Grötz KA, Wüstenberg P, Kohnen R, Al-Nawas B, Henneicke-von Zepelin H, Bockisch A, et al. Prophylaxis of radiogenic sialadenitis and mucositis by coumarin/troxerutine in patients with head and neck cancer - a prospective, randomized, placebo-controlled, double-blind study. British Journal of Oral and Maxillofacial Surgery. 2001;39(1):34-9. 190. Haddad P, Karimi M. A randomised, double-blind, placebo-controlled trial of concomitant pilocarpine with head and neck irradiation for prevention of radiation-induced xerostomia. Radiotherapy and Oncology. 2002;64(1):29-32. 191. Haddad R, Sonis S, Posner M, Wirth L, Costello R, Braschayko P, et al. Randomized phase 2 study of concomitant chemoradiotherapy using weekly carboplatin/paclitaxel with or without daily subcutaneous amifostine in patients with locally advanced head and neck cancer. Cancer. 2009;115(19):4514-23. 192. Jaguar GC, Lima EN, Kowalski LP, Pellizzon AC, Carvalho AL, Boccaletti KW, et al. Double blind randomized prospective trial of bethanechol in the prevention of radiation-induced salivary gland dysfunction in head and neck cancer patients. Radiotherapy and Oncology. 2015;115(2):253-6. 193. Jellema AP, Slotman BJ, Muller MJ, Leemans CR, Smeele LE, Hoekman K, et al. Radiotherapy alone, versus radiotherapy with amifostine 3 times weekly, versus radiotherapy with amifostine 5 times weekly: a prospective randomized study in squamous cell head and neck cancer. Cancer. 2006;107(3):544-53. 194. Jham BC, Teixeira IV, Aboud CG, Carvalho AL, Coelho Mde M, Freire AR. A randomized phase III prospective trial of bethanechol to prevent radiotherapy-induced salivary gland damage in patients with head and neck cancer. Oral oncology. 2007;43(2):137-42. 195. Lajtman Z, Krajina Z, Krpan D, Vincelj J, Borcic V, Popovic-Kovacic J. Pilocarpine in the prevention of postirradiation xerostomia. Acta Medica Croatica. 2000;54(2):65-7. 196. Lanzós I, Herrera D, Santos S, O'Connor A, Peña C, Lanzós E, et al. Mucositis in irradiated cancer patients: effects of an antiseptic mouthrinse. Medicina Oral, Patología Oral y Cirugía Bucal. 2010;15(5):e732-e8. 197. Lin KS, Jen YM, Chao TY, Lin YS, Wang LH, Lin CJ, et al. Prevention of acute radiation- associated toxicity by traditional chinese medicine Tianwang Buxin mini-pills in patients with head and neck cancer. Journal of Medical Sciences (Taipei, Taiwan). 2014;34:152-60. 198. Lozada-Nur F, Schoelch M, Fu K, Muscoplat C, Trivedi M, Smith C, et al. A pilot study to evaluate the effect of pilocarpine tablets on salivary flow and mucositis in head and neck cancer

298 patients during radiotherapy. Proceedings of the American Society of Clinical Oncology. 1998;17:399-. 199. Patni N, Patni S, Bapna A, Somani N, Gupta A, Ratnam BV. The role of amifostine in prophylaxis of radiotherapy induced mucositis and xerostomia in head and neck cancer. Journal of Clinical Oncology. 2004;22(14 Suppl):5568-. 200. Peng LH, Sun Y, Zhang XL, Zhang Q, Fu S. Efficacy of amifostine on patients with head and neck squamous cell carcinomas undergoing radiation therapy and chemotherapy and clinical observation on its hematologic toxicity. Journal of China Pharmacy. 2006;17(5):1166-8. 201. Pimentel MJ, Filho MM, Araújo M, Gomes DQ, Da Costa LJ. Evaluation of radioprotective effect of pilocarpine ingestion on salivary glands. Anticancer Research. 2014;34(4):1993-9. 202. Reshma K, Kamalaksh S, Bindu YS, Pramod K, Asfa A, Amritha D, et al. Tulasi (Ocimum Sanctum) as radioprotector in head and neck cancer patients undergoing radiation therapy. Biomedicine. 2012;32:39-44. 203. Sangthawan D, Watthanaarpornchai S, Phungrassami T. Randomized double blind, placebo-controlled study of pilocarpine administered during head and neck irradiation to reduce xerostomia. Journal of the Medical Association of Thailand. 2001;84(2):195-203. 204. Vacha P, Fehlauer F, Mahlmann B, Marx M, Hinke A, Sommer K, et al. Randomized phase III trial of postoperative radiochemotherapy +/- amifostine in head and neck cancer. Is there evidence for radioprotection? Strahlentherapie und Onkologie. 2003;179(6):385-9. 205. Valdez IH, Wolff A, Atkinson JC, Macynski AA, Fox PC. Use of pilocarpine during head and neck radiation therapy to reduce xerostomia and salivary dysfunction. Cancer. 1993;71(5):1848-51. 206. Veerasarn V, Phromratanapongse P, Suntornpong N, Lorvidhaya V, Sukthomya V, Chitapanarux I, et al. Effect of amifostine to prevent radiotherapy-induced acute and late toxicity in head and neck cancer patients who had normal or mild impaired salivary gland function. Journal of the Medical Association of Thailand. 2006;89(12):2056-67. 207. Warde P, O'Sullivan B, Aslanidis J, Kroll B, Lockwood G, Math M, et al. A phase III placebo-controlled trial of oral pilocarpine in patients undergoing radiotherapy for head and neck cancer. International journal of radiation oncology, biology, physics. 2002;54(1):9-13. 208. Watanabe T, Ishihara M, Matsuura K, Mizuta K, Itoh Y. Polaprezinc prevents oral mucositis associated with radiochemotherapy in patients with head and neck cancer. International Journal of Cancer. 2010;127(8):1984-90. 209. Mercadante V, Al Hamad A, Lodi G, Porter S, Fedele S. Interventions for the management of radiotherapy-induced xerostomia and hyposalivation: A systematic review and meta-analysis. Oral oncology. 2017;66:64-74. 210. National Guideline Clearinghouse. 2013 (Revised) Criteria for Inclusion of Clinical Practice Guidelines in NGC. [accessed 04/05/2017] https://guideline.gov/help-and- about/summaries/inclusion-criteria. 211. AGREE Next Steps Consortium (2009). The AGREE II Instrument [Electronic version]. Accessed 15/05/2017] http://www.agreetrust.org. 212. Alberta Health Services. Oral and Dental Care Management in Head and Neck Cancer. Available from http://www.albertahealthservices.ca/info/cancerguidelines.aspx [accessed 11 July 2017]. 213. Bensinger W, Schubert M, Ang KK, Brizel D, Brown E, Eilers JG, et al. NCCN Task Force Report. prevention and management of mucositis in cancer care. Journal of the National Comprehensive Cancer Network : JNCCN. 2008;6 Suppl 1:S1-21; quiz S2-4. 214. Buglione M, Cavagnini R, Di Rosario F, Sottocornola L, Maddalo M, Vassalli L, et al. Oral toxicity management in head and neck cancer patients treated with chemotherapy and radiation: Dental pathologies and osteoradionecrosis (Part 1) literature review and consensus statement. Critical reviews in oncology/hematology. 2016;97:131-42. 215. Butterworth C, McCaul L, Barclay C. Restorative dentistry and oral rehabilitation: United Kingdom National Multidisciplinary Guidelines. The Journal of laryngology and otology. 2016;130(S2):S41-S4. 216. Califano R, Tariq N, Compton S, Fitzgerald DA, Harwood CA, Lal R, et al. Expert Consensus on the Management of Adverse Events from EGFR Tyrosine Kinase Inhibitors in the UK. Drugs. 2015;75(12):1335-48.

299 217. Cohen EE, LaMonte SJ, Erb NL, Beckman KL, Sadeghi N, Hutcheson KA, et al. American Cancer Society Head and Neck Cancer Survivorship Care Guideline. CA: a cancer journal for clinicians. 2016;66(3):203-39. 218. Davies A, Bagg J, Laverty D, Sweeney P, Filbet M, Newbold K, et al. Salivary gland dysfunction ('dry mouth') in patients with cancer: a consensus statement. European journal of cancer care. 2010;19(2):172-7. 219. De Sanctis V, Bossi P, Sanguineti G, Trippa F, Ferrari D, Bacigalupo A, et al. Mucositis in head and neck cancer patients treated with radiotherapy and systemic therapies: Literature review and consensus statements. Critical reviews in oncology/hematology. 2016;100:147-66. 220. Edmonds K, Hull D, Spencer-Shaw A, Koldenhof J, Chrysou M, Boers-Doets C, et al. Strategies for assessing and managing the adverse events of sorafenib and other targeted therapies in the treatment of renal cell and hepatocellular carcinoma: recommendations from a European nursing task group. European journal of oncology nursing : the official journal of European Oncology Nursing Society. 2012;16(2):172-84. 221. Fogh S, Yom SS. Symptom management during the radiation oncology treatment course: a practical guide for the oncology clinician. Semin Oncol. 2014;41(6):764-75. 222. Hensley ML, Hagerty KL, Kewalramani T, Green DM, Meropol NJ, Wasserman TH, et al. American Society of Clinical Oncology 2008 clinical practice guideline update: use of chemotherapy and radiation therapy protectants. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2009;27(1):127-45. 223. Hutchinson CT, Suntharalingam M, Strome SE. What are the best management strategies for radiation-induced xerostomia? The Laryngoscope. 2014;124(2):359-60. 224. Lacouture ME, Anadkat MJ, Bensadoun RJ, Bryce J, Chan A, Epstein JB, et al. Clinical practice guidelines for the prevention and treatment of EGFR inhibitor-associated dermatologic toxicities. Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer. 2011;19(8):1079-95. 225. Mirabile A, Airoldi M, Ripamonti C, Bolner A, Murphy B, Russi E, et al. Pain management in head and neck cancer patients undergoing chemo-radiotherapy: Clinical practical recommendations. Critical reviews in oncology/hematology. 2016;99:100-6. 226. The Royal College of Surgeons of England/The British Society for Disability and Oral Health. The Oral Management of Oncology Patients Requiring Radiotherapy, Chemotherapy and / or Bone Marrow Transplantation: Clinical Guidelines. Available from https://www.rcseng.ac.uk/ [accessed 11 July 2017]. 227. Sandherr M, Hentrich M, von Lilienfeld-Toal M, Massenkeil G, Neumann S, Penack O, et al. Antiviral prophylaxis in patients with solid tumours and haematological malignancies--update of the Guidelines of the Infectious Diseases Working Party (AGIHO) of the German Society for Hematology and Medical Oncology (DGHO). Ann Hematol. 2015;94(9):1441-50. 228. Snowden JA, Ahmedzai SH, Ashcroft J, D'Sa S, Littlewood T, Low E, et al. Guidelines for supportive care in multiple myeloma 2011. Br J Haematol. 2011;154(1):76-103. 229. Tolentino Ede S, Centurion BS, Ferreira LH, Souza AP, Damante JH, Rubira-Bullen IR. Oral adverse effects of head and neck radiotherapy: literature review and suggestion of a clinical oral care guideline for irradiated patients. Journal of applied oral science : revista FOB. 2011;19(5):448-54. 230. Potting C, Mistiaen P, Poot E, Blijlevens N, Donnelly P, van Achterberg T. A review of quality assessment of the methodology used in guidelines and systematic reviews on oral mucositis. Journal of clinical nursing. 2009;18(1):3-12. 231. Grimshaw J, McAuley LM, Bero LA, Grilli R, Oxman AD, Ramsay C, et al. Systematic reviews of the effectiveness of quality improvement strategies and programmes. Quality & safety in health care. 2003;12(4):298-303. 232. Weir MC, Grimshaw JM, Mayhew A, Fergusson D. Decisions about lumping vs. splitting of the scope of systematic reviews of complex interventions are not well justified: a case study in systematic reviews of health care professional reminders. Journal of clinical epidemiology. 2012;65(7):756-63. 233. Spielberger RT, Stiff P, Emmanouilides C, Yanovich S, Bensinger W, Hedrick E, et al., editors. Efficacy of recombinant human keratinocyte growth factor (rHuKGF) in reducing mucositis in patients with hematologic malignancies undergoing autologous peripheral blood progenitor cell transplantation (auto-PBPCT) after radiation-based conditioning - results of a phase 2 trial. Proceedings of the American Society of Clinical Oncology; 2001.

300 234. Akobeng AK. Principles of evidence based medicine. Archives of disease in childhood. 2005;90(8):837-40. 235. Tugwell P, Boers M. OMERACT conference on outcome measures in rheumatoid arthritis clinical trials: introduction. The Journal of rheumatology. 1993;20(3):528-30. 236. Tovey D: Impact of Cochrane Reviews [editorial], The Cochrane Library.; 2010. http://www.thecochranelibrary.com/details/editorial/756937/The-Impact-of-Cochrane-Reviews-by- Dr-David-Tovey.html. 237. Kirkham JJ, Dwan KM, Altman DG, Gamble C, Dodd S, Smyth R, et al. The impact of outcome reporting bias in randomised controlled trials on a cohort of systematic reviews. Bmj. 2010;340:c365. 238. Robenolt J, Trovato J, Thompson J, Gordon S, de Vera M. Effect of oral cryotherapy on mucositis-related pain and patient functioning in hematopoeitic stem cell transplant recipients receiving high-dose melphalan. Biology of Blood and Marrow Transplantation. 2010;16(2 Suppl 2):S322-S. 239. Xourafas V, Heras P, Hatzopoulos A, Karagiannis S, editors. The role of cryotherapy on fentanyl use in breast cancer patients. European Journal of Cancer, Supplement Proceedings of the Joint ECCO 15 - 34th ESMO Multidisciplinary Congress Berlin Germany; 2009. 240. Turhal NS, Erdal S, Karacay S. Efficacy of treatment to relieve mucositis-induced discomfort. Supportive Care in Cancer. 2000;8(1):55-8. 241. Antonadou D, Athanassiou E, Synodinou M, Koliarakis N, Panoussaki K, Karageorgis P, et al. Evaluation of the efficacy of granulocyte macrophage colony stimulating factor (GM-CSF) in the prevention of radiation induced mucositis. Radiotherapy and Oncology. 1998;48 Suppl 1:S39- S. 242. Elsaid A, Farouk M. Significance of anemia and role of erythropoietin in radiation induced mucositis in head and neck cancer patients [abstract]. International journal of radiation oncology, biology, physics. 2001;51(3 Suppl 1):368-. 243. Grzegorczyk-Jazwinska A, Dwilewicz-Trojaczek J, Kozak I, Karakulska-Prystupiuk E, Gorska R. Effect of locally applied G-CSF on oral mucositis after autogeneic and allogeneic stem cell transplantation. Acta Haematologica Polonica. 2006;37(2):225-40. 244. Koga Y, Tsurusawa M, Watanabe T, Gosho M, Mori T, Mitsui T, et al., editors. Randomized study of granulocyte colony stimulating factor for childhood B-cell non-Hodgkin lymphoma: A report from the Japanese pediatric leukemia/lymphoma study Group B-NHL03 study. British Journal of Haematology; 2015: Blackwell Publishing Ltd. 245. Patte C, Laplanche A, Bertozzi AI, Baruchel A, Frappaz D, Schmitt C, et al. Granulocyte colony-stimulating factor in induction treatment of children with non-Hodgkin's lymphoma: a randomized study of the French Society of Pediatric Oncology. Journal of Clinical Oncology. 2002;20(2):441-8. 246. Schuster MW, Anaissie E, Hurd D, Bensinger W, Mason J, McCarty J, et al. Final analysis of the phase II, randomized, double-blind, placebo-controlled trial of single-dose velafermin (CG53135-05) for the prevention of oral mucositis. Journal of Supportive Oncology. 2007;5(4 Suppl 2):58-9. 247. Schuster MW, Mehta J, Waller EK, Rifkin RM, Micallef I, Hurd D, et al. A randomized placebo controlled phase II trial of prevention of severe oral mucositis using single dose velafermin in patients receiving myeloablative therapy and autologous hematopoietic stem cell transplant (AHSCT) [abstract]. Blood. 2007;110(11):615-. 248. Shea TC, Kewalramani T, Mun Y, Jayne G, Dreiling LK. Evaluation of single-dose palifermin to reduce oral mucositis in fractionated total-body irradiation and high-dose chemotherapy with autologous peripheral blood progenitor cell transplantation. Journal of Supportive Oncology. 2007;5(4 Suppl 2):60-2. 249. Anné PR. Phase II trial of subcutaneous amifostine in patients undergoing radiation therapy for head and neck cancer. Seminars in Oncology. 2002;29(6 Suppl 19):80-3. 250. Bagga P, Anand AK, Raina A, Choudhary PS, Chaudhoory AR. Evaluation of pilocarpine for the prevention of radiation induced xerostomia in head and neck cancers. Annals of Oncology. 2007;18(Suppl 9):ix177-ix. 251. Bakowski MT, Macdonald E, Mould RF, Cawte P, Sloggem J, Barrett A, et al. Double blind controlled clinical trial of radiation plus razoxane (ICRF 159) versus radiation plus placebo in the treatment of head and neck cancer. International journal of radiation oncology, biology, physics. 1978;4(1-2):115-9.

301 252. Belcaro G, Cesarone MR, Genovesi D, Ledda A, Vinciguerra G, Ricci A, et al. Pycnogenol may alleviate adverse effects in oncologic treatment. Panminerva Medica. 2008;50(3):227-34. 253. Borg M, Krishnan S, Olver L, Stein B, Chatterton B, Coates E, et al. Randomised double- blind trial of amifostine vs placebo for radiation-induced xerostomia in patients with head and neck cancer. ANZ Journal of Surgery. 2007;77(12):A3-A. 254. Bourhis J, de Crevoisier R, Abdulkarim B, Deutsch E, Lusinchi A, Luboinski B, et al. A randomized study of very accelerated radiotherapy with and without amifostine in head and neck squamous cell carcinoma. International journal of radiation oncology, biology, physics. 2000;46(5):1105-8. 255. Braaksma M, Levendag P. Tools for optimal tissue sparing in concomitant chemoradiation of advanced head and neck cancer: subcutaneous amifostine and computed tomography-based target delineation. Seminars in Oncology. 2002;29(6 Suppl 19):63-70. 256. Braaksma M, van Agthoven M, Nijdam W, Uyl-de Groot C, Levendag P. Costs of treatment intensification for head and neck cancer: concomitant chemoradiation randomised for radioprotection with amifostine. European journal of cancer. 2005;41(14):2102-11. 257. Chambers MS, Posner MR, Jones CU, Weber RS, Vitti R. Two phase III clinical studies of cevimeline for post-radiation xerostomia in patients with head and neck cancer. Journal of Clinical Oncology. 2005;23(16 Suppl):5503-. 258. Demiroz C, Ozcan L, Cebelli G, Karadag O, Ozsahin EM. The effect of amifostine on acute and late radiation side effects in head and neck cancer patients. Turkiye Klinikleri Journal of Medical Sciences. 2012;32:1207-16. 259. Fallahi B, Beiki D, Abedi SM, Saghari M, Fard-Esfahani A, Akhzari F, et al. Does vitamin E protect salivary glands from I-131 radiation damage in patients with thyroid cancer? Nuclear Medicine Communications. 2013;34(8):777-86. 260. Fan Z, Lin H, Liu Z, He Q, Zhang S, Li W. Protective effect of amifostine in irradiation mucositis of the oral cavity and dry mouth after nasapharyngeal carcinoma. Military Medical Journal of Southeast China. 2011;13(2):146-8. 261. Franzén L, Henriksson R, Littbrand B, Zackrisson B. Effects of sucralfate on mucositis during and following radiotherapy of malignancies in the head and neck region. A double-blind placebo-controlled study. Acta Oncologica (Stockholm, Sweden). 1995;34(2):219-23. 262. Goyal S, Sharma DN, Julka PK, Rath GK. Effect of oral pilocarpine on xerostomia and quality of life in patients receiving curative radiotherapy for head and neck cancers. Journal of Clinical Oncology. 2007;25(18 Suppl):6088-. 263. Gu J, Zhu S, Li X, Wu H, Li Y, Hua F. Effect of amifostine in head and neck cancer patients treated with radiotherapy: a systematic review and meta-analysis based on randomized controlled trials. PloS one. 2014;9(5):e95968-e. 264. Johnson DJ, Scott CB, Marks JE, Seay TE, Atkins JN, Berk LB, et al. Assessment of quality of life and oral function of patients participating in a phase II study of radioprotection of oral and pharyngeal mucosa by the prostaglandin E(1) analog misoprostol (RTOG 96-07). International journal of radiation oncology, biology, physics. 2002;54(5):1455-9. 265. Karacetin D, Yücel B, Leblebicioglu B, Aksakal O, Maral O, Incekara O. A randomized trial of amifostine as radioprotector in the radiotherapy of head and neck cancer. Journal of BUON. 2004;9(1):23-6. 266. Koukourakis MI, Kyrias G, Kakolyris S, Kouroussis C, Frangiadaki C, Giatromanolaki A, et al. Subcutaneous administration of amifostine during fractionated radiotherapy: a randomized phase II study. Journal of Clinical Oncology. 2000;18(11):2226-33. 267. Kumarchandra R, Shenoy K, Bindu YS, Kadaba P, Chandrashekar R. Ocimum Sanctum as a radioprotector in head and neck cancer patients undergoing radiation therapy. Free Radical Biology and Medicine. 2010;49 Suppl:S64-S. 268. Manoor Maiya V, Vaid N, Basu S, Vatyam S, Hegde S, Deshmukh S, et al. The use of xylitol for the prevention of xerostomia in patients receiving intensity modulated radiation therapy for head and neck cancers. International journal of radiation oncology, biology, physics. 2014;90(1 Suppl):S562-S. 269. Mateos JJ, Setoain X, Ferre J, Rovirosa A, Navalpotro B, Martin F, et al. Salivary scintigraphy for assessing the protective effect of pilocarpine in head and neck irradiated tumours. Nuclear Medicine Communications. 2001;22(6):651-6. 270. Mitine C, Chaltin M, Salembier C, Merlo P, Beauduin M, editors. Head and neck radiotherapy: does pilocarpine hydrochloride reduce radiation-induced xerostomia? Proceedings of

302 the management of head and neck tumors: what are the challenges for the third millennium; 1999 Dec 3-4; Brussels (Belgium); 1999: European Laryngological Society. 271. Mix MD, Jameson M, Tills M, Dibaj S, Groman A, Jaggernauth W, et al. Randomized double-blind, placebo-controlled, multicenter phase 2 trial of selenomethionine as a modulator of efficacy and toxicity of chemoradiation in locally-advanced squamous cell carcinoma of the head and neck. International journal of radiation oncology, biology, physics. 2013;87(2 Suppl):S466-S7. 272. Nicolatou-Galitis O, Sotiropoulou-Lontou A, Velegraki A, Pissakas G, Kolitsi G, Kyprianou K, et al. Oral candidiasis in head and neck cancer patients receiving radiotherapy with amifostine cytoprotection. Oral oncology. 2003;39:397-401. 273. Norberg-Spaak LE, Lundquist PG, Berndtson M, Klintenberg C. Can pilocarpine treatment during irradiation prevent salivary gland damage? Swedish Dental Journal. 1996;20:234-. 274. Norberg-Spaak LE. Can pilocarpine treatment during irradiation prevent salivary gland damage. Clinical Otolaryngology and Allied Sciences. 1997;22:88-. 275. Nyárády Z, Németh A, Bán A, Mukics A, Nyárády J, Ember I, et al. A randomized study to assess the effectiveness of orally administered pilocarpine during and after radiotherapy of head and neck cancer. Anticancer Research. 2006;26(2B):1557-62. 276. Park J, McGuire DB, Kang H. Effects of cold sterile normal saline (CSNS) mouth care in head and neck cancer (HNC) patients undergoing concurrent chemoradiotherapy (CCRT). Supportive Care in Cancer. 2012;20(1 Suppl):S246-S7. 277. Park KN, Son YI, Baek CH. Protection of radiotherapy-induced xerostomia with vitamin E+C complex. Otolaryngology - Head and Neck Surgery. 2012;147(2 Suppl):P174-P. 278. Peters K, Mücke R, Hamann D, Ziegler PG, Fietkau R. Supportive use of amifostine in patients with head and neck tumours undergoing radio-chemotherapy. Is it possible to limit the duration of the application of amifostine? Strahlentherapie und Onkologie. 1999;175 Suppl 4:23-6. 279. Qian Y. Effects of Traditional Chinese Medicine on Salivary Glands of Patients with Head and Neck Cancer During Radiotherapy [Dissertation]: China National Knowledge Infrastructure (CNKI); 2003. 280. Resubal JR, Calaguas MJ. The effect of zilongjin(r) in patients with head and neck cancer needing radiotherapy with and without chemotherapy. Radiotherapy and Oncology. 2011;99(Suppl 1):S336-S. 281. Rieger JM, Jha N, Lam Tang JA, Harris J, Seikaly H. Functional outcomes related to the prevention of radiation-induced xerostomia: oral pilocarpine versus submandibular salivary gland transfer. Head & neck. 2012;34(2):168-74. 282. Rischin D, Peters LJ, O'Sullivan B, Giralt J, Fisher R, Yuen K, et al. Tirapazamine, cisplatin, and radiation versus cisplatin and radiation for advanced squamous cell carcinoma of the head and neck (TROG 02.02, HeadSTART): a phase III trial of the Trans-Tasman Radiation Oncology Group. Journal of Clinical Oncology. 2010;28(18):2989-95. 283. Rudat V, Münter M, Rades D, Grötz K, Haberkorn U, Brenner W. The effect of amifostine or IMRT to preserve the parotid function after radiotherapy of the head and neck region measured by quantitative salivary gland scintigraphy. Journal of Clinical Oncology. 2005;23(16 Suppl):5502- 284. Schönekäs KG, Wagner W, Prott FJ. Amifostine - a radioprotector in locally advanced head and neck tumours. Strahlentherapie und Onkologie. 1999;175 Suppl 4:27-9. 285. Sharma A, Rath GK, Chaudhary SP, Thakar A, Mohanti BK, Bahadur S. Lactobacillus brevis CD2 lozenges reduce radiation- and chemotherapy-induced mucositis in patients with head and neck cancer: a randomized double-blind placebo-controlled study. European journal of cancer. 2012;48(6):875-81. 286. Strnad V, Sauer R, Krafft T, Lerch S. Preliminary results of a randomised study using WR- 2721 in radiation therapy alone in patients with head and neck cancer. European journal of cancer. 1997;33(Suppl 8):S189-S. 287. Thorstad WL, Haughey B, Chao KS. Pilot study of subcutaneous amifostine in patients undergoing postoperative intensity modulated radiation therapy for head and neck cancer: preliminary data. Seminars in Oncology. 2003;30(6 Suppl 18):96-100. 288. Uchiyama Y, Murakami S, Kakimoto N, Nakatani A, Furukawa S. Effectiveness of cepharanthin in decreasing interruptions during radiation therapy for oral cancer. Oral Radiology. 2005;21(1):41-4. 289. Zale M, Chambers MS, Khan Z. Pilocarpine effects on xerostomia associated with radiation therapy. Journal of dental research. 1993;72(Abs Spec Issue):375-.

303 290. Zimmerman RP, Mark RJ, Tran LM, Juillard GF. Concomitant pilocarpine during head and neck irradiation is associated with decreased posttreatment xerostomia. International journal of radiation oncology, biology, physics. 1997;37(3):571-5. 291. Yu DS, Zhao W, Wu XL, Huang ZY, Huang RB. Amifostine in prevention of radiation- induced xerostomia. Journal of Sun Yat-Sen University (Medical Sciences). 2009;30(4S):86-7, 95. 292. Reeves BC, Deeks JJ, Higgins JPT, Wells GA. Chapter 13: Including non-randomized studies. In: Higgins JPT, Green S (editors), Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 (updated March 2011). The Cochrane Collaboration, 2011. Available from www.cochrane-handbook.org. 293. Sinha IP, Smyth RL, Williamson PR. Using the Delphi technique to determine which outcomes to measure in clinical trials: recommendations for the future based on a systematic review of existing studies. PLoS medicine. 2011;8(1):e1000393.

304

THE PREVENTION OF ORAL SIDE EFFECTS OF CANCER TREATMENT

VOLUME II of II

A thesis submitted to The University of Manchester for the degree of Doctor of Philosophy in the Faculty of Biology, Medicine and Health

2017

PHILIP RILEY

SCHOOL OF MEDICAL SCIENCES

Division of Dentistry

305 APPENDIX 1: Oral cryotherapy review - electronic search strategies

1) Cochrane Oral Health Group Trials Register search strategy

1 ((neoplasm* or cancer* or tumor* or tumour*):ti,ab) AND (INREGISTER) 2 ((leukaemi* or leukemi*):ti,ab) AND (INREGISTER) 3 ((malignan* or neutropeni* or carcino* or adenocarcinoma* or lymphoma*):ti,ab) AND (INREGISTER) 4 ((radioth* or radiat* or irradiat*):ti,ab) AND (INREGISTER) 5 (("bone marrow transplant*"):ti,ab) AND (INREGISTER) 6 ((chemo*):ti,ab) AND (INREGISTER) 7 (#1 or #2 or #3 or #4 or #5 or #6) AND (INREGISTER) 8 ((stomatitis or mucositis):ti,ab) AND (INREGISTER) 9 ((oral and mucos*):ti,ab) AND (INREGISTER) 10 ((mycosis or mycotic):ti,ab) AND (INREGISTER) 11 (mIAS:ti,ab) AND (INREGISTER) 12 (#8 or #9 or #10 or #11) AND (INREGISTER) 13 (cryotherap*:ti,ab) AND (INREGISTER) 14 ((ice or freez* or cold*):ti,ab) AND (INREGISTER) 15 (#13 or #14) AND (INREGISTER) 16 (#7 and #12 and #15) AND (INREGISTER)

2) Cochrane Central Register of Controlled Trials (CENTRAL) search strategy

#1 [mh neoplasms] #2 [mh leukemia] #3 [mh lymphoma] #4 [mh radiotherapy] #5 [mh "antineoplastic agents"] #6 (neoplasm* or cancer* or tumor* or tumour*) #7 (leukaemi* or leukemi*) #8 (malignan* or neutropeni* or carcino* or adenocarcinoma* or lymphoma*) #9 "bone marrow transplant*" #10 chemo* #11 {or #1-#10} #12 [mh stomatitis] #13 [mh ^"candidiasis, oral"] #14 stomatitis #15 mucositis #16 (oral near/6 mycos*) #17 (mycosis or mycotic) #18 mIAS:ti,ab #19 {or #12-#18} #20 [mh cryotherapy] #21 cryotherap* #22 (cold* or ice or freez*) #23 {or #20-#22} #24 #11 and #19 and #23

3) MEDLINE (Ovid) search strategy

1. exp NEOPLASMS/ 2. exp LEUKEMIA/ 3. exp LYMPHOMA/ 4. exp RADIOTHERAPY/ 5. exp Antineoplastic agents/ 6. Bone Marrow Transplantation/ 7. neoplasm$.mp. 306 8. cancer$.mp. 9. (leukaemi$ or leukemi$).mp. 10. (tumour$ or tumor$).mp. 11. malignan$.mp. 12. neutropeni$.mp. 13. carcino$.mp. 14. adenocarcinoma$.mp. 15. lymphoma$.mp. 16. (radioth$ or radiat$ or irradiat$).mp. 17. (bone adj marrow adj5 transplant$).mp. 18. chemo$.mp. 19. or/1-18 20. exp STOMATITIS/ 21. Candidiasis, Oral/ 22. stomatitis.mp. 23. mucositis.mp. 24. (oral adj6 mucos$).mp. 25. (mycosis or mycotic).mp. 26. mIAS.ti,ab. 27. or/20-26 28. Cryotherapy/ 29. cryotherap$.mp. 30. (cold or freez$ or ice).mp. 31. or/28-30 32. 19 and 27 and 31

4) EMBASE (Ovid) search strategy

1. exp NEOPLASM/ 2. exp LEUKEMIA/ 3. exp LYMPHOMA/ 4. exp RADIOTHERAPY/ 5. exp Antineoplastic agent/ 6. Bone Marrow Transplantation/ 7. neoplasm$.mp. 8. cancer$.mp. 9. (leukaemi$ or leukemi$).mp. 10. (tumour$ or tumor$).mp. 11. malignan$.mp. 12. neutropeni$.mp. 13. carcino$.mp. 14. adenocarcinoma$.mp. 15. lymphoma$.mp. 16. (radioth$ or radiat$ or irradiat$).mp. 17. (bone adj marrow adj5 transplant$).mp. 18. chemo$.mp. 19. or/1-18 20. exp STOMATITIS/ 21. Thrush/ 22. stomatitis.mp. 23. mucositis.mp. 24. (oral adj6 mucos$).mp. 25. (mycosis or mycotic).mp. 26. mIAS.ti,ab. 27. or/20-26 28. Cryotherapy/ 29. cryotherap$.mp. 30. (cold or freez$ or ice).mp.

307 31. or/28-30 32. 19 and 27 and 31

The above subject search was linked to the Cochrane Oral Health Group filter for identifying RCTs in EMBASE via Ovid:

1. random$.ti,ab. 2. factorial$.ti,ab. 3. (crossover$ or cross over$ or cross-over$).ti,ab. 4. placebo$.ti,ab. 5. (doubl$ adj blind$).ti,ab. 6. (singl$ adj blind$).ti,ab. 7. assign$.ti,ab. 8. allocat$.ti,ab. 9. volunteer$.ti,ab. 10. CROSSOVER PROCEDURE.sh. 11. DOUBLE-BLIND PROCEDURE.sh. 12. RANDOMIZED CONTROLLED TRIAL.sh. 13. SINGLE BLIND PROCEDURE.sh. 14. or/1-13 15. (exp animal/ or animal.hw. or nonhuman/) not (exp human/ or human cell/ or (human or humans).ti.) 16. 14 NOT 15

5) CANCERLIT via PubMed search strategy

#31 (#19 and #26 and #30) #30 (#27 or #28 or #29) #29 ((cold* or freez* or ice)) #28 cryotherap* #27 cryotherapy [mh:noexp] #26 (#20 or #21 or #22 or #23 or #24 or #25) #25 mIAS [tiab] #24 (mycosis or mycotic) #23 mucositis #22 stomatitis #21 oral candidiasis [mh:noexp] #20 stomatitis [mh:exp] #19 (#1 or #2 or #3 or #4 or #5 or #6 or #7 or #8 or #9 or #10 or #11 or #12 or #13 or #14 or #15 or #16 or #17 or #18) #18 chemo* #17 "bone marrow transplant* #16 (radioth* or radiat* or irradiat*) #15 lymphoma* #14 adenocarcinoma* #13 carcino* #12 neutropeni* #11 malignan* #10 (tumor* or tumour*) #9 (leukaemi* or leukemi*) #8 cancer* #7 neoplasm* #6 bone marrow transplantation [mh:noexp] #5 antineoplastic agents [mh:exp] #4 radiotherapy [mh:exp] #3 lymphoma [mh:exp] #2 leukemia [mh:exp] #1 neoplasm [mh:exp]

308 6) CINAHL via EBSCO search strategy

S32 S19 and S27 and S31 S31 S28 or S29 or S30 S30 (cold* or ice or freez*) S29 cryotherap* S28 (mh "cryotherapy") S27 S20 or S21 or S22 or S23 or S24 or S25 or S26 S26 TI mIAS OR AB mIAS S25 (mycosis or mycotic) S24 (oral N6 mucos*) S23 mucositis S22 stomatitis S21 (mh "candidiasis, oral") S20 (mh "stomatitis+") S19 S1 or S2 or S3 or S4 or S5 or S6 or S7 or S8 or S9 or S10 or S11 or S12 or S13 or S14 or S15 or S16 or S17 or S18 S18 chemo* S17 "bone marrow transplant*" S16 (radioth* or radiat* or irradiat*) S15 lymphoma* S14 adenocarcinoma* S13 carcino* S12 neutropeni* S11 malignan* S10 (tumor* or tumour*) S9 (leukaem* or leukem*) S8 cancer* S7 neoplasm* S6 (MH "Bone marrow transplantation") S5 (MH "Antineoplastic Agents+") S4 (mh "radiotherapy+") S3 (mh "lymphoma+") S2 (mh "leukemia+") S1 (mh "neoplasms+")

7) US National Institutes of Health trials registry (ClinicalTrials.gov) and the WHO International Clinical Trials Registry Platform search strategies mucositis and cryotherapy mucositis and ice mucositis and cold

309 APPENDIX 2: Oral cryotherapy review - characteristics of studies and risk of bias tables

Characteristics of included studies

Askarifar 2015

Methods Trial design: parallel (2 arms) Location: Iran Number of centres: 1 Study duration: April to September 2013 Participants Inclusion criteria: full consciousness; suffering Hodgkin or non-Hodgkin lymphoma or multiple myeloma; good oral health; isolation in a separate room; undergoing similar basic chemotherapy; undergoing first course of chemotherapy; undergoing autologous BMT Exclusion criteria: patient dissatisfaction; loss of consciousness; susceptible to other diseases that could potentially disrupt treatment; use of analgesics continuously prior to start of study; receiving combined therapies such as radiotherapy; fever; neutropenia; mucositis prior to the treatment; respiratory diseases; oral infections; systemic diseases affecting oral health (especially periodontal tissues); more than 2 weeks interval between chemotherapy and transplantation; changes in treatment protocol during the study Cancer type: haematological (Hodgkin: Gp A: 31%; Gp B: 46%; non-Hodgkin: Gp A: 13%; Gp B: 23%; multiple myeloma: Gp A: 56%; Gp B: 31%) Cancer treatment: melphalan for Hodgkin and non-Hodgkin lymphoma; melphalan, cytarabine, etoposide, and lomustine for multiple myeloma ("There were no differences in terms of...treatment regimen") Any other potentially important prognostic factors: "There were no differences in terms of...educational status"; smokers: Gp A: 13%; Gp B: 31% Age at baseline (years): Gp A: 43 (range 19 to 66); Gp B: 39.8 (range 21 to 62) Gender: Gp A: 56% male; Gp B: 62% male Number randomised: 33 (Gp A: 17; Gp B: 16) Number evaluated: 29 (Gp A: 16; Gp B: 13) Interventions Comparison: cryotherapy versus normal saline Gp A: prior to BMT, ice cubes held in mouth 5 min before start of chemotherapy, held for 30-min periods with maximum 20-min breaks between each period, until 5 min after completion of chemotherapy Gp B: prior to BMT, 30 to 50 cc of saline mouthwash used 30 min before start of chemotherapy, then again every half-hour, until 6 hours after completion of chemotherapy Compliance: not reported Duration of treatment (intended): not reported but probably variable depending on chemotherapy regimen Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed on days 3, 7, 14 and 21 and reported as a mean on each separate assessment day); we requested maximum score experienced per participant over the whole study period but the authors provided incidence of each grade on days 3, 7 and 14  Neutrophil rate (not an outcome of this review) Notes Sample size calculation: based on detection of MD of 0.51, with 80% power at 5% significance level, and accounting for 40% attrition (14 per group required) Adverse effects: not reported Funding: "financial support of Tabriz University of Medical Sciences" Declarations/conflicts of interest: "Authors declare no conflict of interest in this study"

310 Data handling by review authors: the authors provided us with data on the incidence of each oral mucositis grade on days 3, 7 and 14, and we used the data for day 7 in our meta-analyses as the incidence of grades > 0 was highest and therefore probably most closely equates to the maximum score experienced per participant (as reported in the other studies in the meta-analyses) Other information of note: the information on this study is obtained from a pre- publication copy of the study report provided to us by the authors, and also from correspondence with the authors

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "The patients randomly were allocated into control and generation (selection intervention groups using size-2 random blocks and based on bias) a 1:1 allocation ratio random numbers were generated by "Random software Allocation" software"

Comment: computer generated randomisation so probably done adequately Allocation Low risk Quote: "The patients randomly were allocated into control and concealment intervention groups using size-2 random blocks and based on (selection bias) a 1:1 allocation ratio random numbers were generated by "Random software Allocation" software"

Correspondence: "the allocation was performed by a person who was not involved in sampling and analysis"

Comment: appears to be third-party randomisation which should have ensured that the allocation sequence was not manipulated Blinding of High risk Not possible to blind participants and personnel to allocated participants and groups personnel (performance bias) Blinding of outcome High risk The subjective elements in the scale used to measure oral assessment (detection mucositis could have introduced bias into the assessments bias) Incomplete outcome High risk 12% of randomised participants were not included in the data (attrition bias) analyses (Gp A: 6%; Gp B: 19%). All drop-outs were due to fever but this could be a risk of bias if the fever was linked to oral mucositis Selective reporting Low risk Data for outcomes of this review were reported appropriately (reporting bias) Other bias Low risk No other sources of bias are apparent

Cascinu 1994

Methods Trial design: parallel (2 arms) Location: Pesaro, Italy Number of centres: 1 Study duration: not reported Participants Inclusion criteria: first ever course of chemotherapy Exclusion criteria: not reported 311 Cancer type: Gp A: 98% gastrointestinal, 2% prostrate; Gp B: 98% gastrointestinal, 2% prostrate Cancer treatment: 5FU, different dosages and co-treatments (LV, IFN, VP16) equally distributed between groups due to stratification Any other potentially important prognostic factors: performance status (EOCG): Gp A: 0 = 50%, 1 = 32%, 2 = 18%; Gp B: 0 = 50%, 1 = 35%, 2 = 15%; denture wearers equally distributed between groups due to stratification Age at baseline (years): Gp A: median 60 (range 38 to 73); Gp B: median 58 (range 44 to 72) Gender: Gp A: 68% male; Gp B: 70% male Number randomised: 84 (Gp A: 44; Gp B: 40) Number evaluated: 84 (Gp A: 44; Gp B: 40) Interventions Comparison: cryotherapy versus no treatment Gp A: ice chips placed in mouth 5 min before 5FU and continuously swished around, then replenished before the previous ice had completely melted, for total 30 min Gp B: 5FU only All participants were asked to remove dentures Compliance: all Gp A participants received cryotherapy in the first cycle but 2 participants "noted an 'ice cream' headache which caused them to refuse this technique after the second and third cycle of chemotherapy, respectively" Duration of treatment (intended): not reported (variable and dependent on number of cycles of cancer treatment) Outcomes  Oral mucositis: global assessment of the physician's judgement and participants' description on a 0 to 4 scale (very similar to WHO scale and NCI common toxicity criteria) based on methods of Mahood 1991 (assessed after each cycle and reported as first cycle only and all cycles, maximum score reported)  Duration of oral mucositis (not an outcome of this review)  Other adverse effects of cancer treatment (not an outcome of this review) Notes Sample size calculation: not reported Adverse effects: 2 participants in the cryotherapy group "noted an 'ice cream' headache" Funding: not reported Declarations/conflicts of interest: not reported Data handling by review authors: data reported as all cycles (so double-counting is a problem) and first cycle only, so we used the data for first cycle only Other information of note: mean oral mucositis score reported by smoking status for each group for the first cycle only (smokers had higher mean oral mucositis score than non-smokers in both groups)

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "randomised to a control arm or to receive generation (selection cryotherapy" bias) Correspondence: "Randomisation using cards from a computer generated list in sealed envelopes was performed by a person not involved with the care or evaluation of the patient"

Comment: adequate method of random sequence generation Allocation concealment Low risk Quote: "randomised to a control arm or to receive (selection bias) cryotherapy"

312

Correspondence: "Randomisation using cards from a computer generated list in sealed envelopes was performed by a person not involved with the care or evaluation of the patient"

Comment: third-party randomisation and use of sealed envelopes should have ensured that the allocation sequence was not manipulated Blinding of participants High risk Not possible to blind participants and personnel to allocated and personnel groups (performance bias) Blinding of outcome High risk The subjective elements in the scale used to measure oral assessment (detection mucositis could have introduced bias into the assessments bias) Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately. Although only the oral mucositis outcome was mentioned in the methods, this is more likely to be related to reporting quality rather than bias as the study pre- dates the CONSORT statement Other bias Low risk No other sources of bias are apparent

Gori 2007

Methods Trial design: parallel (2 arms) Location: various locations in Italy Number of centres: "multicentre" but unclear how many centres; co-ordinated by the Institute of Hematology and Medical Oncology, University of Bologna Study duration: October 2004 to January 2006 Participants Inclusion criteria: undergoing allogeneic HSCT and MTX-containing GVHD prophylaxis; minimum age 8 years Exclusion criteria: clinical evidence of oral mucositis; participants not receiving at least 3 administrations of MTX following HSCT Cancer type: haematological (types were equally distributed between groups) Cancer treatment: pre-transplant radio/chemotherapy generally comparable between groups; total body irradiation: Gp A: 30.6%; Gp B: 28.3% Any other potentially important prognostic factors: stem cell donor related: Gp A: 45.1%; Gp B: 58.3%; stem cell source: Gp A: marrow = 33.9%, peripheral blood = 66.1%; Gp B: marrow = 28.3%, peripheral blood = 71.7%; folinic acid rescue: Gp A: 43.5%; Gp B: 38.3% Age at baseline (years): Gp A: median 35.5 (range 9 to 59); Gp B: median 40 (range 8 to 66) Gender: Gp A: 51.6% male; Gp B: 50% male Number randomised: 130 (not reported by group) Number evaluated: 122 (Gp A: 62; Gp B: 60) Interventions Comparison: cryotherapy versus no treatment Gp A: after allogeneic HSCT, ice chips (mineral water) or popsicles placed in mouth for minimum 60 min starting from the time of low-dose MTX administration (20 mg/m2 on day +1, 15 mg/m2 on days +3, +6 and +11) as an IV infusion lasting 5 min (± 2), and replenished when melted Gp B: after allogeneic HSCT, MTX as above Compliance: "Six patients enrolled in the cryotherapy arm did not actually complete cryotherapy as planned because of refusal or poor tolerance. However, the 313 exclusion of these patients did not change the results" Duration of treatment (intended): 4 occasions (minimum of 60 min) on 4 separate days (days 1, 3, 6 and 11) Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed once daily for 20 to 30 days, maximum score reported)  Duration of moderate to severe (grade 2 to 4) and severe (grade 3 to 4) oral mucositis (not an outcome of this review) Notes Sample size calculation: based on previous study, 90% power at 5% significance (unclear whether required sample size was achieved) Adverse effects: not reported, only refers to the 6 participants who did not complete cryotherapy due to "refusal or poor tolerance" Funding: "We thank the Italian HSCT Nurses Group (GITMO) for sponsoring the study" Declarations/conflicts of interest: not reported Data handling by review authors: data is maximum oral mucositis score across all cycles of MTX Other information of note: univariate and multivariate analyses showed severe (grade 3 to 4) oral mucositis was significantly associated with total body irradiation and lack of folinic acid rescue

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "After giving their informed consent, patients were generation (selection included in a preformed randomization list that was updated bias) by the coordinating center. Randomization was performed at the ratio of 1 patient per arm with no further stratifications"

Comment: adequate method of random sequence generation Allocation concealment Unclear Quote: "After giving their informed consent, patients were (selection bias) risk included in a preformed randomization list that was updated by the coordinating center. Randomization was performed at the ratio of 1 patient per arm with no further stratifications"

Comment: co-ordinating centre mentioned, but unclear whether or not they allocated participants remotely from this centre (central randomisation by a third party) Blinding of participants High risk Not possible to blind participants and personnel to allocated and personnel groups (performance bias) Blinding of outcome High risk The subjective elements in the scale used to measure oral assessment (detection mucositis could have introduced bias into the assessments bias) Incomplete outcome Low risk 6% of the participants were not included in the analyses but data (attrition bias) this attrition was not reported by group. However, the amount of attrition was low and reasons are fully reported Selective reporting Low risk Data for the primary outcome of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Heydari 2012

314 Methods Trial design: parallel (2 arms) Location: Mashhad, Iran Number of centres: 2 Study duration: March 2007 to April 2008 Participants Inclusion criteria: able to undergo one of the chemotherapy regimens described in the study at a standard dose; normal laboratory levels (including complete blood counts); normal kidney and hepatic function; participant or carer able to read and write Exclusion criteria: previous chemotherapy; not undergoing one of the combined courses of chemotherapy described in the study; treated with head and neck radiotherapy; diabetic Cancer type: Gp A: 55% colorectal, 45% breast; Gp B: 45% colorectal, 55% breast Cancer treatment:  MAYO (mean infusion time 20 min): 5FU (425 mg/m2) and LV (25 mg/m2) for 5 days, repeated every 28 days: Gp A: 55%; Gp B: 45%  CAF (mean infusion time 25 to 35 min): cyclophosphamide (500 mg/m2), adriamycin (50 mg/m2) and 5FU (500 mg/m2) on 1st day of cycle, repeated every 21 days: Gp A: 30%; Gp B: 42.5%  CMF (mean infusion time 25 to 35 min): cyclophosphamide (600 mg/m2), MTX (40 mg/m2) and 5FU (600 mg/m2) on 1st day of cycle, repeated every 28 days: Gp A: 15%; Gp B: 12.5% Any other potentially important prognostic factors: no statistically significant differences between groups in the following factors: tooth status, smoking status, mouthwash use, brushing habit, BMI, educational status Age at baseline (years): Gp A: mean 59.5 (SD 12.35); Gp B: mean 63.25 (SD 15.06) Gender: 40% male overall and reports that there were no statistically significant differences between groups Number randomised: 80 (Gp A: 40; Gp B: 40) Number evaluated: 80 (Gp A: 40; Gp B: 40) Interventions Comparison: cryotherapy versus no treatment Gp A: ice chips placed in mouth 5 min before chemotherapy until 5 min after and continuously swished around, then replenished before the previous ice had completely melted Gp B: chemotherapy only All participants were asked to remove dentures Compliance: well tolerated, no discontinuation of therapy, and most participants kept their mouths constantly cool for most of the chemotherapy session Duration of treatment (intended): mean duration of cryotherapy was 20 to 45 min for a session; those receiving MAYO regimen (see above) had cryotherapy for each of the 5 days of treatment, whilst those receiving CAF/CMF regimen (see above) had cryotherapy on the single day of treatment Outcomes  Oral mucositis: WHO 0 to 4 scale assessed separately by participants and clinicians (first cycle-only reported, assessed daily by participants or on days 1, 5, 14 and 21 for the 5-day regimen (MAYO) and days 7, 14 and 21 for the single-day regimens (CAF/CMF), maximum score reported)

Obtained from correspondence:  Interruptions to cancer treatment: (assessed over first 2 cycles, reported as both event (dichotomous) data and continuous data 315 in the form of mean number of days of interruption)  Oral pain: 1 to 5 scale relating to duration of pain experience (1 = never, 2 = one day of week, 3 = 2 to 3 days of week, 4 = most of week, 5 = 7 days of week) Notes Sample size calculation: not reported Adverse effects: 8 (20%) of participants in the cryotherapy group complained of chills Funding: "This work was supported by the department of research, Mashhad University of Medical Science" Declarations/conflicts of interest: not reported Data handling by review authors: clinician judgement of oral mucositis was preferred over participant judgement as we deemed that this may be more objective and less biased; oral pain was measured in a different way to the other studies measuring intensity/severity of pain and therefore it is not appropriate to meta- analyse using standardised mean difference, so we have presented the data in an additional table Other information of note: not reported

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "Randomization was performed by the use of a generation (selection random numbers table" bias) Comment: adequate method of random sequence generation Allocation concealment Low risk Quote: "Randomization was performed by the use of a (selection bias) random numbers table"

Correspondence: "We designed a list that numbered from 100 to 180. Then an external person involved assigning a letter (A, AB, B, and BA) to the each number randomly. The entire investigator was blinded about number and letters. As patients enrolled for study, the external person enters the patient's code in the list. In summary, we used an external person to allocate patient to the intervention or control group"

Comment: third-party randomisation should have ensured that the allocation sequence was not manipulated Blinding of participants High risk Not possible to blind participants and personnel to allocated and personnel groups (performance bias) Blinding of outcome High risk The subjective elements in the scale used to measure oral assessment (detection mucositis could have introduced bias into the assessments bias) Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported appropriately (reporting bias) Other bias Low risk No other sources of bias are apparent

Kakoei 2013

316 Methods Trial design: parallel (2 arms) Location: Kerman, Iran Number of centres: 1 Study duration: not reported Participants Inclusion criteria: partial or complete exposure of head and neck to radiation; minimum radiation dose of 2500 to 3000 cGy (trial registry says Gy) per session; beginning radiotherapy at the start of the study and continuing constantly for the following 2 weeks Exclusion criteria: existing oral mucositis; systemic disease or medication affecting oral condition; less than 15 or more than 55 years of age Cancer type: head and neck (not reported by group) Cancer treatment: radiotherapy to the head and neck Any other potentially important prognostic factors: no statistically significant difference between groups in smoking status or education level Age at baseline (years): Gp A: mean 42.9 (SD 14.9); Gp B: mean 49.1 (SD 15.4) Gender: 57.5% male overall and reports that there were no statistically significant differences between groups Number randomised: 40 (Gp A: 20; Gp B: 20) Number evaluated: 40 (Gp A: 20; Gp B: 20) Interventions Comparison: cryotherapy versus no treatment Gp A: ice cubes placed in mouth and sucked for 5 min before radiotherapy and for a further 5 min after the session Gp B: standard oral care Both groups received standard oral care (use of soft toothbrush, nonabrasive toothpaste and dental floss twice daily) Compliance: only states "no lapse during the study" Duration of treatment (intended): 10 min per day for 2 weeks Outcomes  Oral mucositis: 0 to 4 scale assessed separately by participants and clinicians (very similar to WHO scale and NCI common toxicity criteria - based on methods of Mahood 1991) (assessed on days 1, 7 and 14 and reported as a mean on each separate assessment day)  Oral pain: scale not mentioned (assessed days 1, 7 and 14 and reported as a mean on each separate assessment day) Notes Sample size calculation: 80% power at 5% significance level to detect a 40% difference in treatment effect (as there were no drop-outs, it is assumed that this was achieved) Adverse effects: not reported but presumably none were expected due to the 5-min periods of cryotherapy Funding: "This study was financially supported by the Office of Vice Chancellor for Research of Kerman University of Medical Sciences" Declarations/conflicts of interest: not reported Data handling by review authors: we report MD and 95% CI for mucositis severity in an additional table; physician-judged mucositis rating was preferred over participant judgement as we deemed that this may be more objective and less biased; we used the data on day 14 due to the highest control group mean; we were unable to use the oral pain data as the scale was not described Other information of note: not reported

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "The participants were divided into experimental and generation (selection control groups using block randomization technique with the bias) formula of AABB, ABAB, ABBA, BBAA, BABA, and

317 BAAB"

Comment: random sequence appears to have been adequately generated Allocation concealment Unclear Quote: "The participants were divided into experimental and (selection bias) risk control groups using block randomization technique with the formula of AABB, ABAB, ABBA, BBAA, BABA, and BAAB"

Comment: insufficient information to determine whether or not the random sequence was adequately concealed Blinding of participants High risk Not possible to blind participants and personnel to allocated and personnel groups (performance bias) Blinding of outcome High risk The subjective elements in the scale used to measure oral assessment (detection mucositis could have introduced bias into the assessments bias) Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting High risk Oral pain not mentioned in the trial registry record and not (reporting bias) described in the methods of the published trial report. It is possible that this was reported because it showed favourable results for cryotherapy Other bias Low risk No other sources of bias are apparent

Katranci 2012

Methods Trial design: parallel (2 arms) Location: Gaziantep, Turkey Number of centres: 1 Study duration: not reported Participants Inclusion criteria: due to receive first course of chemotherapy; healthy oral mucosa; no dental problems Exclusion criteria: receiving more than 1 combination chemotherapy course or antineoplastic drug treatment with half-life of 30 min or more; discomfort in the mouth; head-neck cancer Cancer type: gastric 33.3%; colon 33.3%; rectal 16.9%; pancreatic 9.9%; unknown 6.6% (equal numbers per group); stage of disease equally distributed between groups Cancer treatment: 5FU and LV Any other potentially important prognostic factors: education level, denture wearers, toothbrushing habits, smoking status, nutrition, dry mouth, lack of appetite and systemic disease all equally distributed between groups Age at baseline (years): not reported Gender: 50% male in both groups Number randomised: 60 (Gp A: 30; Gp B: 30) Number evaluated: 60 (Gp A: 30; Gp B: 30) Interventions Comparison: cryotherapy versus routine care Gp A: ice chips placed in mouth 5 min before 5FU + LV, during treatment and within 15 min after treatment, for total 30 min; continuously swished around, then replenished before the previous ice had completely melted; whole procedure repeated for 5 consecutive days Gp B: routine care All participants were asked to remove dentures

318 Compliance: "Oral cryotherapy was tolerated well by the patients. The majority of the patients reported that they managed to keep the oral cavity constantly cooled most of the time that the chemotherapy treatment was administered. Patients who experienced discomfort during the cryotherapy application continued their treatment after a maximum 30-60 s break" Duration of treatment (intended): 30 min per day for 5 days (first cycle only) Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed and reported on days 7, 14 and 21 - i.e. not reported as maximum score per participant) Notes Sample size calculation: not reported Adverse effects: "The patients completed the procedure quite comfortably, without any problems during the application" Funding: not reported Declarations/conflicts of interest: "None declared" Data handling by review authors: data were reported separately on the 3 assessment days rather than a maximum score per person over the whole assessment period. We used the data on the day with the highest incidence of grades > 0 (day 14), although we are not sure how valid this is because there was still a high incidence of severe oral mucositis at day 21 Other information of note: not reported

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "Randomisation was performed by MedCalc software generation (selection to give equal chance to assign each intervention group" bias) Comment: computer generated randomisation so probably done adequately Allocation concealment Unclear Quote: "Randomisation was performed by MedCalc software (selection bias) risk to give equal chance to assign each intervention group"

Comment: insufficient information to determine whether or not the random sequence was adequately concealed Blinding of participants High risk Not possible to blind participants and personnel to allocated and personnel groups (performance bias) Blinding of outcome High risk The subjective elements in the scale used to measure oral assessment (detection mucositis could have introduced bias into the assessments bias) Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported appropriately (reporting bias) Other bias Low risk No other sources of bias are apparent

Lilleby 2006

Methods Trial design: parallel (2 arms) Location: Seattle, USA Number of centres: 1 Study duration: August 2003 to June 2005 Participants Inclusion criteria: minimum age 18 years with multiple myeloma; scheduled to 319 receive single-agent high-dose melphalan (200 mg/m2) followed by autologous PBSCT 2 days later Exclusion criteria: previous autologous PBSCT Cancer type: haematological (multiple myeloma) Cancer treatment: high-dose melphalan (200 mg/m2) followed by autologous PBSCT Any other potentially important prognostic factors: not reported Age at baseline (years): Gp A: median 59 (range 51 to 71); Gp B: median 57 (range 33 to 72) Gender: Gp A: 76.2% male; Gp B: 63.2% male Number randomised: 41 (Gp A: 21; Gp B: 20) Number evaluated: 40 (Gp A: 21; Gp B: 19) (above figures for age and gender are for evaluated participants) Interventions Comparison: cryotherapy versus saline rinse Gp A: 2 days before stem cell infusion, 1 ounce of ice chips held in mouth 30 min prior to beginning single-agent high-dose melphalan (200 mg/m2) infusion, replenished when melted, procedure continued for 6 hours after the end of the 30- min melphalan infusion Gp B: 2 days before stem cell infusion, 1 ounce of room temperature normal saline swished around the mouth and spat out 30 min prior to beginning single-agent high- dose melphalan (200 mg/m2) infusion, procedure repeated every 30 min for 6 hours after the end of the 30-min melphalan infusion All participants instructed not to eat or drink anything extremely hot or cold during cryotherapy/saline treatment Compliance: 14 participants had at least 5 hours of cryotherapy, and 2 had at least 2 hours, whilst 5 did not report the duration. Some participants stopped using the ice chips due to their coldness. Average frequency of use: cryotherapy: 1 cup/hour; saline: 1 to 4 rinses/hour Duration of treatment (intended): 7 hours Outcomes  Oral mucositis: NCI-CTC 0 to 4 scale (assessed until 30 days after cryotherapy/saline administration, maximum score reported)  Normalcy of diet: duration of TPN (assessed until 30 days after cryotherapy/saline administration)  Duration of IV narcotic use (assessed until 30 days after cryotherapy/saline administration)  Duration of hospitalisation (assessed until 30 days after cryotherapy/saline administration)  Weight loss (not an outcome of this review)  First day 30% of calorific needs met (not an outcome of this review)

Patient-reported events:  Mouth and throat pain: 0 to 10 scale (assessed daily by questionnaire until 30 days after cryotherapy/saline administration)  Adverse effects of cancer treatment: difficulties swallowing, drinking, eating, talking, sleeping and taste disturbance (not outcomes of this review) Notes Sample size calculation: required sample size was achieved ("sample size of 40 was chosen to provide 91% power to observe a statistically significant difference (at the 2-sided significance level of 0.05) in the probability of grades 3–4 mucositis under the assumption that the true probabilities of severe mucositis are 0.25 for patients receiving ice chips and 0.75 for patients receiving normal saline") Adverse effects: not reported, only refers to some participants that stopped using the ice chips due to their coldness Funding: "This work was supported by Friends of Jose Carreras International Leukemia Foundation Presidential Award, NCI P01 CA-18029" Declarations/conflicts of interest: not reported Data handling by review authors: for patient-reported oral pain, we used the

320 mean, number of participants and P value to calculate a single SD to be used for both groups (we used the overall mean pain scores rather than the number of days of pain or the mean of the highest value); only medians and ranges presented for the outcomes days of TPN, IV narcotics and hospitalisation, so we present the results, as reported in the study report, in an additional table Other information of note: not reported

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "patients...were randomized to receive either ice chips generation (selection or room temperature normal saline rinses" bias) Correspondence: "randomisation was accomplished using a computer program designed by one of our statisticians"

Comment: adequate method of random sequence generation Allocation concealment Low risk Quote: "patients...were randomized to receive either ice chips (selection bias) or room temperature normal saline rinses"

Correspondence: "Central randomization from the protocol office"

Comment: central randomisation should have ensured that the allocation sequence was not manipulated Blinding of participants High risk Not possible to blind participants and personnel to allocated and personnel groups (performance bias) Blinding of outcome High risk The subjective elements in the scale used to measure oral assessment (detection mucositis could have introduced bias into the assessments bias) Incomplete outcome Low risk Only 1 randomised participant, from the control group, was data (attrition bias) not included in the analyses. The participant withdrew consent because he wanted to use ice chips Selective reporting Low risk Although the data for some outcomes of this review were not (reporting bias) presented in a way amenable to meta-analysis, this is unlikely to be due to bias Other bias Low risk No other sources of bias are apparent

Mahood 1991

Methods Trial design: cross-over (2 arms) Location: USA Number of centres: unclear (multicentre) Study duration: not reported Participants Inclusion criteria: first ever course of chemotherapy Exclusion criteria: not reported Cancer type: not reported but must be solid due to chemotherapy regimen Cancer treatment: 5FU and LV Any other potentially important prognostic factors: not reported Age at baseline (years): not reported Gender: not reported Number randomised: 95 (Gp A: 50; Gp B: 45) 321 Number evaluated: 93 (Gp A: 50; Gp B: 43) Interventions Comparison: cryotherapy versus no treatment Gp A: ice chips placed in mouth 5 min before receiving 5FU (425 mg/m2) and LV (20 mg/m2) by IV over a few minutes, and continuously swished around, then replenished before the previous ice had completely melted, for total 30 min, whole procedure repeated for 5 consecutive days Gp B: 5FU (425 mg/m2) and LV (20 mg/m2) only for 5 consecutive days

All participants were asked to remove dentures Compliance: not clearly reported. Only states "well tolerated by most patients" Duration of treatment (intended): 30 min per day for 5 days (first cycle only) Outcomes  Oral mucositis: physician judgement and participant judgement both on a 0 to 4 scale (very similar to WHO scale and NCI common toxicity criteria) (physician's judgement assessed by historical means approximately 1 month after treatment initiation, maximum score reported)  Duration of oral mucositis (not an outcome of this review) Notes Sample size calculation: not reported Adverse effects: "A few patients noted mild, temporary mouth numbness or an "ice cream" headache which rapidly resolved after cessation of cryotherapy. Also, some patients ascribed nausea to the oral ice chips (the nausea may have actually been from the 5FU)" Funding: "supported in part by Public Health Service grants...and Community Clinical Oncology Program grants" Declarations/conflicts of interest: not reported Data handling by review authors: we only used the data from the first treatment cycle (rather than cross-over data); physician judgement of oral mucositis was preferred over participant judgement as we deemed that this may be more objective and less biased Other information of note: mean oral mucositis score reported by smoking status for each group for the first cycle only (smokers had statistically significantly lower mean oral mucositis score than non-smokers, but data not available for all participants)

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear Quote: "Prior to therapy, patients were stratified by age and generation (selection risk whether or not they had dentures. They were then bias) randomized to a control arm or to receive cryotherapy"

Comment: insufficient information to determine method of random sequence generation Allocation concealment Unclear Quote: "Prior to therapy, patients were stratified by age and (selection bias) risk whether or not they had dentures. They were then randomized to a control arm or to receive cryotherapy"

322 bias) Incomplete outcome Low risk Only 2 randomised participants, both from the control group, data (attrition bias) were not included in the physician-judged oral mucositis analyses. We do not believe that this could pose a risk of bias significant enough to have led to a distortion of the true intervention effect Selective reporting Low risk Data for outcomes of this review were reported appropriately (reporting bias) Other bias Low risk No other sources of bias are apparent

Rocke 1993

Methods Trial design: cross-over (2 arms) Location: USA Number of centres: unclear (multicentre) Study duration: not reported Participants Inclusion criteria: first course of chemotherapy Exclusion criteria: not reported Cancer type: not reported but must be solid due to chemotherapy regimen Cancer treatment: 5FU and LV, different dosages taken orally or by IV, equally distributed between groups due to stratification Any other potentially important prognostic factors: participants were stratified by smoking status, dentures and institution/centre; gum condition not used for stratification but Gp B (60 min cryotherapy) had worse gums at baseline (P = 0.02) Age at baseline (years): Gp A: median 65 (range 24 to 79); Gp B: median 65 (range 25 to 85) Gender: Gp A: 55% male; Gp B: 51% male Number randomised: 179 (Gp A: 90; Gp B: 89) Number evaluated: 178 (Gp A: 89; Gp B: 89) Interventions Comparison: 30 min of cryotherapy versus 60 min of cryotherapy Gp A: ice chips placed in mouth 5 min before receiving 5FU and LV, and continuously swished around, then replenished before the previous ice had completely melted, for total 30 min, whole procedure repeated for 5 consecutive days Gp B: as above but for total 60 min All participants were asked to remove dentures Compliance: well tolerated with high degree of compliance. Only a few participants stopped cryotherapy early (due to nausea, headache or chills). Many participants in the 60-min group were unhappy with the duration of cryotherapy, indicating that 30 min of cryotherapy is preferred Duration of treatment (intended): 30 or 60 min per day for 5 days (first cycle only) Outcomes  Oral mucositis: physician judgement and participant judgement both on a 0 to 4 scale (very similar to WHO scale and NCI common toxicity criteria) (physician's judgement assessed by historical means approximately 1 month after treatment initiation, maximum score reported)  Duration of oral mucositis (not an outcome of this review) Notes Sample size calculation: not reported Adverse effects: only a few participants stopped cryotherapy early (due to nausea, headache or chills) Funding: "supported in part by Public Health Service grants...from the National Cancer Institute, Department of Health and Human Services" Declarations/conflicts of interest: not reported Data handling by review authors: we only used the data from the first treatment cycle (rather than cross-over data); physician judgement of oral mucositis was 323 preferred over participant judgement as we deemed that this may be more objective and less biased Other information of note: in exploratory analyses, age over 65 years was statistically significantly (P < 0.001) associated with severity of mucositis

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear Quote: "Patients were randomized to receive cryotherapy for generation (selection risk either 30 or 60 minutes" bias) Comment: insufficient information to determine method of random sequence generation Allocation concealment Unclear Quote: "Patients were randomized to receive cryotherapy for (selection bias) risk either 30 or 60 minutes"

Comment: insufficient information to determine whether or not the random sequence was adequately concealed Blinding of participants High risk Not possible to blind participants and personnel to allocated and personnel groups (performance bias) Blinding of outcome High risk The subjective elements in the scale used to measure oral assessment (detection mucositis could have introduced bias into the assessments bias) Incomplete outcome Low risk Only 1 randomised participant, from the cryotherapy group, data (attrition bias) was not included in the analyses due to an unrelated medical problem Selective reporting Low risk Data for outcomes of this review were reported appropriately (reporting bias) Other bias Low risk No other sources of bias are apparent

Salvador 2012

Methods Trial design: parallel (2 arms) Location: Toronto, Canada Number of centres: 1 Study duration: May to December 2007 Participants Inclusion criteria: minimum 18 years of age; able to read and understand instructions on oral cryotherapy and basic oral care in English; diagnosis of multiple myeloma; due to receive high-dose melphalan (200 mg/m2); due to receive growth factors as part of treatment protocol; no pre-existing oral disease Exclusion criteria: previous radiotherapy to the head and neck region; amyloidosis (when abnormal proteins collect together and build up in tissues/organs) involving the heart, kidneys, or tongue; receiving investigational drugs during the trial period Cancer type: haematological (multiple myeloma) Cancer treatment: high-dose melphalan (200 mg/m2) followed by autologous SCT Any other potentially important prognostic factors: college/university educated: Gp A: 43%; Gp B: 16%; smokers: Gp A: 17%; Gp B: 9%; drinks alcohol: Gp A: 35%; Gp B: 27% Age at baseline (years): Gp A: mean 56 (SD 8.9) (range 43 to 72); Gp B: mean 62 (SD 7.7) (range 43 to 72) Gender: Gp A: 61% male; Gp B: 55% male Number randomised: 46 (Gp A: 23; Gp B: 23) 324 Number evaluated: 45 (Gp A: 23; Gp B: 22) Interventions Comparison: cryotherapy versus no treatment Gp A: on day -1 (the day before autologous SCT), ice chips held in mouth 5 min prior to receiving high-dose melphalan (200 mg/m2), replenished before melted, procedure continued until 5 min after melphalan infusion, for total 60 min. Basic oral care also received (described below) from day -1 Gp B: on day -1, basic oral care began (described below) All participants received high-dose melphalan on day -1 and SCT the following day (day 0). On average, participants stayed in hospital for 14 days after SCT. During the engraftment period, patients also received similar treatment and care protocols: prophylactic antimicrobial, antiviral and antacid (day 1), and growth factor (granulocyte colony-stimulating factor) (day 7); and basic oral care protocol (day - 1). Oral care protocol consisted of regular oral assessment and documentation, patient education on OM, and training in oral self care (toothbrushing with Toothette sponges dipped in sodium bicarbonate mouthwash, mouthrinsing with sodium bicarbonate mouthwash, and application of moisturiser to lips and oral cavity). Compliance: all participants were able to perform the basic oral care procedures and all participants in the cryotherapy group complied with the intervention Duration of treatment (intended): basic oral care from day -1 to day 12 (14 days); oral cryotherapy for 60 min on day -1 Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed on days 3, 6, 9 and 12 and reported as a mean on each separate assessment day - not reported for day 3 as OM had not yet developed); we requested maximum score experienced per participant over the whole study period and the author provided this data  Oral pain: 0 to 10 VAS (assessed on days 3, 6, 9 and 12 and reported as an overall mean score)  Duration of hospital stay (mean number of days, measured as total number of days from admission to discharge - all participants were treated as inpatients)  Normalcy of diet: functional intake of food and fluids on 1 (solids) to 5 (nothing by mouth) scale (assessed on days 3, 6, 9 and 12 and reported in a mixed-effects regression analysis)  Time to onset, duration and time to resolution of oral mucositis (not an outcome of this review)  Amount of opioid analgesics used (not an outcome of this review) Notes Sample size calculation: 17 participants per group needed to detect an effect size of 1 (presumably the authors mean a MD of 1 on the WHO scale) at 80% power and 5% significance Adverse effects: 4 participants (17.4%) in the cryotherapy group experienced side effects (teeth sensitivity and chills) Funding: "The authors acknowledge the financial support of the University Health Network Nursing Research and Canadian Nurses Foundation for the successful implementation of the project" Declarations/conflicts of interest: "The authors have nothing to disclose and declare no conflicts of interest" Data handling by review authors: the authors provided the maximum score experienced per participant over the whole study period; we were unable to use the data on normalcy of diet Other information of note: although the authors report that the MD of 0.71 in OM severity (at day 9) was statistically significant, they state that it is not clinically significant because most participants only had OM grades of 0 to 1; conversely, the authors report that the difference in days of hospital stay was not statistically significant, but that the difference of approximately 1 day is clinically significant; we note an error in Table 2 where the mean in the control group at day 6 should be 0.5 rather than 0.05

Risk of bias table 325

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "Participants were allocated based on a randomization generation (selection process performed by the biostatistics staff at the study site, bias) using consecutively numbered sealed envelopes containing a computer-generated random number list patient assignment"

Comment: adequate method of random sequence generation Allocation concealment Low risk Quote: "Participants were allocated based on a randomization (selection bias) process performed by the biostatistics staff at the study site, using consecutively numbered sealed envelopes containing a computer-generated random number list patient assignment"

Comment: these methods should have ensured that the allocation sequence was not manipulated Blinding of participants High risk Not possible to blind participants and personnel to allocated and personnel groups (performance bias) Blinding of outcome High risk Although described as blinded, the subjective elements in the assessment (detection scale used to measure oral mucositis could have introduced bias) bias into the assessments Incomplete outcome Low risk Only 1 randomised participant, from the control group, was data (attrition bias) not included in the analyses due to an adverse event related to the melphalan on the day of its administration Selective reporting Low risk Data for outcomes of this review were reported appropriately (reporting bias) Other bias Low risk No other sources of bias are apparent

Sorensen 2008

Methods Trial design: parallel (3 arms) Location: Denmark Number of centres: unclear (appears to be multicentre) Study duration: 2001 to 2005 Participants Inclusion criteria: gastric or colorectal cancers; due to receive first course of chemotherapy; healthy oral mucosa Exclusion criteria: history of head and neck radiotherapy; symptoms of any infections; history of dental or mouth discomfort when consuming hot/cold food and drinks Cancer type: approximately 95% colorectal cancer in each group, with the remainder having gastric cancer Cancer treatment: 5FU and LV Any other potentially important prognostic factors: participants were stratified by age, smoking status and dentures; performance status score was equally distributed between groups Age at baseline (years): Gp A: 40 or older = 97%, median 62 (range 36 to 84); Gp B: 40 or older = 95%, median 61 (range 30 to 81); Gp C: 40 or older = 92%, median 62 (range 28 to 82) Gender: Gp A: 60% male; Gp B: 53% male; Gp C: 53% male Number randomised: 225 (Gp A: 75; Gp B: 75; Gp C: 75) Number evaluated: reported in the study: 206 (Gp A: 67; Gp B: 66; Gp C: 73); data available for OM incidence: 197 (Gp A: 63; Gp B: 64; Gp C: 70) Interventions Comparison: cryotherapy versus saline rinse (placebo version of 326 chlorhexidine) versus chlorhexidine rinse Gp A: crushed ice placed in mouth 10 min before receiving 5FU (425 mg/m2) and LV (20 mg/m2) by IV, and for 35 min after the start of infusion (total 45 min), whole procedure repeated for 5 consecutive days every 4 weeks Gp B: same cancer treatment but saline mouthwash (with same taste additives as Gp C), 10 ml swished around the mouth for 1 min, 3 times per day for 21 days Gp C: same cancer treatment but chlorhexidine 0.1% mouthwash, same schedule as Gp B All participants were asked to remove dentures; participants were "instructed to continue prophylaxis in case OM occurred, which was treated according to the discretion of the respective investigators" Compliance: (complete) Gp A: 87%; Gp B: 80%; Gp C: 75%; (partial) Gp A: 13%; Gp B: 20%; Gp C: 25% Duration of treatment (intended): 45 min per day for 5 days per chemotherapy cycle (cryotherapy); 3 min per day for 21 days per chemotherapy cycle (chlorhexidine and saline rinses) Outcomes  Oral mucositis: NCI-CTC (2.0) 0 to 4 scale (first cycle-only reported at day 28, participants kept daily record and self-evaluated OM by questionnaire on days 14 and 28, physician also scored OM on days 14 and 28, participant score was reported as there were no statistically significant differences between participant and physician scores, maximum score reported)  Duration of oral mucositis (not an outcome of this review)  Compliance (not an outcome of this review)  Haematologic toxicity (not an outcome of this review) Notes Sample size calculation: considering a 15% decrease in incidence of grade 3 to 4 OM as being clinically meaningful, with 80% power and at the 5% significance level, 75 patients required per group. Therefore, considering drop-outs, required sample size was not achieved Adverse effects: taste disturbance: Gp A: 24/67 (36%); Gp B: 25/66 (38%); Gp C: 35/73 (48%); headache: Gp A: 14/67 (21%); Gp B: 9/66 (14%); Gp C: 10/73 (14%) Funding: "Supported by a grant from the National University Hospitals Research Foundation in Denmark" Declarations/conflicts of interest: not reported Data handling by review authors: the study authors report that data were available on 206 participants, however, for OM incidence, some of this number of participants are not given a grade and are listed as 'NA' (abbreviation not explained). We subtracted the 'NA' participants from the total number of participants and have addressed this problem under 'Incomplete outcome data (attrition bias)' Other information of note: smoking status, performance status and being aged 40 or older (underpowered - only 11 participants less than 40) were not statistically significantly associated with severity of OM

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear Quote: "Patients were stratified...and randomized after generation (selection risk informed consent to 1 of 3 prophylactic regimens" bias) Comment: insufficient information to determine method of random sequence generation Allocation concealment Unclear Quote: "Patients were stratified...and randomized after (selection bias) risk informed consent to 1 of 3 prophylactic regimens"

Comment: insufficient information to determine whether or 327 not the random sequence was adequately concealed Blinding of participants High risk Not possible to blind participants and personnel to allocated and personnel groups (performance bias) Blinding of outcome High risk The subjective elements in the scale used to measure oral assessment (detection mucositis could have introduced bias into the assessments bias) Incomplete outcome High risk For OM incidence, 12% of randomised participants were not data (attrition bias) included in the analysis (Gp A: 16%; Gp B: 15%; Gp C: 9%). Reasons are not reported but there is a potential risk of bias if this differential drop-out is related to outcomes/prognosis Selective reporting Low risk Data for outcomes of this review were reported appropriately (reporting bias) Other bias Low risk No other sources of bias are apparent

Svanberg 2007

Methods Trial design: parallel (2 arms) Location: Uppsala, Sweden Number of centres: 1 Study duration: January 2002 to August 2004 Participants Inclusion criteria: more than 18 years of age; able to communicate in Swedish; about to receive BMT Exclusion criteria: not reported Cancer type: 1 participant per group had testicular cancer, the remainder had haematological cancers Cancer treatment: type of chemotherapy prior to BMT was fairly equally distributed between groups; participants were stratified by autologous BMT (both groups 79.5%) versus allogeneic URD BMT (both groups 20.5%) Any other potentially important prognostic factors: tobacco use equally distributed between groups (Gp A: 17.9%; Gp B: 15.4%) Age at baseline (years): Gp A: mean 49.8 (SD 14.4); Gp B: mean 54.3 (SD 11) Gender: Gp A: 66.7% male; Gp B: 48.7% male Number randomised: 78 (Gp A: 39; Gp B: 39) Number evaluated: 78 (Gp A: 39; Gp B: 39) Interventions Comparison: cryotherapy versus no treatment Gp A: prior to BMT (unclear how far in advance); ice chips placed in mouth or rinsing with ice-cold water starting 5 min before receiving chemotherapy by IV, and replenished for the duration of the chemotherapy session Gp B: chemotherapy prior to BMT All participants received standard oral care during BMT Compliance: 58% to 75% of the participants reported that they kept the mouth constantly cooled for the entire duration of the chemotherapy; 71% to 100% did so more than half the time Duration of treatment: not reported (variable and dependent on type of chemotherapy) Outcomes  Oral mucositis: OMAS (assessed once daily for 21 days after start of chemotherapy by nurse, mean score reported on each day, reported separately for autologous/allogeneic BMT but no overall score reported)  Oral mucositis: WHO 0 to 4 scale (incidence of severe mucositis i.e. grade 3 to 4, assessed once daily for 21 days after start of chemotherapy) (outcome reported in secondary publication); we requested maximum score experienced per participant over the whole study period and the author provided this data  Oral pain: 0 to 10 VAS (assessed twice daily (morning and afternoon) for 21

328 days after start of chemotherapy, reported verbally to nurse)  Duration of opioid use (mean number of days, assessed over period of 31 days after start of chemotherapy from medical records)  Duration of hospital stay (outcome reported in secondary publication) (mean number of days, assessed over period of 31 days after start of chemotherapy)  Normalcy of diet: incidence of TPN and duration of TPN (outcome reported in secondary publication) (we used mean number of days, assessed over period of 31 days after start of chemotherapy)  Weight loss (not an outcome of this review)  Duration of fever (not an outcome of this review)  Antibiotic use (not an outcome of this review)  Total dose of opioids (not an outcome of this review)  White blood cell counts and c-reactive protein levels (not outcomes of this review) Notes Sample size calculation: 36 participants per group needed for 80% power at 5% significance level (based on outcome 'duration of IV opioids') Adverse effects: 7 (18%) of participants found cryotherapy unpleasant, with 4 of those (10%) finding it very unpleasant due to shooting pain from the teeth Funding: "This study was in part supported by FoU funds, Uppsala läns landsting, Sweden" Declarations/conflicts of interest: not reported Data handling by review authors: we were unable to use the OMAS score but the authors provided full data for oral mucositis on the WHO scale; oral pain was not reported adequately so we were unable to report on this outcome Other information of note: not reported

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "A random assignment to EXP or CTR group was generation (selection performed in blocks of six outside of the clinic by an bias) independent researcher"

Comment: if using block-randomisation, it could be assumed that the independent researcher would have done this adequately Allocation concealment Low risk Quote: "A random assignment to EXP or CTR group was (selection bias) performed in blocks of six outside of the clinic by an independent researcher"

Comment: third-party randomisation should have ensured that the allocation sequence was not manipulated Blinding of participants High risk Not possible to blind participants and personnel to allocated and personnel groups (performance bias) Blinding of outcome High risk The subjective elements in the scale used to measure oral assessment (detection mucositis could have introduced bias into the assessments bias) Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting High risk Oral mucositis using OMAS and oral pain inadequately (reporting bias) reported. Data are presented in the text for mucositis scores on day 10 (autologous patients) and day 16 (allogeneic

329 patients), and this appears to be based on statistical significance. The study authors have since provided full data on the WHO scale. Therefore, our rating for this domain is based on the fact that the text only states that there was no significant differences for oral pain Other bias Low risk 45 staff members assessed mucositis using the OMAS instrument. Calibration is not mentioned. However, we were unable to use any data for this outcome so we can discount any potential bias

Toro 2013

Methods Trial design: parallel (3 arms) Location: San Antonio, Texas, USA Number of centres: 1 Study duration: first subject randomised August 2009; last subject randomised January 2013 Participants Inclusion criteria: minimum 18 years of age; diagnosis of multiple myeloma and scheduled to receive high-dose melphalan (70 to 100 mg/m2/day) for 2 days, as a single agent, for conditioning regimen prior to HSCT Exclusion criteria: received palifermin (Kepivance) in the past 30 days; received any investigational drug in the past 30 days; received radiation therapy in the past 30 days; oral mucositis at the time of randomisation; altered mental status precluding understanding of the informed consent process and/or completion of the necessary assessments Cancer type: haematological (multiple myeloma) Cancer treatment: high-dose melphalan, prior to autologous HSCT Any other potentially important prognostic factors: no differences between groups in race/ethnicity, Karnofsky performance score, diabetes, denture wearing, smoking status Age at baseline (years): Gp A: median 62 (range 39 to 75); Gp B: median 61.5 (range 43 to 70) Gender: both groups 95% male Number randomised: 78 (Gp A: 40; Gp B: 38) (numbers are for the 2 groups of interest to this review) Number evaluated: 78 (Gp A: 40; Gp B: 38) Interventions Comparison: cryotherapy plus saline rinse versus saline rinse The 3rd group involved using supersaturated calcium phosphate rinse (Caphosol®) but we have excluded this arm because the co-intervention of cryotherapy plus saline may confound the comparison. The comparison of this 3rd group with the saline group will be included in our review of basic oral care interventions for the prevention of oral mucositis in cancer patients Gp A: on day -2 and -1 (HSCT was on day 0), approximately 1 ounce of crushed ice held in the mouth 15 min prior to the initiation of melphalan infusion, replenished as soon as it had completely melted, this procedure continued during the melphalan infusion and for 90 min after the end of the infusion. After completion of cryotherapy, standard care was followed until the end of the study (see Gp B below) Gp B: standard care (0.9% sodium chloride irrigation solution): rinsing of mouth twice, with 1 ounce (30 ml) of room temperature 0.9% NaCl (normal saline), 4 times daily after admission and until end of study Compliance: 100% compliance with cryotherapy but 3 subjects from the saline group refused to follow protocol but were kept in study Duration of treatment (intended): 2 consecutive days (exact duration of cryotherapy sessions unclear) Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed daily until resolution or hospital 330 discharge up to a maximum of 30 days, maximum score reported)  Oral pain: scale not reported  Narcotic use: unclear whether or not reported as duration of opioid use or amount of opioid use  Quality of life: Patient-Reported Oral Mucositis Symptom (PROMS) scale (assessed daily until resolution of oral mucositis)  Duration of oral mucositis (not an outcome of this review) Notes Sample size calculation: "This study will achieve a power of 80% to detect a 0.30 reduction in the proportion of subjects experiencing mucositis relative to saline with 55 subjects per arm and relative to Caphosol with 43 subjects per arm, using 2- sided pairwise Pearson chi-square testing with the Bonferroni corrected significance level of 0.017 (PASS, Version 08.0.8, NCSS Kaysville, Utah 2008). Assuming no loss to follow-up and complete data, the required sample size for this study is therefore 55 subjects per arm, giving a total required sample size of 165 subjects. The study was terminated early because of ethical concerns after an interim analysis" Adverse effects: no serious adverse events Funding: "There was no outside funding for this study" Declarations/conflicts of interest: "No conflict of interest for all authors" Data handling by review authors: the oral mucositis incidence of each WHO grade was presented as percentages so we converted this to numbers of persons experiencing each grade Other information of note: only abstracts available but the authors have provided extra information through email correspondence to allow inclusion in this review. Study report is currently being written up and secondary outcome data have not yet been analysed, so we will include these in the next update of the review

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "were randomized to the above mentioned groups" generation (selection bias) Correspondence: "The computer program used to generate these sheets makes block stratified assignments with user selected block size"

Comment: adequate method of random sequence generation Allocation concealment Low risk Quote: "were randomized to the above mentioned groups" (selection bias) Correspondence: "The random sequence list was kept at the principal investigator's office and only when a subject had signed the informed consent and agrees to participate in any of the three arms of the study, was the research co-ordinator able to look at the list and assigned the treatment"

Comment: these methods should have ensured that the allocation sequence was not manipulated Blinding of participants High risk Not possible to blind participants and personnel to allocated and personnel groups (performance bias) Blinding of outcome High risk The subjective elements in the scale used to measure oral assessment (detection mucositis could have introduced bias into the assessments bias) Incomplete outcome Low risk All randomised participants were included in the analyses 331 data (attrition bias) Selective reporting Low risk This domain is not yet applicable as the study report is (reporting bias) currently being written up Other bias Low risk No other sources of bias are apparent

Zhang 2011

Methods Trial design: parallel (4 arms) Location: Beijing, China Number of centres: 1 Study duration: June 2009 to May 2010 Participants Inclusion criteria: patients due to receive high-dose MTX Exclusion criteria: not reported Cancer type: osteosarcoma (bone cancer) Cancer treatment: MTX plus vincristine plus LV Any other potentially important prognostic factors: not reported Age at baseline (years): not reported Gender: not reported Number randomised: 147 (Gp A: 52; Gp B: 66; Gp C: 29) (numbers are for the 3 groups of interest to this review) Number evaluated: 147 (Gp A: 52; Gp B: 66; Gp C: 29) Interventions Comparison: cryotherapy versus LV versus high dose LV The 4th group involved using the standard LV dose once OM symptoms appeared. This constitutes treatment rather than prevention and therefore we excluded this arm Gp A: from the start of MTX, 100 ml ice water to be used for gargling (amount per gargle and frequency of gargling not specified), for 3 consecutive days Gp B: from the start of MTX, 3 mg LV dissolved in 100 ml water to be used for gargling per day, for 3 consecutive days (amount per gargle not specified) Gp C: from the start of MTX, 200 mg LV dissolved in 40 ml water to be used for gargling per day (10 ml 4 times daily), for 3 consecutive days All participants received MTX (10 g/m2 administered by IV over 4 to 6 hours), vincristine (2 mg), and LV (12 mg every 6 hours, beginning 6 to 8 hours after finishing the MTX IV drop - it is not clear from the translation of the paper how long this occurred for or indeed if was just a single dose 6 to 8 hour after MTX). From 1 day prior to the start of MTX, participants had diuresis and alkalinising for 3 consecutive days, plus oral sodium bicarbonate (1 g) and allopurinol (200 mg) both 3 times daily for 4 consecutive days Compliance: not reported Duration of treatment (intended): 3 days (actual length of time the ice water or LV was held in the mouth is not reported) Outcomes  Oral mucositis: WHO and NCI-CTC (3.0) 0 to 4 scale (assessed 1 and 10 days after start of chemotherapy - reported as incidence of any OM over the study period and also by severity on the 4th day after MTX)  Duration of oral mucositis (not an outcome of this review)  MTX concentration in saliva (not an outcome of this review) Notes Sample size calculation: not reported Adverse effects: not reported Funding: not reported/not obtained from translation Declarations/conflicts of interest: not reported/not obtained from translation Data handling by review authors: we have used the data from Table 1 for the analysis of no mucositis versus any mucositis as this appears to be the incidence of any mucositis over the whole study period, and is therefore not at risk of bias; we have used the data on severity at day 4 but it may be at risk of bias due to selective reporting and readers should take this into consideration when interpreting the

332 results Other information of note: it is unclear why the authors report OM severity at day 4, after they state that symptoms usually occur 5 to 7 days after chemotherapy. No participants had severe OM (grades 3 or 4) on day 4. If any participants developed severe OM after day 4, then the study will not reflect this and the incidence of severe OM in this type of study will have been underestimated. We would recommend that authors report the maximum grade experienced by each participant in future publications, or at least report a range of appropriate time points

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "patients were randomly allocated to 4 groups using a generation (selection random number table" bias) Comment: random sequence appears to have been adequately generated Allocation concealment Unclear Quote: "patients were randomly allocated to 4 groups using a (selection bias) risk random number table"

Blinding of outcome High risk The subjective elements in the scales used to measure oral assessment (detection mucositis could have introduced bias into the assessments bias) Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting High risk Oral mucositis was measured on the day of chemotherapy (reporting bias) and on day 10, but severity is reported on day 4 Other bias Low risk No other sources of bias are apparent

Footnotes

5FU = fluorouracil; autologous = patients' own cells; allogeneic = cells from donor; BMI = body mass index; BMT = bone marrow transplantation; CI = confidence interval; EOCG = Eastern Oncology Co-operative Group; GVHD = graft-versus-host disease; HSCT = haematopoietic stem cell transplantation; IFN = interferon; IV = intravenous; LV = leucovorin; MD = mean difference; MTX = methotrexate; NCI-CTC = National Cancer Institute common toxicity criteria; OM = oral mucositis; OMAS = oral mucositis assessment scale; PBSCT = peripheral blood stem cell transplantation; SCT = stem cell transplantation; SD = standard deviation; TPN = total parenteral nutrition; URD = unrelated donor; VAS = visual analogue scale; VP16 = vepesid/etoposide; WHO = World Health Organization

Characteristics of excluded studies

Aisa 2005

Reason for exclusion Not RCT (historical controls) 333

Baydar 2005

Reason for exclusion Cross-over study. First-period data not presented

Castelino 2011

Reason for exclusion Plain versus flavoured ice. Cross-over study. First-period data not presented de Paula Eduardo 2014

Reason for exclusion Cryotherapy plus laser versus laser. We cannot exclude the possibility of interaction between interventions and therefore we would not be confident in stating that any improved/reduced effect is due to the cryotherapy

Karagozoglu 2005

Reason for exclusion Not RCT (participants were allocated by alternation)

Mori 2006

Reason for exclusion Not RCT (historical controls)

Nikoletti 2005

Reason for exclusion Plain versus flavoured ice versus standard care. Cross-over study. First-period data not presented

Ohyama 1994

Reason for exclusion Translated from Japanese: case series of 5 cancer patients

Papadeas 2007

Reason for exclusion Not RCT (allocation based on date of birth)

Sato 1997

Reason for exclusion Unclear if RCT. Author was contacted during previous update of this Cochrane review but we received no response

Sato 2006

Reason for exclusion Not RCT (no mention of random allocation to groups)

Footnotes

RCT = randomised controlled trial

334 Characteristics of studies awaiting classification

CTRI/2013/08/003906

Methods Trial design: parallel (2 arms) Location: Chandigarh, India Number of centres: 1 Study duration: 6 months (dates unclear from trials registry) Participants Inclusion criteria: aged from 18 to 80 years; males or females newly diagnosed with head and neck cancer; radical radiotherapy planned; normal mucosa Exclusion criteria: palliative radiotherapy planned; coming for booster dose of radiation Cancer type: head and neck Cancer treatment: radiotherapy to the head and neck Any other potentially important prognostic factors: unclear from trials registry Age at baseline (years): unclear from trials registry Gender: unclear from trials registry Number randomised: 60 (Gp A: 30; Gp B: 30) Number evaluated: unclear from trials registry Interventions Comparison: cryotherapy versus no treatment Gp A: ice cubes held in mouth for 4 min before radiotherapy and for a further 4 min after the session Gp B: standard oral care All participants received standard oral care Compliance: unclear from trials registry Duration of treatment: unclear from trials registry Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed on days 5, 10, 15 and 20)  Tumour size and response (not an outcome of this review) Notes Emailed study investigators 29/06/2015 for publication details or full unpublished study data

Response: "This was an M.Sc dissertation by a student from the institute of nursing and I was one of the guides. I shall try and contact this student for complete details and then get back to you. With regards, Sushmita Ghoshal" Status: still awaiting further details and therefore unable to include these valuable data in the review

Lu 2013 (98)

Methods Trial design: parallel (2 arms) Location: Suzhou, China Number of centres: 1 Study duration: unclear from abstract Participants Inclusion criteria: unclear from abstract Exclusion criteria: unclear from abstract Cancer type: unclear from abstract Cancer treatment: unclear from abstract (followed by HSCT) Any other potentially important prognostic factors: unclear from abstract Age at baseline (years): unclear from abstract Gender: unclear from abstract Number randomised: 37 (number per group unclear from abstract) Number evaluated: unclear from abstract 335 Interventions Comparison: full cryotherapy versus partial cryotherapy Gp A: oral cryotherapy from the beginning of chemotherapy infusion until the end of infusion Gp B: oral cryotherapy starting half way through chemotherapy infusion until the end of infusion All participants received basic oral care Compliance: unclear from abstract Duration of treatment: unclear from abstract Outcomes  Oral mucositis: NCI-CTC 0 to 4 scale  Duration of oral mucositis (not an outcome of this review) Notes Unable to obtain email addresses for study investigators. Awaiting publication of full trial report

NCT01653106

Methods Trial design: parallel (2 arms) Location: Columbus, Ohio, USA Number of centres: 1 Study duration: April 2012 to March 2015 Participants Inclusion criteria: minimum 18 years of age; diagnosed with multiple myeloma and due to receive autologous STC Exclusion criteria: any other medical condition (including mental illness or substance abuse) which may interfere with ability to give informed consent, co- operate, and participate in the study, or which may interfere with the interpretation of the results Cancer type: haematological (multiple myeloma) Cancer treatment: high-dose melphalan followed by autologous SCT Any other potentially important prognostic factors: unclear from trials registry Age at baseline (years): unclear from trials registry Gender: unclear from trials registry Number randomised: 146 (Gp A: 73; Gp B: 73) Number evaluated: unclear from trials registry Interventions Comparison: 2 hours of cryotherapy versus 6 hours of cryotherapy Gp A: 1 ounce of shaved ice placed in mouth starting 15 min before receiving chemotherapy, and replenished for 2 hours Gp B: 1 ounce of shaved ice placed in mouth starting 15 min before receiving chemotherapy, and replenished for 6 hours Compliance: unclear from trials registry Duration of treatment: 2 hours or 6 hours Outcomes  Oral mucositis: NCI-CTC (4.0) 0 to 4 scale (incidence of severe mucositis i.e. grade 3 to 4, assessed over 21 days)  Melphalan pharmacokinetic modelling (not an outcome of this review) Notes Emailed study investigators 24/09/2015 for publication details or full unpublished study data

Robenolt 2010 (238)

Methods Trial design: parallel (2 arms) Location: USA Number of centres: unclear from abstract Study duration: unclear from abstract Participants Inclusion criteria: unclear from abstract Exclusion criteria: unclear from abstract Cancer type: unclear from abstract 336 Cancer treatment: high-dose melphalan (> 140 mg/m2) either alone or as part of the BEAM regimen, prior to autologous HSCT Any other potentially important prognostic factors: unclear from abstract Age at baseline (years): unclear from abstract Gender: unclear from abstract Number randomised: 40 (number per group unclear from abstract) Number evaluated: unclear from abstract Interventions Comparison: cryotherapy versus no treatment Gp A: oral cryotherapy 5 min before the start of chemotherapy infusion, continuing until 30 min after the completion of infusion Gp B: chemotherapy only Compliance: unclear from abstract Duration of treatment: unclear from abstract Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed over 14 days)  Oral pain: scale not reported (assessed over 14 days)  Patient functioning: swallowing, eating, and talking (not outcomes of this review) Notes Emailed study investigators 29/09/2015 for publication details or full unpublished study data

Xourafas 2009 (239)

Methods Trial design: parallel (2 arms) Location: unclear from abstract Number of centres: unclear from abstract Study duration: unclear from abstract Participants Inclusion criteria: minimum 18 years of age Exclusion criteria: unclear from abstract Cancer type: breast Cancer treatment: super intensive chemotherapy (various types) Any other potentially important prognostic factors: unclear from abstract Age at baseline (years): unclear from abstract Gender: females Number randomised: 150 (number per group unclear from abstract) Number evaluated: unclear from abstract Interventions Comparison: cryotherapy versus control (unclear from abstract) Gp A: details of regimen not described in abstract Gp B: details of regimen not described in abstract Compliance: unclear from abstract Duration of treatment (intended): unclear from abstract Outcomes  Duration of opioid use (fentanyl)  Oral pain (assessed verbally daily) Notes Emailed study investigators 29/09/2015 for publication details or full unpublished study data

Footnotes autologous = patients' own cells; HSCT = haematopoietic stem cell transplantation; NCI-CTC = National Cancer Institute common toxicity criteria; SCT = stem cell transplantation; WHO = World Health Organization

337 Characteristics of ongoing studies

NCT01789658

Study name A randomized controlled trial of cryotherapy for prevention and reduction of severity of oral mucositis in children undergoing hematopoietic stem cell transplantation Methods Trial design: parallel (2 arms) Location: unclear from trials registry Number of centres: unclear from trials registry Study duration: October 2012 to estimated September 2015 Participants Inclusion criteria: aged between 4 and 18 years; due to receive chemotherapy as conditioning treatment prior to HSCT in Sweden; sufficient knowledge of Swedish to understand the protocols Exclusion criteria: not reported Cancer type: haematological (multiple myeloma) Cancer treatment: chemotherapy prior to HSCT Any other potentially important prognostic factors: unclear from trials registry Age at baseline (years): 4 to 18 Gender: unclear from trials registry Number randomised: 50 (estimated enrolment - not reported by group) Number evaluated: unclear from trials registry Interventions Comparison: cryotherapy versus standard oral care Gp A: ice chips/ice cream/ice water used during chemotherapy infusion, replenished continuously (children receiving a 24-hour infusion instructed to use cryotherapy for 1 hour, 4 times per day) Gp B: standard oral care Compliance: unclear from trials registry Duration of treatment: unclear from trials registry Outcomes  Oral mucositis: 1) WHO 0 to 4 scale (assessed daily until HSCT, for expected average of 20 days); 2) ChiMES 0 to 23 scale (assessed daily by parent and child until HSCT, for expected average of 20 days)  Oral pain (assessed daily by nurse until HSCT, for expected average of 20 days)  Quality of life: emotional and psychological status using Beck youth inventories for depression and anxiety (assessed daily until HSCT, for expected average of 20 days)  Duration and dose of opioid use (during hospitalisation in connection with HSCT, for expected average of 25 days)  Normalcy of diet: duration of TPN (during hospitalisation in connection with HSCT, for expected average of 25 days)  Duration of hospital stay (in connection with HSCT)  Pain (general: not an outcome of this review)  Weight loss (not an outcome of this review)  Duration of antibiotic treatment (not an outcome of this review)  Duration of febrile neutropenia (not an outcome of this review)  C-reactive protein level (not an outcome of this review)  S-Albumin (not an outcome of this review) Starting date October 2012 Contact Gustaf Ljungman ([email protected]); Tove Kamsvåg Magnusson information ([email protected])

NCT02326675

338 Study name Randomized controlled, open-label study on the use of cryotherapy in the prevention of chemotherapy-induced mucositis in stem cell transplant patients Methods Trial design: parallel (2 arms) Location: Florida, USA Number of centres: unclear from trials registry Study duration: March 2015 to estimated March 2019 Participants Inclusion criteria: minimum 18 years of age; due to receive etoposide chemotherapy (minimum dose of 30 mg/kg) as conditioning treatment prior to autologous SCT Exclusion criteria: prior radiation to head and neck; known oropharynx involvement in malignancy; history of non-compliance or lack capacity to give informed consent Cancer type: unclear from trials registry Cancer treatment: etoposide (minimum dose of 30 mg/kg) prior to autologous SCT Any other potentially important prognostic factors: unclear from trials registry Age at baseline (years): unclear from trials registry Gender: unclear from trials registry Number randomised: 48 (estimated enrolment - not reported by group) Number evaluated: unclear from trials registry Interventions Comparison: cryotherapy versus no treatment Gp A: ice chips/other very cold or frozen food used starting 15 min before starting etoposide infusion for a 30-min period, then 3 saline rinses over 15-min period. This alternating cycle is repeated until 30 min after completion of etoposide infusion (total 150 min) Gp B: at start of etoposide infusion 3 saline rinses over 15-min period followed by 30-rest period (no treatment). This alternating cycle is repeated until 30 min after completion of etoposide infusion (total 150 min) All participants will receive standard oral care (saline rinses) Compliance: unclear from trials registry Duration of treatment: 150 min Outcomes  Oral mucositis: scale not reported (assessed starting on day +1 post-transplant until discharged from hospital, for expected average of 21 to 28 days)  Time to onset of oral mucositis (not an outcome of this review)  Duration of oral mucositis (not an outcome of this review)  Compliance: number of participants that tolerate cryotherapy during each infusion (not an outcome of this review but we will record the information in the 'Characteristics of included studies' tables) Starting date March 2015 Contact Christina Cline ([email protected]); Leslie Pettiford ([email protected]) information

Footnotes autologous = patients' own cells; ChiMES = Children's International Mucositis Evaluation Scale; HSCT = haematopoietic stem cell transplantation; SCT = stem cell transplantation; TPN = total parenteral nutrition; WHO = World Health Organization

339 APPENDIX 3: Oral mucositis measurement scales

Oral mucositis assessment scale (OMAS) (as reported in Freytes 2004 (113))

Oral cavity ulceration Oral cavity erythema Grade 0 no lesion Grade 0 none Grade 1 lesion < 1 cm2 Grade 1 not severe Grade 2 lesion 1 to 3 cm2 Grade 2 severe Grade 3 lesion > 3 cm2

Ulceration and erythema are measured at 9 different sites (upper and lower labial mucosa, right and left buccal mucosa, right and left lateral and ventral portions of the tongue, floor of the mouth, soft palate (fauces), and hard palate (gingival)) and summated to give a total mucositis score. Therefore score could range from 0 to 45.

Radiation Therapy Oncology Group (RTOG) scale (physician-rated objective gross score as reported in Chi 1995 (109))

Grade 0 none Grade 1 erythematous sores Grade 2 patchy mucositis (< 1/2 mucosa) Grade 3 confluent fibrinous mucositis (> 1/2 mucosa) Grade 4 haemorrhage and necrosis

Radiation Therapy Oncology Group (RTOG) scale (as reported in Wu 2009 (136))

Grade none 0 Grade injection; may experience mild pain not requiring analgesia 1 Grade patchy mucositis that may produce an inflammatory serosanguineous discharge; 2 may experience moderate pain requiring analgesia Grade confluent fibrinous mucositis; may include severe pain requiring narcosis 3 Grade ulceration, hemorrhage, or necrosis 4

Cancer and Leukemia Group B (CALGB) scale (as reported in Turhal 2000 (240))

Grade 0 no mucositis Grade 1 painless ulcers, erythema or mild soreness Grade 2 painful erythema, edema, or ulcers, but can eat Grade 3 painful erythema, edema, or ulcers, and cannot eat Grade 4 requires parenteral or enteral support

340 National Cancer Institute common toxicity criteria (NCI-CTC) (as reported in Dazzi 2003 (111))

Grade 0 no mucositis Grade 1 painless ulcers, erythema or mild soreness in the absence of lesions Grade 2 painful erythema, edema or ulcers, but can eat or swallow Grade 3 painful erythema, edema or ulcers requiring intravenous hydration Grade 4 severe ulceration or requires parenteral or enteral nutritional support

World Health Organization (WHO) scale (as reported in Vadhan-Raj 2010 (134))

Grade 0 no mucositis Grade 1 soreness ± erythema Grade 2 erythema, ulcers. Patient can swallow solid diet Grade 3 ulcers, extensive erythema. Patient cannot swallow solid diet Grade 4 mucositis to the extent that alimentation is not possible

341 APPENDIX 4: Oral cryotherapy review - study data not included in analyses

Outcome Study Description of data Cryotherapy Control Result WHO 0 to 4 scale; physician-judged rating used; day n = 20 n = 20 14 data used due to highest control group mean MD -0.25 (95% CI -0.72 to 0.22); P Kakoei Oral mucositis (should be noted that on days 1 and 7 the cryotherapy = 0.29 (there is no evidence of a 2013 mean 0.95 (SD mean 1.2 group mean is higher but with no statistically difference in mucositis severity) 0.58) (SD 0.89) significant difference) n = 40 n = 40 MD -1.93 (95% CI -2.37 to -1.49); Heydari 1 to 5 scale: 1 = never, 2 = 1 day of week, 3 = 2 to 3 P < 0.00001 (cryotherapy Oral pain 2012 days of week, 4 = most of week, 5 = 7 days of week mean 1.71 (SD mean 3.64 statistically significantly reduced 0.74) (SD 1.20) the duration of pain experience) n = 19 n = 21 P = 0.04 (cryotherapy statistically Lilleby Days of TPN (we use the P value quoted in the study Normalcy of diet median 5.5 significantly reduced the number of 2006 report) median 2 (range 0 to days of TPN) (range 0 to 15) 21) n = 19 n = 21 P = 0.11 (there is no evidence of a Number of days Lilleby Days of hospitalisation (we use the P value quoted in median 14 difference in the number of days of in hospital 2006 the study report) median 9 (range 0 to hospitalisation) (range 0 to 22) 30) Number of days n = 21 n = 19 P = 0.0003 (cryotherapy statistically Lilleby Days of IV narcotics (we use the P value quoted in of treatment with significantly reduced the number of 2006 the study report) opioid analgesics median 0 median 5.5 days of IV narcotics)

342 (range 0 to 10) (range 0 to 13)

Footnotes

CI = confidence interval; IV = intravenous; MD = mean difference; n = number of participants analysed; N/A = not applicable; SD = standard deviation; TPN = total parenteral nutrition

343 APPENDIX 5: Oral cryotherapy review - secondary outcome analyses

Analysis 1.4: Cryotherapy versus control; Outcome: Interruptions to cancer treatment

Analysis 1.5: Cryotherapy versus control; Outcome: Interruptions to cancer treatment (days of interruption) 344

Analysis 1.6: Cryotherapy versus control; Outcome: Oral pain (0 to 10 scale)

Analysis 1.7: Cryotherapy versus control; Outcome: Normalcy of diet (days of total parenteral nutrition)

345

Analysis 1.8: Cryotherapy versus control; Outcome: Duration of hospitalisation (days)

Analysis 1.9: Cryotherapy versus control; Outcome: Duration of opioid use (days)

346 APPENDIX 6: Cytokines and growth factors review - electronic search strategies

1) Cochrane Oral Health's Trials Register search strategy

1 ((neoplasm* or cancer* or leukaemi* or leukemi* or tumour* or tumor* or malignan* or neutropeni* or neutropeni* or carcino* or adenocarcinoma* or lymphoma*):ti,ab) AND (INREGISTER) 2 ((radioth* or radiat* or irradiat*):ti,ab) AND (INREGISTER) 3 ((("bone marrow" and transplant*) or "hematopoietic stem cell transplant*" or "haematopoietic stem cell transplant*" or HSCT):ti,ab) AND (INREGISTER) 4 (chemo*:ti,ab) AND (INREGISTER) 5 (#1 or #2 or #3 or #4) AND (INREGISTER) 6 ((stomatitis or mucositis):ti,ab) AND (INREGISTER) 7 ((oral and mucos*):ti,ab) AND (INREGISTER) 8 ((mycosis or mycotic):ti,ab) AND (INREGISTER) 9 (mIAS:ti,ab) AND (INREGISTER) 10 (#6 or #7 or #8 or #9) AND (INREGISTER) 11 (("growth factor*" or "growth substance*" or "immunologic factor*"):ti,ab) AND (INREGISTER) 12 ("fibroblast growth factor*":ti,ab) AND (INREGISTER) 13 ((keratinocyte* or cytokine*):ti,ab) AND (INREGISTER) 14 ((palifermin* or KGF or FGF):ti,ab) AND (INREGISTER) 15 ((keprivance or velafermin or repifermin):ti,ab) AND (INREGISTER) 16 (glycoprotein*:ti,ab) AND (INREGISTER) 17 ((colony-stimulat* or "macrophage-granulocyte inducer*" or "myeloid cell-growth inducer*" or "protein inducer MGI"):ti,ab) AND (INREGISTER) 18 ((GM-CSF or G-CSF):ti,ab) AND (INREGISTER) 19 ((molgramostim or Growgen-GM or Leucomax or Molcass or Gramostim or Leucocitim or Mielogen or Meustim or Bagomol or Gramal):ti,ab) AND (INREGISTER) 20 ((rhEGF or "recombinant epithelial growth factor*" or "epidermal growth factor*" or EGF):ti,ab) AND (INREGISTER) 21 (("platelet-derived growth factor*" or PDGF or "platelet lysate"):ti,ab) AND (INREGISTER) 22 (("transforming growth factor*" or "bone-derived transforming growth factor*" or "milk growth factor*" or "platelet transforming growth factor*" or TGF-beta or TGFBeta):ti,ab) AND (INREGISTER) 23 (("hepatocyte growth factor*" or hepatopoietin or "scatter factor"):ti,ab) AND (INREGISTER) 24 ((somatomedin* or "insulin-like growth factor*" or "sulfation factor*" or Mecasermin or Increlex or Iplex or IGF-1 or IGF1):ti,ab) AND (INREGISTER) 25 (erythropoietin*:ti,ab) AND (INREGISTER) 26 ((thrombopoietin* or "mpl Ligand" or "megakaryocyte colony stimulating factor*" or "megakaryocyte growth and development factor*" or "MGDF factor" or "myeloproliferative leukemia virus oncogene ligand" or "thrombocytopoiesis-stimulating factor" or thrombocytopoietin*):ti,ab) AND (INREGISTER) 27 (("interleukin 11" or "adipogenesis inhibitory factor*" or IL-11 or IL11):ti,ab) AND (INREGISTER) 28 ((ghrelin* or "GHRL protein" or obestatin):ti,ab) AND (INREGISTER) 29 ((ATL-104 or ATL104):ti,ab) AND (INREGISTER) 30 ((whey or "milk derived protein" or "milk derived growth factor"):ti,ab) AND (INREGISTER) 31 (("glucagon-like peptide 2" or "amino acid*" or proglucagon or GLP-2 or teduglutide or Gattex or Revestive):ti,ab) AND (INREGISTER) 32 (("trefoil factor" or "carcinoembryonic antigen cell adhesion molecule 1" or glutathione or isethion):ti,ab) AND (INREGISTER) 33 (("vascular endothelial growth factor*" or VEGF*):ti,ab) AND (INREGISTER) 34 (("targeted therap*" or "targeted agent*"):ti,ab) AND (INREGISTER)

347 35 ("biologic* therap*":ti,ab) AND (INREGISTER) 36 ((#11 or #12 or #13 or #14 or #15 or #16 or #17 or #18 or #19 or #20 or #21 or #22 or #23 or #24 or #25 or #26 or #27 or #28 or #29 or #30 or #31 or #32 or #33 or #34 or #35)) AND (INREGISTER) 37 (#5 and #10 and #36) AND (INREGISTER)

2) Cochrane Central Register of Controlled Clinical Trials (CENTRAL) search strategy

#1 [mh neoplasms] #2 [mh radiotherapy] #3 [mh "antineoplastic agents"] #4 [mh "^bone marrow transplantation"] #5 [mh ^"hematopoietic stem cell transplantation"] #6 MeSH descriptor: [Antineoplastic Combined Chemotherapy Protocols] explode all trees #7 (neoplasm* or cancer* or leukaemi* or leukemi* or tumour* or tumor* or malignan* or neutropeni* or neutropeni* or carcino* or adenocarcinoma* or lymphoma*) #8 (radioth* or radiat* or irradiat*) #9 (("bone marrow" and transplant*) or "hematopoietic stem cell transplant*" or "haematopoietic stem cell transplant*" or HSCT) #10 chemo* #11 {or #1-#10} #12 [mh Stomatitis] #13 [mh ^"oral candidiasis"] #14 stomatitis #15 mucositis #16 (oral near/6 mucos*) #17 (mycosis or mycotic) #18 mIAS #19 {or #12-#18} #20 ("growth factor*" or "growth substance*" or "immunologic factor*") #21 [mh ^"Fibroblast growth factor 7"] #22 "fibroblast growth factor*" #23 (keratinocyte* or cytokine*) #24 (palifermin* or KGF or FGF) #25 (keprivance or velafermin or repifermin) #26 glycoprotein* #27 [mh "colony-stimulating factors"] #28 (colony-stimulat* or "macrophage-granulocyte inducer*" or "myeloid cell-growth inducer*" or "protein inducer MGI") #29 (GM-CSF or G-CSF) #30 (molgramostim or Growgen-GM or Leucomax or Molcass or Gramostim or Leucocitim or Mielogen or Meustim or Bagomol or Gramal) #31 [mh ^"Epidermal growth factor"] #32 (rhEGF or "recombinant epithelial growth factor*" or "epidermal growth factor*" or EGF) #33 [mh ^"Platelet-derived growth factor"] #34 ("platelet-derived growth factor*" or PDGF or "platelet lysate") #35 [mh ^"Transforming growth factor beta"] #36 ("transforming growth factor*" or "bone-derived transforming growth factor*" or "milk growth factor*" or "platelet transforming growth factor*" or TGF-beta or TGFBeta) #37 [mh ^"Hepatocyte growth factor"] #38 ("hepatocyte growth factor*" or hepatopoietin or "scatter factor") #39 [mh Somatomedin] #40 (somatomedin* or "insulin-like growth factor*" or "sulfation factor*" or Mecasermin or Increlex or Iplex or IGF-1 or IGF1) #41 erythropoietin* #42 [mh ^Thrombopoietin]

348 #43 (thrombopoietin* or "mpl Ligand" or "megakaryocyte colony stimulating factor*" or "megakaryocyte growth and development factor*" or "MGDF factor" or "myeloproliferative leukemia virus oncogene ligand" or "thrombocytopoiesis-stimulating factor" or thrombocytopoietin*) #44 [mh ^Ghrelin] #45 [mh "Interleukin 11"] #46 ("interleukin 11" or "adipogenesis inhibitory factor*" or IL-11 or IL11) #47 (ghrelin* or "GHRL protein" or obestatin) #48 (ATL-104 or ATL104) #49 (whey or ("milk derived" next (protein or "growth factor"))) #50 [mh ^"Glucagon-like peptide 2"] #51 ("glucagon-like peptide 2" or "amino acid*" or proglucagon or GLP-2 or teduglutide or Gattex or Revestive) #52 [mh ^Glutathione] #53 ("trefoil factor" or "carcinoembryonic antigen cell adhesion molecule 1" or glutathione or isethion) #54 [mh ^"Vascular endothelial growth factors"] #55 ("vascular endothelial growth factor*" or VEGF*) #56 [mh ^"Molecular targeted therapy"] #57 (targeted near/3 (therap* or agent*)) #58 (biologic* next therap*) #59 {or #20-#58} #60 #11 and #19 and #59

3) MEDLINE Ovid search strategy

1. exp NEOPLASMS/ 2. exp RADIOTHERAPY/ 3. exp Antineoplastic agents/ 4. Anti-neoplastic combined chemotherapy protocols/ 5. Bone Marrow Transplantation/ 6. Hematopoietic Stem Cell Transplantation/ 7. (neoplasm$ or cancer$ or leukaemi$ or leukemi$ or tumour$ or tumor$ or malignan$ or neutropeni$ or carcino$ or adenocarcinoma$ or lymphoma$).ti,ab. 8. (radioth$ or radiat$ or irradiat$).ti,ab. 9. ((bone adj marrow adj5 transplant$) or "hematopoietic stem cell transplant$" or "haematopoietic stem cell transplant$" or HSCT).ti,ab. 10. chemo$.ti,ab. 11. or/1-10 12. exp STOMATITIS/ 13. Candidiasis, Oral/ 14. stomatitis.ti,ab. 15. mucositis.ti,ab. 16. (oral adj6 mucos$).ti,ab. 17. (mycosis or mycotic).ti,ab. 18. mIAS.ti,ab. 19. or/12-18 20. ((growth adj factor$) or (growth adj substance$) or (immunologic adj factor$)).ti,ab. 21. Fibroblast growth factor 7/ 22. "fibroblast growth factor$".ti,ab. 23. (keratinocyte$ or cytokine$).ti,ab. 24. (palifermin$ or KGF or FGF).ti,ab. 25. (kepivance or velafermin or repifermin).ti,ab. 26. glycoprotein$.ti,ab. 27. exp Colony-stimulating factors/ 28. ("colony-stimulat$" or "macrophage-granulocyte inducer$" or "myeloid cell-growth inducer$" or "protein inducer MGI").ti,ab.

349 29. (GM-CSF or G-CSF).ti,ab. 30. (molgramostim or Growgen-GM or Leucomax or Molcass or Gramostim or Leucocitim or Mielogen or Meustim or Bagomol or Gramal).ti,ab. 31. Epidermal growth factor/ 32. (rhEGF or "recombinant epithelial growth factor$" or "epidermal growth factor$" or EGF).ti,ab. 33. Platelet-derived growth factor/ 34. ("platelet-derived growth factor$" or PDGF or "platelet lysate").ti,ab. 35. Transforming Growth Factor beta/ 36. ("transforming growth factor$" or "bone-derived transforming growth factor$" or "milk growth factor$" or "platelet transforming growth factor" or TGF-beta or TGFbeta).ti,ab. 37. Hepatocyte growth factor/ 38. ("hepatocyte growth factor$" or hepatopoietin or "scatter factor").ti,ab. 39. exp Somatomedin/ 40. (somatomedin$ or "insulin-like growth factor$" or "sulfation factor" or Mecasermin or Increlex or Iplex or IGF-1 or IGF1).ti,ab. 41. erythropoietin$.ti,ab. 42. Thrombopoietin/ 43. (thrombopoietin$ or "mpl Ligand" or "megakaryocyte colony stimulating factor$" or "megakaryocyte growth and development factor$" or "MGDF factor" or "myeloproliferative leukemia virus oncogene ligand" or "thrombocytopoiesis-stimulating factor" or thrombocytopoietin$).ti,ab. 44. Ghrelin/ 45. Interleukin 11/ 46. ("interleukin 11" or "Adipogenesis Inhibitory Factor$" or IL-11 or IL11).ti,ab. 47. (ghrelin$ or "GHRL protein" or obestatin).ti,ab. 48. (ATL-104 or ATL104).ti,ab. 49. (whey or ("milk derived" adj (protein or growth factor))).ti,ab. 50. Glucagon-Like Peptide 2/ 51. ("glucagon-like peptide 2" or (amino adj acid$) or proglucagon or GLP-2 or teduglutide or Gattex or Revestive).ti,ab. 52. Glutathione/ 53. ("trefoil factor" or "carcinoembryonic antigen cell adhesion molecule 1" or glutathione or isethion).ti,ab. 54. Vascular endothelial growth factors/ 55. ("vascular endothelial growth factor$" or VEGF$).ti,ab. 56. Molecular targeted therapy/ 57. (targeted adj3 (therap$ or agent$)).ti,ab. 58. (biologic$ adj therap$).ti,ab. 59. or/20-58 60. 11 and 19 and 59

The above subject search was linked to the Cochrane Highly Sensitive Search Strategy (CHSSS) for identifying randomised trials (RCTs) in MEDLINE: sensitivity maximising version (2008 revision) as referenced in Chapter 6.4.11.1 and detailed in box 6.4.c of the Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 (updated March 2011).

1. randomized controlled trial.pt. 2. controlled clinical trial.pt. 3. randomized.ab. 4. placebo.ab. 5. drug therapy.fs. 6. randomly.ab. 7. trial.ab. 8. groups.ab. 9. or/1-8

350 10. exp animals/ not humans.sh. 11. 9 not 10

4) Embase Ovid search strategy

1. exp NEOPLASM/ 2. exp RADIOTHERAPY/ 3. exp Antineoplastic agent/ 4. Bone Marrow Transplantation/ 5. Hematopoietic Stem Cell Transplantation/ 6. (neoplasm$ or cancer$ or leukaemi$ or leukemi$ or tumour$ or tumor$ or malignan$ or neutropeni$ or carcino$ or adenocarcinoma$ or lymphoma$).ti,ab. 7. (radioth$ or radiat$ or irradiat$).ti,ab. 8. ((bone adj marrow adj5 transplant$) or "hematopoietic stem cell transplant$" or "haematopoietic stem cell transplant$" or HSCT).ti,ab. 9. chemo$.ti,ab. 10. or/1-9 11. exp STOMATITIS/ 12. Thrush/ 13. stomatitis.ti,ab. 14. mucositis.ti,ab. 15. (oral adj6 mucos$).ti,ab. 16. (mycosis or mycotic).ti,ab. 17. mIAS.ti,ab. 18. or/11-17 19. ((growth adj factor$) or (growth adj substance$) or (immunologic adj factor$)).ti,ab. 20. "fibroblast growth factor$".ti,ab. 21. (keratinocyte$ or cytokine$).ti,ab. 22. (palifermin$ or KGF or FGF).ti,ab. 23. (kepivance or velafermin or repifermin).ti,ab. 24. glycoprotein$.ti,ab. 25. exp Colony-stimulating factor/ 26. ("colony-stimulat$" or "macrophage-granulocyte inducer$" or "myeloid cell-growth inducer$" or "protein inducer MGI").ti,ab. 27. (GM-CSF or G-CSF).ti,ab. 28. (molgramostim or Growgen-GM or Leucomax or Molcass or Gramostim or Leucocitim or Mielogen or Meustim or Bagomol or Gramal).ti,ab. 29. Epidermal growth factor/ 30. (rhEGF or "recombinant epithelial growth factor$" or "epidermal growth factor$" or EGF).ti,ab. 31. Platelet derived growth factor/ 32. ("platelet-derived growth factor$" or PDGF or "platelet lysate").ti,ab. 33. Transforming Growth Factor beta/ 34. ("transforming growth factor$" or "bone-derived transforming growth factor$" or "milk growth factor$" or "platelet transforming growth factor" or TGF-beta or TGFbeta).ti,ab. 35. Scatter factor/ 36. ("hepatocyte growth factor$" or hepatopoietin or "scatter factor").ti,ab. 37. exp Somatomedin/ 38. (somatomedin$ or "insulin-like growth factor$" or "sulfation factor" or Mecasermin or Increlex or Iplex or IGF-1 or IGF1).ti,ab. 39. erythropoietin$.ti,ab. 40. Thrombopoietin/ 41. (thrombopoietin$ or "mpl Ligand" or "megakaryocyte colony stimulating factor$" or "megakaryocyte growth and development factor$" or "MGDF factor" or "myeloproliferative leukemia virus oncogene ligand" or "thrombocytopoiesis-stimulating factor" or thrombocytopoietin$).ti,ab. 42. Ghrelin/

351 43. Interleukin 11/ 44. ("interleukin 11" or "Adipogenesis Inhibitory Factor$" or IL-11 or IL11).ti,ab. 45. (ghrelin$ or "GHRL protein" or obestatin).ti,ab. 46. (ATL-104 or ATL104).ti,ab. 47. (whey or ("milk derived" adj (protein or growth factor))).ti,ab. 48. ("glucagon-like peptide 2" or (amino adj acid$) or proglucagon or GLP-2 or teduglutide or Gattex or Revestive).ti,ab. 49. Glutathione/ 50. ("trefoil factor" or "carcinoembryonic antigen cell adhesion molecule 1" or glutathione or isethion).ti,ab. 51. Vasculotropin/ 52. ("vascular endothelial growth factor$" or VEGF$).ti,ab. 53. Molecularly targeted therapy/ 54. (targeted adj3 (therap$ or agent$)).ti,ab. 55. (biologic$ adj therap$).ti,ab. 56. or/19-55 57. 10 and 18 and 56

The above subject search was linked to adapted version of the Cochrane Embase Project filter for identifying RCTs in EMBASE via OVID (see www.cochranelibrary.com/help/central- creation-details.html for information).

1. Randomized controlled trial/ 2. Controlled clinical study/ 3. Random$.ti,ab. 4. randomization/ 5. intermethod comparison/ 6. placebo.ti,ab. 7. (compare or compared or comparison).ti. 8. ((evaluated or evaluate or evaluating or assessed or assess) and (compare or compared or comparing or comparison)).ab. 9. (open adj label).ti,ab. 10. ((double or single or doubly or singly) adj (blind or blinded or blindly)).ti,ab 11. double blind procedure/ 12. parallel group$1.ti,ab. 13. (crossover or cross over).ti,ab. 14. ((assign$ or match or matched or allocation) adj5 (alternate or group$1 or intervention$1 or patient$1 or subject$1 or participant$1)).ti,ab. 15. (assigned or allocated).ti,ab. 16. (controlled adj7 (study or design or trial)).ti,ab. 17. (volunteer or volunteers).ti,ab. 18. trial.ti. 19. or/1-18 20. (exp animal/ or animal.hw. or nonhuman/) not (exp human/ or human cell/ or (human or humans).ti.) 21. 19 not 20

5) CINAHL EBSCO (Cumulative Index to Nursing and Allied Health Literature) search strategy

S53 S10 and S18 and S52 S52 S19 or S20 or S21 or S22 or S23 or S24 or S25 or S26 or S27 or S28 or S29 or S30 or S31 or S32 or S33 or S34 or S35 or S36 or S37 or S38 or S39 or S40 or S41 or S42 or S43 or S44 or S45 or S46 or S47 or S48 or S49 or S50 or S51 S51 (biologic* N1 therap*) S50 (targeted N3 (therap* or agent*)) S49 ("vascular endothelial growth factor*" or VEGF*)

352 S48 (mh Vascular endothelial growth factors) S47 ("trefoil factor" or "carcinoembryonic antigen cell adhesion molecule 1" or glutathione or isethion) S46 (mh Glutathione) S45 ("glucagon-like peptide 2" or "amino acid*" or proglucagon or GLP-2 or teduglutide or Gattex or Revestive) S44 (mh Glucagon-like peptide 2) S43 (whey or ("milk derived" N1 (protein or "growth factor"))) S42 (ATL-104 or ATL104) S41 (ghrelin* or "GHRL protein" or obestatin) S40 ("interleukin 11" or "adipogenesis inhibitory factor*" or IL-11 or IL11) S39 (mh Ghrelin) S38 (thrombopoietin* or "mpl Ligand" or "megakaryocyte colony stimulating factor*" or "megakaryocyte growth and development factor*" or "MGDF factor" or "myeloproliferative leukemia virus oncogene ligand" or "thrombocytopoiesis-stimulating factor" or thrombocytopoietin*) S37 erythropoietin* S36 (somatomedin* or "insulin-like growth factor*" or "sulfation factor*" or Mecasermin or Increlex or Iplex or IGF-1 or IGF1) S35 ("hepatocyte growth factor*" or hepatopoietin or "scatter factor") S34 ("transforming growth factor*" or "bone-derived transforming growth factor*" or "milk growth factor*" or "platelet transforming growth factor*" or TGF-beta or TGFBeta) S33 (mh Transforming growth factor beta) S32 ("platelet-derived growth factor*" or PDGF or "platelet lysate") S31 (mh Platelet-derived growth factor) S30 (rhEGF or "recombinant epithelial growth factor*" or "epidermal growth factor*" or EGF) S29 (MH "Epidermal Growth Factors") S28 (molgramostim or Growgen-GM or Leucomax or Molcass or Gramostim or Leucocitim or Mielogen or Meustim or Bagomol or Gramal) S27 (GM-CSF or G-CSF) S26 (colony-stimulat* or "macrophage-granulocyte inducer*" or "myeloid cell-growth inducer*" or "protein inducer MGI") S25 (mh colony-stimulating factors) S24 glycoprotein* S23 (keprivance or velafermin or repifermin) S22 (palifermin* or KGF or FGF) S21 (keratinocyte* or cytokine*) S20 "fibroblast growth factor*" S19 ("growth factor*" or "growth substance*" or "immunologic factor*") S18 S11 or S12 or S13 or S14 or S15 or S16 or S17 S17 mIAS S16 (mycosis or mycotic) S15 (oral N6 mucos*) S14 mucositis S13 stomatitis S12 (MH "Candidiasis, Oral") S11 (mh stomatitis+) S10 S1 or S2 or S3 or S4 or S5 or S6 or S7 or S8 or S9 S9 chemo* S8 (("bone marrow" and transplant*) or "hematopoietic stem cell transplant*" or "haematopoietic stem cell transplant*" or HSCT) S7 (radioth* or radiat* or irradiat*) S6 (neoplasm* or cancer* or leukaemi* or leukemi* or tumour* or tumor* or malignan* or neutropeni* or neutropeni* or carcino* or adenocarcinoma* or lymphoma*) S5 (mh hematopoietic stem cell transplantation) S4 (mh bone marrow transplantation) S3 (mh antineoplastic agents+)

353 S2 (mh radiotherapy+) S1 (mh neoplasms+)

The above subject search was linked to Cochrane Oral Health's filter for CINAHL via EBSCO. S1 MH Random Assignment or MH Single-blind Studies or MH Double-blind Studies or MH Triple-blind Studies or MH Crossover design or MH Factorial Design S2 TI ("multicentre study" or "multicenter study" or "multi-centre study" or "multi center study") or AB ("multicentre study" or "multicenter study" or "multi centre study" or "multicenter study") or SU ("multicentre study" or "multicenter study" or "multi-centre study" or "multi-center study")

S3 TI random* or AB random* S4 AB "latin square" or TI "latin square" S5 TI (crossover or cross-over) or AB (crossover or cross-over) or SU (crossover or cross-over) S6 MH Placebos S7 AB (singl* or doubl* or trebl* or tripl*) or TI (singl* or doubl* or trebl* or tripl*) S8 TI blind* or AB mask* or AB blind* or TI mask* S9 S7 and S8 S10 TI Placebo* or AB Placebo* or SU Placebo* S11 MH Clinical Trials S12 TI (Clinical AND Trial) or AB (Clinical AND Trial) or SU (Clinical AND Trial) S13 S1 or S2 or S3 or S4 or S5 or S6 or S9 or S10 or S11 or S12

6) CANCERLIT PubMed search strategy

We searched CANCERLIT by limiting a PubMed search with the Cancer subject filter and the following search strategy.

#60 (#11 and #19 and #59) #59 (#20 or #21 or #22 or #23 or #24 or #25 or #26 or #27 or #28 or #29 or #30 or #31 or #32 or #33 or #34 or #35 or #36 or #37 or #38 or #39 or #40 or #41 or #42 or #43 or #44 or #45 or #46 or #47 or #48 or #49 or #50 or #51 or #52 or #53 or #54 or #55 or #56 or #57 or #58) #58 "biologic* therap*" #57 ("targeted therap*" or "targeted agent*") #56 Molecular targeted therapy [mh:noexp] #55 (("vascular endothelial growth factor*" or VEGF*)) #54 Vascular endothelial growth factors [mh:noexp] #53 (("trefoil factor" or "carcinoembryonic antigen cell adhesion molecule 1" or glutathione or isethion)) #52 Glutathione [mh:noexp] #51 (("glucagon-like peptide 2" or "amino acid*" or proglucagon or GLP-2 or teduglutide or Gattex or Revestive)) #50 Glucagon-like peptide 2 [mh:noexp] #49 ((whey or "milk derived protein" or "milk derived growth factor")) #48 ((ATL-104 or ATL104)) #47 ((ghrelin* or "GHRL protein" or obestatin)) #46 (("interleukin 11" or "adipogenesis inhibitory factor*" or IL-11 or IL11)) #45 Interleukin 11 [mh:exp] #44 Ghrelin [mh:noexp] #43 ((thrombopoietin* or "mpl Ligand" or "megakaryocyte colony stimulating factor*" or "megakaryocyte growth and development factor*" or "MGDF factor" or "myeloproliferative leukemia virus oncogene ligand" or "thrombocytopoiesis-stimulating factor" or thrombocytopoietin*)) #42 Thrombopoietin [mh:noexp] #41 erythropoietin* #40 ((somatomedin* or "insulin-like growth factor*" or "sulfation factor*" or Mecasermin or Increlex or Iplex or IGF-1 or IGF1))

354 #39 Somatomedin [mh:exp] #38 (("hepatocyte growth factor*" or hepatopoietin or "scatter factor")) #37 Hepatocyte growth factor [mh:noexp] #36 (("transforming growth factor*" or "bone-derived transforming growth factor*" or "milk growth factor*" or "platelet transforming growth factor*" or TGF-beta or TGFBeta)) #35 Transforming growth factor beta [mh:noexp] #34 (("platelet-derived growth factor*" or PDGF or "platelet lysate")) #33 Platelet-derived growth factor [mh:noexp] #32 ((rhEGF or "recombinant epithelial growth factor*" or "epidermal growth factor*" or EGF)) #31 Epidermal growth factor [mh:noexp] #30 ((molgramostim or Growgen-GM or Leucomax or Molcass or Gramostim or Leucocitim or Mielogen or Meustim or Bagomol or Gramal)) #29 ((GM-CSF or G-CSF)) #28 ((colony-stimulat* or "macrophage-granulocyte inducer*" or "myeloid cell-growth inducer*" or "protein inducer MGI")) #27 colony-stimulating factors [mh:exp] #26 glycoprotein* #25 ((keprivance or velafermin or repifermin)) #24 ((palifermin* or KGF or FGF)) #23 ((keratinocyte* or cytokine*)) #22 "fibroblast growth factor*" #21 Fibroblast growth factor 7 [mh:noexp] #20 (("growth factor*" or "growth substance*" or "immunologic factor*")) #19 (#12 or #13 or #14 or #15 or #16 or #17 or #18) #18 mIAS #17 ((mycosis or mycotic)) #16 "oral mucos*" #15 mucositis #14 stomatitis #13 oral candidiasis [mh:noexp] #12 Stomatitis [mh:exp] #11 (#1 or #2 or #3 or #4 or #5 or #6 or #7 or #8 or #9 or #10) #10 chemo* #9 ((("bone marrow" and transplant*) or "hematopoietic stem cell transplant*" or "haematopoietic stem cell transplant*" or HSCT)) #8 ((radioth* or radiat* or irradiat*)) #7 ((neoplasm* or cancer* or leukaemi* or leukemi* or tumour* or tumor* or malignan* or neutropeni* or neutropeni* or carcino* or adenocarcinoma* or lymphoma*)) #6 Antineoplastic Combined Chemotherapy Protocols [mh:noexp] #5 "hematopoietic stem cell transplantation" [mh:noexp] #4 bone marrow transplantation [mh:noexp] #3 antineoplastic agents [mh:exp] #2 radiotherapy [mh:exp] #1 neoplasms [mh:exp]

This search was linked with the Cochrane Highly Sensitive Search Strategy (CHSSS) for identifying randomised trials in MEDLINE: sensitivity maximising version (2008 revision) as referenced in Chapter 6.4.11.1 and detailed in box 6.4.a of the Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 (updated March 2011).

#1 randomized controlled trial [pt] #2 controlled clinical trial [pt] #3 randomized [tiab] #4 placebo [tiab] #5 drug therapy [sh] #6 randomly [tiab] #7 trial [tiab]

355 #8 groups [tiab] #9 #1 or #2 or #3 or #4 or #5 or #6 or #7 or #8 #10 animals [mh] not humans [mh] #11 #9 NOT #10

7) US National Institutes of Health Ongoing Trials Register (ClinicalTrials.gov) search strategy oral mucositis and cytokines oral mucositis and growth factor oral mucositis and palifermin oral mucositis and GM-CSF

8) World Health Organization International Clinical Trials Registry Platform search strategy oral mucositis and cytokines oral mucositis and growth factor oral mucositis and palifermin oral mucositis and GM-CSF

356 APPENDIX 7: Cytokines and growth factors review - characteristics of studies and risk of bias tables

Characteristics of included studies

Antoun 2009

Methods Trial design: parallel (2 arms) Location: Institut Gustave Roussy, Villejuif, France Number of centres: 1 Study duration: February 2005 to September 2006 Trials registry number: none/unknown Participants Inclusion criteria: adults with metastatic colorectal adenocarcinoma (grade 3 to 4); life expectancy greater than 3 months; receiving 5FU-based chemotherapy Exclusion criteria: HIV; pregnant or lactating; unlikely to comply with interventions; participation in another trial in the previous 12 months (unless regarding chemotherapeutic protocols); undergone a total colectomy; state of subocclusion; chronic inflammatory diseases of the digestive tract; radiation enteropathy Cancer type: metastatic colorectal adenocarcinoma (grade 3 to 4) Cancer treatment: 5FU-based chemotherapy Age at baseline (years): median 60 (not reported by group) Gender: not reported Number randomised: 22 (not reported by group) Number evaluated: 13 (Gp A: 9; Gp B: 4) Interventions Comparison: TGF-beta(2) versus placebo Gp A: nutritional supplement of proteins, carbohydrates, fats, vitamins and minerals, with TGF-beta(2) (2 ng/mg protein); formulas were in powder form, mixed with cool previously boiled water at 0.23 g/mL (100 kcl/100 mL); during each cycle participants received 750 to 1000 mL per day plus any other food desired; formula administered for 2 days before, 2 days during, and 3 days after chemotherapy (7 days/cycle) Gp B: same as above without the TGF-beta(2) Compliance: "Nine randomised patients who never ate the formula were excluded from the study" (not reported by group) Duration of treatment: "3 months (test or control formula), for a minimum of one and a maximum of eight cycles of treatment" Outcomes  Oral mucositis: WHO 0 to 4 scale (no details reported on who assessed this, or when it was assessed; only reports incidence of any mucositis)  Chemotherapy-induced diarrhoea (not an outcome of this review) Notes Sample size calculation: not reported Funding: "This study was funded by Nestec Ltd" - Nestlé (manufacturer of the intervention) Declarations/conflicts of interest: 6 of the 9 authors were either consultants (1) or employees (5) of Nestlé Data handling by review authors: reported in additional table Other information of note: "Due to low accrual of patients (22 patients were enrolled and randomised in 18 months), the study was prematurely stopped"

Risk of bias table

Authors' Bias Support for judgement judgement 357 Random sequence Unclear Quote: "randomly assigned" generation (selection risk bias) Comment: insufficient information to determine method of random sequence generation Allocation Unclear Quote: "randomly assigned" concealment (selection risk bias) Comment: insufficient information to determine whether or not the random sequence was adequately concealed Blinding of Low risk Quote: "double-blind" and "The test formula differed only participants and by containing an additional..." personnel (performance bias) Comment: the use of a placebo should have ensured that blinding was successful Blinding of outcome Low risk Quote: "double-blind" assessment (detection bias) Comment: it is not clear who was blinded. There are subjective elements to the assessment of oral mucositis using this scale, requiring the patient's assessment of pain/soreness and their ability to swallow but, as the participants were unaware of their group allocation, the assessment of oral mucositis can be considered to be blinded Incomplete outcome High risk Overall attrition was 41% although it was not reported by data (attrition bias) group Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Blazar 2006

Methods Trial design: parallel (7 arms) dose-ranging study Location: Universities of Minnesota and Michigan, USA Number of centres: 2 Study duration: not reported Trials registry number: none/unknown Participants Inclusion criteria: aged 3 to 65 years; diagnosed with haematological malignancy (including myelodysplastic syndromes); ECOG score of 0 to 2; eligible for allogeneic HSCT after conditioning treatment with chemotherapy with or without TBI Exclusion criteria: received previous allogeneic HSCT; due to receive a T-cell- depleted donor graft; active chronic skin disease; pre-existent inflammatory bowel disease; uncontrolled (antibiotic-resistant) bacterial infection; hepatitis; HIV Cancer type: haematologic: ALL (Gp A: 12%; Gp B: 3%); AML (Gp A: 35%; Gp B: 39%); CML (Gp A: 10%; Gp B: 26%); MDS (Gp A: 9%; Gp B: 19%); NHL (Gp A: 19%; Gp B: 3%); Hodgkin's (Gp A: 1%; Gp B: 0%); Other (Gp A: 14%; Gp B: 10%) Cancer treatment: both centres had allogeneic HSCT on day 0 but differed in conditioning regimen and GVHD prophylaxis as follows:  Minnesota centre (n = 54): conditioning with cyclophosphamide (60 mg/kg per day for 2 days) on days -7 and -6 and TBI (total dose 13.2 Gy, fractionated as 165 cGy twice daily for 4 days) on days -4 to -1; GVHD prophylaxis with methotrexate (15 mg/m², IV bolus on day +1, and 10

358 mg/m², IV bolus on days 3, 6, and 11) and cyclosporine A (starting from day -3); Gp A: 69%; Gp B: 31%  Michigan centre (n = 46): conditioning with busulfan (1 mg/kg per dose given 4 times daily for 4 days) on days -8 to -5 and cyclophosphamide (60 mg/kg per day for 2 days) on days -3 and -2; GVHD prophylaxis with methotrexate (15 mg/m², IV bolus on day +1, and 10 mg/m², IV bolus on days 3, 6, and 11) and tacrolimus or cyclosporine A (starting from day -3); Gp A: 70%; Gp B: 30% Both centres received G-CSF (filgrastim) 5 µg/kg per day from 24 hours after HSCT until neutrophil recovery Age at baseline (years): Gp A: median 46 (range 7 to 65); Gp B: median 46 (range 7 to 63) Gender: both groups 58% male Number randomised: 100 (Gp A: 69; Gp B: 31) Number evaluated: 96 (Gp A: 65; Gp B: 31) Interventions Comparison: KGF (palifermin) versus placebo (4 KGF arms and 3 placebo arms were each combined into a single arm) Gp A: KGF  (n = 8): 40 µg/kg per day in 6 doses on days -11, -10, -9 and 0, 1, 2 (total dose = 240 µg/kg)  (n = 10): 60 µg/kg per day in 6 doses on days -11, -10, -9 and 0, 1, 2 (total dose = 360 µg/kg)  (n = 14): 60 µg/kg per day in 9 doses on days -11, -10, -9 and 0, 1, 2 and 7, 8, 9 (total dose = 540 µg/kg)  (n = 37): 60 µg/kg per day in 12 doses on days -11, -10, -9 and 0, 1, 2 and 7, 8, 9 and 14, 15, 16 (total dose = 720 µg/kg) Gp B: placebo with matching schedule to either the 6, 9 or 12 dose regimen Mode of administration not described but presumably IV as in other KGF studies Compliance: Gp A: 20 did not receive all study doses (17 of these were replaced to allow a full assessment of safety); Gp B: 2 did not receive all study doses (1 replaced) Duration of treatment: varied from 13 days (6 doses) to 27 days (12 doses) - see above Outcomes  Oral mucositis: WHO 0 to 4 scale (measured 3 times per week during hospitalisation by designated observers, maximum score reported)  Adverse effects (assessed daily during study period using WHO and NCI- CTC toxicities scale)  Incidence and severity of acute GVHD (not an outcome of this review)  Overall survival (not an outcome of this review)  Incidence of transplantation-related toxicity (not an outcome of this review)  Time to marrow engraftment (not an outcome of this review) Notes Sample size calculation: not reported Funding: government grants from NIH and FDA, and also supported by Amgen (pharmaceutical industry) Declarations/conflicts of interest: not reported Data handling by review authors: the data for incidence of mucositis was not reported separately for each dose and therefore it was not possible to include head-to-head comparisons of different dosages in this review; the data for incidence of mucositis was presented in subgroups of those that did or did not receive the final methotrexate infusion on day 11 but we used the overall data in our meta-analyses (the study authors report that there was no difference between these subgroups) Other information of note: the study authors report a greater decrease in incidence of grade 3 to 4 (severe) oral mucositis due to palifermin in the Minnesota participants (who received a more mucotoxic conditioning regimen)

359 than in the Michigan participants

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear Quote: "patients...were randomly assigned..." generation (selection risk bias) Comment: insufficient information to determine method of random sequence generation Allocation Unclear Quote: "patients...were randomly assigned..." concealment (selection risk bias) Comment: insufficient information to determine whether or not the random sequence was adequately concealed Blinding of Low risk Quote: "double-blind, placebo-controlled" participants and personnel Comment: the use of a placebo should have ensured that (performance bias) blinding was successful Blinding of outcome Low risk Quote: "double-blind, placebo-controlled" assessment (detection bias) Comment: it is not clear who was blinded. There are subjective elements to the assessment of oral mucositis using this scale, requiring the patient's assessment of pain/soreness and their ability to swallow but, as the participants were unaware of their group allocation, the assessment of oral mucositis can be considered to be blinded Incomplete outcome Low risk Quote: "Data for all patients randomly assigned and who data (attrition bias) received a transplant were used in all other analyses (intent- to-treat)"

Comment: although 18 participants (Gp A: 17; Gp B: 1) were replaced to allow a full assessment of safety, it seems that the originally randomised participants were included in the analyses.

Overall attrition was 4% (Gp A: 6%; Gp B: 0%) for the oral mucositis incidence outcome. The reasons were unclear but this proportion of attrition is unlikely to have biased the results Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Blijlevens 2013

Methods Trial design: parallel (3 arms) Location: Europe (Italy, France, The Netherlands, Ireland, Germany, UK, Denmark, Austria, Switzerland, Czech Republic, Sweden, Hungary, Belgium, Finland) Number of centres: 39 Study duration: December 2006 to February 2009 Trials registry number: NCT00434161

360 Participants Inclusion criteria: aged between 18 and 70 years; due to receive high-dose melphalan; ECOG score of 0 to 2 (or 3, if reason was due to multiple myeloma); at least 2 x 10⁶ CD34+ cells per kg; corrected carbon monoxide diffusing capacity 50% or higher of predicted; absolute neutrophil count at least 1.5 x 10⁹/L and platelets at least 100 x 10⁹/L; total bilirubin 2 mg/dL or lower; aspartate aminotransferase and/or alanine aminotransferase 4 x institutional upper limit of normal or lower Exclusion criteria: not reported Cancer type: multiple myeloma Cancer treatment: one-day administration of high-dose melphalan (200mg/m²) on day -2, followed by auto-SCT on day 0 Age at baseline (years): Gp A: median 55 (range 32 to 69); Gp B: median 58 (range 40 to 68); Gp C: median 58 (range 41 to 68) Gender: Gp A: 54% male; Gp B: 55% male; Gp C: 58% male Number randomised: 281 (Gp A: 109; Gp B: 115; Gp C: 57) Number evaluated: 281 (Gp A: 109; Gp B: 115; Gp C: 57) Interventions Comparison: KGF versus placebo Gp A: KGF (60 µg/kg) daily IV on days -6, -5, and -4, then placebo on days 0 (the day of auto-SCT), 1, and 2 (total dose = 180 µg/kg) Gp B: KGF (60 µg/kg) daily IV on days -6, -5, -4, 0, 1, and 2 (total dose = 360 µg/kg) Gp C: placebo daily IV on days -6, -5, -4, 0, 1, and 2 Compliance: Gp A: 8% discontinued; Gp B: 12% discontinued; Gp C: 4% discontinued; (point of discontinuation or number of treatments not stated for any group) Duration of treatment: 6 treatment days (over 9 days) Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed daily by nurses and physicians from day -2 to day 32, maximum score reported) (duration of grade 2 to 4 and 3 to 4 oral mucositis also measured but not outcomes of this review)  Oral pain: OMDQ 5-point scale for mouth and throat soreness (higher = worse pain) (assessed daily by participants from day -2 to day 32, data reported as AUC, not used)  Quality of life: EQ-5D 0 (worst imaginable health) to 10 (best imaginable health) scale, incorporating mobility, self-care, usual activities, pain/discomfort and anxiety/depression (assessed daily by participants from day -2 to day 32, mean reported only at day 7 and with no SD, no usable data)  Normalcy of diet (measured as incidence of TPN) (duration of TPN also measured but not used for analysis in review)  Adverse events (NCI-CTC version 3.0 toxicity scale)  Number of days in hospital  Number of days of treatment with opioid analgesics (incidence of opioid analgesic use also measured and reported but not an outcome of this review)  Febrile neutropenia (not an outcome of this review)  Significant infections (not an outcome of this review)  Anti-infective (IV) drug use (not an outcome of this review)  Blood product use (not an outcome of this review)  Nonopioid analgesic use (not an outcome of this review) Notes Sample size calculation: 275 participants required at 95% power and 5% significance to detect an odds ratio of at least 3.5 between placebo and KGF in grade 2 to 4 oral mucositis Funding: sponsored by Swedish Orphan Biovitrum (pharmaceutical industry); KGF and placebo manufactured and packaged by Amgen (pharmaceutical industry) Declarations/conflicts of interest: two authors were employees of the sponsors;

361 the remaining authors declared no competing financial interests Data handling by review authors: we combined the two KGF groups to make a single pairwise comparison against placebo; we also made a separate comparison of the two KGF regimens Other information of note: not reported

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "Randomization was performed by using an generation (selection interactive-voice-response-system before planned bias) admission"

Comment: large multicentre trial using high-tech randomisation method - likely to be done properly Allocation Low risk Quote: "Randomization was performed by using an concealment (selection interactive-voice-response-system before planned bias) admission"

Comment: large multicentre trial using high-tech randomisation method - likely to be done properly Blinding of Low risk Quote: "double-blind, placebo-controlled" and "Study participants and drug...packaged...in identical vials" personnel (performance bias) Comment: the use of a placebo should have ensured that blinding was successful Blinding of outcome Low risk Quote: "double-blind, placebo-controlled" assessment (detection bias) Comment: it is not clear who was blinded. There are subjective elements to the assessment of lower grades of oral mucositis using the WHO scale, requiring the patient's assessment of pain/soreness and their ability to swallow. Higher grades have more objective elements so may not be affected by potential lack of blinding of the assessor. This would be the same for other subjective and objective outcomes Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately (although QoL reported with no SD or P- values, this doesn't affect the ROB judgement for other outcomes) Other bias Low risk No other sources of bias are apparent

Bradstock 2014

Methods Trial design: parallel (2 arms) Location: Australia Number of centres: 23 Study duration: recruitment from September 2006 to April 2010 Trials registry number: ACTRN012605000095662 (mentioned in trial report) Participants Inclusion criteria: aged 15 to 60 years with newly diagnosed and previously

362 untreated (except for hydroxycarbamide for high presenting white blood cell count) acute myeloid leukaemia - all subtypes except t(15;17) or variants, or core-binding factor AML (t(8;21) or inv(16) or variants); ECOG score of 0 to 3; no history of cancer (other than basal cell skin cancer or carcinoma of the cervix in situ, or other localised cancer treated by surgical excision only more than 5 years earlier without evidence of recurrence in the intervening period) Exclusion criteria: not reported Cancer type: acute myeloid leukaemia Cancer treatment: induction chemotherapy consisting of: idarubicin 9 mg/m² daily IV infusion on days 1 to 3; etoposide 75 mg/m² daily IV infusion on days 1 to 7; cytarabine 3 g/m² 12-hourly IV infusion on days 1, 3, 5, and 7 All participants received G-CSF (pegfilgrastim) 6 mg subcutaneously on day 8 Age at baseline (years): Gp A: mean 46 (SD 12; range 17 to 60); Gp B: mean 44 (SD 12; range 16 to 60) Gender: Gp A: 61% male; Gp B: 67% male Number randomised: 160 (Gp A: 79; Gp B: 81) Number evaluated: 151 (Gp A: 73; Gp B: 78) Interventions Comparison: KGF (palifermin) versus placebo Gp A: KGF (60 µg/kg) daily IV on days -3, -2, -1 prior to chemotherapy and for 3 days after completion of chemotherapy (total dose = 360 µg/kg) Gp B: same schedule with placebo

Compliance: received all 3 pre-chemotherapy doses: Gp A: 97%; Gp B: 100%; received all 3 post-chemotherapy doses: Gp A: 95%; Gp B: 96% Duration of treatment: 6 treatment days (over 14 days) Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed daily by investigators and specifically trained site personnel from the first day of chemotherapy and until the earlier of the date of discharge or day 28 after the start of chemotherapy, maximum score reported) (duration of grade 3 to 4 oral mucositis also measured but not an outcome of this review)  Adverse events (NCI-CTC version 2.0 toxicity scale)  Incidence of severe gastrointestinal toxicities related to the induction chemotherapy (not an outcome of this review)  Complete response to chemotherapy (not an outcome of this review) Notes Sample size calculation: 128 per group required to detect a reduction in grade 3 to 4 mucositis from 22% to 10% at 70% power and 5% significance Funding: "This study was funded in part from Project Grant 302133 from the National Health and Medical Research Council of Australia" (government); KGF and placebo provided by Amgen (pharmaceutical industry) Declarations/conflicts of interest: "The authors have no conflicts of interest to declare" Data handling by review authors: N/A Other information of note: not reported

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "Eligible patients were randomized 1:1 using a generation (selection block randomization technique and stratification by bias) participating centre to receive placebo or palifermin"

Comment: large multicentre trial using block randomisation and stratification - likely to be done properly

363 Allocation Unclear Quote: "Eligible patients were randomized 1:1 using a concealment (selection risk block randomization technique and stratification by bias) participating centre to receive placebo or palifermin"

Comment: insufficient information to determine whether or not the random sequence was adequately concealed Blinding of Low risk Quote: "placebo-controlled" and "Both investigators and participants and patients were blinded to the randomization outcome" personnel (performance bias) Comment: the use of a placebo should have ensured that blinding was successful Blinding of outcome Low risk Quote: "placebo-controlled" and "Both investigators and assessment (detection patients were blinded to the randomization outcome" bias) Comment: investigators assessed oral mucositis, which would have partly relied on patient's assessment of pain/soreness and their ability to swallow; both investigators and patients were blinded Incomplete outcome Low risk Overall attrition was 6% (Gp A: 8%; Gp B: 4%) for the oral data (attrition bias) mucositis incidence outcome. The reasons were similar between groups and this proportion of attrition is unlikely to have biased the results Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Brizel 2008

Methods Trial design: parallel (2 arms) Location: Australia, Canada and USA Number of centres: 22 Study duration: September 1999 to May 2001 Trials registry number: none/unknown Participants Inclusion criteria: adults with newly diagnosed stage III/IVa or IVb squamous carcinoma of the oral cavity, oropharynx, nasopharynx, hypopharynx, and larynx (or with unknown primary and extensive neck disease) undergoing CRT intended to be curative; Karnofsky performance score of 60 or higher; haemoglobin 10 g/dL or higher; white blood cell count 3.5 x 10⁹/L or higher or absolute neutrophil count 1.5 x 10⁹/L or higher; platelet count 100 x 10⁹/L or higher; serum bilirubin 1.5 mg/dL or lower; serum creatinine lower than 2.0 mg/dL (plus a 24-hour unrinary creatinine clearance 50 mL/min in those aged 60 years or older) Exclusion criteria: previous RT to the head and neck; previous surgery for the primary tumour (not including biopsy); previous CT; allergy to Escherichia coli- derived products; participation in any other investigational study within the 30 days prior to this study; refusal to use adequate contraception during the study; pregnant or breastfeeding Cancer type: head and neck: oral (Gp A: 12%; Gp B: 6%); oropharynx/nasopharynx (Gp A: 61%; Gp B: 66%); hypopharynx/larynx (Gp A: 27%; Gp B: 28%) Cancer treatment:  Radiotherapy: standard (once daily 2-Gy fractions, 5 days per week; total 70 Gy over 7 weeks) or hyperfractionated (twice daily 1.25-Gy fractions with 6 hour interval, 5 days per week with an 8 to 9 day break after 3 weeks; total

364 72 Gy over 6.5 weeks)  Chemotherapy: cisplatin (20 mg/m²/day) as IV bolus injection and 5FU (1000 mg/m²/day) as continuous infusion on the first 4 days of the first and fifth weeks of radiotherapy Age at baseline (years): Gp A: mean 54 (SD 10; range 25 to 80); Gp B: mean 56 (SD 10; range 42 to 75) Gender: Gp A: 82% male; Gp B: 84% male Number randomised: 101 (Gp A: 69; Gp B: 32) Number evaluated: 97 (Gp A: 65; Gp B: 32) Interventions Comparison: KGF (palifermin) versus placebo Gp A: KGF 60 µg/kg once weekly by IV bolus injection starting on the Friday before CRT began (on the following Monday), then each Friday after completion of RT for 7 weeks, and two more doses after completion of CRT i.e. 10 doses in total (total dose = 600 µg/kg) Gp B: same schedule with matching placebo Compliance: 99 participants (Gp A: 67; Gp B: 32) received at least one dose of their allocated intervention; 69 participants completed the full course (Gp A: 47; Gp B: 22); mean number of doses (Gp A: 8.4; Gp B: 9.1) Duration of treatment: 9 weeks (10 doses) Outcomes  Oral mucositis: NCI-CTC (2.0) 0 to 4 scale (measured weekly by a radiation oncologist for the first 12 weeks, reported as incidence of grade 2 to 4 i.e. moderate to severe, and grade 3 to 4 i.e. severe) (duration, time to onset, and cumulative radiotherapy dose at onset of grade 2+ and 3+ oral mucositis also measured but not outcomes of this review)  Interruptions to cancer treatment (unscheduled radiotherapy breaks: reported as any breaks and breaks longer than 4 days)  Normalcy of diet (measured as incidence of supplemental nutrition by a gastrostomy tube)  Adverse effects (reports collected throughout the 20-week study)  Opioid analgesic use (reported as incidence; number of days of treatment with opioid analgesics is an outcome of this review and therefore we did not use this data)  Dysphagia (not an outcome of this review)  Xerostomia (not an outcome of this review)  Antibiotic use (not an outcome of this review)  Tumour response rate (not an outcome of this review)  Progression-free and overall survival (assessed in a longer-term follow-up study) (not an outcome of this review) Notes Sample size calculation: based on a previous study, 99 participants required to detect a 30% difference in the duration of grade 2 or higher oral mucositis with 80% power provided that the mean duration in the placebo arm was 56 days Funding: "Supported by Amgen Inc" (pharmaceutical industry) Declarations/conflicts of interest: multiple and involving: employment or leadership positions with the funders (Amgen); consultant or advisory roles with the funders and other pharmaceutical companies; stock ownership with the funders; honoraria from the funders and other pharmaceutical companies; research funding from the funders and other pharmaceutical companies Data handling by review authors: the data for incidence of mucositis was presented in subgroups of those that received standard or hyperfractionated RT but we used the overall data in our meta-analyses; for the interruptions to radiotherapy outcome, we used the data for breaks longer than 4 days as this could be pooled with other studies in this comparison Other information of note: the study authors report a greater decrease in incidence due to palifermin in the hyperfractionated subgroup than in the standard subgroup (see figure 3A in the study report)

365

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear Quote: "Patients were randomly assigned..." generation (selection risk bias) Comment: insufficient information to determine method of random sequence generation Allocation Unclear Quote: "Patients were randomly assigned..." concealment (selection risk bias) Comment: insufficient information to determine whether or not the random sequence was adequately concealed Blinding of Low risk Quote: "double blinded" and "Palifermin...or matching participants and placebo was administered by intravenous bolus" personnel (performance bias) Comment: the use of a placebo should have ensured that blinding was successful Blinding of outcome Low risk Quote: "double blinded" and "Palifermin...or matching assessment (detection placebo was administered by intravenous bolus" bias) Comment: it is not clear who was blinded. There are subjective elements to the assessment of oral mucositis using this scale, requiring the patient's assessment of pain/soreness and their ability to swallow but, as the participants were unaware of their group allocation, the assessment of oral mucositis can be considered to be blinded. The other outcomes are objective and therefore unlikely to be affected by any potential lack of blinding of the outcome assessor(s) Incomplete outcome Low risk Overall attrition was 4% (Gp A: 6%; Gp B: 0%) for the oral data (attrition bias) mucositis incidence outcome. The reasons were unclear but this proportion of attrition is unlikely to have biased the results Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Cartee 1995

Methods Trial design: parallel (5 arms) dose-ranging study Location: Duke University Medical Centre, Durham, North Carolina, USA Number of centres: 1 Study duration: not reported Trials registry number: none/unknown Participants Inclusion criteria: premenopausal or perimenopausal patients with histologically confirmed metastatic breast cancer who had chemotherapy for inoperable or metastatic disease; performance status of 0 or 1 (CALGB criteria) Exclusion criteria: metastatic disease involving the central nervous system; pregnant Cancer type: stage IV breast Cancer treatment: AFM regimen (21-day cycle): 5FU (500 mg/m²/day) continuous infusion on days 1 to 5; adriamycin (25 mg/m²) IV bolus on days 3 to

366 5; methotrexate (250 mg/m²) IV on day 15 (if oral mucositis less than grade 3) All participants received G-CSF (filgrastim) subcutaneously (5 µg/kg/day) on days 7 to 13 and from day 16 until resolution of neutropenia Age at baseline (years): mean 44 (not reported by group) Gender: 49 female; 1 male (not reported by group) Number randomised: 50 (not reported by group) Number evaluated: 45 (Gp A: 36; Gp B: 9) Interventions Comparison: GM-CSF (molgramostim) versus placebo Gp A: GM-CSF  (n = 9 analysed): 15 ml of mouthwash (0.01 µg/ml) gently swirled in the mouth for 2 mins before expectorating it; 4 times daily (after mealtimes, even if a meal was not eaten, and at bedtime) after routine oral care procedures; no eating/drinking for 15 mins after using mouthwash; beginning within 24 hours of the start of 5FU and continued for the 21-day AFM cycle (total dose = 12.6 µg)  (n = 9 analysed): same schedule with 0.1 µg/ml mouthwash (total dose = 126 µg)  (n = 9 analysed): same schedule with 1 µg/ml mouthwash (total dose = 1260 µg)  (n = 9 analysed): same schedule with 10 µg/ml mouthwash (total dose = 12,600 µg) Gp B: (n = 9 analysed) same schedule with matching placebo mouthwash Compliance: (not reported by group) mouthwash therapy was discontinued if the participant experienced oral mucositis of grade 3 or above; 30 participants took at least 80% of their prescribed doses; 11 participants discontinued mouthwash therapy within 3 days prior to day 15; 4 participants discontinued mouthwash therapy between day 15 and day 21 Duration of treatment: 21 days (1st treatment cycle of AFM) Outcomes  Oral mucositis: CALGB 0 to 4 scale (measured on days 1 to 5, 8 to 10, 15 and 22, reported as incidence of grade 3 to 4 i.e. severe) (duration of grade 3 to 4 oral mucositis also measured but not an outcome of this review)  Adverse effects (assessed during study period)  Blood measurements (platelet, WBC, granulocyte, lymphocyte) (not an outcome of this review)  Myelosuppression (not an outcome of this review)  GM-CSF plasma concentrations (not an outcome of this review) Notes Sample size calculation: this was done but the numbers required are not reported Funding: "supported in part by National Cancer Institute (Bethesda, MD) grant number PO1-47741-A4" Declarations/conflicts of interest: not reported Data handling by review authors: we combined the four GM-CSF groups to make a single pairwise comparison against placebo and, in order to make a head- to-head comparison of doses, we grouped the two lower doses (0.01 and 0.1 µg/ml) together and grouped the two higher doses (1 and 10 µg/ml) together to make pairwise groups for comparison Other information of note: not reported

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "Patients were randomized by the Duke Cancer generation (selection Center Protocol Office according to a block randomization

367 bias) scheme and assigned a unique identifier number which designated the GM-CSF dose level to be received"

Comment: method of sequence generation not fully described but was done by a dedicated specialist centre so was probably done adequately Allocation Low risk Quote: "Patients were randomized by the Duke Cancer concealment (selection Center Protocol Office" and "The patient supply of bias) mouthwash was labelled to correspond with the assigned identifier number and dispensed by the Pharmacy. The patient assignment information was maintained by the Pharmacy"

Comment: the entire randomisation process was performed by third party so the random sequence is unlikely to have been manipulated Blinding of Low risk Quote: "double blind, placebo controlled" participants and personnel Comment: the use of a placebo should have ensured that (performance bias) blinding was successful Blinding of outcome Low risk Quote: "double-blind, placebo-controlled" and "The patient assessment (detection assignment information was maintained by the Pharmacy" bias) Comment: it is not clear who was blinded. There are subjective elements to the assessment of oral mucositis using this scale, requiring the patient's assessment of pain/soreness and their ability to swallow but, as the participants were unaware of their group allocation, the assessment of oral mucositis can be considered to be blinded Incomplete outcome Unclear Overall attrition was 10%. Reasons for attrition fully data (attrition bias) risk reported. If all participants would have developed severe oral mucositis or dropped out due to severe oral mucositis and were all from one particular group, this would have biased the results. However, attrition was not reported by group, so it is unclear

Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Cesaro 2013

Methods Trial design: parallel (2 arms) Location: Italy Number of centres: 4 Study duration: May 2007 to June 2011 Trials registry number: EudraCT 2007-001430-14 (mentioned in trial report) Participants Inclusion criteria: aged 0 to 17 years with leukaemia, lymphoma or solid tumour; due to receive a first autologous PBSCT Exclusion criteria: not reported Cancer type: leukaemia/lymphoma (Gp A: 22%; Gp B: 17%); solid (Gp A:

368 78%; Gp B: 83%) (neuroblastoma, Ewing sarcoma/peripheral neutroectodermal tumour, medulloblastoma, Wilms tumour, central nervous system tumour) Cancer treatment: all participants had autologous PBSCT on day 0 but differed in conditioning regimen as follows:  Chemotherapy: multiple regimens involving 1 to 4 chemotherapy drugs; the most common regimen was busulfan 16 mg/kg with melphalan 140 mg/m² (Gp A: 53%; Gp B: 37%)  Radiotherapy: only 4 participants (2 in each arm) had TBI 12 to 14.4 Gy prior to their chemotherapy Age at baseline (years): Gp A: median 11.1 (range 1.7 to 17.4); Gp B: median 11.9 (range 1.6 to 17.2) Gender: Gp A: 66% male; Gp B: 59% male Number randomised: 61 (Gp A: 32; Gp B: 29) Number evaluated: 61 (Gp A: 32; Gp B: 29) Interventions Comparison: G-CSF (pegfilgrastim) versus G-CSF (filgrastim) Gp A: pegfilgrastim single dose (100 µg/kg; maximum 6 mg) injected on day 3 Gp B: filgrastim (5 µg/kg per day; maximum 300 µg per day) injected by 9 or more doses starting on day 3 (total dose = at least 45 µg/kg) Mode of administration not decribed but presumably subcutaneously as in other G-CSF studies Compliance: all participants received their allocated intervention with no discontinuations Duration of treatment: Gp A: 1 day; Gp B: 9 or more days Outcomes  Oral mucositis: WHO 0 to 4 scale (reported as incidence of any mucositis and grade 2 to 4 i.e. moderate to severe) (duration of any mucositis also measured but not an outcome of this review)  Normalcy of diet (measured as incidence of TPN) (duration of TPN also measured but not used for analysis in review)  Adverse events  Number of days in hospital (reported as median and range, unable to use data)  Polymorphonuclear cell recovery (not an outcome of this review)  Time to platelet engraftment (not an outcome of this review)  Incidence of febrile neutropenia and proven infection (not an outcome of this review)  Duration of IV antibiotics (not an outcome of this review)  Survival (not an outcome of this review) Notes Sample size calculation: based on the non-inferiority of pegfilgrastim versus filgrastim in speeding the recovery of polymorphonuclear cells Funding: "The authors have no support or funding to report" Declarations/conflicts of interest: "The authors have declared that no competing interests exist" Data handling by review authors: N/A Other information of note: G-CSF administration only began after the chemotherapy and PBSCT were completed, by which point oral mucositis may have already begun to develop

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "A computer-generated randomisation list was generation (selection drawn up at Data Office Centre of AIEOP in Bologna, bias)

369 Italy, by a statistician not involved in patient management"

Comment: adequate method used Allocation Low risk Quote: "The list was stored by sequentially numbered concealment (selection sealed envelopes that was concealed to investigators until bias) the completion of recruitment. The local investigator...assigned each eligible patient to randomization list by phoning to AIEOP Data Office Centre"

Comment: ideal method of concealment used Blinding of High risk Treatment regimens were different so blinding not possible participants and personnel (performance bias) Blinding of outcome Unclear It would be possible to blind the outcome assessor for oral assessment (detection risk mucositis, but it was not mentioned bias) Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Chi 1995

Methods Trial design: cross-over (2 arms) Location: Cancer Centre and Department of Otolaryngology, Veterans General Hospital, Taiwan, Republic of China Number of centres: 1 Study duration: not reported Trials registry number: none/unknown Participants Inclusion criteria: diagnosed stage IV SCC of head and neck, previously untreated or locally recurrent after previous surgery or radiotherapy or both; ECOG score of 2 or above; adequate bone marrow, liver and renal function Exclusion criteria: concurrent medical illness; local radiotherapy to oropharynx region in the previous 3 months Cancer type: head and neck: nasopharyngeal (Gp A: 44%; Gp B: 27%); tongue (Gp A: 22%; Gp B: 36%); hypopharynx (Gp A: 11%; Gp B: 18%); buccal (Gp A: 11%; Gp B: 9%); tonsilor (Gp A: 11%; Gp B: 9%) Cancer treatment: PFL regimen (21-day cycle): cisplatin (20 mg/m²/day), 5FU (800 mg/m²/day) and leucovorin (90 mg/m²/day) IV for days 1 to 4; cycle repeated every 3 weeks (study consisted of two cycles) Age at baseline (years): Gp A: median 44 (range 36 to 62); Gp B: median 49 (range 40 to 66) Gender: Gp A: 89% male; Gp B: 91% male Number randomised: 20 (Gp A: 9; Gp B: 11) - figures for first cycle Number evaluated: 20 (Gp A: 9; Gp B: 11) Interventions Comparison: GM-CSF versus no treatment Gp A: GM-CSF (4 µg/kg) subcutaneously from day 5 to 14 (total dose = 40 µg) Gp B: no treatment Compliance: not reported Duration of treatment: 10 days (days 5 to 14 of 21-day cycle)

370 Outcomes  Oral mucositis: RTOG 0 to 4 scale (measured on days 5 to 21 by both physician's objective gross score and participant's subjective functional score, reported as area under the curve and also in the text as incidence of severe gross mucositis i.e. grade 3 to 4) (duration of grade 2 to 4 and 3 to 4 oral mucositis also measured but not an outcome of this review)  Adverse effects (assessed across both cycles) Notes Sample size calculation: not reported Funding: "supported in part by Department of Health, Taiwan, Republic of China, research grant no. DOH 83-HR-202" and "GM-CSF (supplied by Schering Plough Corp, Kenilworth, NJ)" (pharmaceutical industry) Declarations/conflicts of interest: not reported Data handling by review authors: as stated in the methods section, we would only include first-period data from cross-over studies due to potential for period effects (which were reported in this study). The only usable data in this study were reported in the text as incidence of severe gross mucositis for the first cycle. Other information of note: the authors report a significant period effect of GM- CSF (P < 0.01), whereby the benefits continued into the second cycle. GM-CSF administration only began after the 4-day chemotherapy was completed, by which point oral mucositis may have already begun to develop

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear risk Quote: "Patients were randomized to receive GM-CSF or generation (selection no therapy" bias) Comment: insufficient information to determine method of random sequence generation Allocation Unclear risk Quote: "Patients were randomized to receive GM-CSF or concealment (selection no therapy" bias) Comment: insufficient information to determine whether or not the random sequence was adequately concealed Blinding of High risk Comparison with no treatment so blinding not possible participants and personnel (performance bias) Blinding of outcome Unclear risk It would be possible to blind the outcome assessor, as the assessment (detection data we used was assessed by a physician using an bias) objective scale. However, it was not mentioned Incomplete outcome Low risk All randomised participants appear to be included in the data (attrition bias) analyses Selective reporting Low risk Although most of the data was not usable in this review, (reporting bias) this does not seem to be due to selective reporting Other bias Low risk No other sources of bias are apparent

Crawford 1999

Methods Trial design: parallel (2 arms) Location: USA Number of centres: 14

371 Study duration: recruitment from May 1988 to November 1989 Trials registry number: none/unknown Participants Inclusion criteria: newly diagnosed small-cell lung cancer meeting standard criteria for end-organ function; ECOG score of 0 to 2 Exclusion criteria: previous radiotherapy; other serious medical illnesses precluding participation Cancer type: small-cell lung cancer Cancer treatment: CAE regimen (21-day cycle): cyclophosphamide (1000 mg/m²) and doxorubicin (50 mg/m²) on day 1; etoposide (120 mg/m²) on days 1 to 3; all by IV; repeated for up to 6 cycles Age at baseline (years): Gp A: mean 61 (SD 10; range 31 to 78); Gp B: mean 62 (SD 8; range 31 to 80) Gender: Gp A: 65% male; Gp B: 63% male Number randomised: 211 (Gp A: 101; Gp B: 110) Number evaluated: 195 (Gp A: 93; Gp B: 102) - figures for first cycle Interventions Comparison: G-CSF (r-metHuG-CSF) (filgrastim) versus placebo Gp A: G-CSF (230 µg/m²) self-administered subcutaneously on days 4 to 17 (total dose = 3220 µg/m²) Gp B: as above but with placebo G-CSF stopped if postnadir neutrophil count exceeded 10 x 10⁹/litre after day 12; participants kept receiving their allocated intervention until they experienced fever with neutropenia, then they received unblinded G-CSF (230 µg/m²) in subsequent cycles; participants in the G-CSF group who experienced fever with neutropenia were allowed 25% reduction in chemotherapy dosages in subsequent cycles Compliance: not reported Duration of treatment: 14 days during a 21-day cycle Outcomes  Oral mucositis: WHO 0 to 4 scale (measured weekly, reported as incidence of any mucositis) (duration and time to onset of oral mucositis also measured but not an outcome of this review)  Adverse effects (assessed over the six cycles)  Number of days in hospital (reported graphically in secondary trial report with no SD or P-value, no usable data)  Incidence, duration and severity of fever with neutropenia (not an outcome of this review)  Incidence and duration of antibiotic use (not an outcome of this review) Notes Sample size calculation: based on a difference of 20% in the incidence of fever with neutropenia over the six cycles Funding: "The study was designed, coordinated, and analyzed in conjunction with Amgen, the supplier of the G-CSF" Declarations/conflicts of interest: not reported but some authors were employed by Amgen (pharmaceutical industry) Data handling by review authors: for oral mucositis, we only used the data from the first cycle due to the reasons listed above (under 'Interventions') Other information of note: G-CSF administration only began after the 3-day chemotherapy was completed, by which point oral mucositis may have already begun to develop

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear Quote: "The patients were randomly assigned to generation (selection risk chemotherapy followed by study drug (either placebo or G-

372 bias) CSF)"

Comment: insufficient information to determine method of random sequence generation Allocation Unclear Quote: "The patients were randomly assigned to concealment (selection risk chemotherapy followed by study drug (either placebo or G- bias) CSF)"

Comment: insufficient information to determine whether or not the random sequence was adequately concealed Blinding of Low risk Quote: "Placebo was supplied in matching vials for double participants and blinding" personnel (performance bias) Comment: the use of a placebo should have ensured that blinding was successful Blinding of outcome Low risk Quote: "Placebo was supplied in matching vials for double assessment (detection blinding" bias) Comment: it is not clear who was blinded. There are subjective elements to the assessment of oral mucositis using this scale, requiring the patient's assessment of pain/soreness and their ability to swallow but, as the participants were unaware of their group allocation, the assessment of oral mucositis can be considered to be blinded Incomplete outcome Low risk Overall attrition was 8% (Gp A: 8%; Gp B: 7%) for the oral data (attrition bias) mucositis incidence outcome. The reasons were reported and similar between groups, and this proportion of attrition is unlikely to have biased the results Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Dazzi 2003

Methods Trial design: parallel (2 arms) Location: Ravenna, Italy Number of centres: 1 Study duration: recruitment from July 1997 to February 2002 Trials registry number: none/unknown Participants Inclusion criteria: 14 years of age or older; hospitalised for high-dose chemotherapy with autologous PBSCT Exclusion criteria: not reported Cancer type: breast (Gp A: 17.5%; Gp B: 27.5%); Ewing's sarcoma (Gp A: 41.5%; Gp B: 28.5%); osteosarcoma (Gp A: 13%; Gp B: 14%); NHL (Gp A: 15%; Gp B: 11.5%); germ cell tumours (Gp A: 11%; Gp B: 16%); small-cell lung (Gp A: 2%; Gp B: 0%); soft tissue sarcoma (Gp A: 0%; Gp B: 2.5%) Cancer treatment: high-dose chemotherapy with autologous PBSCT; chemotherapy regimens were categorised into high risk and low risk and this was used as a stratification factor for randomisation, therefore high- and low-risk participants were equally distributed across groups All participants received subcutaneous G-CSF (300 µg/day) until haematopoietic reconstitution Age at baseline (years): Gp A: median 29 (range 15 to 57); Gp B: median 29

373 (range 17 to 61) Gender: Gp A: 59% male; Gp B: 55% male Number randomised: 90 (Gp A: 46; Gp B: 44) Number evaluated: 90 (Gp A: 46; Gp B: 44) Interventions Comparison: GM-CSF versus placebo Gp A: GM-CSF mouthwash 150 µg/day in 100 cm³ of sterile water taken in 4 doses per day; mouthrinsing performed for 1 min each time; treatment started on the day after the completion of chemotherapy and continued until bone marrow recovery (absolute neutrophil count > 500/mm³) or resolution of mucositis if still persistent after bone marrow recovery (total dose = variable) Gp B: as above but with placebo (sterile water) All participants received 0.2% oral chlorhexidine and amphotericin B Compliance: all but seven participants regularly completed mouthwashes: one participant in the placebo group had none due to persistent vomiting; six (four in placebo group and two in GM-CSF group) started treatment but interrupted it early due to nausea and vomiting Duration of treatment: variable and dependent on bone marrow recovery/resolution of mucositis Outcomes  Oral mucositis: NCI-CTC 0 to 4 scale (measured daily by the physicians, reported as incidence of mucositis and incidence of grade 3 to 4 i.e. severe) (duration of grade 3 to 4 oral mucositis also measured but not an outcome of this review)  Oral pain: 0 to 10 VAS (self-evaluated daily, reported as mean worst score experienced)  Number of days of treatment with opioid analgesics (also reported as incidence; we did not use this data) Notes Sample size calculation: 90 participants required to detect 25% minimal difference in the rate of severe mucositis at 90% power and 5% significance Funding: no external funding (from correspondence with authors) Declarations/conflicts of interest: not reported Data handling by review authors: the data for incidence of mucositis was presented in subgroups of those at low or high risk of mucositis but we used the overall data in our meta-analyses Other information of note: there does not appear to be any difference in risk of any or severe mucositis between the low- and high-risk subgroups. GM-CSF administration only began after the chemotherapy was completed, by which point oral mucositis may have already begun to develop

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote from correspondence with authors: "The generation (selection randomization list was centralized" bias) Comment: centralised randomisation method - likely to be done properly Allocation Low risk Quote from correspondence with authors: "The concealment (selection randomization list was centralized" bias) Comment: centralised randomisation method - likely to be done properly Blinding of Low risk Quote: "double blind, randomized, placebo-controlled participants and study" and "The color, odor, texture and taste of both

374 personnel solutions were virtually identical" (performance bias) Comment: the use of a placebo should have ensured that blinding was successful Blinding of outcome Low risk Quote: "double blind, randomized, placebo-controlled assessment (detection study" bias) Comment: it is not clear who was blinded. There are subjective elements to the assessment of oral mucositis using this scale, requiring the patient's assessment of pain/soreness and their ability to swallow but, as the participants were unaware of their group allocation, the assessment of oral mucositis can be considered to be blinded Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Fink 2011

Methods Trial design: parallel (2 arms) Location: Department of Hematology and Oncology, University Hospital Freiburg, Germany Number of centres: 1 Study duration: March 2006 to December 2010 Trials registry number: EudraCT 2008-001833-87; DRKS00000043 Participants Inclusion criteria: adults aged 18 to 75 years with either: high-grade non- Hodgkin's lymphoma with high-risk syndrome (> 2 risk factors according to age- adapted IPI = international prognostic index) in the first complete remission; Hodgkin's lymphoma in the first recurrence; recurrence of follicular lymphoma; primary therapy of a coat-cell lymphoma (MCL) in stage II-IV; due to receive BEAM chemotherapy followed by autologous PBSCT; Karnofsky performance score more than 60%; life expectancy more than 3 months Exclusion criteria: previous therapy using palifermin; severe concomitant diseases with organ failure; pregnancy, lactation, positive pregnancy test; hypersensitivity to one of the trial drugs; severe psychiatric illness; HIV disease or immunologic deficiency; known central nervous system involvement Cancer type: haematologic: diffuse large-cell lymphoma (Gp A: 33%; Gp B: 42%); B-cell type acute lymphocytic leukaemia (Gp A: 10%; Gp B: 6%); T-cell non-Hodgkin's lymphoma (Gp A: 13%; Gp B: 11%); follicular/mantle cell lymphoma (Gp A: 27%; Gp B: 28%); Hodgkin's lymphoma (Gp A: 17%; Gp B: 14%) Cancer treatment: prior to receiving autologous PBSCT on day 0, participants received BEAM conditioning regimen from day -8 to -2: carmustine (BCNU) 300 mg/m²; etoposide 800 mg/m²; cytosine arabinoside 1600 mg/m²; melphalan 140 mg/m² Age at baseline (years): (ITT population) Gp A: median 50 (range 22 to 71); Gp B: median 55 (range 22 to 73) Gender: (ITT population) Gp A: 57% male; Gp B: 61% male Number randomised: 73 (Gp A: 37; Gp B: 36) Number evaluated: ITT: 66 (Gp A: 30; Gp B: 36); PP: 54 (Gp A: 22; Gp B: 32) Interventions Comparison: KGF (palifermin) plus best supportive care versus best

375 supportive care alone Gp A: KGF (60 µg/kg) by IV daily for 3 days (days -10, -9, -8) prior to conditioning regimen and autologous PBSCT and then for 3 days after (days 0, 1, 2) (total dose = 360 µg/kg) Gp B: best supportive care ("effective oral hygiene like teeth brushing, oral rinsing") beginning on day -8 (the day of hospital admission for BEAM conditioning) Compliance: Gp A: 7/37 withdrew before therapy started, 3/37 had a different conditioning regimen to that specified in the study protocol (unclear if they still received KGF), 5/37 either had no KGF or did not receive all doses; Gp B: 4/36 had a different conditioning regimen to that specified in the study protocol (unclear if they still received control intervention) Duration of treatment: 6 treatment days (over 13 days) Outcomes  Oral mucositis: WHO 0 to 4 scale (measured daily during hospital stay by trained nurses, study assistant, or treating physician, maximum score reported) (duration of oral mucositis also measured but not an outcome of this review)  Normalcy of diet (measured as incidence of TPN)  Adverse events  Number of days in hospital (medians reported, unable to use data)  Number of days of treatment with opioid analgesics (medians reported, unable to use data)  Survival (not an outcome of this review)  Febrile neutropenia (not an outcome of this review) Notes Sample size calculation: 76 participants required to detect 30% difference in the rate of severe mucositis at 80% power and 5% significance Funding: Amgen (pharmaceutical industry) Declarations/conflicts of interest: not reported Data handling by review authors: data for ITT population used Other information of note: not reported

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear Quote: "randomly assigned" generation (selection risk bias) Comment: insufficient information to determine method of random sequence generation Allocation Unclear Quote: "randomly assigned" concealment (selection risk bias) Comment: insufficient information to determine whether or not the random sequence was adequately concealed Blinding of High risk Quote: "The randomization result was known to the patient participants and as well as to the practitioners before the start of therapy" personnel (performance bias) Blinding of outcome High risk Quote: "The randomization result was known to the patient assessment (detection as well as to the practitioners before the start of therapy" bias) Incomplete outcome High risk Overall attrition was 10% (Gp A: 19%; Gp B: 0%) for the data (attrition bias) ITT population. All 7 participants died before threapy started. Although this reason is not related to the outcomes,

376 the balance created by randomisation may have been lost Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Freytes 2004

Methods Trial design: parallel (3 arms) Location: USA Number of centres: 8 Study duration: not reported Trials registry number: none/unknown Participants Inclusion criteria: aged 18 years or older; due to receive autologous HSCT with a conditioning regimen with a high propensity for producing mucositis (typically 50% incidence of NCI-CTC grade 3 or 4 when receiving standard mucositis management); Karnofsky performance score of 70%; free of acute or significant chronic dental or periodontal disease at baseline examination Exclusion criteria: previous HSCT; visible oral ulcerations at screening; pregnant or breastfeeding; childbearing potential or not using adequate contraception; history of allergy to Escherichia coli-derived products; posterior subcapsular cataract identified at screening; history of thyroid disease prior to receiving chemotherapy (except for hypothyroidism adequately controlled with replacement therapy); history or clinical evidence of active significant acute or chronic diseases that may affect evaluation or interpretation of the effects of the study medication on mucositis; following medications: interleukin 11, topical steroids, sucralfate, hydrogen peroxide, pilocarpine, misoprostol, oral chlorhexidine rinses, or any agent that would affect the assessment of changes in the appearance of mucositis during the study Cancer type: lymphoma (Gp A: 64%; Gp B: 71%; Gp C: 57%); other haematologic malignancy (Gp A: 36%; Gp B: 29%; Gp C: 43%) Cancer treatment: prior to receiving autologous HSCT, participants received the following conditioning regimens:  CBV (cyclophosphamide, etoposide, and carmustine) (Gp A: 50%; Gp B: 64%; Gp C: 43%)  melphalan monotherapy (Gp A: 7%; Gp B: 21%; Gp C: 14%)  melphalan combination (Gp A: 21%; Gp B: 14%; Gp C: 28%)  cyclophosphamide + TBI (total body irradiation) (Gp A: 7%; Gp B: 0%; Gp C: 14%)  thiotepa + TBI (Gp A: 7%; Gp B: 0%; Gp C: 0%)  cyclophosphamide + busulfan (Gp A: 7%; Gp B: 0%; Gp C: 0%) Age at baseline (years): Gp A: mean 54 (SD 10); Gp B: mean 47 (SD 10); Gp C: mean 51 (SD 15) Gender: Gp A: 79% male; Gp B: 64% male; Gp C: 79% male Number randomised: 42 (Gp A: 14; Gp B: 14; Gp C: 14) Number evaluated: 42 (Gp A: 14; Gp B: 14; Gp C: 14) Interventions Comparison: KGF-2 (repifermin) versus placebo Gp A: KGF-2 (25 µg/kg) by IV daily for 3 days prior to conditioning regimen and autologous HSCT and then for 10 days after (total dose = 325 µg/kg) Gp B: KGF-2 (50 µg/kg) as above (total dose = 650 µg/kg) Gp C: placebo as above Compliance: not reported Duration of treatment: 13 treatment days (over a longer period dependent on conditioning regimen) Outcomes  Oral mucositis: NCI-CTC 0 to 4 scale (incidence of grade 2, 3 or 4, assessed

377 prior to conditioning regimen, on day of HSCT, then 3 times per week until resolution of mucositis)  Oral mucositis: OMAS 0 to 45 scale (reported as mean worst score and mean 3 worst scores, NCI-CTC data used for analysis)  Oral and orophayngeal pain: 0 (no pain) to 10 (worse pain) scale (assessed on days reported above, reported as mean worst score experienced)  Normalcy of diet: ability to eat 1 to 4 score, where 1 = normal, 2 = only soft solids, 3 = only liquids, 4 = no solids or liquids (assessed on days reported above, reported as mean worst score)  Adverse events (assessed from the start of the intervention until 28 days after the final dose, reported as events with a statistically significant difference between groups or that occurred in at least 50% of participants in any group and differed between groups by at least 10%)  Number of days of treatment with opioid analgesics (assessed as reported for oral mucositis outcome, reported as mean number of days due to mucositis pain, and mean number of days due to all pain; we used the former although acknowledge other studies typically do not specify whether or not they are reporting usage due to mucositis pain)  Pain on swallowing (not an outcome of this review)  Laboratory parameters (not outcomes of this review)  Immunogenicity (not an outcome of this review)  Electrocardiogram abnormalities, chest x-ray assessments and ophthalmologic examinations (not outcomes of this review) Notes Sample size calculation: not reported Funding: not reported Declarations/conflicts of interest: not reported Data handling by review authors: we combined the two KGF-2 groups to make a single pairwise comparison against placebo and we also made a comparison of the two different KGF-2 dosages against each other Other information of note: not reported

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote from correspondence with authors: "The method to generation (selection implement the random allocation was by central telephone" bias) Comment: centralised randomisation method - likely to be done properly Allocation Low risk Quote from correspondence with authors: "The method to concealment (selection implement the random allocation was by central telephone" bias) Comment: centralised randomisation method - likely to be done properly Blinding of Low risk Quote: "double-blinded, placebo-controlled" participants and personnel Comment: the use of a placebo and identical schedule of (performance bias) treatment for all three arms should have ensured that blinding was successful Blinding of outcome Low risk Quote: "double-blinded, placebo-controlled" assessment (detection bias) Comment: it is not clear who was blinded. There are subjective elements to the assessment of oral mucositis

378 using the NCI-CTC scale, requiring the patient's assessment of pain/soreness and their ability to swallow but, as the participants were unaware of their group allocation, the assessment of oral mucositis can be considered to be blinded. This would be the same for other subjective outcomes. The objective outcomes are unlikely to be affected by any potential lack of blinding of the outcome assessor(s) Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Gholizadeh 2016

Methods Trial design: parallel (2 arms) Location: Iran Number of centres: not reported Study duration: not reported Trials registry number: IRCT2013021812510N1 (mentioned in trial report) Participants Inclusion criteria: previously untreated acute lymphoblastic leukaemia patients; aged between 5 and 18 years Exclusion criteria: any other systemic disease; presence of oral mucositis or other oral lesions prior to chemotherapy; history of dermatology or respiratory hypersensitivity; acute lymphoblastic leukaemia recurrence Cancer type: acute lymphoblastic leukaemia (ALL) Cancer treatment: induction chemotherapy protocol consisted of standard risk B-precursor ALL (COG)/dexamethasone, vincristine, L-asparaginase, intrathecal (methotrexate + ara-C + hydrocortisone). The intensification protocol was dexamethasone, vincristine, L-asparaginase/ dexamethasone, cyclophosphamide/6-thioguanine + cytarabine + intrathecal methotrexate Age at baseline (years): Gp A: mean 8.8 (SD 2.5); Gp B: mean 8.4 (SD 2.2); overall range: 5 to 18 Gender: Gp A: 49% male; Gp B: 49% male Number randomised: 90 (Gp A: 45; Gp B: 45) Number evaluated: 90 (Gp A: 45; Gp B: 45) Interventions Comparison: KGF (palifermin) versus chlorhexidine Gp A: KGF (60 µg/kg) by IV bolus daily for 3 days prior to chemotherapy regimen and then for 3 days after (total dose = 360 µg/kg) Gp B: chlorhexidine (concentration not reported) mouthwash used for 1 min once daily for 3 days prior to chemotherapy regimen and then for 3 days after Compliance: not reported Duration of treatment: 6 treatment days (over an unspecified longer period) Outcomes  Oral mucositis: WHO 0 to 4 scale (measured after 1 and 2 weeks and reported separately for each timepoint)  Adverse events Notes Sample size calculation: not reported Funding: not reported Declarations/conflicts of interest: "There is no conflict of interest in relation to this study" Data handling by review authors: we report the data at 2 weeks as they represent the maximum oral mucositis score experienced better than those at 1 week

379 Other information of note: not reported

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "The patients were randomly assigned to the generation (selection palifermin or control group by using the table of random bias) numbers"

Comment: adequate method used Allocation Unclear Quote: "The patients were randomly assigned to the concealment (selection risk palifermin or control group by using the table of random bias) numbers"

Comment: insufficient information to determine whether or not the random sequence was adequately concealed Blinding of High risk Comparison with chlorhexidine so blinding not possible participants and personnel (performance bias) Blinding of outcome Low risk Quote: "Each patient was evaluated for oral lesions one and assessment (detection two weeks after the chemotherapy completion by the same bias) specialist who was blind to the type of treatment" and "This limited use of chlorhexidine was to prevent the adverse effects like tooth discoloration and temporally taste changes"

Comment: grade 1 on this scale would require the unblinded participant's assessment of soreness but other aspects of the scale are more objective and were assessed by a blinded assessor. Also, blinding may not have been broken by staining/discolouration due to limited use of chlorhexidine Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Henke 2011

Methods Trial design: parallel (2 arms) Location: Australia, Canada, and Europe (Austria, France, Germany, Italy, Spain, UK) Number of centres: 38 Study duration: recruitment from January 2005 to August 2007 Trials registry number: NCT00131638; 2004-002016-28 (EudraCT number) Participants Inclusion criteria: more than 18 years old; resected for pathohistologically documented high-risk stage 2 to 4B SCC of the oral cavity, oropharynx, hypopharynx, or larynx; ECOG score of 0 to 2; at least two of nine areas of the oral or oropharyngeal mucosa due to receive at least 50 Gy RT Exclusion criteria: tumours of the lips, paranasal sinuses, salivary glands, or

380 unknown primary site; metastatic disease; history of chronic pancreatitis or acute pancreatitis within the last year; prior RT to the head and neck region or prior chemotherapy; previous treatment on this study or with other KGFs Cancer type: head and neck: oropharynx (Gp A: 47%; Gp B: 48%); oral cavity (Gp A: 32%; Gp B: 27%); larynx (Gp A: 11%; Gp B: 15%); hypopharynx (Gp A: 10%; Gp B: 10%); other (Gp A: 1%; Gp B: 1%) Cancer treatment: after R0 or R1 resection:  Radiotherapy: standard fractionation of once daily 2-Gy fractions, 5 days per week; total 60 Gy (for R0 resection) over 6 weeks, or 66 Gy (for R1 resection) over 7 weeks, both with allowable range of ± 15%  Chemotherapy: cisplatin (100 mg/m²) by IV after appropriate hydration on days 1 and 22 (for R0 resection), or days 1, 22 and 43 (for R1 resection) Age at baseline (years): Gp A: mean 56 (SD 8); Gp B: mean 57 (SD 9) Gender: Gp A: 85% male; Gp B: 80% male Number randomised: 186 (Gp A: 92; Gp B: 94) Number evaluated: 186 (Gp A: 92; Gp B: 94) Interventions Comparison: KGF (palifermin) versus placebo Gp A: KGF (120 µg/kg) 3 days prior to start of, and then once per week during radiochemotherapy, i.e. 7 doses for those with R0 resection, 8 doses for those with R1 resection (total dose = 840 or 960 µg/kg respectively) Gp B: same schedule with placebo Mode of administration not described but presumably IV as in other KGF studies Compliance: 78% of participants in KGF group completed all planned doses compared to 86% in placebo group Duration of treatment: 7 or 8 treatment days (over 7 or 8 weeks), depending on R0 or R1 resection respectively Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed twice weekly by trained evaluators during radiochemotherapy and then until either mucositis had reduced to grade 2 or lower or week 15, whichever occured first, maximum score reported) (duration and time to onset of grade 3 to 4 oral mucositis also measured but not outcomes of this review)  Interruptions to cancer treatment: incidence of 5 or more missed consecutive RT fractions; incidence of chemotherapy delays/discontinuations  Oral pain: OMWQ-HN 0 (no soreness) to 4 (extreme soreness) scale for mouth and throat soreness  Normalcy of diet (measured as incidence of supplemental feeding by TPN, PEG, nasogastric tube, or IV hydration) (broken down by overall supplemental feeding and also where due to OM; we used the latter although acknowledge other studies don't specify reason for supplemental feeding)  Adverse events: reported as those with a difference in incidence of at least 5% between arms  Use of opioid analgesics (total dose reported but not an outcome of this review)  Xerostomia (not an outcome of this review)  Weight change (not an outcome of this review)  Laboratory assessments (not an outcome of this review)  Survival (not an outcome of this review) Notes Sample size calculation: assuming 60% of placebo group would develop grade 3 to 4 mucositis, 90 per group required to detect a reduction of at least 25% at 90% power and 5% significance Funding: "This study was supported by Amgen" (Amgen also named as sponsor on trials registry - pharmaceutical industry) Declarations/conflicts of interest: some authors had both employment or leadership positions and stock ownership within Amgen Data handling by review authors: N/A

381 Other information of note: study originally randomised participants to 3 arms (180 µg/kg once per week for 7 weeks, 180 µg/kg once per week for 4 weeks followed by placebo for the next 3 doses, or placebo throughout) but, after one serious adverse event of respiratory insufficiency reported in one of the first 10 participants, the data monitoring committee decided to restart the study using 120 µg/kg doses, excluding the 17 randomised participants from the efficacy assessments. The arm with KGF for 4 weeks followed by placebo was stopped due to slow recruitment, after enrolment of 38 participants, and the results analysed in a separate appendix

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "Random assignment was made by a centralized generation (selection interactive voice response system" bias) Comment: large multicentre trial using high-tech randomisation method - likely to be done properly Allocation Low risk Quote: "Random assignment was made by a centralized concealment (selection interactive voice response system" bias) Comment: large multicentre trial using high-tech randomisation method - likely to be done properly Blinding of Low risk Quote: "double-blind, randomized, placebo-controlled" participants and personnel Comment: the use of a placebo should have ensured that (performance bias) blinding was successful Blinding of outcome Low risk Quote: "double-blind, randomized, placebo-controlled" assessment (detection bias) Comment: it is not clear who was blinded. There are subjective elements to the assessment of lower grades of oral mucositis using the WHO scale, requiring the patient's assessment of pain/soreness and their ability to swallow. Higher grades have more objective elements so may not be affected by potential lack of blinding of the assessor. This would be the same for other subjective and objective outcomes Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Hosseinjani 2017

Methods Trial design: parallel (2 arms) Location: Tehran University of Medical Sciences, Tehran, Iran Number of centres: 1 Study duration: February 2014 to March 2015 Trials registry number: IRCT2015042518842N8 (mentioned in trial report) Participants Inclusion criteria: aged 18 years or older with non-Hodgkin's lymphoma, Hodgkin's disease or multiple myeloma; due to receive autologous HSCT;

382 adequate cardiac, pulmonary, renal and hepatic function Exclusion criteria: Karnofsky performance score less than 70%; participation in another study using an unlicensed product Cancer type: haematologic: non-Hodgkin's lymphoma (Gp A: 25%; Gp B: 25%); Hodgkin's disease (Gp A: 23%; Gp B: 23%); multiple myeloma (Gp A: 53%; Gp B: 53%) Cancer treatment: prior to receiving autologous HSCT, participants received the following conditioning regimens:  Hodgkin's/non-Hodgkin's: high-dose combination chemotherapy (carboplatin 750 mg/m² IV daily for 2 days, etoposide 300 mg/m² IV daily for 2 days, cytarabine 300mg/m²/dose IV two doses in each day for 2 days and melphalan 140 mg/m² IV for 1 day  Multiple myeloma: high-dose melphalan (100 mg/m² IV daily for 2 days) Age at baseline (years): Gp A: mean 43 (SD 14); Gp B: mean 45 (SD 16) Gender: Gp A: 55% male; Gp B: 48% male Number randomised: 80 (Gp A: 40; Gp B: 40) Number evaluated: 80 (Gp A: 40; Gp B: 40) Interventions Comparison: Erythropoietin (recombinant human) versus placebo Gp A: 50 IU/ml erythropoietin mouthwash in aqueous vehicle (sodium benzoate, sodium citrate, citric acid, sodium hydroxide, sugar and distilled water) supplied in glass bottle stored at 4°C, 15 ml 4 times daily, starting from the first day of conditioning chemotherapy until 14 days after HSCT or until discharge from hospital (i.e. neutrophil recovery), whichever occurred first, oral intake not permitted for 1 hour following mouthwashing Gp B: same schedule with placebo (aqueous vehicle-only) All participants received oral hygiene care in addition to 20 drops of nystatin every 3 hours, mouthwashes containing 10 ml chlorhexidine 0.02% plus 10 ml diluted povidone iodine every 3 hours Compliance: "However, it was a limitation of our study that EPO mouthwash administration might be affected by patients' low compliance" (no data reported) Duration of treatment: variable and dependent on neutrophil recovery Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed daily by single trained pharmacist starting from the first day of conditioning chemotherapy and then until either 21 days after HSCT or mucositis had resolved, whichever occured first, maximum score reported) (duration and time to onset reported but not outcomes of this review)  Number of days in hospital  Use of opioid analgesics (incidence reported but not an outcome of this review)  Blood measurements (not an outcome of this review)  Incidence and duration of fever (not an outcome of this review) Notes Sample size calculation: 40 per group required assuming a 30% decrease in incidence of grade 2 to 4 mucositis at 5% significance and 80% power Funding: "There was no applicable funding source for the clinical trial" Declarations/conflicts of interest: "The authors have no conflict of interests to report" Data handling by review authors: there is a discrepency in the incidence of grade 2 to 4 mucositis between Figure 2 and Table 2 (the latter has one extra event per group). However, this does not change the effect estimate. We have used the data in Table 2 as it reports numbers of participants along with percentages Other information of note: not reported

Risk of bias table

383 Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "Patients were randomly allocated...in a blocked generation (selection randomization schedule" and "Both patient randomization bias) and drug preparation were performed in the pharmaceutical laboratory of Pharmacy Department"

Comment: method of random sequence generation not described but done by university hospital pharmacy and therefore probably done adequately Allocation Low risk Quote: "Both patient randomization and drug preparation concealment (selection were performed in the pharmaceutical laboratory of bias) Pharmacy Department"

Comment: not explicitly described but pharmacy-controlled randomisation should have ensured concealment of the random sequence from those recruiting participants Blinding of Low risk Quote: "double-blind, randomized, placebo-controlled" and participants and "The study participants, the attending physician and the personnel outcome assessor were all blind to the treatment (performance bias) assignment" and "There were no differences in colour, flavour, taste or container of the study drug and the placebo"

Comment: the use of a placebo should have ensured that blinding was successful Blinding of outcome Low risk Quote: "double-blind, randomized, placebo-controlled" and assessment (detection "The study participants, the attending physician and the bias) outcome assessor were all blind to the treatment assignment"

Comment: all parties were blinded Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Jagasia 2012

Methods Trial design: parallel (2 arms) Location: USA (16 sites) and Australia (4 sites) Number of centres: 20 Study duration: December 2005 to November 2008 Trials registry number: NCT00189488 (mentions obsolete number in trial report: NCT00964899) Participants Inclusion criteria: aged 18 years or older with haematologic malignancy (including myelodysplastic syndromes) and due to receive allogeneic SCT (marrow or PBPC) after a conditioning regimen; Karnofsky performance score of 70% or more; related donor or HLA-matched unrelated donor identical at 6/6 HLA-A, -B and -DRB1 loci (molecular typing of class I and class II for unrelated donors) Exclusion criteria: other malignancies; prior SCT; previous use of KGF; active

384 infection or oral mucositis; congestive heart failure (NYHA class III or IV); use of a T-cell depleted graft for GVHD prophylaxis; inadequate renal, liver or pulmonary function; pregnant or breast feeding; refusal to use adequate contraception during study; participation in another investigational device or drug trial in previous 30 days Cancer type: haematologic: leukaemia (Gp A: 71%; Gp B: 79%); myelodysplastic syndrome (Gp A: 16%; Gp B: 12%); non-Hodgkin's lymphoma (Gp A: 12%; Gp B: 8%); multiple myeloma (Gp A: 0%; Gp B: 1%); Hodgkin's disease (Gp A: 1%; Gp B: 0%) Cancer treatment: prior to receiving allogeneic SCT on day 0, participants received one of the following conditioning regimens from day -11 to -2:  cyclophosphamide plus TBI with or without etoposide  TBI plus etoposide  melphalan plus TBI > 11 Gy  busulfan plus cyclophosphamide  busulfan plus melphalan (fully ablative doses)  fludarabine plus melphalan (fully ablative doses) Participants received methotrexate (with a calcineurin inhibitor - either cyclosporine or tacrolimus) for GVHD prophylaxis on days 1, 3 and 6 (planned), and on day 11 (if toxicity allowed) at doses of 15, 10, 10 and 10 mg/m² respectively Age at baseline (years): Gp A: median 42 (range 18 to 62); Gp B: median 44 (range 18 to 64) Gender: Gp A: 52% male; Gp B: 63% male Number randomised: 155 (Gp A: 77; Gp B: 78) Number evaluated: 155 (Gp A: 77; Gp B: 78) Interventions Comparison: KGF (palifermin) versus placebo Gp A: KGF (60 µg/kg) by IV bolus daily for 3 days prior to start of conditioning therapy, then a single 180 µg/kg dose after conditioning, but often 1 or 2 days before SCT (total dose = 360 µg/kg) Gp B: same schedule with placebo Compliance: received at least one dose: Gp A: 99%; Gp B: 96%; received all doses: Gp A: 92%; Gp B: 88% Duration of treatment: 4 treatment days (over roughly 14 days) Outcomes  Oral mucositis: WHO 0 to 4 scale (reported as incidence of grade 2 to 4 i.e. moderate to severe, and grade 3 to 4 i.e. severe, assessed daily by trained evaluators from day -11 (first day of conditioning) and then until hospital discharge or day 28, whichever occured first) (duration also measured but not an outcome of this review)  Normalcy of diet (measured as incidence of TPN)  Adverse events  Use of opioid analgesics (incidence reported but not an outcome of this review)  Incidence and severity of acute GVHD (not an outcome of this review) Notes Sample size calculation: based on GVHD (not met due to early stopping) Funding: "This study was supported by research funding from Amgen Inc. Jonathan Latham of PharmaScribe, LLC received funding from Amgen Inc. to provide assistance with the preparation of the manuscript. Xuesong Guan of Amgen Inc. provided assistance with statistical analyses" (pharmaceutical industry) Declarations/conflicts of interest: some authors were employees and stockholders of Amgen and some received compensation from Amgen for consultation Data handling by review authors: N/A Other information of note: planned sample size was 200 participants but the

385 study was stopped due to slow recruitment

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear Quote: "Subjects were randomly assigned..." generation (selection risk bias) Comment: insufficient information to determine method of random sequence generation Allocation Unclear Quote: "Subjects were randomly assigned..." concealment (selection risk bias) Comment: insufficient information to determine whether or not the random sequence was adequately concealed Blinding of Low risk Quote: "double-blind, placebo-controlled" participants and personnel Comment: the use of a placebo should have ensured that (performance bias) blinding was successful Blinding of outcome Low risk Quote: "double-blind, placebo-controlled" assessment (detection bias) Comment: it is not clear who was blinded but grades 2 to 4 on the WHO scale are sufficiently objective and unlikely to be affected by any lack of blinding of the assessors Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Katano 1995

Methods Trial design: parallel (2 arms) Location: Saga Medical School, Saga, Japan Number of centres: 1 Study duration: not reported Trials registry number: none/unknown Participants Inclusion criteria: breast cancer patients Exclusion criteria: not reported Cancer type: breast: primary advanced (Gp A: 2; Gp B: 5); inflammatory (Gp A: 4; Gp B: 1); recurrent (Gp A: 1; Gp B: 1) Cancer treatment: preoperative IA high-dose adriamycin (10 to 40 mg every 2 to 3 days to a total dose of 70 to 170 mg) Age at baseline (years): Gp A: mean 53 (SD 11; range 38 to 69); Gp B: mean 52 (SD 10; range 45 to 69) Gender: all female Number randomised: 14 (Gp A: 7; Gp B: 7) Number evaluated: 14 (Gp A: 7; Gp B: 7) Interventions Comparison: G-CSF versus no treatment Gp A: G-CSF (125 µg) by daily subcutaneous injection until leukocyte counts > 8000/mm³; timing in relation to chemotherapy not specifically reported, but the group was further divided into two subgroups where one (n = 4) received G-CSF during/as an adjunct to the chemotherapy, and the other (n = 3) received G-CSF afterwards (after the leukocyte counts were likely to drop below 2000/mm³)

386 Gp B: no treatment Compliance: not reported Duration of treatment: variable and dependent on leukocyte recovery (to > 8000/mm³) Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed once 1 to 7 days prior to chemotherapy and then every day afterwards by a single experienced examiner, reported as incidence of grade 2 to 4) (duration of grade 2 to 4 also measured but not an outcome of this review)  Blood measurements (not an outcome of this review)  Alopecia (not an outcome of this review)  Adult respiratory distress syndrome (not an outcome of this review)  Fever (not an outcome of this review) Notes Sample size calculation: not reported Funding: "G-CSF (Neutrogin) was provided by Chugai Pharmaceutical" Declarations/conflicts of interest: not reported Data handling by review authors: the data for incidence of mucositis was presented in subgroups of those receiving G-CSF during or after chemotherapy but we used the overall data in our meta-analyses Other information of note: both cases of mucositis were in the subgroup who received G-CSF after chemotherapy. Oral mucositis may have already begun to develop in this subgroup

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear Quote: "randomized into two groups" generation (selection risk bias) Comment: insufficient information to determine method of random sequence generation Allocation Unclear Quote: "randomized into two groups" concealment (selection risk bias) Comment: insufficient information to determine whether or not the random sequence was adequately concealed Blinding of High risk Comparison with no treatment so blinding not possible participants and personnel (performance bias) Blinding of outcome Unclear It would be possible to blind the outcome assessor, as the assessment (detection risk data we used was assessed by an examiner looking for bias) erythema and ulcers. However, it was not mentioned Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Kim 2017

Methods Trial design: parallel (2 arms) Location: Seoul National University Hospital, Seoul, South Korea Number of centres: 1 Study duration: recruitment from March 2009 to August 2014

387 Trials registry number: NCT00845819 (mentioned in trial report) Participants Inclusion criteria: aged 18 years or older with haematologic malignancy; due to receive intensive chemotherapy followed by autologous or allogeneic HSCT; normal oral cavity (grade 0 mucositis); ECOG score of 0 to 2 Exclusion criteria: received chemotherapy, radiotherapy or surgery within previous 3 weeks; history of allergy to the intervention or similar drugs; participation in other clinical trials within the previous 4 weeks with the potential to affect study results Cancer type: haematologic: multiple myeloma (Gp A: 57%; Gp B: 55%); lymphoma (Gp A: 36%; Gp B: 38%); other (Gp A: 7%; Gp B: 7%) Cancer treatment: prior to receiving autologous (Gp A: 96%; Gp B: 96%) or allogeneic HSCT, participants received the following conditioning regimens:  high-dose melphalan (Gp A: 57%; Gp B: 57%)  mitoxantrone-etoposide-cytarabine-melphalan (Gp A: 15%; Gp B: 23%)  busulfan-etoposide-cytarabine-melphalan (Gp A: 19%; Gp B: 16%)  other (Gp A: 9%; Gp B: 4%) Age at baseline (years): Gp A: median 53 (range 18 to 65); Gp B: median 51 (range 19 to 65) Gender: Gp A: 49% male; Gp B: 54% male Number randomised: 138 (Gp A: 69; Gp B: 69) Number evaluated: 136 (Gp A: 67; Gp B: 69) Interventions Comparison: EGF (recombinant human) versus placebo Gp A: EGF (50 µg/ml) daily by oral spray, applied twice daily, sprayed (6 sprays per application) over the entire oral mucosa and then swallowed, no oral intake for 30 mins afterwards; starting on first day of conditioning therapy and continuing until absolute neutrophil count recovered more than 1000 µl for 3 days and mucositis had resolved Gp B: placebo as above Compliance: median patient compliance rate: Gp A: 93% (range 35 to 100%); Gp B: 92% (range 18 to 100%) Duration of treatment: variable and dependent on neutrophil recovery/resolution of mucositis Outcomes  Oral mucositis: NCI-CTC (version 3.0) 0 to 4 scale (assessed daily during study period by researchers, reported as incidence of grade 2 to 4 i.e. moderate to severe, and grade 3 to 4 i.e. severe) (duration and time to onset reported but not outcomes of this review)  Oral pain: mouth and throat soreness 0 to 10 scale (reported as AUC median/range and only for those who had grade 2 to 4 mucositis - data not usable)  Quality of life: modified OMDQ (reported as AUC median/range and only for those who had grade 2 to 4 mucositis - data not usable)  Normalcy of diet (use of total parenteral nutrition)  Adverse events (NCI CTC version 3.0)  Number of days in hospital (listed as an outcome but not reported anywhere in the results - data not usable)  Number of days of treatment with opioid analgesics (reported as median/range and only for those who had grade 2 to 4 mucositis - data not usable)  Incidence of febrile neutropenia (not an outcome of this review)  Blood infections (not an outcome of this review)  Antibiotic use (not an outcome of this review)  Clinical laboratory measurements (not outcomes of this review) Notes Sample size calculation: 62 participants per group required to detect 27% difference in incidence of grade 2 to 4 mucositis with 80% power and 5%

388 significance Funding: multiple government grants; Daewoong Pharmaceutical Company (Seoul, Korea) only supplied interventions but provided no further funding and had no involvement with data collection, analysis or manuscript writing Declarations/conflicts of interest: none apparent Data handling by review authors: N/A Other information of note: not reported

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "Patients were randomly assigned...using a generation (selection computer-generated randomization protocol, by the bias) Medical Research Collaborating Center, Seoul National University Hospital"

Comment: adequate method used Allocation Low risk Quote: "Patients were randomly assigned...using a concealment (selection computer-generated randomization protocol, by the bias) Medical Research Collaborating Center, Seoul National University Hospital"

Comment: although concealment not explicitly mentioned, use of centralised/third party randomisation - likely to be done properly Blinding of Low risk Quote: "placebo-controlled, double-blind" and "clinicians, participants and patients, and investigators responsible for assessing personnel outcomes and analyzing data were masked to treatment (performance bias) assignments"

Comment: the use of a placebo should have ensured that blinding was successful Blinding of outcome Low risk Quote: "placebo-controlled, double-blind" and "clinicians, assessment (detection patients, and investigators responsible for assessing bias) outcomes and analyzing data were masked to treatment assignments"

Comment: all parties were blinded Incomplete outcome Low risk Only 2 of 138 randomised participants were not included in data (attrition bias) the analysis Selective reporting Low risk Although most of the data was not usable in this review, (reporting bias) this does not seem to be due to selective reporting Other bias Low risk No other sources of bias are apparent

Le 2011

Methods Trial design: parallel (2 arms) Location: Canada, USA and Europe (Hungary, Poland, Austria, Germany, Italy, Czech Republic) Number of centres: 46 Study duration: recruitment started August 2005, 4-month follow-up finished September 2007

389 Trials registry number: NCT00101582; 2005-000213-35 (EudraCT number) Participants Inclusion criteria: newly diagnosed, unresected stage 3 to 4B SCC of the oral cavity, oropharynx, nasopharynx, hypopharynx, or larynx; no evidence of secondary malignancy; at least two of nine areas of the oral or oropharyngeal mucosa due to receive more than 50 Gy RT Exclusion criteria: not reported Cancer type: head and neck: oropharynx (Gp A: 59%; Gp B: 54%); oral cavity (Gp A: 5%; Gp B: 10%); larynx (Gp A: 17%; Gp B: 10%); hypopharynx (Gp A: 15%; Gp B: 23%); nasopharynx (Gp A: 4%; Gp B: 3%) Cancer treatment:  Radiotherapy: standard fractionation of once daily 2-Gy fractions, 5 days per week; total 70 Gy over 7 weeks  Chemotherapy: cisplatin (100 mg/m²) by IV infusion on days 1, 22 and 43 Age at baseline (years): Gp A: mean 56 (SD 9); Gp B: mean 55 (SD 8) Gender: Gp A: 84% male; Gp B: 85% male Number randomised: 188 (Gp A: 94; Gp B: 94) Number evaluated: 188 (Gp A: 94; Gp B: 94) Interventions Comparison: KGF (palifermin) versus placebo Gp A: KGF (180 µg/kg) by IV bolus over 30 to 60 seconds, 3 days prior to start of, and then once per week during radiochemotherapy, i.e. 8 doses (total dose = 1440 µg/kg) Gp B: same schedule with placebo Compliance: 93% (SD 19%) of planned KGF doses were administered compared to 96% (SD 14%) in placebo group Duration of treatment: 8 treatment days (over 8 weeks) Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed twice weekly by trained evaluators during radiochemotherapy and then until either mucositis had reduced to grade 2 or lower or week 15, whichever occured first, maximum score reported) (duration and time to onset of grade 3 to 4 oral mucositis also measured but not outcomes of this review)  Interruptions to cancer treatment: incidence of 5 or more missed consecutive RT fractions; incidence of chemotherapy delays/discontinuations  Oral pain: OMWQ-HN 0 (no soreness) to 4 (extreme soreness) scale for mouth and throat soreness (assessed twice weekly by trained evaluators during radiochemotherapy)  Normalcy of diet (measured as incidence of supplemental feeding by TPN, PEG, nasogastric tube, or IV hydration) (duration of supplemental feeding also reported but not used for analysis)  Adverse events: NCI-CTC (version 3.0) reported separately for those related to study drugs  Use of opioid analgesics (total dose reported but not an outcome of this review)  Xerostomia (not an outcome of this review)  Survival (not an outcome of this review)  Laboratory assessments (not an outcome of this review)  Antipalifermin antibodies (not an outcome of this review) Notes Sample size calculation: assuming 60% of placebo group would develop grade 3 to 4 mucositis, 90 per group required to detect a reduction of at least 25% at 90% power and 5% significance Funding: "Supported by Amgen" (Swedish Orphan Biovitrum named as sponsor on trials registry, Amgen named as collaborator - both pharmaceutical industry) Declarations/conflicts of interest: some authors had both employment or leadership positions and stock ownership within Amgen; some authors had received research funding from Amgen Data handling by review authors: N/A

390 Other information of note: not reported

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "A centralized randomization system assigned generation (selection patients to either palifermin or placebo in a 1:1 ratio" bias) Comment: large multicentre trial using centralised randomisation method - likely to be done properly Allocation Low risk Quote: "A centralized randomization system assigned concealment (selection patients to either palifermin or placebo in a 1:1 ratio" bias) Comment: large multicentre trial using centralised randomisation method - likely to be done properly Blinding of Low risk Quote: "placebo-controlled, double-blind" participants and personnel Comment: the use of a placebo should have ensured that (performance bias) blinding was successful Blinding of outcome Low risk Quote: "placebo-controlled, double-blind" assessment (detection bias) Comment: it is not clear who was blinded. There are subjective elements to the assessment of lower grades of oral mucositis using the WHO scale, requiring the patient's assessment of pain/soreness and their ability to swallow. Higher grades have more objective elements so may not be affected by potential lack of blinding of the assessor. This would be the same for other subjective and objective outcomes Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Linch 1993

Methods Trial design: parallel (5 arms) Location: UK Number of centres: 12 Study duration: recruitment from August 1989 to July 1990 Trials registry number: none/unknown Participants Inclusion criteria: adults due to receive BMT after conditioning Exclusion criteria: myeloid malignancies Cancer type: mixed haematologic and solid (not reported by group): Hodgkin's disease (29%); Non-Hodgkin's lymphoma (33%); multiple myeloma (Gp A: 20%); ALL (15%); solid (3%) Cancer treatment: prior to receiving BMT (autologous 84%; allogeneic 16%), participants received a conditioning regimen which consisted of chemotherapy only (71%) or with TBI (29%) Age at baseline (years): median 36 (range 17 to 64) (not reported by group) Gender: 69% male (not reported by group)

391 Number randomised: 121 (Gp A: 96; Gp B: 25); Gp A represents 4 arms with different dosages Number evaluated: 121 (Gp A: 96; Gp B: 25) Interventions Comparison: G-CSF versus placebo Gp A: G-CSF (2, 5, 10 or 15 µg/kg) by 30-min IV daily starting from the day after BMT transplant and continuing until neutrophil count was > 1.0 x 10⁹/L for 3 consecutive days or until day 28, whichever occurred first Gp B: as above but with placebo Compliance: not reported Duration of treatment: variable and dependent on neutrophil recovery Outcomes  Oral mucositis: unspecified scale (frequency of measurement not mentioned, no usable data)  Adverse events  Number of days in hospital (median reported, no usable data)  Neutropenia-related outcomes (not outcomes of this review)  Antibiotic use (not an outcome of this review)  Fever (not an outcome of this review)  Sepsis (not an outcome of this review)  Blood product use (not an outcome of this review) Notes Sample size calculation: not reported Funding: "Financial support for this trial was provided by Chugai Rhone Poulene" (pharmaceutical industry) Declarations/conflicts of interest: one author was employed by the funders Data handling by review authors: oral mucositis reported narratively in additional table Other information of note: G-CSF administration only began after the conditioning and BMT transplant were completed, by which point oral mucositis may have already begun to develop (although timescale of conditioning/BMT transplant not reported)

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "Patients were randomised...in blocks of five by a generation (selection computer-generated randomisation schedule" bias) Comment: adequate method used Allocation Unclear Quote: "Patients were randomised...in blocks of five by a concealment (selection risk computer-generated randomisation schedule" bias) Comment: insufficient information to determine whether or not the random sequence was adequately concealed Blinding of Low risk Quote: "single blind...vehicle-controlled" participants and personnel Comment: the use of a placebo (the 'vehicle') should have (performance bias) ensured that blinding was successful Blinding of outcome High risk Quote: "single blind...vehicle-controlled" assessment (detection bias) Comment: the quote implies that outcome assessment was not blinded Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias)

392 Selective reporting High risk Oral mucositis not mentioned in the methods section as one (reporting bias) of the study endpoints. It is only mentioned in the results narratively as there being no difference between any group, but with no data or P-value Other bias Low risk No other sources of bias are apparent

Lucchese 2016a

Methods Trial design: parallel (2 arms) Location: Vita-Salute San Raffaele University Hospital, Milan, Italy Number of centres: 1 Study duration: conducted from April 2009 to January 2015 Trials registry number: none/unknown Participants Inclusion criteria: children aged 7 years or older with B-cell lineage acute lymphoblastic leukaemia; scheduled to receive autologous HSCT after a conditioning regimen; Karnofsky performance score of 70 or more; were to have at least 1.5 x 10⁶ CD34+ cells reinfused per kilogram available for transplant; adequate cardiac, pulmonary, renal and hepatic function Exclusion criteria: not reported Cancer type: B-cell lineage acute lymphoblastic leukaemia Cancer treatment: prior to receiving autologous HSCT (on day 0), participants received a conditioning regimen which consisted of TBI delivered in 8 fractions over 3 days (-3 to -1) with at least 6 hours between fractions, followed by chemotherapy on day -1 (type and dose of radiotherapy and chemotherapy not reported) Age at baseline (years): Gp A: median 11 (range 7 to 16); Gp B: median 11 (range 7 to 16) Gender: Gp A: 52% male; Gp B: 44% male Number randomised: 60 (Gp A: 30; Gp B: 30) Number evaluated: 54 (Gp A: 27; Gp B: 27) Interventions Comparison: KGF (palifermin) versus placebo Gp A: KGF (60 µg/kg) by IV, on days -6 (3 days prior to start of conditioning regimen), -5 and -4, and on days 0 (the day of HSCT), 1, and 2 after transplant (total dose = 360 µg/kg) Gp B: same schedule with placebo Compliance: not reported Duration of treatment: 6 treatment days (over 9 days) Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed daily by same clinician from day -7 to day 28, or until severe mucositis had reduced to grade 0, 1, or 2, data in text and figure 2 do not agree with data in table 3) (duration of grade 2 to 4 and 3 to 4 oral mucositis also measured but not an outcome of this review)  Oral pain: OMDQ 5-point scale for mouth and throat soreness (higher = worse pain) (assessed daily by participant, reported as AUC, not used)  Normalcy of diet (methods states incidence of supplemental feeding by TPN but only duration is reported, yet text states enteral and table states parenteral, not used for analysis) (patient-reported difficulty eating and drinking also assessed daily, both using OMDQ 0 (no difficulty) to 5 (unable to do) scale, but the means are reported as whole numbers, data not used)  Adverse events: NCI-CTC (version 4.0)  Opioid analgesic use (reported as quantity per day; number of days of treatment with opioid analgesics is an outcome of this review but only medians were reported, and therefore we did not use this data)  GVHD incidence and severity (not an outcome of this review)

393  Fever with neutropenia (not an outcome of this review)  HSV incidence (not an outcome of this review)  Candidiasis incidence (not an outcome of this review)  Superinfections incidence (not an outcome of this review)  Blood measurements (not an outcome of this review) Notes Sample size calculation: numbers required not reported but was estimated at 80% power and 5% significance Funding: "This work was performed with Departmental funding only" Declarations/conflicts of interest: the authors report that they have no conflict of interest (supplemental material on journal website) Data handling by review authors: we emailed the lead author June 2017 for clarification of the oral mucositis data but, until we receive a response, we are unable to use that data Other information of note: unclear definitions of ulcerative (should be grade 2 to 4) and severe (should be grade 3 to 4): "...ulcerative OM (WHO grades 3 and 4), incidence and duration of severe OM (WHO grades 3 and 4)..."

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "A centralized randomization system assigned generation (selection patients either to palifermin or conventional treatment in a bias) 1:1 ratio...... The statistician gave randomization list to the pharmacy, so the patient and the clinical research team (who assessed outcomes) were blinded to the study treatment"

Comment: randomisation 'system' used and done by a statistician, likely to be done properly Allocation Low risk Quote: "A centralized randomization system assigned concealment (selection patients either to palifermin or conventional treatment in a bias) 1:1 ratio...... The statistician gave randomization list to the pharmacy, so the patient and the clinical research team (who assessed outcomes) were blinded to the study treatment"

Comment: centralised randomisation with pharmacy assigning participants to groups Blinding of Low risk Quote: "double-blind, placebo-controlled" and "The participants and statistician gave randomization list to the pharmacy, so the personnel patient and the clinical research team (who assessed (performance bias) outcomes) were blinded to the study treatment. The pharmacy provided the research team with the blinded study medication"

Comment: the use of a placebo should have ensured that blinding was successful Blinding of outcome Low risk Quote: "The statistician gave randomization list to the assessment (detection pharmacy, so the patient and the clinical research team bias) (who assessed outcomes) were blinded to the study treatment. The pharmacy provided the research team with the blinded study medication"

394 Comment: outcome assessors were clearly blinded Incomplete outcome Low risk Overall attrition was same in both groups (10%) with the data (attrition bias) same reason given Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Lucchese 2016b

Methods Trial design: parallel (2 arms) Location: Vita-Salute San Raffaele University Hospital, Milan, Italy Number of centres: 1 Study duration: conducted from April 2010 to January 2014 Trials registry number: none/unknown Participants Inclusion criteria: children aged 9 to 15 years with B-cell lineage acute lymphoblastic leukaemia; scheduled to receive allogeneic HSCT after a conditioning regimen Exclusion criteria: not reported Cancer type: B-cell lineage acute lymphoblastic leukaemia Cancer treatment: prior to receiving allogeneic HSCT, participants received a conditioning regimen of a total of 12 Gy TBI delivered in 8 fractions of 1.5 Gy twice daily over 4 days Age at baseline (years): Gp A: median 12 (range 8 to 15); Gp B: median 12 (range 8 to 15) Gender: Gp A: 50% male; Gp B: 55% male Number randomised: 46 (Gp A: 24; Gp B: 22) Number evaluated: 46 (Gp A: 24; Gp B: 22) Interventions Comparison: KGF (palifermin) versus placebo Gp A: KGF (60 µg/kg) by IV, for 3 days prior to start of conditioning regimen and for 3 days after completion (total dose = 360 µg/kg) Gp B: same schedule with placebo Compliance: not reported Duration of treatment: 6 treatment days (over unclear period of time - not fully described) Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed by the clinical research team at baseline, days 1, 2, 3 (it is not clear when this is in relation to the cancer treatment or study treatment) and at the end of the cycle - "All of these results were confirmed at the last follow-up (60 days)" - it is not clear how they have been confirmed if, for example, OM has resolved by 60 days; it is not clear whether or not there were multiple cycles, incidence of each grade does not add up or make sense or match the data in table 3) (duration of grade 2 to 4 and 3 to 4 oral mucositis also measured but not an outcome of this review)  Normalcy of diet (duration of supplemental feeding by TPN, not used for analysis) (patient-reported difficulty eating and drinking assessed using OMDQ 0 (no difficulty) to 5 (unable to do) scale, but the means are reported as whole numbers, data not used)  Adverse events  Opioid analgesic use (number of days of treatment with opioid analgesics (morphine) reported but unclear if mean/median and no SD or P-value reported; unable to use data; also reported as quantity per day but not an outcome of this review) Notes Sample size calculation: not reported Funding: "This work has been performed with departmental funding only"

395 Declarations/conflicts of interest: "The authors certify that there is no conflict of interest with any financial organization regarding the material discussed in the manuscript" Data handling by review authors: we emailed the lead author June 2017 for clarification of the oral mucositis data but, until we receive a response, we are unable to use that data Other information of note: "Patients who had severe (grade 3 or 4) OM during blinded cycles received open-label palifermin at the same dosages as in the other group. The research team assessed patients for OM at baseline before the cycle; at days 1, 2, 3 and at the end of the transplant cycle." - nowhere else in the report suggests that there were multiple cycles of treatment; unclear definitions of ulcerative mucositis i.e. described correctly in one section as grades 2 to 4, but described in another section as grades 1 to 4

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "No stratification was performed and two distinct generation (selection computer-generated a randomization lists. The statistician bias) gave both randomization lists to the pharmacy, so the patient and the clinical research team (who assessed outcomes) were blinded to the study treatment"

Comment: adequate method done by a statistician (although it is unclear why there were 2 lists) Allocation Low risk Quote: "No stratification was performed and two distinct concealment (selection computer-generated a randomization lists. The statistician bias) gave both randomization lists to the pharmacy, so the patient and the clinical research team (who assessed outcomes) were blinded to the study treatment"

Comment: randomisation 'system' used and done by a statistician, likely to be done properly Blinding of Low risk Quote: "double-blind, placebo-controlled" and "The participants and statistician gave both randomization lists to the pharmacy, personnel so the patient and the clinical research team (who assessed (performance bias) outcomes) were blinded to the study treatment. The pharmacy provided the research team with the blinded study medication"

Comment: the use of a placebo should have ensured that blinding was successful Blinding of outcome Low risk Quote: "The statistician gave both randomization lists to the assessment (detection pharmacy, so the patient and the clinical research team bias) (who assessed outcomes) were blinded to the study treatment. The pharmacy provided the research team with the blinded study medication"

Comment: outcome assessors were clearly blinded Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately

396 Other bias Low risk No other sources of bias are apparent

Makkonen 2000

Methods Trial design: parallel (2 arms) Location: Turku University Central Hospital and Helsinki University Central Hospital, Finland Number of centres: 2 Study duration: recruitment from November 1994 to August 1996 Trials registry number: none/unknown Participants Inclusion criteria: scheduled to receive total target dose of at least 56 Gy to the oropharyngeal mucosa Exclusion criteria: prior chemotherapy or radiotherapy; concurrent use of anticholinergic drugs; autoimmune thrombocytopenic purpura; WHO performance status higher than 2 Cancer type: head and neck: mobile tongue (Gp A: 30%; Gp B: 20%); oral cavity, other (Gp A: 40%; Gp B: 25%); oropharynx (Gp A: 20%; Gp B: 15%); nasopharynx (Gp A: 5%; Gp B: 10%); supraglottic larynx (Gp A: 5%; Gp B: 15%); hypopharynx (Gp A: 0%; Gp B: 15%) Cancer treatment:  conventional radiotherapy (Gp A: 50%; Gp B: 50%): 1.9 to 2 Gy daily fractions, 5 fractions per week  hyperfractionated radiotherapy (Gp A: 50%; Gp B: 50%): 1.6 Gy fractions twice per day with minimum 6 hours between fractions, to a total dose of 38.4 Gy; planned break of 9 to 12 days during which the mucosa healed in order to allow further doses to a total of 56 to 73 Gy; overall treatment time was 5 to 6 weeks Total target dose: Gp A: median 65 Gy (range 56 to 68); Gp B: median 66 Gy (range 58 to 73); oral surgery prior to RT: Gp A: 20%; Gp B: 30%; oral surgery after radiotherapy: 23 (not reported by group) Age at baseline (years): Gp A: mean 63 (SD 10; range 47 to 87); Gp B: mean 59 (SD 13; range 33 to 79) Gender: Gp A: 60% male; Gp B: 55% male Number randomised: 40 (Gp A: 20; Gp B: 20) Number evaluated: Unclear; results presented as percentages Interventions Comparison: GM-CSF (molgramostim) plus sucralfate versus sucralfate alone Both groups rinsed mouth with sucralfate (1 g) suspension for 1 minute before swallowing, 6 times per day, starting after 1 week of RT and continued until the end of RT (including weekends and planned/unplanned treatment breaks). Both groups advised to rinse their mouths with saline solution in between the sucralfate rinses if required Gp A: after cumulative radiation dose of 10 Gy, GM-CSF (150 to 300 µg - dependent on body weight) by daily subcutaneous injection until the last day of RT; not given over weekends or during planned/unplanned treatment breaks; mean total dose = 3398 µg (range 300 to 7200 µg) Gp B: no other treatment Oral pain was treated with anti-inflammatory analgesics and with local anaesthetic mouthwashes (lidocaine hydrochloride, Xylocain 0.5%) Compliance: not reported Duration of treatment: variable and dependent on duration of RT Outcomes  Oral mucositis: 0 to 2 scale where 0 = no mucositis, 1 = erythema and edema without ulcers but food intake/use of dental prosthesis was not affected, and 2 = one or more ulcers or bleeding or food intake/use of dental prosthesis was affected (assessed weekly during RT and 1 and 6 months after

397 completion of RT, reported at start of RT and weekly for 4 weeks after; no usable data)  Interruptions to cancer treatment (only 2 due to mucositis but not reported by group; data not usable)  Oral pain: 0 to 10 VAS and also 1 to 4 scale where 1 = no pain, 2 = mild, 3 = moderate, and 4 = severe (assessed daily at midday during RT) (data not usable - 1 to 4 scale reported as "no difference" and VAS reported on a graph with no SD or P-values)  Adverse events (due to GM-CSF, sucralfate, or both)  Saliva flow rates (not an outcome of this review)  Salivary lactoferrin (not an outcome of this review)  Blood measurements (not an outcome of this review)  Body weight (not an outcome of this review)  Overall survival (not an outcome of this review) Notes Sample size calculation: not reported Funding: GM-CSF and sucralfate were both provided by pharmaceutical industry (Schering-Plough Corporation and Orion-Farmos Pharmaceuticals respectively) Declarations/conflicts of interest: not reported Data handling by review authors: no usable data Other information of note: GM-CSF administration began after cumulative radiation dose of 10 Gy, by which point oral mucositis may have already begun to develop

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "assignment to the treatment groups was carried out generation (selection via a phone call to the randomization center located at the bias) Finnish Cancer Registry, Helsinki"

Comment: use of a specialist randomisation centre Allocation Low risk Quote: "assignment to the treatment groups was carried out concealment (selection via a phone call to the randomization center located at the bias) Finnish Cancer Registry, Helsinki"

Comment: third party/remote randomisation would have ensured concealment of the random sequence from those recruiting participants Blinding of High risk Blinding not possible as no placebo was used participants and personnel (performance bias) Blinding of outcome Unclear It would be possible to blind the outcome assessor for oral assessment (detection risk mucositis as the scale was fairly objective, but it was not bias) mentioned. It would not be possible to blind oral pain assessment as this was done by the participant who was not blinded (however, we could not use the pain data so there was no bias for this outcome) Incomplete outcome Unclear It is not clear how many participants were included in the data (attrition bias) risk analyses Selective reporting High risk Poor reporting of oral mucositis and oral pain

398 (reporting bias) Other bias Low risk No other sources of bias are apparent

McAleese 2006

Methods Trial design: parallel (2 arms) Location: Royal Marsden Hospital, London, UK Number of centres: 1 Study duration: recruitment from September 1997 to October 2000 Trials registry number: NCT00004256 Participants Inclusion criteria: histologically proven T1 N0 or T2 N0 glottic carcinoma; due to receive RT with a 16-fraction 3-week regimen; WHO performance status 0 or 1 Exclusion criteria: renal or hepatic impairment; serious infections requiring antibiotics; taking corticosteroids or likely to require them; known allergy to GM-CSF Cancer type: laryngeal Cancer treatment: accelerated radiotherapy: once-daily fractions of 3.125 Gy, to a total dose of 50 Gy in 16 fractions over 21 days Age at baseline (years): Gp A: median 60 (range 48 to 79); Gp B: median 65 (range 32 to 70) Gender: Gp A: 93% male; Gp B: 86% male Number randomised: 29 (Gp A: 15; Gp B: 14) Number evaluated: 29 (Gp A: 15; Gp B: 14) Interventions Comparison: GM-CSF (molgramostim) versus no treatment Gp A: GM-CSF (150 µg) by daily subcutaneous injection, starting on day 14 of RT and continuing for 14 days i.e. for the last week of RT and for 1 week after completion of RT (total dose = 2100 µg) Gp B: no treatment Compliance: 2 participants did not complete their course of GM-CSF Duration of treatment: 14 days Outcomes  Oral mucositis: RTOG 0 to 4 scale (measured weekly by one of two independent observers using physician's objective criteria (see Appendix 3), reported graphically over time and by maximum score experienced in the text) (time to mucositis healing also measured but not an outcome of this review)  Normalcy of diet: use of feeding tubes  Adverse events  Analgesic use ("No differences were detected in...analgesic usage" - don't know if this is number of days or whether or not opioids)  Dysphagia and odynophagia (not outcomes of this review)  Candida infection (not an outcome of this review)  Laryngeal oedema (not an outcome of this review)  Moist or dry desquamation (not an outcome of this review)  Weight change (not an outcome of this review)  Skin erythema (not an outcome of this review) Notes Sample size calculation: 17 per group needed at 90% power and 5% significance to detect reduction from an anticipated 60% incidence of severe mucositis to 10% Funding: not reported Declarations/conflicts of interest: not reported Data handling by review authors: N/A Other information of note: imbalance in stage distribution, with more T2 patients in the GM-CSF group. Consequently more of this group were treated

399 with larger fields. GM-CSF administration only began after 14 days of radiotherapy, by which point oral mucositis may have already begun to develop

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear Quote: "were randomly assigned to the active or control generation (selection risk arms" bias) Comment: insufficient information to determine method of random sequence generation Allocation Unclear Quote: "were randomly assigned to the active or control concealment (selection risk arms" bias) Comment: insufficient information to determine whether or not the random sequence was adequately concealed Blinding of High risk Blinding not possible as no placebo was used participants and personnel (performance bias) Blinding of outcome Low risk Quote: "At each visit one of two independent observers, assessment (detection blinded to group allocation, scored mucositis..." bias) Comment: a physician's objective version of the RTOG scale was used. Use of feeding tube is also an objective outcome Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Analgesic use reported as "No differences" with no data, (reporting bias) however mucositis fully reported Other bias Low risk No other sources of bias are apparent

Meropol 2003

Methods Trial design: parallel (7 arms) Location: USA Number of centres: 10 Study duration: not reported Trials registry number: none/unknown Participants Inclusion criteria: 18 years or older; metastatic colon or rectal adenocarcinoma; scheduled to receive palliative 5FU and leucovorin chemotherapy; ECOG performance status 0 to 2 (ambulatory at least 50% of waking hours); life expectancy of at least 4 months; free of lesions and no recent history (within 30 days before baseline examination) of oral ulceration, herpes simplex, oral candidiasis, severe gingivitis, or the presence of active or chronic mucositis, xerostomia, or diarrhoea; absolute neutrophil count ≥ 1.5 x 109/L; platelet count ≥ 100 x 109/L; serum creatinine ≤ 2.0 mg/dL; serum bilirubin ≤ 2.0 mg/dL; serum aspartate amino transferase less than five times the upper limit of normal; absence of other serious concurrent medical illness Exclusion criteria: received chemotherapy, radiotherapy, or other investigational drugs within 4 weeks of enrolment (6 weeks for chemotherapy

400 with mitomycin or nitrosoureas); any unresolved adverse event from previous therapy; major surgery within 2 weeks before study entry; history of insulin- dependent diabetes mellitus; pregnant or breast feeding; of child-bearing potential and not using adequate contraceptive precautions; previous hypersensitivity reaction to leucovorin calcium or Escherichia coli-derived material Cancer type: metastatic colorectal Cancer treatment: (palliative) leucovorin 20 mg/m² by rapid IV injection, followed immediately by 5FU 425 mg/m² by rapid IV injection for 5 consecutive days on days 4 to 8 of each 28-day cycle Age at baseline (years): Gp A: mean 62 (SD 10; range 44 to 84); Gp B: mean 66 (SD 13; range 41 to 86) Gender: Gp A: 57% male; Gp B: 59% male Number randomised: 81 (Gp A: 54; Gp B: 27) Number evaluated: 81 (Gp A: 54; Gp B: 27) Interventions Comparison: KGF (recombinant human) versus placebo Gp A: KGF  (n = 7): 1 µg/kg per day by IV bolus on days 1 to 3 (total dose = 3 µg/kg)  (n = 8): 10 µg/kg as above (total dose = 30 µg/kg)  (n = 10): 20 µg/kg as above (total dose = 60 µg/kg)  (n = 11): 40 µg/kg as above (total dose = 120 µg/kg)  (n = 8): 60 µg/kg as above (total dose = 180 µg/kg)  (n = 10): 80 µg/kg as above (total dose = 240 µg/kg) Gp B: placebo as above Compliance: three participants stopped KGF treatment due to adverse reactions involving the skin (one in 60 and two in 80 µg/kg group) Duration of treatment: 3 days Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed on days 1, 4, 8, 15 and 28, reported as incidence of grade 2 to 4; first cycle data only) (duration of mucositis also measured but not an outcome of this review)  Oral pain: self-assessed daily questionnaire including mouth and throat pain assessed on both a 5-point ordinal scale (reported graphically, no usable data) and a 0 to 10 VAS (reported as AUC, not used)  Adverse events: WHO 0 to 4 scale (reported as events of any grade that differed by at least 10% between any KGF group and placebo; events were not reported if they occurred in less than 10% of KGF group; also reported as incidence of grade 3 to 4 events)  Blood measurements (not an outcome of this review) Notes Sample size calculation: not reported Funding: "Supported by a grant from Amgen, Inc" (pharmaceutical industry) Declarations/conflicts of interest: not reported Data handling by review authors: we combined the six KGF groups to make a single pairwise comparison against placebo; in order to make a head-to-head comparison of doses we grouped the three lower doses (1, 10 and 20 µg/kg) together and grouped the three higher doses (40, 60 and 80 µg/kg) together to make pairwise groups for comparison Other information of note: not reported

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear Quote: "This study was a multicenter, randomized, double- generation (selection risk blinded, placebo-controlled, phase I study"

401 bias) Comment: insufficient information to determine method of random sequence generation Allocation Unclear Quote: "This study was a multicenter, randomized, double- concealment (selection risk blinded, placebo-controlled, phase I study" bias) Comment: insufficient information to determine whether or not the random sequence was adequately concealed Blinding of Low risk Quote: "double-blinded, placebo-controlled" participants and personnel Comment: the use of a placebo should have ensured that (performance bias) blinding was successful; incidence of adverse events does not appear to differ enough between groups to cause unblinding Blinding of outcome Low risk Quote: "double-blinded, placebo-controlled" assessment (detection bias) Comment: it is not clear who was blinded but the outcome is grade 2 to 4 WHO-scale mucositis and it is unlikely that this would be affected by any lack of blinding Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Nemunaitis 1995

Methods Trial design: parallel (2 arms) Location: USA and Canada Number of centres: 7 Study duration: 1 November 1990 to 1 July 1993 Trials registry number: none/unknown Participants Inclusion criteria: any age; undergoing allogeneic BMT (from HLA-identical sibling) for haematological malignancy Exclusion criteria: due to receive methotrexate as part of GVHD prophylaxis; chronic lymphocytic leukaemia; chronic myelogenous leukaemia in blast crisis; acute lymphoblastic leukaemia or acute myelogenous leukaemia with failure at first induction or progressed beyond a second relapse; previously received cytokines; HIV; life expectancy less than 7 days Cancer type: haematologic malignancy: lymphoid malignancy (Gp A: 21%; Gp B: 16%); acute myeloid malignancy (Gp A: 34%; Gp B: 27%); chronic myeloid malignancy (Gp A: 32%; Gp B: 30%); multiple myeloma (Gp A: 6%; Gp B: 5%); myelodysplastic syndrome (Gp A: 6%; Gp B: 11%); aplastic anaemia (Gp A: 2%; Gp B: 11%) Cancer treatment: prior to receiving allogeneic BMT (from HLA-identical sibling), participants received the following conditioning regimens: busulfan and cyclophosphamide with TBI (Gp A: 9%; Gp B: 16%) or without TBI (Gp A: 32%; Gp B: 34%); busulfan, cyclophosphamide, cytosine arabinoside, and methotrexate with TBI (Gp A: 23%; Gp B: 13%) or without TBI (Gp A: 8%; Gp B: 11%); cyclophosphamide and steroid with TBI (Gp A: 0%; Gp B: 2%); cyclophosphamide, cytosine arabinoside, and steroid with TBI (Gp A: 23%; Gp B: 18%); etoposide with TBI (Gp A: 0%; Gp B: 2%); etoposide, cytosine arabinoside, and cyclophosphamide with TBI (Gp A: 2%; Gp B: 0%); etoposide, cytosine arabinoside, cyclophosphamide, and asparaginase with TBI (Gp A: 0%;

402 Gp B: 2%); etoposide, methotrexate, cytosine arabinoside, and steroid with TBI (Gp A: 4%; Gp B: 4%) All participants received cyclosporine and methylprednisolone for GVHD prophylaxis Age at baseline (years): Gp A: mean 32; Gp B: mean 34 Gender: Gp A: 60% male; Gp B: 54% male Number randomised: 109 (Gp A: 53; Gp B: 56) Number evaluated: 109 (Gp A: 53; Gp B: 56) Interventions Comparison: GM-CSF (yeast-derived recombinant human) versus placebo Gp A: GM-CSF (250 µg/m²) by 4-hour IV infusion starting on day 0 (the day of BMT) to day 20 (total dose = 5250 µg/m²) Gp B: as above with placebo Compliance: Gp A: 13 participants (25%) stopped their intervention early: 11 due to adverse events (1 rash, 7 bone pain, 2 acute dyspnea, 1 seizure); 2 (with no apparent toxicity) due to participant or physician request; Gp B: 9 participants (16%) stopped their intervention early: 6 due to adverse events (2 rash, 2 bone pain, 1 elevated liver enzymes, 1 infections); 2 (on day 19) due to miscalculation of the number of days to receive the intervention; 1 (on day 9) due to severe mucositis and veno-occlusive disease of the liver Duration of treatment: 21 days Outcomes  Oral mucositis: assessment methods are not clear but reported on 0 to 4 scale which is probably WHO (assessment frequency and duration also unclear, reported in text as incidence of grade 2 to 4 and grade 3 to 4)  Adverse events: 0 to 4 scale which is probably WHO (reported as incidence of any grade of event and incidence of grade 3 to 4 events with a greater than 10% frequency)  Number of days in hospital (reported as median values; unable to use data)  GVHD incidence and severity (not an outcome of this review)  Veno-occlusive disease of the liver (not an outcome of this review)  Blood measurements (not an outcome of this review)  Infection: bacterial, fungal and viral (not an outcome of this review)  Duration of IV antibiotics (not an outcome of this review)  Fever (not an outcome of this review)  Survival: overall, disease-free, relapse (not an outcome of this review) Notes Sample size calculation: not reported Funding: "product provided by Immunex" (pharmaceutical). Also mention of "sponsoring company" which is likely to be Immunex Declarations/conflicts of interest: not reported but four authors employed by Immunex Data handling by review authors: N/A Other information of note: GM-CSF administration only began after the conditioning regimen, by which point oral mucositis may have already begun to develop.

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear Quote: "Assignment to treatment was made via a generation (selection risk randomization schema prepared by Almedica Corporation" bias) Comment: insufficient information to determine method of random sequence generation Allocation Low risk Quote: "Blinded numbered vials containing placebo or

403 concealment (selection rhGM-CSF were provided to each participating centre. The bias) pharmacists, principal investigators, patients, support care personnel and sponsoring company were blinded to the study medication for the entire course of the study"

Comment: probably done Blinding of Low risk Quote: "double-blind placebo controlled" and "The participants and pharmacists, principal investigators, patients, support care personnel personnel and sponsoring company were blinded to the (performance bias) study medication for the entire course of the study"

Comment: the use of a placebo should have ensured that blinding was successful Blinding of outcome Low risk Quote: "double-blind placebo controlled" and "The assessment (detection pharmacists, principal investigators, patients, support care bias) personnel and sponsoring company were blinded to the study medication for the entire course of the study"

Comment: everyone involved in the study was blinded Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk Five participants in the placebo group received cytokines off study during the first 42 days after BMT but unlikely to bias the results in a meaningful way

Peterson 2009

Methods Trial design: parallel (3 arms) Location: Russia Number of centres: 9 Study duration: recruitment from August 2006 to May 2007 Trials registry number: none/unknown Participants Inclusion criteria: 18 years or older with colorectal cancer (stage I to IV); undergoing chemotherapy as primary cancer therapy; experienced WHO grade 2 or higher oral mucositis during first cycle of chemotherapy, but recovered prior to enrollment (i.e. grade 0); ECOG score of 0 to 2 Exclusion criteria: pregnant (or risk of pregnancy) or lactating; scheduled to receive RT to the head and neck; received other investigational drugs within 14 days of the start of the study; evidence of alcohol and drug abuse; pre-existing conditions such as active fungal or herpetic infection Cancer type: colorectal (stage I to IV) Cancer treatment: all participants received 5FU (97% with leucovorin) but one received capecitabine; 6 participants (3 in low-dose ITF, 3 in placebo) also received oxaliplatin; doses or regimens comparable between groups Age at baseline (years): Gp A: median 62 ; Gp B: median 56 ; Gp C: median 59 Gender: Gp A: 42% male; Gp B: 33% male; Gp C: 48% male Number randomised: 99 (Gp A: 33; Gp B: 33; Gp C: 33) Number evaluated: 99 (Gp A: 33; Gp B: 33; Gp C: 33) Interventions Comparison: ITF (recombinant human) versus placebo Gp A: ITF (10 mg/ml) oral spray, 300 µl (3 sprays) to the oral mucosa 8 times daily, starting on the first day (day 1) of the second chemotherapy cycle for total 14 days (total dose = 336 mg)

404 Gp B: ITF (80 mg/ml) oral spray as above (total dose = 2688 mg) Gp C: placebo oral spray as above Compliance: patient-reported compliance was 97% Duration of treatment: 14 days Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed on days 0, 1, 3, 5, 7, 10, 12, 14, and 21 ± 2, maximum score reported) (duration of mucositis also measured but not an outcome of this review)  Oral mucositis: OMAS 0 to 45 scale (assessed on days reported above; WHO data used for analysis)  Adverse events (assessed on days reported above)  Patient daily self-assessment: including discolouration, mouth and throat pain, preference for solid/semi-solid food, analgesic use, and dysphagia (no usable data for the outcomes relevant to this review) Notes Sample size calculation: based on previous study, 80% power to detect 40% difference in incidence of WHO grade 2 or above Funding: The GI Company (pharmaceutical) Declarations/conflicts of interest: some authors had roles with the funding pharmaceutical company (and other companies) both compensated and uncompensated Data handling by review authors: we combined the two ITF groups to make a single pairwise comparison against placebo; we also made a separate comparison of the two ITF dosages Other information of note: not reported

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear Quote: "The patients were randomly assigned" generation (selection risk bias) Comment: insufficient information to determine method of random sequence generation Allocation Unclear Quote: "The patients were randomly assigned" concealment (selection risk bias) Comment: insufficient information to determine whether or not the random sequence was adequately concealed Blinding of Low risk Quote: "double-blind, placebo-controlled" participants and personnel Comment: the use of a placebo should have ensured that (performance bias) blinding was successful Blinding of outcome Low risk Quote: "double-blind, placebo-controlled" assessment (detection bias) Comment: it is not clear who was blinded. There are subjective elements to the assessment of oral mucositis using this scale, requiring the patient's assessment of pain/soreness and their ability to swallow but, as the participants were unaware of their group allocation, the assessment of oral mucositis can be considered to be blinded Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately

405 Other bias Low risk No other sources of bias are apparent

Rosen 2006

Methods Trial design: parallel (2 arms) Location: USA and Australia Number of centres: 15 Study duration: not reported Trials registry number: none/unknown Participants Inclusion criteria: 18 years or older with advanced (Duke's D) colon or rectal adenocarcinoma; scheduled to receive bolus 5FU and low-dose leukovorin (Mayo regimen) as first-line or subsequent therapy; normal oral cavity examination (no ulceration, herpes simplex, candidiasis, or severe gingivitis); ECOG score of 0 to 2; life expectancy of 4 months or more; absolute neutrophil count of 1.5 x 10⁹/L or higher; platelet count of 100 x 10⁹/L or higher; normal renal and hepatic function Exclusion criteria: previous radiotherapy or chemotherapy within 4 weeks or major surgery within 2 weeks of study day 1; insulin-dependant diabetes; known allergy to leucovorin; known hypersensitivity to Escherichia coli-derived material Cancer type: advanced colorectal (Duke's D) Cancer treatment: low-dose leukovorin (20 mg/m²) by IV immediately followed by 5FU (425 mg/m²) by rapid IV once daily for 5 consecutive days on days 4 to 8 of each 28-day cycle; 2 cycles (5FU dose could be decreased by 20% during cycle 2 if moderately severe toxicities occured in cycle 1) Age at baseline (years): Gp A: mean 65 (SD 11.2); Gp B: mean 65 (SD 11.1) Gender: Gp A: 57% male; Gp B: 72% male Number randomised: 65 (not reported by group) Number evaluated: 64 (Gp A: 28; Gp B: 36) Interventions Comparison: KGF (palifermin) versus placebo Gp A: KGF (40 µg/kg) by IV for 3 consecutive days (days 1 to 3 of each 28-day cycle) before the start of chemotherapy (total dose = 120 µg/kg) Gp B: as above with placebo Compliance: not reported (only reports "palifermin was generally well tolerated") Duration of treatment: 3 days Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed on days 0, 1, 4, 8, 12, 15, and 28, maximum score reported)  Oral pain: patient daily self-assessment using OMDQ 5-point scale for mouth and throat soreness (higher = worse pain, reported as AUC, not used)  Adverse events  Diarrhoea (related to the cancer therapy - not an outcome of this review)  Laboratory assessments (not an outcome of this review)  Neutropenia (not an outcome of this review)  Antibody assessments (not an outcome of this review)  Survival (not an outcome of this review) Notes Sample size calculation: not reported Funding: "Supported by Amgen Inc" (pharmaceutical industry) Declarations/conflicts of interest: all authors were linked to the funding pharmaceutical company in terms of employment, consultancy, stock ownership, honoraria, and receipt of research funding Data handling by review authors: data were reported separately for the chemotherapy cycles 1 and 2 - we used the data for cycle 1 Other information of note: not reported

406

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear Quote: "Patients were randomly assigned (by center and generation (selection risk prior chemotherapy)" bias) Comment: insufficient information to determine method of random sequence generation Allocation Unclear Quote: "Patients were randomly assigned (by center and concealment (selection risk prior chemotherapy)" bias) Comment: insufficient information to determine method of random sequence generation Blinding of Low risk Quote: "double-blind, placebo-controlled" participants and personnel Comment: the use of a placebo should have ensured that (performance bias) blinding was successful Blinding of outcome Low risk Quote: "double-blind, placebo-controlled" assessment (detection bias) Comment: it is not clear who was blinded. There are subjective elements to the assessment of oral mucositis using this scale, requiring the patient's assessment of pain/soreness and their ability to swallow but, as the participants were unaware of their group allocation, the assessment of oral mucositis can be considered to be blinded Incomplete outcome Low risk Only one randomised participant was not included in the data (attrition bias) analyses Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Saarilahti 2002

Methods Trial design: parallel (2 arms) Location: Helsinki University Central Hospital, Finland Number of centres: 1 Study duration: recruitment from October 1999 to April 2001 Trials registry number: none/unknown Participants Inclusion criteria: undergone radical surgery for head and neck cancer and scheduled to receive postoperative RT to a total dose of 50 Gy or more to the oral and oropharyngeal mucosa Exclusion criteria: prior chemotherapy or RT; chronic autoimmune or inflammatory disease; WHO performance score > 2 Cancer type: head and neck: mobile tongue (Gp A: 33%; Gp B: 37%); floor of mouth (Gp A: 19%; Gp B: 16%); tonsil (Gp A: 29%; Gp B: 32%); oral cavity other (Gp A: 19%; Gp B: 16%) Cancer treatment: mean time from radical surgery to start of RT: Gp A: 39 days (range 20 to 73); Gp B: 38 days (range 20 to 56); conventionally fractionated RT to a total dose of 50 to 60 Gy in 2-Gy daily fractions, 5 times weekly (on weekdays) for 5 to 6 weeks, to the primary site and locoregional

407 lymph nodes Age at baseline (years): Gp A: median 56 (range 43 to 87); Gp B: median 60 (range 24 to 72) Gender: Gp A: 57% male; Gp B: 79% male Number randomised: 40 (Gp A: 21; Gp B: 19) Number evaluated: 40 (Gp A: 21; Gp B: 19) Interventions Comparison: GM-CSF versus sucralfate Gp A: starting after a cumulative RT dose of 10 Gy (after 1 week of RT); mouth was cleaned with water, then GM-CSF (37.5 µg per 25 ml rinse) mouthwash rinsed around the mouth for 3 mins then swallowed, 4 times daily after meals, on RT-days (weekdays), until the end of RT (total dose dependent on duration of RT) Gp B: as above but with sucralfate (1 g per 25 ml rinse) (total dose dependent on duration of RT) Compliance: not reported (only reports "both mouthwashes were well tolerated, and none of the patients reported any adverse effects related to the mouthwashes") Duration of treatment: 4 to 5 weeks (dependent on duration of RT) Outcomes  Oral mucositis: RTOG 0 to 4 scale (assessed before RT, weekly during RT, and at 1, 2, and 4 weeks after RT, reported graphically as mean score over time but authors provided maximum incidence data on request)  Interruptions to cancer treatment (RT interruptions due to OM)  Oral pain: 0 (no pain) to 10 (worse pain) VAS (assessed as reported above, reported graphically as mean score over time – no mean maximum score and no SDs, unable to use data)  Normalcy of diet: use of PEG feeding tube  Adverse events  Incidence of hospitalisation (number of days in hospital is an outcome of this review and therefore we did not use this data)  Opioid analgesic use (reported as incidence; number of days of treatment with opioid analgesics is an outcome of this review and therefore we did not use this data)  Laboratory parameters (not outcomes of this review)  Use of antimycotic agents (not an outcome of this review)  Use of antibiotics (not an outcome of this review)  Weight loss (not outcomes of this review) Notes Sample size calculation: not reported Funding: not reported Declarations/conflicts of interest: not reported Data handling by review authors: we used oral mucositis incidence (maximum score experienced per patient) data provided by the authors; all participants experienced mucositis (grade 1 or above) as they had already received a week of RT before the intervention started, therefore we felt an analysis of incidence of any grade of mucositis should not be included Other information of note: GM-CSF/sucralfate administration only began after 1 week of RT, by which point oral mucositis may have already begun to develop

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "Randomization was done using computer- generation (selection generated random digits" bias)

408 Comment: adequate method used Allocation Low risk Quote: "they were assigned to a treatment group by way of concealment (selection a telephone call to the randomization office" bias) Comment: third party/remote randomisation would have ensured concealment of the random sequence from those recruiting participants Blinding of Low risk Quote: "double-blind" and "Both solutions looked alike, participants and and neither the investigators nor the patients were aware of personnel the contents of the solutions given. The drug vials were (performance bias) marked with a study code that prevented identification of the allocation group"

Comment: adequate methods to ensure blinding Blinding of outcome Low risk Quote: "double-blinded, placebo-controlled" assessment (detection bias) Comment: nobody involved in the study was aware of group allocation Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Schneider 1999

Methods Trial design: parallel (2 arms) Location: USA Number of centres: not reported Study duration: recruitment from January 1995 to April 1996 Trials registry number: none/unknown Participants Inclusion criteria: 18 years or older with histologically proven head and neck malignancy; mucosa of the oropharynx and/or oral cavity to be included in RT portal; Karnofsky performance score of 60 or more; no known hypersensitivity to E coli-derived proteins; haemoglobin and platelet counts 100 x 10⁹/L or higher; absolute neutrophil count higher thah 1.5 x 10⁹/L; normal blood pressure; agreement to bring dentition to acceptable level prior to RT (teeth with periapical abscesses or severe periodontal disease extracted, followed by a 10- day healing period allowance) Exclusion criteria: lactating/pregnant/females not taking effective form of contraception; scheduled to receive chemotherapy or radiosensitising drugs during the planned RT; history of myeloid malignancy; underlying collagen vascular disease; active rheumatoid arthritis Cancer type: head and neck: nasopharynx (Gp A: 0; Gp B: 1); oropharynx (Gp A: 2; Gp B: 1); tongue (Gp A: 2; Gp B: 1); larynx (Gp A: 2; Gp B: 1); Other unknown (Gp A: 2; Gp B: 2) Cancer treatment: radiotherapy in 1.8 to 2 Gy standard daily fractions from Monday to Friday, to total dose 50 Gy or more Age at baseline (years): not reported Gender: Gp A: 100% male; Gp B: 83% male Number randomised: 14 (Gp A: 8; Gp B: 6) Number evaluated: 14 (Gp A: 8; Gp B: 6) Interventions Comparison: G-CSF (r-metHuG-CSF) (filgrastim) versus placebo

409 Gp A: G-CSF (starting at 3 µg/kg and titrated to keep neutrophil count between 10 x 10⁹/L and 30 x 10⁹/L) by daily subcutaneous injection, starting on the first day of RT (weekdays), until the end of RT (total dose dependent on duration of RT and neutrophil counts) Gp B: placebo as above Compliance: not reported (only reports "Filgrastim was well tolerated") Duration of treatment: variable and dependent on duration of RT Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed on weekly by single examiner, reported as incidence of any mucositis and grade 3 or higher)  Oral mucositis: Hickey 0 to 3 scale (not used)  Interruptions to cancer treatment (delays or dose reductions, no usable data)  Weight change (not an outcome of this review)  Blood measurements (not an outcome of this review) Notes Sample size calculation: 54 required to detect 30% decrease in incidence of grade 2 or 3 mucositis at 80% power at the 5% significance level Funding: "Financial support was provided through a grant from Amgen Inc" (pharmaceutical industry) Declarations/conflicts of interest: not reported Data handling by review authors: interruptions to cancer treatment outcome only reported narratively Other information of note: study was stopped after an interim analysis. Authors state that "owing to administrative obstacles, completion of the trial is not possible", rather than due to previously stated early stopping rules

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "The patients were randomized equally between the generation (selection two treatment groups" and "The study material and bias) randomization list were held by the UCLA Pharmacy for the duration of the study"

Comment: method of random sequence generation not described but done by UCLA Pharmacy and therefore probably done adequately Allocation Low risk Quote: "Amgen Inc...prepared and packaged all drug and concealment (selection placebo in identical containers, with the only designator bias) being the randomization number. The study material and randomization list were held by the UCLA Pharmacy for the duration of the study"

Comment: pharmacy-controlled randomisation would have ensured concealment of the random sequence from those recruiting participants Blinding of Low risk Quote: "double-blind, placebo-controlled" and "Amgen participants and Inc...prepared and packaged all drug and placebo in personnel identical containers, with the only designator being the (performance bias) randomization number"

Comment: the use of a placebo should have ensured that blinding was successful Blinding of outcome Low risk Quote: "double-blind, placebo-controlled" and "A single assessment (detection examiner...who was blinded to the patients' randomization

410 bias) status...scored mucositis on a weekly basis"

Comment: both participants and outcome assessor were blinded Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk Although the study reports an interim analysis only, the study was stopped due to administrative reasons and is unlikely to introduce bias

Spielberger 2004

Methods Trial design: parallel (2 arms) Location: USA Number of centres: 13 Study duration: recruitment from March 2001 to October 2002 Trials registry number: NCT00041665 Participants Inclusion criteria: 18 years or older; Karnofsky performance score of 70 or more; scheduled to undergo autologous HSCT after conditioning regimen of fractionated TBI plus etoposide and cyclophosphamide for haematological cancers; adequate cardiac, pulmonary, renal, and hepatic function Exclusion criteria: not reported Cancer type: haematologic: NHL (Gp A: 68%; Gp B: 65%); Hodgkins (Gp A: 20%; Gp B: 22%); multiple myeloma (Gp A: 10%; Gp B: 8%); leukaemia (Gp A: 2%; Gp B: 5%) Cancer treatment: prior to autologous HSCT, participants received the following conditioning regimen:  Radiotherapy: prior to chemotherapy, TBI of total 12 Gy in 6, 8, or 10 fractions over 3 or 4 days, with at least 6 hours between fractions  Chemotherapy: IV etoposide (60 mg/kg) the day after the last RT fraction (four days before HSCT) and cyclophosphamide (100 mg/kg) 2 days later (2 days before HSCT) All participants received G-CSF (filgrastim) 5 µg/kg per day from HSCT until neutrophil recovery Age at baseline (years): Gp A: median 48 (range 18 to 69); Gp B: median 49 (range 19 to 68) Gender: Gp A: 56% male; Gp B: 68% male Number randomised: 214 (Gp A: 107; Gp B: 107) Number evaluated: 212 (Gp A: 106; Gp B: 106) Interventions Comparison: KGF (palifermin) versus placebo Gp A: KGF (60 µg/kg) by IV for 3 consecutive days starting 3 days before RT, followed by 3 more doses after HSCT starting on the day of HSCT (total dose = 360 µg/kg) Gp B: placebo as above Compliance: 212 participants received at least one dose of their allocated intervention and 205 participants (Gp A: 103; Gp B: 102) "completed the study" (it is not clear whether or not this means that they received all doses) Duration of treatment: 6 treatment days over 13 to 14 days Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed daily by trained assessors starting 8 days before HSCT and for 28 days after HSCT or until severe mucositis had returned to grade 0 to 2, maximum score reported) (duration of grade 3 to 4 oral mucositis also measured but not an outcome of this

411 review)  Oral mucositis: RTOG 0 to 4 scale (assessed as above but only used for reporting duration of severe mucositis)  Oral mucositis: WCCNR 0 to 3 scale (assessed as above but only used for reporting duration of severe mucositis)  Oral pain: patient daily self-assessment (day -12 to 28) using OMDQ 5-point scale for mouth and throat soreness (higher = worse pain, reported as AUC, not used)  Quality of life: physical, functional, emotional, and social/family well-being domains of the FACT general questionnaire (days -12, -1, 7, 10, 14, and 28, reported as AUC, not used)  Normalcy of diet (measured as incidence of TPN)  Adverse events: 1 (mild) to 5 (fatal) scale  Number of days of treatment with opioid analgesics (reported as median and range, unable to use data)  Dysphagia (not an outcome of this review)  Febrile neutropenia (not an outcome of this review)  Infections (not an outcome of this review)  Survival (not an outcome of this review)  Laboratory results (not an outcome of this review) Notes Sample size calculation: 210 participants required to detect a mean difference in duration of severe oral mucositis of at least 3 days (SD 6.6) at 90% power and 5% significance Funding: "Funded by Amgen" (pharmaceutical industry) Declarations/conflicts of interest: the majority of authors were linked to the funding pharmaceutical company in terms of employment, consulting fees, lecture fees, receipt of research funding, and ownership of equity Data handling by review authors: N/A Other information of note: not reported

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear Quote: "patients were randomly assigned in a 1:1 ratio" generation (selection risk bias) Comment: insufficient information to determine method of random sequence generation Allocation Unclear Quote: "patients were randomly assigned in a 1:1 ratio" concealment (selection risk bias) Comment: insufficient information to determine method of random sequence generation Blinding of Low risk Quote: "placebo-controlled, double-blind" participants and personnel Comment: the use of a placebo should have ensured that (performance bias) blinding was successful Blinding of outcome Low risk Quote: "placebo-controlled, double-blind" assessment (detection bias) Comment: it is not clear who was blinded. There are subjective elements to the assessment of lower grades of oral mucositis using the WHO scale, requiring the patient's assessment of pain/soreness and their ability to swallow. Higher grades have more objective elements so may not be

412 affected by potential lack of blinding of the assessor Incomplete outcome Low risk Only two randomised participants (one per group) were not data (attrition bias) included in the analyses Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Su 2006

Methods Trial design: parallel (2 arms) Location: USA Number of centres: 1 Study duration: recruitment from January 1992 to December 1996 Trials registry number: none/unknown Participants Inclusion criteria: 18 years or older with histologically proven AJCC stage II to IV squamous cell carcinoma of the head and neck; undergone gross complete resection with negative pathological surgical margins and medically fit to begin RT within 8 weeks of surgery; Karnofsky performance score of 80% or more; adequate haematologic and serum metabolic laboratory indices Exclusion criteria: nasopharyngeal cancer; concurrent active malignancy other than localised; nonmelanoma skin cancer; previous RT to head and neck; previous chemotherapy; positive serum β-human chorionic gonadotropin in women of procreative potential; need for tube feeding at the start of RT Cancer type: stage II to IV squamous cell carcinoma of the head and neck: hypopharynx (Gp A: 11%; Gp B: 5%); larynx (Gp A: 16%; Gp B: 23%); oral cavity (Gp A: 26%; Gp B: 27%); oropharynx (Gp A: 21%; Gp B: 32%); unknown primary (Gp A: 26%; Gp B: 14%) Cancer treatment: postoperative external beam RT in once-daily fractions of 1.8 Gy, 5 days per week, to total dose of 63 Gy at primary site and involved neck (54 Gy to regional lymphatics at risk for subclinical metastasis; spinal cord dose limited to 45 Gy) Age at baseline (years): Gp A: median 67 (interquartile range 59 to 75); Gp B: median 61 (interquartile range 54 to 67) Gender: Gp A: 79% male; Gp B: 73% male Number randomised: 41 (Gp A: 19; Gp B: 22) Number evaluated: 40 (Gp A: 19; Gp B: 21) Interventions Comparison: G-CSF versus placebo Gp A: G-CSF (3 µg/kg) by daily subcutaneous injection, 7 days per week, starting 3 days before the start of RT and continued to end of RT (total dose = dependent on duration of RT) Gp B: placebo as above Compliance: not reported Duration of treatment: variable and dependent on duration of RT Outcomes  Oral mucositis: unknown 0 to 3 scale, where 0 = none, 1 = erythema, 2 = patchy mucositis, and 3 = confluent mucositis (assessed twice weekly by the treating radiation oncologist, maximum score reported)  Interruptions to cancer treatment: RT interruptions  Normalcy of diet (use of PEG feeding tube, as defined by 10% or higher weight loss from pre-RT baseline)  Adverse events (NCI-CTC)  Blood measurements (not an outcome of this review)  Survival (not an outcome of this review) Notes Sample size calculation: planned sample sample size was 66 per group to detect absolute difference of 20% in PEG tube use (from 10% in one group to 30% in

413 the other) at 80% power and 5% significance Funding: "This study was supported by NIH grant #CA69913" (government) Declarations/conflicts of interest: "No benefits in any form have been or will be received from a commercial party related directly or indirectly to the subject of this article" Data handling by review authors: although the scale for assessing oral mucositis does not completely match the 0 to 4 scales (such as WHO, etc), it was possible to dichotomise it for use in the 'any mucositis' meta-analysis, and we also used grade 3 (confluent mucositis) as incidence of severe mucositis Other information of note: planned to enrol 132 participants but stopped due to slow recruitment

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "Patients were randomized...by randomly permuted generation (selection blocks...A randomization list was prepared by the Memorial bias) Sloan-Kettering Cancer Center Biostatistics Service and held by the pharmacy"

Comment: specialist centre used Allocation Low risk Quote: "A randomization list was prepared by the concealment (selection Memorial Sloan-Kettering Cancer Center Biostatistics bias) Service and held by the pharmacy. Investigators did not have access to this list, ensuring that allocation could not be predicted before registration or changed afterwards"

Comment: central/remote randomisation would have ensured concealment of the random sequence from those recruiting participants Blinding of Low risk Quote: "double-blind, placebo-controlled" and "All patients participants and and treating physicians were blinded to treatment group personnel assignment" (performance bias) Comment: the use of a placebo should have ensured that blinding was successful Blinding of outcome Low risk Quote: "double-blind, placebo-controlled" and "All patients assessment (detection and treating physicians were blinded to treatment group bias) assignment"

Comment: the treating physician was the outcome assessor Incomplete outcome Low risk Only one randomised participant was not included in the data (attrition bias) analyses Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Vadhan-Raj 2010

Methods Trial design: parallel (2 arms) Location: University of Texas M.D. Anderson Cancer Center, USA Number of centres: 1

414 Study duration: recruitment from December 2005 to February 2008; last follow-up in May 2008 Trials registry number: NCT00267046 (mentioned in trial report) Participants Inclusion criteria: aged 15 to 65 years with sarcoma; due to start chemotherapy at the centre; Karnofsky performance score of 80% or more; adequate bone marrow, renal, and hepatic function Exclusion criteria: history of pelvic RT; clinically significant cardiac disease; undergone surgery within the previous 2 weeks Cancer type: sarcoma Cancer treatment: doxorubicin (total dosage 90 mg/m²) administered by continuous IV infusion over 72 hours, and ifosfamide (total dosage 10 g/m²) administered by 3-hour IV infusion for 4 days; participants with osteosarcoma (Gp A: 2; Gp B: 1) received the same dosage of doxorubicin but with intra- arterial cisplatin (120 mg/m²); all participants received G-CSF (pegfilgrastim) the day after chemotherapy; cycle repeated every 21 days; planned 6 cycles Age at baseline (years): Gp A: median 42 (range 17 to 63); Gp B: median 39 (range 15 to 64) Gender: Gp A: 53% male; Gp B: 50% male Number randomised: 48 (Gp A: 32; Gp B: 16) Number evaluated: 48 (Gp A: 32; Gp B: 16) Interventions Comparison: KGF (palifermin) versus placebo Gp A: KGF (180 µg/kg) by IV as single dose 3 days before the start of each chemotherapy cycle Gp B: placebo as above Compliance: the proportion of participants that completed all 6 blinded cycles (i.e. they took their allocated intervention) was higher in Gp A (63%) than Gp B (31%) Duration of treatment: 1 day per 3 week cycle; planned total 6 days of intervention over 18 weeks Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed for each cycle before chemotherapy, days 10, 12, 14 and at the end of the cycle, reported as incidence of grade 2 to 4 and grade 3 to 4, reported separately for blinded phase (first 2 cycles) and then the open-label phase) (duration of mucositis also measured but not an outcome of this review)  Oral pain: 0 to 10 (10 being worst) scale (assessed by questionnaire for each cycle before chemotherapy, days 10, 12, 14 and at the end of the cycle, reported for blinded phase, reported as median maximum score; unable to use data)  Quality of life: 1 to 7 (7 being worst) scale (assessed by questionnaire for each cycle before chemotherapy, days 10, 12, 14 and at the end of the cycle, reported for blinded phase, reported as median maximum score; unable to use data)  Normalcy of diet: eating and drinking assessed separately on 0 to 4 (4 being most difficult) scales (assessed by questionnaire for each cycle before chemotherapy, days 10, 12, 14 and at the end of the cycle, reported for blinded phase, reported as median maximum score; unable to use data)  Adverse events  Opioid analgesic use (reported as quantity per cycle; number of days of treatment with opioid analgesics is an outcome of this review and therefore we did not use this data)  Multiple patient-reported outcomes (overall health, sleeping, dysphagia, talking, brushing teeth, throat pain, rectal soreness) (not outcomes of this review)  Blood and laboratory measurements (not outcomes of this review)  Weight loss (not an outcome of this review)

415  Survival (not an outcome of this review) Notes Sample size calculation: 48 participants required to detect absolute difference of 50% (from 26% in Gp A to 76% in Gp B) in grade 2 to 4 mucositis at 88% power and 5% significance Funding: "Amgen provided the palifermin and partial funding for the study" (pharmaceutical industry) Declarations/conflicts of interest: principal investigator is a member of the Amgen board Data handling by review authors: we only report data from the first 2 cycles (blinded phase) as very few participants received placebo in the remaining cycles Other information of note: not reported

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "Two distinct computer-generated randomization generation (selection lists were prepared by the University of Texas M.D. bias) Anderson Cancer Center, Department of Biostatistics, one for the 20 patients who consented to pharmacokinetic sampling and the other for the 28 patients who did not. For the pharmacokinetics cohort, the treatment allocation ratio was 4 patients receiving palifermin to 1 receiving placebo, in blocks of 5; for the other cohort, the ratio was 4 patients receiving palifermin to 3 receiving placebo, in blocks of 14"

Comment: adequate method used Allocation Low risk Quote: "The statistician provided both randomization lists concealment (selection to the pharmacy, so the patient and the clinical research bias) team (who assessed outcomes) were blinded to the study treatment. At patient enrolment, the research team notified the pharmacy, which assigned the patient the next sequential slot and treatment from the appropriate randomization list on the basis of whether he or she had consented to pharmacokinetic sampling. The pharmacy provided the research team with the blinded study medication. Upon completion of the study, pharmacy provided the statistician with the 2 randomization lists, including individual patient treatment assignments, for analysis"

Comment: pharmacy-controlled randomisation would have ensured concealment of the random sequence from those recruiting participants Blinding of Low risk Quote: "double-blind, placebo-controlled" and "the patient participants and and the clinical research team (who assessed outcomes) personnel were blinded to the study treatment" and "Blinding might (performance bias) not have been maintained because of adverse effects of palifermin"

Comment: some personnel may not have been blinded due to increased adverse effects, although adverse effects from palifermin may have been difficult to isolate from adverse effects of cancer treatment

416 Blinding of outcome Low risk Quote: "double-blind, placebo-controlled" and "the patient assessment (detection and the clinical research team (who assessed outcomes) bias) were blinded to the study treatment" and "Blinding might not have been maintained because of adverse effects of palifermin" and "patients were assessed at each cycle by both research and clinical teams, including those without direct knowledge of the protocol"

Comment: it seems unlikely that lack of blinding would affect outcomes as the higher grades of oral mucositis assessed in this study are more objective; also, some were assessed by personnel without knowledge of the protocol Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent van der Lelie 2001

Methods Trial design: parallel (2 arms) Location: The Netherlands Number of centres: 1 Study duration: recruitment from May 1997 to August 1999 Trials registry number: none/unknown Participants Inclusion criteria: malignant disease and due to have myeloablative treatment followed by autologous or allogeneic BMT or PBSCT Exclusion criteria: not reported Cancer type: haematologic: lymphoma (Gp A: 28%; Gp B: 50%); acute leukaemia (Gp A: 28%; Gp B: 28%); CML (Gp A: 22%; Gp B: 6%); multiple myeloma (Gp A: 22%; Gp B: 17%) Cancer treatment: prior to autologous or allogeneic BMT or PBSCT, participants received the following conditioning regimens:  BEAM (Gp A: 28%; Gp B: 44%): carmustine (BCNU) 300 mg/m² on day - 6; etoposide (VP16) and cytosine arabinoside (Ara-C) 200 mg/m² of each on days -5, -4, -3, -2; melphalan 140 mg/m² on day -1  CYTBI (Gp A: 56%; Gp B: 44%): cyclophosphamide 60 mg/kg on days -5 and -4; TBI 4.5 Gy on days -2 and -1  BUCY (Gp A: 17%; Gp B: 11%): busulfan 4 mg/kg on days -7, -6, -5, -4; cyclophosphamide 60 mg/kg on days -3 and -2 Lymphoma patients received BEAM; other patients received CYTBI or BUCY; nearly all participants received autologous or allogeneic PBSCT Age at baseline (years): Gp A: median 47 (range 25 to 63); Gp B: median 48 (range 18 to 62) Gender: Gp A: 61% male; Gp B: 39% male Number randomised: 36 (Gp A: 18; Gp B: 18) Number evaluated: 36 (Gp A: 18; Gp B: 18) Interventions Comparison: GM-CSF versus placebo Gp A: GM-CSF (300 µg daily dose) gel, 5 ml twice daily (early in the morning and before going to sleep) kept in the mouth for as long as possible and then swallowed; no oral intake for 1 hour afterwards; starting on day 1 (the day after the day of transplant) and continued until neutrophil recovery (typically up to 14 days after transplant) Gp B: placebo as above

417 All participants followed hospital's standard protocol for mouth care: toothbrushing after meals, rinsing with 0.9% saline, and if there was inflammation, 0.12% chlorhexidine rinse 6 times daily Compliance: 8 participants (Gp A: 4; Gp B: 4) did not complete the study; they all cited the main reason being nausea and the taste of the gel Duration of treatment: variable and dependent on neutrophil recovery Outcomes  Oral mucositis: WHO 0 to 4 scale (assessed daily by dentists, reported in text as incidence of grade 3 to 4)  Oral mucositis: 8 (all 8 sites normal) to 24 (all 8 sites severely affected) oral assessment score (assessed daily by dentists, reported graphically as median score over 14 days, unable to use data)  Oral pain: 0 (no pain) to 100 (worst pain) VAS (assessed daily by participant, reported graphically as median score over 14 days, unable to use data)  Normalcy of diet: incidence of total parenteral nutrition (started when patients were unable to eat for longer than 3 days)  Number of days in hospital (reported as median, unable to use data)  Opioid analgesic use (reported as incidence; number of days of treatment with opioid analgesics is an outcome of this review and therefore we did not use this data)  Blood measurements (not an outcome of this review)  Fever (not an outcome of this review)  Infection (not an outcome of this review)  Antibiotic use (not an outcome of this review) Notes Sample size calculation: not reported ("GM-CSF was supplied by the sponsor for 18 patients so that 36 patients could be included in the study. This should be enough to demonstrate a clinically significant difference") Funding: "We thank Novartis and Schering-Plough for supplying the GM-CSF" (pharmaceutical industry) Declarations/conflicts of interest: not reported Data handling by review authors: N/A Other information of note: GM-CSF administration only began after the conditioning treatment had been completed, by which point oral mucositis may have already begun to develop

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear Quote: "the patients were randomized to receive GM-CSF generation (selection risk or placebo" bias) Comment: insufficient information to determine method of random sequence generation Allocation Unclear Quote: "the patients were randomized to receive GM-CSF concealment (selection risk or placebo" bias) Comment: insufficient information to determine method of random sequence generation Blinding of Low risk Quote: "double-blind placebo-controlled" and "There was participants and no difference in taste or appearance between the placebo personnel and the GM-CSF" (performance bias) Comment: the use of a placebo should have ensured that

418 blinding was successful Blinding of outcome Low risk Quote: "double-blind placebo-controlled" and "There was assessment (detection no difference in taste or appearance between the placebo bias) and the GM-CSF"

Comment: it is not clear who was blinded. There are subjective elements to the assessment of lower grades of oral mucositis using the WHO scale, requiring the patient's assessment of pain/soreness and their ability to swallow. Higher grades have more objective elements so may not be affected by potential lack of blinding of the assessor Incomplete outcome Low risk All randomised participants were included in the analyses data (attrition bias) Selective reporting Low risk Data for outcomes of this review were reported (reporting bias) appropriately Other bias Low risk No other sources of bias are apparent

Wu 2009

Methods Trial design: parallel (4 arms) Location: South Korea Number of centres: 6 Study duration: recruitment from January 2007 to August 2007 Trials registry number: none/unknown Participants Inclusion criteria: 18 years or older with histological evidence of head and neck cancer; due to receive at least 5 weeks of primary RT, primary chemoradiotherapy, or postoperative RT; agreement to have complete medical history and physical examination; ECOG score of 0 to 2; white blood counts 3 x 10³/L or higher; platelet counts 100 x 10³/L or higher Exclusion criteria: oral ulcers; herpes simplex virus; severe periodontal disease; serum creatine levels greater than 2 mg/dL; ALT/AST values greater than 200 IU/L; cytotoxic chemotherapy or RT within 3 weeks of the start of the study; systemic or topical corticosteroids within 30 days of the start of the study; participated in another clinical study within 30 days of the start of the study Cancer type: head and neck: nasopharynx (Gp A: 36%; Gp B: 32%); oropharynx (Gp A: 24%; Gp B: 25%); oral cavity (Gp A: 19%; Gp B: 25%); hypopharynx (Gp A: 2%; Gp B: 4%); Other (Gp A: 19%; Gp B: 14%) Cancer treatment:  Radiotherapy: conventional fractionation once daily in 2 (± 0.25) Gy fractions, 5 times per week, for at least 5 weeks  Chemotherapy: concurrent cisplatin was allowed (just over 50% per group received this) Age at baseline (years): Gp A: median 56 (range 18 to 75); Gp B: median 51 (range 18 to 77) Gender: Gp A: 69% male; Gp B: 57% male Number randomised: 113 (Gp A: 85; Gp B: 28) Number evaluated: 103 (Gp A: 76; Gp B: 27) for incidence of moderate to severe mucositis; 94 (Gp A: 70; Gp B: 24) for incidence of severe mucositis Interventions Comparison: EGF (recombinant human) versus placebo Gp A: EGF  (n = 29): 10 µg daily by oral spray, applied twice daily, sprayed over the entire oral mucosa and then swallowed, no oral intake for 30 mins afterwards; starting on first day of RT and continuing for 5 weeks  (n = 29): 50 µg daily as above

419  (n = 27): 100 µg daily as above Gp B: placebo as above All participants gargled with chlorhexidine to maintain oral hygiene Compliance: not reported Duration of treatment: 5 weeks Outcomes  Oral mucositis: RTOG 0 to 4 scale (assessed weekly by radiation oncologists, reported as incidence of grade 2 to 4 and grade 3 to 4 at week 4 or 5; unclear reporting for grade 3 to 4, unable to use data) (time to develop mucositis also assessed but not an outcome of this review)  WHO oral toxicity grade (assessed weekly, not clear whether this is the WHO 0 to 4 scale for oral mucositis, no data reported)  Interruptions to cancer treatment (incidence of 3 or more consecutive days of interruption to RT)  Oral pain (assessed weekly, no description of scale used, no data reported)  Averse events (no data reported; narrative only)  Opioid analgesic use (assessed weekly, not clear if duration, quantity or incidence of use was assessed, no data reported)  Weight (not an outcome of this review)  Laboratory measurements (not an outcome of this review) Notes Sample size calculation: allowing for 10% attrition, 26 participants per group required Funding: "This study was supported by a grant from the National R&D Program for Cancer Central, Ministry of Health, Welfare and Family Affairs, Republic of Korea (0620270)" (government) and "The EGF and placebo treatments were supplied free of charge" (presumably pharmaceutical industry) Declarations/conflicts of interest: not reported Data handling by review authors: we combined the three EGF groups to make a single pairwise comparison against placebo; no formal statistical analysis was undertaken for the different EGF dosages Other information of note: not reported

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear Quote: "We assigned patients randomly to 4 arms" generation (selection risk bias) Comment: insufficient information to determine method of random sequence generation Allocation Unclear Quote: "We assigned patients randomly to 4 arms" concealment (selection risk bias) Comment: insufficient information to determine method of random sequence generation Blinding of Low risk Quote: "double-blind, placebo-controlled" and "Three participants and concentrations of EGF and a placebo containing the drug personnel delivery vehicle were manufactured, packaged, and (performance bias) supplied in a double-blind fashion"

Comment: the use of a placebo should have ensured that blinding was successful Blinding of outcome Low risk Quote: "double-blind, placebo-controlled" assessment (detection bias) Comment: it is not clear who was blinded. There are

420 subjective elements to the assessment of most grades of oral mucositis using the RTOG scale, requiring the patient's assessment of pain. Grade 4 is more objective so may not be affected by potential lack of blinding of the assessor Incomplete outcome Low risk Overall attrition was 9% (Gp A: 11%; Gp B: 4%). Slight data (attrition bias) differential between groups but the reasons were reported and mostly unrelated to interventions/outcomes. Unlikely to have biased the results Selective reporting High risk Several outcomes were assessed but not properly reported (reporting bias) ("No secondary endpoint showed any difference between the placebo and study groups") Other bias Low risk No other sources of bias are apparent

Footnotes

5FU = fluorouracil; AJCC = American Joint Committee on Cancer; ALL = acute lymphoblastic leukaemia; allogeneic = cells from donor; AML = acute myelogenous leukaemia; AUC = area under the curve; autologous = patients' own cells; BMT = bone marrow transplantation; CALGB = Cancer and Leukaemia Group B; cGy = centigray; CML = chronic myelogenous leukaemia; CRT = chemoradiotherapy; CT = chemotherapy; ECOG = Eastern Co-operative Oncology Group; EGF = epidermal growth factor; EQ-5D = European Quality Of Life Utility Scale; FACT = Functional Assessment of Cancer Therapy; G-CSF = granulocyte-colony stimulating factor; GM-CSF = granulocyte-macrophage colony-stimulating factor; GVHD = graft-versus-host disease; Gy = gray; HLA = human leukocyte antigen; HSCT = haematopoietic stem cell transplantation; IA = intra-arterial; ITF = intestinal trefoil factor; ITT = intention-to- treat; IV = intravenous; KGF = keratinocyte growth factor; MDS = myelodysplastic syndrome; N/A = not applicable; NCI-CTC = National Cancer Institute common toxicity criteria; NHL = non-Hodgkin's lymphoma; OMAS = oral mucositis assessment scale; OMDQ = oral mucositis daily questionnaire; OMWQ-HN = Oral Mucositis Weely Questionnaire - Head and Neck cancer; PBPC = peripheral blood progenitor cell; PBSCT = peripheral blood stem cell transplantation; PEG = percutaneous endoscopic gastrostomy; PP = per protocol; RT = radiotherapy; RTOG = Radiation Therapy Oncology Group; SCC = squamous cell carcinoma; SCT = stem cell transplantation; SD = standard deviation; SMD = standardised mean difference; TBI = total-body irradiation; TGF = transforming growth factor; TPN = total parenteral nutrition; VAS = visual analogue scale; WBC = white blood cell; WHO = World Health Organisation; WCCNR = Western Consortium for Cancer Nursing Research

Characteristics of excluded studies

Antin 2002

Reason for exclusion Recombinant human interleukin-11 versus placebo. Study stopped early due to adverse event triggering preset stopping rule - data only for 10 patients in rhIL group and 3 placebo de Koning 2007

Reason for exclusion TGF-beta(2) versus placebo. Cross-over study with no first period data reported (see 'Types of studies')

Foncuberta 2001

421

Reason for exclusion TGF-beta(3) versus placebo. Patients assigned sequentially, not randomised

Gebbia 1994

Reason for exclusion G-CSF versus thymopentin versus G-CSF plus thymopentin versus placebo. Oral mucositis not mentioned (unknown if measured)

Gladkov 2013

Reason for exclusion G-CSF (lipegfilgrastim) versus G-CSF (pegfilgrastim). Oral mucositis not mentioned (unknown if measured). Insufficient information (abstract)

Gordon 1993

Reason for exclusion GM-CSF versus no treatment. Unclear if randomised or not. Only 13 participants. Insufficient information (abstract)

Horsley 2007

Reason for exclusion KGF (palifermin) versus standard care. No random allocation

Hunter 2009

Reason for exclusion ATL-104 versus placebo. Not RCT - this study combines patients who were in cohorts with increasing doses of mouthrinse to assess safety, with an RCT

Ifrah 1999

Reason for exclusion GM-CSF versus placebo. Primary outcome was survival/cure, with mucositis as a toxicity. Unclear if mucositis was oral or gastrointestinal

Iwase 1997

Reason for exclusion G-CSF versus no treatment. No mention of randomisation and no description of when intervention given in relation to cancer treatment

Jones 1996

Reason for exclusion GM-CSF versus placebo. Unclear if mucositis was oral or gastrointestinal

Karthaus 1998

Reason for exclusion G-CSF versus placebo. Only 8 patients, 32 chemo cycles and results presented assuming independent

422 Kubo 2016

Reason for exclusion G-CSF (pegfilgrastim) versus G-CSF (filgrastim). Incidence of stomatitis reported in adverse events table (JapicCTI-111394)

Lee 2016

Reason for exclusion G-CSF (pegteograstim) versus G-CSF (pegfilgrastim). Incidence of stomatitis reported in adverse events table (NCT01328938)

Legros 1997

Reason for exclusion GM-CSF versus placebo. Unclear if mucositis was oral or gastrointestinal

Limaye 2013

Reason for exclusion AG013 versus placebo. AG013 is composed of recombinant Lactococcus lactis engineered to secrete human Trefoil Factor 1. Randomised at first but participants not developing oral mucositis in chemotherapy cycle 1 were excluded from the next phase where oral mucositis was measured, so randomisation was lost (NCT00938080)

Nabholtz 2002

Reason for exclusion G-CSF (leridistim) versus G-CSF (filgrastim). Incidence of stomatitis reported in adverse events table

NCT00360971

Reason for exclusion KGF (palifermin) versus placebo. Study terminated at 21 participants (298 planned) due to positive preliminary results from other palifermin studies)

NCT00626639

Reason for exclusion KGF (palifermin) versus placebo. Study terminated at 5 participants due to departure of PI and slow enrolment

Pettengell 1992

Reason for exclusion G-CSF versus no treatment. Unclear if mucositis was oral or gastrointestinal

Ryu 2007

Reason for exclusion GM-CSF versus placebo. Some participants (6%) had oral mucositis at baseline (NCT00008398)

Throuvalas 1995

423 Reason for exclusion GM-CSF versus no treatment. Probably not RCT - described as comparative study. Only 10 participants. Insufficient information (abstract)

Tsurusawa 2016

Reason for exclusion G-CSF versus no treatment. Incidence of stomatitis reported in adverse events table (UMIN ID: 000000675)

Vitale 2014

Reason for exclusion KGF (palifermin) versus no treatment. From full text it is not RCT - retrospective and allocation of KGF/no KGF based on doctor's decision

Footnotes

G-CSF = granulocyte-colony stimulating factor; GM-CSF = granulocyte-macrophage colony- stimulating factor; KGF = keratinocyte growth factor; RCT = randomised controlled trial; TGF = transforming growth factor

Characteristics of studies awaiting classification

ACTRN12606000083594

Methods Multicentre, double blind, randomised placebo-controlled trial Participants 18 years or older with lymphoma and due to undergo high-dose BEAM chemotherapy and autologous stem cell transplantation as inpatient Interventions Whey growth factor extract at 13.5 mg/ml in sterile saline versus placebo (sterile saline) Outcomes  Oral mucositis: WHO 0 to 4 scale  Normalcy of diet: incidence of enteral/parenteral feeding  Duration of opiate analgesics  Quality of life: OMDQ  Adverse events Notes Funding: commercial sector/Industry (TGR Biosciences, Australia)

Contact: correspondence with pharmaceutical company suggests no benefit

Antonadou 1998 (241)

Methods Multicentre, randomised controlled trial (no placebo) Participants Head and neck cancer; receiving continuous course of radiotherapy for 6 to 7 weeks Interventions GM-CSF (subcutaneous) versus no treatment Outcomes Signs and symptoms of oral mucositis: erythema, pain and dysphagia each measured on 0 to 3 scale

424 Notes Abstract only (we were unable to link this abstract to a full text report)

Some P-values suggest benefit for GM-CSF at some timepoints.

"GM-CSF reduces the incidence and severity of radiation mucositis and allows the completion of the XRT treatment without protraction"

Elsaid 2001(242)

Methods Randomised controlled trial (no placebo) Participants Anaemic participants with head and neck cancer receiving radiotherapy once daily (1.8 or 2 Gy) to doses of 66 to 70 Gy Interventions Recombinant human erythropoietin versus no treatment Outcomes Incidence of grade 3 mucositis and dermatitis Notes Abstract only (we were unable to link this abstract to a full text report)

Higher rate of mucositis in the no-treatment group (5.9% versus 0%)

Grzegorczyk 2006 (243)

Methods Randomised placebo-controlled trial Participants Adults (aged 19 to 68 years) undergoing haematopoietic stem cell transplantation Interventions G-CSF versus placebo Outcomes  Oral mucositis: WHO 0 to 4 scale  Oral pain: 0 to 10 VAS  Neutrophil counts Notes Translation provided insufficient information. Discrepancy between graph legends and descriptions Unable to contact author

Koga 2015 (244)

Methods Randomised controlled trial (no placebo) Participants Children with B-cell non-Hodgkin lymphoma (B-NHL) receiving chemotherapy Interventions G-CSF versus no treatment Outcomes  Oral mucositis (no further details)  Duration of hospitalisation  Incidence of febrile neutropenia  Infections  Time to neutrophil recovery  Costs Notes Abstract only (we were unable to link this abstract to a full text report)

"G-CSF+ patients showed a positive impact on the meantime to neutrophil recovery and hospital stay, but they had no impact in incidences of febrile neutropenia, infections, stomatitis, and total cost"

NCT00293462

425 Methods Double blind, randomised placebo-controlled trial (one of the three arms is a cross-over) Participants 18 years or older with head and neck cancer due to receive conventional or hyperfractionated radiotherapy or intensity-modulated radiotherapy (IMRT) with or without concurrent chemotherapy Interventions GM-CSF mouthwash versus salt and soda mouthwash versus both (cross-over arm) Outcomes  Oral mucositis: scale/s unclear  Quality of life: 0 to 10 scale  Oral pain: 0 to 10 scale  Functional status by Karnofsky performance status scale Notes Trials registry number: NCT00293462

Funding: University of California, San Francisco; National Cancer Institute (NCI)

NCT00393822

Methods Double blind, randomised placebo-controlled trial Participants 18 years or older with resected colon cancer (American Joint Committee on Cancer Stage 2B or 3) and due to receive 5-FU and leucovorin Interventions KGF versus placebo Outcomes  Oral mucositis: WHO 0 to 4 scale - incidence and duration of moderate to severe (grade 2 to 4)  Interruptions to cancer treatment  Mouth and throat pain  Adverse events  Survival  Changes in laboratory values  Serum anti-KGF antibody formation Notes Trials registry number: NCT00393822

Funding: Amgen (pharmaceutical industry)

NCT02303197

Methods Randomised controlled trial (no placebo) Participants Adults (aged 18 to 75 years) with head and neck cancer due to receive radiotherapy Interventions Recombinant human EGF versus ChiNing Outcomes  Oral mucositis: RTOG 0 to 4 scale  Oral pain: VAS  Quality of life: EORTC QLQ-H&N35  Weight change  Safety (blood/urine/kidney/liver/electrocardiogram) Notes Trials registry number: NCT02303197

Funding: Tianjin Medical University Cancer Institute and Hospital

NCT02313792

426

Methods Double blind, randomised placebo-controlled trial Participants 16 years or older due to receive preparative cancer treatment regimen followed by autologous or allogeneic HSCT (cancer type not reported) Interventions KGF versus placebo Outcomes  Oral mucositis: incidence and duration of severe (scale not reported)  Oral pain: VAS  Use of opioid analgesics  Quality of life (scale not reported)  Cost-effectiveness Notes Trials registry number: NCT02313792

Funding: The Catholic University of Korea; Collaborator: "BLNH" (probably pharmaceutical industry)

Patte 2002 (245)

Methods Randomised controlled trial (no placebo) Participants Children with non-Hodgkin's lymphoma due to receive two courses of COPAD induction chemotherapy Interventions G-CSF versus no treatment Outcomes  Mucositis (unclear if oral or gastrointestinal): 0 to 4 scale - incidence of severe (grade 3 to 4)  Duration of hospitalisation  Febrile neutropenia  Severe infections  Blood measurements  Use of IV antifungals or antibiotics  Survival Notes "The incidence of grade 3 and 4 mucositis was similar in both arms" Contact: emailed corresponding author July 2017 to clarify if mucositis is oral or gastrointestinal

Schuster 2007a (246)

Methods Multicentre, double blind, randomised placebo-controlled trial Participants 18 years or older with multiple myeloma or lymphoma due to receive high-dose chemotherapy (with or without TBI) followed by autologous HSCT Interventions Recombinant human FGF-20 (velafermin) (3 arms with different dosages) versus placebo Outcomes  Oral mucositis: WHO 0 to 4 scale - incidence and duration of severe (grade 3 to 4)  Adverse events Notes Abstract only (we were unable to link this abstract to a full text report)

Trials registry number: NCT00104065

Schuster 2007b (247)

Methods Multicentre, double blind, randomised placebo-controlled trial

427 Participants 18 years or older with multiple myeloma or lymphoma undergoing high-dose chemotherapy (with or without TBI) followed by autologous HSCT Interventions Recombinant human FGF-20 (velafermin) (3 arms with different dosages) versus placebo Outcomes  Oral mucositis: WHO 0 to 4 scale - incidence of severe (grade 3 to 4)  Adverse events Notes Abstract only (we were unable to link this abstract to a full text report)

Trials registry number: NCT00323518

Shea 2007 (248)

Methods Randomised controlled trial (no placebo) Participants Aged 18 to 74 years with lymphoma, leukaemia or multiple myeloma; TBI plus high-dose chemotherapy followed by autologous peripheral blood stem cell transplantation Interventions KGF (4 arms with different schedules) Outcomes  Oral mucositis: WHO 0 to 4 scale  Mouth and throat pain: OMDQ  Opioid analgesic use  Adverse events Notes Abstract only (we were unable to link this abstract to a full text report)

Trials registry number: NCT00109031

Spielberger 2001 (233)

Methods Multicentre, double blind, randomised placebo-controlled trial Participants Haematologic malignancies; eligible for TBI plus high-dose chemotherapy followed by autologous peripheral blood stem cell transplantation; aged 12 to 65 years Interventions KGF (7 doses) versus KGF (4 doses + 3 doses of placebo) versus placebo (7 doses) Outcomes  Oral mucositis: WHO 0 to 4 scale - incidence and duration of severe (grade 3 to 4)  Use of opioid analgesics  Quality of life  Febrile neutropenia  Use of IV antifungals or antibiotics  Diarrhoea Notes Abstract only (we were unable to link this abstract to a full text report)

Trials registry number: NCT00004132

Footnotes

5FU = fluorouracil; EGF = epidermal growth factor; EORTC QLQ-H&N35 = European Organisation for Research and Treatment of Cancer, Quality of Life Questionnaire, Head and Neck Module; G-CSF = granulocyte-colony stimulating factor; GM-CSF = granulocyte- macrophage colony-stimulating factor; HSCT = haematopoietic stem cell transplantation; KGF = keratinocyte growth factor; NCI-CTC = National Cancer Institute common toxicity criteria;

428 OMDQ = oral mucositis daily questionnaire; FGF-20 = fibroblast growth factor-20; TBI = total- body irradiation; VAS = visual analogue scale; WHO = World Health Organisation

429 APPENDIX 8: Cytokines and growth factors review - secondary outcome analyses

Analysis 1.4: KGF versus placebo; Outcome: Interruptions to cancer treatment (unscheduled RT breaks of 5 or more days)

Analysis 1.5: KGF versus placebo; Outcome: Interruptions to cancer treatment (chemotherapy delays/discontinuations) 430

Analysis 1.6: KGF versus placebo; Outcome: Oral pain

431

Analysis 1.7: KGF versus placebo; Outcome: Normalcy of diet (use of supplemental nutrition) 432

Analysis 1.8: KGF versus placebo; Outcome: Normalcy of diet (worst ability to eat score - 1 to 4 scale)

Analysis 1.9: KGF versus placebo; Outcome: Number of days in hospital

433

Analysis 1.10: KGF versus placebo; Outcome: Number of days of treatment with opioid analgesics

Analysis 2.1: KGF (dose comparison); Outcome: Oral mucositis (any) 434

Analysis 2.2: KGF (dose comparison); Outcome: Oral mucositis (moderate to severe)

435

Analysis 2.3: KGF (dose comparison); Outcome: Oral mucositis (severe)

Analysis 2.4: KGF (dose comparison); Outcome: Oral pain (maximum score on 0 to 10 VAS)

436

Analysis 2.5: KGF (dose comparison); Outcome: Normalcy of diet (use of total parenteral nutrition)

Analysis 2.6: KGF (dose comparison); Outcome: Normalcy of diet (worst ability to eat score - 1 to 4 scale)

437

Analysis 2.7: KGF (dose comparison); Outcome: Number of days in hospital

Analysis 2.8: KGF (dose comparison); Outcome: Number of days of treatment with opioid analgesics

438

Analysis 4.4: GM-CSF versus control; Outcome: Oral pain (maximum score on 0 to 10 VAS)

Analysis 4.5: GM-CSF versus control; Outcome: Normalcy of diet (use of feeding tube/parenteral nutrition) 439

Analysis 4.6: GM-CSF versus control; Outcome: Number of days of treatment with opioid analgesics

Analysis 5.1: GM-CSF (dose comparison); Outcome: Oral mucositis (severe)

440

Analysis 6.3: GM-CSF versus sucralfate; Outcome: Interruptions to cancer treatment

Analysis 6.4: GM-CSF versus sucralfate; Outcome: Normalcy of diet (use of PEG tube) 441

Analysis 7.4: G-CSF versus placebo; Outcome: Interruptions to cancer treatment (RT interruption)

Analysis 7.5: G-CSF versus placebo; Outcome: Normalcy of diet (use of PEG tube) 442

Analysis 8.1: G-CSF (pegfilgrastim) versus G-CSF (filgrastim); Outcome: Oral mucositis (any)

Analysis 8.2: G-CSF (pegfilgrastim) versus G-CSF (filgrastim); Outcome: Oral mucositis (moderate to severe) 443

Analysis 8.3: G-CSF (pegfilgrastim) versus G-CSF (filgrastim); Outcome: Normalcy of diet (use of supplemental nutrition)

Analysis 9.3: EGF versus placebo; Outcome: Interruptions to cancer treatment (RT breaks > 2 consecutive days) 444

Analysis 9.4: EGF versus placebo; Outcome: Normalcy of diet (use of supplemental nutrition)

Analysis 11.1: ITF (dose comparison); Outcome: Oral mucositis (any) 445

Analysis 11.2: ITF (dose comparison); Outcome: Oral mucositis (moderate to severe)

Analysis 11.3: ITF (dose comparison); Outcome: Oral mucositis (severe) 446

Analysis 12.4: Erythropoietin versus placebo; Outcome: Number of days in hospital

447 APPENDIX 9: Cytokines and growth factors review - adverse events tables

Table 1: Keratinocyte growth factor

Study ID Adverse events results  AEs with incidence ≥ 10% greater in KGF group: Higher rate of skin rash in KGF group (65/69 vs 21/31; RR 1.39, 95% CI 1.08 to 1.79; P = 0.01). Insufficient evidence of a difference in edema, infection, or local pain. Blazar 2006  Grade 3 to 4 (WHO and NCI-CTC 0 to 4 toxicities scales) AEs with higher incidence in KGF group: Insufficient evidence of a difference in skin reactions, diarrhoea, local pain, or cardiac events.  KGF-related AE (NCI-CTC): Higher rate in KGF group (141/220 vs 17/57; RR 2.15, 95% CI 1.43 to 3.24; P = 0.0003). Blijlevens  KGF-related serious AE (no definition of 'serious' given) (NCI-CTC): 2013 Insufficient evidence of a difference (4/220 vs 0/57; P = 0.56).  KGF-related severe AE (NCI-CTC grade 3, 4 or 5) (grade 5 = death): Insufficient evidence of a difference (23/220 vs 0/57; P = 0.08). Bradstock Insufficient evidence of a difference in infection or Grade 3 to 4 (NCI-CTC 0 to 4 2014 toxicity scale) skin rash/desquamation The authors state that most adverse events were considered to be caused by the cancer treatment or the underlying cancer itself and not related to study treatment. Two participants in the palifermin group had serious adverse events considered to Brizel 2008 be related to the intervention: one had increased sputum production; the other had dehydration, dysphagia, pain (including abdominal), pancreatitis, and subsequently had schistosomiasis Total of 28 side effects in palifermin group occurring in 11 of 22 patients (50%) who received at least 4 of the 6 doses. Most frequent (90.9%) were cases of erythema or exanthema, often associated with itching (54.5%). Often (54.5%) a swelling of the oral mucosa including the tongue occurred. In 4 out of 6 patients, this was accompanied by taste disturbance. The severity of side effects were Fink 2011 classified as mild to moderate. The CTC Grade 3 occurred only once in the form of a strong heat sensation.

In one of the 11 cases, there was premature discontinuation of palfermin due to severe facial swelling with eyelid and laryngeal pain as well as painful swelling of the hands following the second injection 25 different adverse events were reported and were mostly not KGF-related. There Freytes was insufficient evidence of a difference for diarrhoea, abdominal pain, infection 2004 or rash "two patients reported knee joint pain, skin rash was observed in one patient, two Gholizadeh patients had abnormal taste, and one showed lingual mucosal thickening" (control 2016 group was chlorhexidine mouthrinse. The authors do not report the events by treatment group)  "Initially, patients were allocated to three arms: palifermin (180 g/kg/wk) throughout radiochemotherapy (ie, for at least seven doses); palifermin (180g/kg/wk) for four doses and then placebo throughout the remainder of radiochemotherapy; or placebo throughout radiochemotherapy. However, Henke 2011 after one serious adverse event of respiratory insufficiency was reported in one of the first 10 patients, the data monitoring committee concluded that the study should be restarted with a lower palifermin dose (120 g/kg/wk)"  "Most patients (97%) experienced at least one adverse event...... One serious adverse event (febrile neutropenia) considered related to study drug was

448 reported for one patient in the palifermin arm" KGF-related AEs with incidence ≥ 5% in KGF group: Higher rate of gastrointestinal disorders in KGF group (18/78 vs 2/73; RR 8.42, 95% CI 2.02 to Jagasia 35.04; P = 0.003). Insufficient evidence of a difference in any AE, tongue coating, 2012 tongue disorder, skin and subcutaneous tissue disorders, rash, pruritus, or erythema. "Study drug–related AEs were reported for 35% of palifermin and 11% of placebo patients. The most frequent study drug–related AEs (palifermin, placebo) were rash (9%, 2%), flushing (5%, 0%), dysgeusia (5%, 1%), nausea (4%, 1%),and vomiting (3%, 1%). None of these events led to study withdrawal. Serious AEs Le 2011 considered related to study treatment were reported for five palifermin patients (5%; one each with necrotic pancreatitis, hypersensitivity, tracheostomy malfunction, peritoneal carcinoma, and convulsion) and two placebo patients (2%; one each with hepatitis/hepatic enzyme increase and cryptogenic organizing pneumonia)" "The administration of palifermin was generally safe and without considerable complications. The only adverse reactions were rashes (lasting for 48–72 hours) Lucchese localized to the face, upper neck and shoulders, erythema, and altered taste 2016a (consistent with the pharmacologic action of palifermin of oral epithelium and skin), most of which were of NCI grade 1 or 2 severity" "The administration of palifermin was mostly safe and without substantial complications. The mean duration of the OM and the number of adverse event was Lucchese significant less in the palifermin group (Tables II, III, Figure 1). The main adverse 2016b episodes were erythema, cutaneous rashes and altered taste and three of the patients in the palifermin group showed a light thickness of the tongue, mouth and palate" "Although the predefined frequency of DLTs attributable to KGF was not reached with KGF doses between 1 and 80 µg/kg/d, there were three adverse reactions involving the skin that required discontinuation of KGF in the 18 patients treated with 60 or 80 µg/kg (Table 5). Overall, skin and oral adverse events (rash, flushing, pruritis, edema, hypoesthesia, paresthesia, tongue disorder [thickening], and alteration in taste sensation) attributed to KGF occurred in 13 of 18 patients Meropol treated with 60 and 80 µg/kg of KGF (eight patients, grade 1; four patients, grade 2003 2; and one patient, grade 3) and in three of 11 patients treated with 40 µg/kg (all grade 1). These events were reported in 16 of 39 patients (41%) dosed with KGF at > 20 µg/kg/d, whereas these symptoms were reported in only two of 21 subjects (10%) treated with placebo. The skin and oral toxicities associated with KGF were generally mild to moderate in severity, with onset approximately 36 hours after the first dose of KGF and resolution 7 to 10 days thereafter"  "As expected based on the pharmacologic activity of palifermin, oral-related AEs were reported more frequently in palifermin than in patients receiving placebo (Table 3). During cycle 1, 50% of patients receiving palifermin experienced an oral-related AE, compared with 33% of patients receiving placebo (P = 0.13). Similarly, 56% of patients receiving palifermin during the second chemotherapy cycle had at least one oral-related AE, compared with Rosen 2006 38% of patients receiving placebo (P = 0.26). The overall incidences of skin- related AEs, reported as a palifermin-related AE in other clinical settings, were comparable between the two treatment groups (Table 3). During cycle 1, skin-related AEs were 56% in the placebo group versus 43% in patients receiving palifermin. During cycle 2, these incidences between the two groups were comparable (palifermin, 52%; placebo, 50%)"  There were no serious KGF-related AEs in either group and either cycle Spielberger "The incidence, frequency, and severity of adverse events were similar in the two 2004 groups, and most were attributable to the underlying cancer, cytotoxic

449 chemotherapy, or total-body irradiation. Those that occurred with an incidence that was at least 5 percentage points higher in the palifermin group than in the placebo group are listed in Table 3. Most of these adverse events were consistent with the pharmacologic action of palifermin on skin and oral epithelium (e.g., rash, pruritus, erythema, paresthesia, mouth and tongue disorders, and taste alteration). All these events were mild to moderate in severity, transient (occurring approximately three days after the third dose of palifermin and lasting approximately three days), and not a cause for the discontinuation of study drug. Serious adverse events considered to be related to treatment occurred in one palifermin recipient (rash) and one placebo recipient (hypotension)"  "Many patients who received palifermin sensed thickening of the oral mucosa and tongue" (first 2 blinded cycles: 72% vs 31%, P = 0.007)  "Treatment with palifermin was well tolerated. Table 3 shows the common Vadhan-Raj adverse effects that occurred during the first 2 blinded cycles, which included 2010 symptoms of thickness of oral mucosa, tongue, and lips (Figure 4); altered taste; flushing; warm sensation; and increased saliva. These adverse effects were mild to moderate and transient in nature. Similar sideeffects were observed during later cycles...but they did not worsen in severity"

Footnotes

AE = adverse event; CI = confidence interval; KGF = keratinocyte growth factor; NCI-CTC = National Cancer Institute common toxicity criteria; RR = risk ratio; WHO = World Health Organisation

Table 2: Granulocyte-macrophage colony-stimulating factor

Study ID Adverse events results Two participants (group allocation not reported) withdrew by day 3 due to intolerance to their mouthwash (dry mouth); one participant receiving GM-CSF (1 Cartee 1995 µg/ml) had mouthwash withdrawn by day 3 due to possible allergic reaction (sensation of fullness in the posterior pharyngeal area) but resolved within 4 hours (the participant was withdrawn but appears to have been included in the analysis) "One patient had fever and chills, and two patients had general malaise and Chi 1995 headache during GM-CSF treatment. No patient had evidence of fluid retention after GM-CSF" Dazzi 2003 Not reported (Sucralfate given to both groups) "Only 2 of the 20 patients treated with GM-CSF and sucralfate did not experience any side effects related to the drugs, but most Makkonen side effects were mild (WHO Grade 1 or 2). The most common side effects were 2000 local skin reactions, fever, bone pain, and mild nausea...In the control group only 1 patient complained of nausea possibly related to the use of sucralfate, and another patient interrupted sucralfate treatment because of the same reason" "12 patients who received GM-CSF had elevated white cell counts (WCC). The range of maximal WCC was 7.2–30.5 (median 19.7). All WCC had returned to normal within 3 weeks of completing injections (median 2 weeks). Three patients McAleese developed influenza-like symptoms with the GM-CSF and in one patient the 2006 injections were stopped because of this symptom. One patient developed an erythematous rash at his injection sites after completing his course of 14 injections (Figure 3). He had a past history of allergy to radiographic contrast medium" Nemunaitis  "The incidences of grades III or IV toxicities between rhGM-CSF or placebo

450 1995 occurring with a >10% frequency included anorexia (38% vs. 36%), nausea (26% vs. 29%), diarrhea (19% vs. 7%), stomatitis (19% vs. 14%) and hypertension (13% vs. 20%)"  "The following events were reported with higher frequency in the rhGM-CSF group compared with placebo: diarrhea (81% vs. 66%), bone pain (21% vs. 5%), abdominal pain (38% vs. 23%), vomiting (70% vs. 57%), pharyngitis (23% vs. 13%), pruritis (23% vs. 13%) and occular hemorrhage (11% vs. 0%)"  "Placebo-treated patients had higher occurrence of unspecified pain (36% vs. 17%), back pain (18% vs. 9%), peripheral edema (21% vs. 15%), hematuria (21% vs. 9%) and pneumonia (7% vs. 0%)" (Comparator was sucralfate) "Both mouthwashes were well tolerated, and none of Saarilahti the patients reported any adverse effects related to the mouthwashes. Adverse 2002 effects commonly associated with subcutaneous GM-CSF administration, such as nausea, vomiting, bone pain, headaches, and fever, were not observed" van der Not reported Lelie 2001

Footnotes

GM-CSF = granulocyte-macrophage colony-stimulating factor

Table 3: Granulocyte-colony stimulating factor

Study ID Adverse events results Cesaro "Both pegfilgrastim and filgrastim were well tolerated and no significant adverse 2013 effects were associated with their use" (G-CSF versus G-CSF) Approximately 20% of participants receiving G-CSF experienced mild to moderate skeletal pain which was resolved by using oral analgesics; 6% of participants in Crawford both groups reported mild generalised rash/itching; three participants experienced 1999 an event thought to be G-CSF-related and which caused them to request withdrawal from the study: abdominal pain, diffuse aches and pains, and a flare-up of pre- existing eczema Katano Not reported 1995 "There was no difference in the overall frequency of adverse clinical or laboratory Linch events between the groups or in the frequency of adverse events thought by the 1993 clinicians to be possibly or probably due to study medication" Schneider Not reported 1999 "In general, toxicities typical of postoperative RT to the head and neck were observed. Additional toxicities attributable to G-CSF and/or daily injections were as follows: elevated WBC requiring G-CSF dose reduction by prospectively planned Su 2006 guidelines occurred in nine patients in the GCSF arm; grade 2–3 bone pain was observed in two patients in the G-CSF arm; three patients refused injection (2 G- CSF, 1 placebo)"

Footnotes

G-CSF = granulocyte-colony stimulating factor

451

Table 4: Epidermal growth factor

Study Adverse events results ID "Adverse events were similar in both groups (Table 3). The most common adverse event in the rhEGF group was nausea (n = 7, 10.4%). The incidence of other adverse Kim events including oral pain, dry mouth, and taste alteration was low. All the adverse 2017 events were mild and transient. No grade 3 or 4 adverse events were noted during the study period." (there were no differences between groups in any adverse event) Wu "The frequency of minor and serious adverse events was similar in all groups. Most 2009 adverse events were related to primary disease status and treatment modalities"

Footnotes rhEGF = recombinant human epidermal growth factor

Table 5: Intestinal trefoil factor

Study ID Adverse events results "Only a minority of patients (six [6.1%] of 99 patients) reported mild to moderate treatment-emergent adverse events on the study. The symptoms included abdominal Peterson pain, diarrhea, oral pain, headache, and hypertension (Table 2). Of these, four were 2009 considered related to study drug: one (3%) was in the placebo group, two (6%) were in the low-dose rhITF group, and one (3%) was in the high-dose rhITF group. The events were isolated and resolved spontaneously without sequelae"

Footnotes rhITF = recombinant human intestinal trefoil factor

452 APPENDIX 10: Salivary gland dysfunction review - electronic search strategies

1) Cochrane Oral Health's Trials Register search strategy

#1 ((radioth* or radiat* or irradiat* or radiochemo*):ti,ab) AND (INREGISTER) #2 ((xerostomi* or "dry mouth" or "salivation disorder*" or saliva* or hypersalivat* or hyposalivat* or xeroses or radioxerost* or "salivary gland hypofunction" or "salivary gland dysfunction" or "dysfunction of the salivary gland*" or "artificial saliva" or "saliva artificial"):ti,ab) AND (INREGISTER) #3 (#1 and #2) AND (INREGISTER)

2) Cochrane Central Register of Controlled Trials (CENTRAL) search strategy

#1 Exp RADIOTHERAPY/ #2 (radioth* OR radiat* OR irradiat* OR radiochemo*) #3 #1 OR #2 #4 Exp XEROSTOMIA/ #5 Exp SALIVARY GLANDS #6 ((parotid or sublingual or submandibular or salivary) AND gland*) #7 ((#5 or #6) AND (hypofunction or dysfunction* or disorder* or function*)) #8 (xerostomi* OR xeroses OR radioxerost* OR (dry* NEAR mouth*)) #9 (hyposalivat* OR hypersalivat* OR sialogogue* or sialagogue*) #10 saliva* #11 (#4 or #7 or #8 or #9 or #10) #12 #3 AND #11

3) MEDLINE Ovid search strategy

1. exp RADIOTHERAPY/ 2. (radioth$ or radiat$ or irradiat$ or radiochemo$) 3. or/1-2 4. exp XEROSTOMIA/ 5. exp SALIVARY GLANDS/ 6. ((parotid or sublingual or submandibular or salivary) AND gland$) 7. ((5 or 6) AND (hypofunction or diysfunction* or disorder* or function*)) 8. (xerostomi* OR xeroses OR radioxerost$ OR (dry$ adj5 mouth$)) 9. (hyposalivat$ OR hypersalivat$ OR sialogogue$ or sialagogue$) 10. saliva$ 11. 4 or 7 or 8 or 9 or 10 12. 3 AND 11

The above subject search was linked to the Cochrane Highly Sensitive Search Strategy (CHSSS) for identifying randomised trials (RCTs) in MEDLINE: sensitivity maximising version (2008 revision) as referenced in Chapter 6.4.11.1 and detailed in Box 6.4.c of the Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 (updated March 2011).

1. randomized controlled trial.pt. 2. controlled clinical trial.pt. 3. randomized.ab. 4. placebo.ab. 5. drug therapy.fs. 6. randomly.ab. 7. trial.ab. 8. groups.ab. 9. or/1-8

453 10. exp animals/ not humans.sh. 11. 9 not 10

4) Embase Ovid search strategy

1. exp RADIOTHERAPY/ 2. (radioth$ or radiat$ or irradiat$ or radiochemo$) 3. or/1-2 4. XEROSTOMIA/ 5. SALIVARY GLAND/ 6. ((parotid or sublingual or submandibular or salivary) AND gland$) 7. ((5 or 6) AND (hypofunction or diysfunction* or disorder* or function*)) 8. (xerostomi* OR xeroses OR radioxerost$ OR (dry$ adj5 mouth$)) 9. (hyposalivat$ OR hypersalivat$ OR sialogogue$ or sialagogue$) 10. saliva$ 11. 4 or 7 or 8 or 9 or 10 12. 3 AND 11

The above subject search was linked to adapted version of the Cochrane Embase Project filter for identifying RCTs in Embase Ovid (see www.cochranelibrary.com/help/central-creation- details.html for information).

1. Randomized controlled trial/ 2. Controlled clinical study/ 3. Random$.ti,ab. 4. randomization/ 5. intermethod comparison/ 6. placebo.ti,ab. 7. (compare or compared or comparison).ti. 8. ((evaluated or evaluate or evaluating or assessed or assess) and (compare or compared or comparing or comparison)).ab. 9. (open adj label).ti,ab. 10. ((double or single or doubly or singly) adj (blind or blinded or blindly)).ti,ab. 11. double blind procedure/ 12. parallel group$1.ti,ab. 13. (crossover or cross over).ti,ab. 14. ((assign$ or match or matched or allocation) adj5 (alternate or group$1 or intervention$1 or patient$1 or subject$1 or participant$1)).ti,ab. 15. (assigned or allocated).ti,ab. 16. (controlled adj7 (study or design or trial)).ti,ab. 17. (volunteer or volunteers).ti,ab. 18. trial.ti. 19. or/1-18 20. (exp animal/ or animal.hw. or nonhuman/) not (exp human/ or human cell/ or (human or humans).ti.) 21. 19 not 20

5) CINAHL EBSCO (Cumulative Index to Nursing and Allied Health Literature) search strategy

S1 MH "Radiotherapy+" S2 (radioth* or radiat* or irradiat* or radiochemo*) S3 S1 or S2 S4 MH "Xerostomia+" S5 MH "Salivary Glands+" S6 ((parotid or sublingual or submandibular or salivary) AND gland*) S7 ((S5 or S6) AND (hypofunction or dysfunction* or disorder or function)) 454 S8 ((xerostomi* or xeroses or radioxerost*) or (dry N5 mouth)) S9 (hyposalivat* or hypersalivat* or sialogogue* or sialagogue*) S10 saliva* S11 S4 or S7 or S8 or S9 or S10 S12 S3 and S11

The above subject search was linked to Cochrane Oral Health's filter for identifying RCTs in CINAHL EBSCO.

S1 MH Random Assignment or MH Single-blind Studies or MH Double-blind Studies or MH Triple-blind Studies or MH Crossover design or MH Factorial Design S2 TI ("multicentre study" or "multicenter study" or "multi-centre study" or "multi-center study") or AB ("multicentre study" or "multicenter study" or "multi-centre study" or "multicenter study") or SU ("multicentre study" or "multicenter study" or "multi-centre study" or "multi-center study") S3 TI random* or AB random* S4 AB "latin square" or TI "latin square" S5 TI (crossover or cross-over) or AB (crossover or cross-over) or SU (crossover or cross-over) S6 MH Placebos S7 AB (singl* or doubl* or trebl* or tripl*) or TI (singl* or doubl* or trebl* or tripl*) S8 TI blind* or AB mask* or AB blind* or TI mask* S9 S7 and S8 S10 TI Placebo* or AB Placebo* or SU Placebo* S11 MH Clinical Trials S12 TI (Clinical AND Trial) or AB (Clinical AND Trial) or SU (Clinical AND Trial) S13 S1 or S2 or S3 or S4 or S5 or S6 or S9 or S10 or S11 or S12

6) LILACS BIREME Virtual Health Library (Latin American and Caribbean Health Science Information database) search strategy

(Mh Radiotherapy or radiotherap$ or radioterapia or irradiat$ or radiochemo$) [Words] and (Mh Xerostomia or xerostom$ or "salivary gland$" or salivat$ or hypersalivat$ or hyposalivat$ or sialogogue$)

The above subject search was linked to the Brazilian Cochrane Center filter for LILACs via BIREME.

((Pt randomized controlled trial OR Pt controlled clinical trial OR Mh randomized controlled trials OR Mh random allocation OR Mh double-blind method OR Mh single-blind method) AND NOT (Ct animal AND NOT (Ct human and Ct animal)) OR (Pt clinical trial OR Ex E05.318.760.535$ OR (Tw clin$ AND (Tw trial$ OR Tw ensa$ OR Tw estud$ OR Tw experim$ OR Tw investiga$)) OR ((Tw singl$ OR Tw simple$ OR Tw doubl$ OR Tw doble$ OR Tw duplo$ OR Tw trebl$ OR Tw trip$) AND (Tw blind$ OR Tw cego$ OR Tw ciego$ OR Tw mask$ OR Tw mascar$)) OR Mh placebos OR Tw placebo$ OR (Tw random$ OR Tw randon$ OR Tw casual$ OR Tw acaso$ OR Tw azar OR Tw aleator$) OR Mh research design) AND NOT (Ct animal AND NOT (Ct human and Ct animal)) OR (Ct comparative study OR Ex E05.337$ OR Mh follow-up studies OR Mh prospective studies OR Tw control$ OR Tw prospectiv$ OR Tw volunt$ OR Tw volunteer$) AND NOT (Ct animal AND NOT (Ct human and Ct animal)))and not (Ct ANIMAL AND NOT (Ct HUMAN and Ct ANIMAL)))

7) ZETOC Conference Proceedings search strategy radiotherap* AND xerostomi* radiotherap* AND saliva* radiotherap* AND sialogog*

455 8) OpenGrey search strategy

The search strategy for OpenSIGLE is below. radiotherapy AND xerostomia radiotherapy AND saliva radiotherapy AND sialogogue

9) US National Institutes of Health Ongoing Trials Register (ClinicalTrials.gov) search strategy

(radiotherapy and salivary) or (radiotherapy and xerostomia)

10) World Health Organization International Clinical Trials Registry Platform search strategy radiotherapy and saliva or radiotherapy and salivary or radiotherapy and xerostomia

456 APPENDIX 11: Salivary gland dysfunction review - characteristics of studies and risk of bias tables

Characteristics of included studies

Abacioglu 1997

Methods Location: Turkey Number of centres: 1 Date of enrolment: July 1996 to January 1997 Participants Inclusion criteria: aged between 18 to 70 years. Histopathologic diagnosis of SCC of head and neck (nasopharynx, larynx, oropharynx, hypopharynx, oral cavity). WHO performance status 0 to 2. Patients to receive primary or postoperative radiation treatment for a minimum of 46 Gy totally and treatment fields to include at least the tail of parotis (1/3), submandibular glands and part of sublingual and minor salivary glands Exclusion criteria: patients with a histopathologic diagnosis other than SCC. Patients with an autoimmune disorder (e.g. Sjögren Syndrome) or diseases effecting saliva secretion. Difficulty in co-operation for saliva collection, understanding the questionnaire and attending the follow-up visits Age (years): pilocarpine: median 55 years, range 38 to 68 years; control: median 50 years, range 30 to 61 years Gender (M:F): pilocarpine 12:0; control 11:1 Cancer type: tumour location: pilocarpine: larynx = 8, nasopharynx = 2 and oral cavity = 2; control: larynx = 7, nasopharynx = 4 and oral cavity = 1 Radiotherapy: pilocarpine: mean dose = 60.2 Gy (range 48 to 70 Gy), number of fractions = 30.1 (mean), treatment time = 44.9 days (mean); control: mean dose = 63.8 Gy (range 50 to 70 Gy), number of fractions = 31.9 (mean), treatment time = 48.2 days (mean) Chemotherapy: none Number randomised: 24 (12 per group) Number evaluated: 24 (no dropouts, although not all participants available at all time points) Interventions Pilocarpine versus no intervention Pilocarpine: 5 mg 3 times daily (4% solution) for 3 months from the beginning of RT Control: no treatment Outcomes  Xerostomia: subjective evaluation scores for xerostomia (0 = no symptoms, 11 = severe xerostomia). Questionnaire included 5 questions  Salivary flow rates: unstimulated and stimulated whole saliva secretion (unstimulated saliva pH measurements also recorded)  Adverse effects: no serious toxicity  Survival data: not reported  Other oral symptoms: not reported  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: before RT, during RT, end of RT and 3 months after start of RT Funding None

457 Trial Not registered nor published registration Sample size Not included calculation presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Not explicit in trial report. Comment from author: generation (selection bias) "randomisation was performed with block randomisation with stratification of treatment fields" Allocation concealment Low risk Comment from author: "sealed envelope were used for (selection bias) concealing" Blinding (performance bias High risk Pilocarpine versus no intervention. Blinding not and detection bias) possible patients/carers Blinding (performance bias High risk Not possible due to 'no intervention' group and and detection bias) subjective assessment of xerostomia outcome assessment Incomplete outcome data Low risk Comment from author: "no dropouts" (attrition bias) Number of participants available for assessment varies by time point, however, those missing for assessment unlikely to influence results Selective reporting Low risk Xerostomia and adverse events reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Antonadou 2002

Methods Location: Greece Number of centres: 1 Date of enrolment: January 1997 to January 1998 Participants Inclusion criteria: histologically proven squamous cell carcinoma of the head and neck. A primary tumour greater than or = T2N0M0, expected survival time greater than or = 12 months, no evidence of metastasis, and no prior chemotherapy or RT. Normal liver and kidney function, adequate bone marrow reserve, no current or previous history of cardiovascular disease and no active systemic infection Exclusion criteria: not reported Age (years): amifostine: mean 53.3 (SD 6.9); control: mean 60.3 (SD 5.5) Gender: amifostine: 13 M, 9 F; control: 16 M, 7 F Cancer type: tumour location: (amifostine/control) nasopharynx = 2/3, oral cavity = 9/11, larynx = 6/6 and oropharynx = 5/3. TNM classification: (amifostine/control) T2 = 6/6, T3 = 13/16, T4 = 3/1, N0 = 12/14 and N1 to 3 = 10/9 Radiotherapy: mean total dose = amifostine: 66.8 Gy (SD 3.2); control: 66.4 Gy (SD 3.4). Treatment duration: mean = amifostine: 49.6 (SD 4.5) days; control: 55.9 (SD 8.9) Chemotherapy: carboplatin (90 mg/m²), once a week before RT in both groups Number randomised: 50 (amifostine 25, control 25)

458 Number evaluated: 45 (amifostine 22, control 23) Interventions Amifostine versus no intervention Amifostine (300 mg/m2), IV 30 minutes before RT on days 1 to 5 of each week. Antimetic treatment administered IV before the amifostine Control: nothing Outcomes  Xerostomia: incidence of late xerostomia (RTOG grade 2 or more - measured on a 0 to 4 scale)  Salivary flow rates: not reported  Adverse effects: haematologic toxicity, nausea, vomiting and transient hypotension  Survival data: progression-free survival at 18 months  Other oral symptoms: incidence of grade 3 or greater acute mucositis and dysphagia  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: xerostomia: 3, 6, 9, 12 and 18 months after RT; haematologic toxicity and acute non-haematological toxicities (mucositis and dysphagia): weekly for 7 weeks during RT, then 1, 2 and 3 months after RT Funding Not reported Trial Not registered registration Sample size Yes calculation presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear Quote: "randomized (1:1)" generation (selection bias) risk Allocation concealment Unclear Insufficient information (selection bias) risk Blinding (performance bias High risk Amifostine versus no intervention. Blinding not and detection bias) possible patients/carers Blinding (performance bias High risk Not possible due to 'no intervention' group and and detection bias) subjective assessment of xerostomia outcome assessment Incomplete outcome data Low risk 45/50 participants evaluated (equal dropouts between (attrition bias) groups). 3 participants dropped out of the amifostine arm. 2 denied further treatment (1 = week 2 and 1 = week 4) and 1 was lost to follow-up. In the control arm, 1 participant died and 1 received palliative treatment because of disease progression Selective reporting Low risk Xerostomia and adverse events reported (reporting bias) Other bias Low risk No other sources of bias are apparent

459 Bardet 2011

Methods Location: France Number of centres: 27 Date of enrolment: March 2001 to January 2006 Participants Inclusion criteria: newly diagnosed head and neck, eligible for radiotherapy. Over 75% of both parotid glands in field. Performance status ≤ 2, no distant metastases, neutrophils ≥ 2000/uL, platelets ≥ 1000,000/uL, creatine < 130 umol/L, aminotransferases ≤ 3 x the upper limit of normal, and ≥ 18 years Exclusion criteria: use of pilocarpine during RT and concomitant CT, second- line treatment, incomplete assessment of salivary gland function Age: intravenous: mean 55.2 range 34 to 78; subcutaneous: mean 56.1 range 36 to 76 Gender: intravenous: 127 M, 16F; subcutaneous: 124 M, 24 F Cancer type: newly diagnosed squamous cell carcinoma of the head and neck, at all stages, and nodal status Radiotherapy: at least 40 Gy of radiation delivered postoperatively Chemotherapy: induction chemotherapy in 42 patients no concurrent chemotherapy Number randomised: 291 (intravenous 143, subcutaneous 148) Number evaluated: 127 (intravenous 67, subcutaneous 60) for xerostomia at 1 year Interventions Intravenous versus subcutaneous amifostine Intravenous: 200 mg/m² daily, administered over 3 minutes, 15 to 30 minutes before RT Subcutaneous: 500 mg at 2 sites, 20 to 60 minutes before RT Outcomes  Xerostomia: grade 2 or above (0 to 4 scale). Physician graded via RTOG before treatment, every 3 months for the 1st year and then every 6 months  Salivary flow rates: unstimulated and stimulated saliva (mg/min)  Adverse effects: nausea, vomiting, hypotension, skin rash, local pain at injection site, fever, asthenia  Survival data: locoregional control, overall survival  Other oral symptoms: dysgeusia (taste disturbance), dysphagia (difficulty in swallowing), dysphonia (difficulty in speaking) - these 3 items were combined with the patients' sensation of mouth dryness and assessed using a patient benefit questionnaire (see QoL); grade 3+ acute mucositis  Other oral signs: not reported  Quality of life: patient benefit questionnaire  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: acute xerostomia measured at 3 months; xerostomia, salivary flow rates and patient benefit questionnaire reported at 6 months, 1, 2 and 3 years; survival reported up to 4 years Funding Externally funded by pharmaceutical company; Schering-Plough, France Trial clinicaltrials.gov/show/NCT00158691ID - 12 registration Sample size Yes calculation presented

Risk of bias table

Bias Authors' Support for judgement

460 judgement Random sequence Unclear Authors claim "randomly assigned". No further details generation (selection bias) risk given Allocation concealment Unclear No details given (selection bias) risk Blinding (performance bias High risk "Lack of double-blind". Patients could not really be and detection bias) considered to be blinded as administration of patients/carers amifostine differed Blinding (performance bias High risk Patient-reported outcome see above and detection bias) outcome assessment Incomplete outcome data High risk Large loss to follow-up. Attrition likely to be related to (attrition bias) outcome Selective reporting Low risk Xerostomia and adverse events reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Brizel 2000

Methods Location: Europe, Canada, USA Number of centres: 35 to 40 (unclear) Date of enrolment: October 1995 to October 1997 33% dropout rate at 12 months Participants Inclusion criteria: patients with newly diagnosed, previously untreated squamous cell head and neck cancer. Inclusion of ≥ 75% of both parotid glands within radiation field and ≥ 40 Gy. Karnofsky Performance Status ≥ 60, granulocyte ≥ 2000 microL and platelet count ≥ 100,000 microL Exclusion criteria: patients with T1N0 or T2N0 carcinomas of the true vocal cords and tumours of the major or minor salivary glands or history of malignancy other than in situ cervix carcinoma within 5 years preceding diagnosis. Pregnant women Age: amifostine: 36 to 76, median = 55 years; control: 28 to 78, median = 56 years Gender: amifostine 123 M, 27 F; control 120 M, 33 F Cancer type: head and neck, various tumour sites, stages and node stages Radiotherapy: amifostine: definitive = 50, postoperative high risk = 70 and postoperative low risk = 28; control: definitive = 52, postoperative high risk = 65 and postoperative low risk = 36. 1.8 to 2.0 Gy , 5 days a week over 5 to 7 weeks for a total dose of 50 to 70 Gy Chemotherapy: none Number randomised: 315 randomised, but 12 never received any treatment (amifostine 150, control 153) Number evaluated: xerostomia at 12 months: 203 (amifostine 97, control 106), all included in analysis for locoregional control, all who received at least 1 dose of amifostine were assessed for toxicity Interventions Amifostine versus no intervention Amifostine: (200 mg/m²) 3 minute intravenous 15-30 minutes before RT Control: nothing Outcomes  Xerostomia: incidence of grade 2+ acute (within 90 days of the start of RT) and chronic xerostomia (0 to 4 scale)  Salivary flow rates: unstimulated and stimulated saliva production - reported as median quantity (g) of saliva and also as number of participants producing > 0.1 g in 5 min ("a clinically relevant volume")

461  Adverse effects: nausea, vomiting, hypotension, allergic response  Survival data: locoregional control, progression-free survival and overall survival at 24 months  Other oral symptoms: oral discomfort, dysgeusia (taste disturbance), dysphagia (difficulty in swallowing), dysphonia (difficulty in speaking) - all included in patient benefit questionnaire (see QoL); grade 3+ acute mucositis  Other oral signs: not reported  Quality of life: patient benefit questionnaire (8 items each on a 10-point scale where higher = better QoL)  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: xerostomia: within 3 months of start of RT, then at 12, 18 and 24 months; salivary flow rates: 12, 18 and 24 months after start of RT; quality of life: 12 months after start of RT Funding Source of funding: Medimmune Oncology Trial Not registered registration Sample size calculation Yes presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "dynamic allocation process" (recognised generation (selection bias) methods referenced) Allocation concealment Low risk Quote: "determined by a phone call from the enrolling (selection bias) institution to the protocol sponsor (US Bioscience)"

Comment: it appears to be central/remote allocation Blinding (performance bias High risk Amifostine versus no intervention and detection bias) patients/carers Blinding (performance bias High risk Open-label, no blinded outcome assessment and detection bias) outcome assessment Incomplete outcome data High risk 315 enrolled and randomised; 12 never received any (attrition bias) treatment or follow-up. Overall attrition 36% Selective reporting Low risk Xerostomia and adverse events reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Brizel 2008

Methods Location: Australia, Canada, USA Number of centres: 22 Date of conduct: September 1999 to May 2001 Participants Inclusion criteria: adults with newly diagnosed head and neck cancer. Patients with unknown primary and extensive neck disease also eligible. Karnofsky

462 Performance Status > 60, haemoglobin > 10 g/dL, plus other similar criteria Exclusion criteria: prior head and neck radiation therapy, prior surgery for primary tumour beyond biopsy, prior chemotherapy, known allergy to Escherichia coli-derived products, participation in another study within the previous 30 days, refusal to use adequate contraception during study, pregnant or breastfeeding Age: palifermin: mean 54 (range 25-80); placebo: mean 56 (range 42-75) Gender: palifermin 55 M, 12 F; placebo 27 M 5 F Cancer type: primary locations: oral cavity, oropharynx/nasopharynx, hypopharynx/larynx Radiotherapy: isocentric 4 to 6 MV photons either standard fractionation (once daily 2 Gy fractions 5 days/week: total primary tumour dose 70 Gy) or hyperfractionation (single 2 Gy fraction followed by a planned 1-week treatment break. Then twice-daily radiation: total dose of 72 Gy/6.5 weeks). Varied by centre Chemotherapy: cisplatin 20 mg/m² per day as IV bolus injection and fluorouracil 1000 mg/m² per day as continuous infusion, both on 1st 4 days of 1st and 5th weeks of RT Number randomised: 101 (69 palifermin, 32 placebo) Number evaluated: varies by outcome but 97 (65 palifermin, 32 placebo) analysed for our primary outcome of xerostomia Interventions Palifermin versus placebo Palifermin: 60 µg/kg by IV bolus injection on study day 1 (Friday) before 1st week of CRT. Subsequent doses administered for 7 consecutive weeks, on each Friday after completion of weekly radiation treatment. 2 additional doses given on weeks 8 and 9 Placebo: as above Follow-up: 5 weeks after end of RT Outcomes  Xerostomia: incidence of grade 2 xerostomia using NCI CTC scale  Salivary flow rates: not reported  Adverse effects: nausea, vomiting, fever, constipation, dehydration, granulocytopenia, fatigue, diarrhoea, insomnia, anaemia, dysaphia, cough, headache, weight decrease, dizziness, anxiety, hypomagnesaemia  Survival data: survival, progression-free survival (up to 75 months)  Other oral symptoms: mucositis (primary outcome of study), dysphagia  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Funding Pharmaceutical trial (Amgen) Trial Not registered registration Sample size Yes calculation presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear risk Quote: "double-blind, randomized, placebo- generation (selection bias) controlled study"

463 Allocation concealment Unclear risk Not stated (selection bias) Blinding (performance bias Low risk Quote: "double-blind randomised placebo-controlled and detection bias) study" patients/carers Blinding (performance bias Low risk Quote: "double-blind randomised placebo-controlled and detection bias) study". However there is a subjective element to the outcome assessment index Incomplete outcome data Low risk Xerostomia data on 97 out of 101 enrolled. Quote: (attrition bias) "3 patients in palifermin group and 1 in the placebo group discontinued study treatment with adverse events not considered related to study treatment". Comment: unclear if 3 of these were missing for xerostomia Selective reporting Low risk Xerostomia and adverse events reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Buentzel 2006

Methods Location: Europe, USA Number of centres: 18 (15 Europe, 3 USA) Date of recruitment: October 1996 to October 1998 Participants Inclusion criteria: at least 18 years of age scheduled for definitive or adjuvant chemoradiotherapy for histologically confirmed squamous cell carcinoma of the head and neck. Postsurgery the surgical wound must be healed but no later than 12 weeks after surgery. Inclusion of at least 75% of each parotid gland within radiation field. Life expectancy 12+ months, Karnofsky Performance Status 60+, adequate function of bone marrow, kidneys and the liver Exclusion criteria: evidence of distant metastatic disease, primary lesion of the parotid gland, or a history of prior malignancy within the past 5 years (other than non-melanomatous skin cancers that are controlled or carcinoma in situ of the cervix). Scheduled to receive hyperfractionated or accelerated radiotherapy, previously treated with chemotherapy or other investigational therapies within 4 weeks of study entry. Pregnant women Age: amifostine: median 57 (range 29-73); placebo: median 58 (range 23-78) Gender: amifostine 54 M, 13 F; placebo 57 M, 8 F Cancer type: head and neck cancer, various primary sites and stages Radiotherapy: standard fractionation (1.8-2.0 Gy per day, 5 days a week) over 6 to 7 weeks for a total dose of 60 to 70 Gy Chemotherapy: carboplatin 70 mg/m² IV over 30 minutes after amifostine and 30 minutes before RT Number enrolled: 132 Number randomised: 132 Number evaluated: 132 (ITT analysis) (67 amifostine; 65 placebo) Interventions Amifostine versus placebo Amifostine: 300 mg/m² IV over 3 minutes (days 1-5 and 21-25 of treatment); 200 mg/m² IV over 3 minutes (days 6-20 and 26-30/35) Placebo (Mannitol): equivalent volume to amifostine Outcomes  Xerostomia: RTOG acute and late radiation morbidity scoring criteria; incidence of grade 2 or higher acute or late xerostomia (0 to 4 scale)  Salivary flow rates: stimulated and unstimulated saliva measurements (not assessed as less than a 3rd of participants had salivary function at 1 year)

464  Adverse effects: nausea, vomiting, allergic response, asthenia  Survival data: locoregional failure rate, progression-free survival and overall survival  Other oral symptoms: RTOG acute and late radiation morbidity scoring criteria: grade 3 or higher acute mucositis  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: acute xerostomia and mucositis measured up to 90 days after start of RT; late xerostomia measured up to 12 months after start of RT Funding Source of funding: MedImmune Oncology Inc grant Trial Not registered registration Sample size Yes calculation presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Assumed. Described as "dynamic allocation generation (selection bias) scheme", similar method to Brizel 2000 Allocation concealment Low risk Fax of baseline data sent to central telephone number (selection bias) for randomisation number. Randomisation number identical to blinded drug container held at pharmacy Blinding (performance bias Low risk Amifostine versus placebo and detection bias) patients/carers Blinding (performance bias Low risk Information provided by author: "blinded drug and detection bias) containers were kept at the pharmacy, treating outcome assessment physicians had no information about the randomization until the end of the follow-up period" Incomplete outcome data Low risk Analyses carried out an ITT basis. Dropouts = 30 (attrition bias) (23% dropout rate). Amifostine group = 21 (1 - never treated, 16 - toxicity, 1 - patient request, 1 - death, 2 - other illness). Placebo group = 10 (1 - never treated, 4 - toxicity, 1 - patient request, 1 - death, 1 - disease progression, 2 - non-compliance) Selective reporting Low risk Xerostomia and adverse events reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Burlage 2008

Methods Location: the Netherlands Number of centres: 2 Date of enrolment: April 1999 - October 2003 Participants Inclusion criteria: biopsy confirmed HNSCC, initial 5% (wt/vol) citric acid-

465 stimulated parotid salivary flow > 0.1 mL/min Exclusion criteria: previous irradiation and/or previous or concurrent chemotherapy, patients with salivary gland tumours, severe cardiovascular disease or chronic obstructive pulmonary disease, pregnant women Age: pilocarpine: 18-60 years 50 participants, > 60 years 35 participants; placebo: 18-60 years 42 participants, > 60 years 42 participants Gender (M:F): pilocarpine 22:63; placebo 13:71 Cancer type: oral cavity (17%), oropharynx (18%), larynx (51%), hypopharynx (7%), nasopharynx (4%), unknown primary (1%) (equally distributed across groups) Submandibular gland removal: both removed: pilocarpine 2%, placebo 5%; 1 removed: pilocarpine 37%, placebo 38% Radiotherapy: clinical target volume of initial field encompassed the primary tumour site with 1.5 cm margin, neck node levels in which pathologic nodes were found and elective node areas on both sides. Conventional fractionation schedule. Received at least 40 Gy in daily 2 Gy fractions Chemotherapy: none Number randomised: 170 (85 per group) Number evaluated: 113 (pilocarpine 55, placebo 58) Interventions Pilocarpine versus placebo Pilocarpine: 5 mg 4 times daily 2 days before start of RT until 14 days after RT Placebo: similar tablets, same schedule Outcomes  Xerostomia: from validated head-and-neck symptom questionnaire on 5- point scale; LENT SOMA  Salivary flow rates: parotid salivary flow using Carlson-Crittenden cups, from left and right hand parotid glands simultaneously under standardised conditions for 10 min. Flow stimulated with 5% (wt/vol) citric acid. Parotid flow complication probability also reported  Adverse effects: not reported  Survival data: locoregional control  Other oral symptoms: eating, swallowing  Other oral signs: not reported  Quality of life: some covered in validated head-and-neck symptom questionnaire on 5-point scale  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: before RT, 6 weeks, 6 months and 12 months postRT Funding Not reported. Conflicts of interest: "none" reported Trial Not registered registration Sample size Yes calculation presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "Randomisation was executed by the hospital generation (selection bias) pharmacist by computer, using random permuted blocks within strata. The randomisation key was opened after the last saliva collection (1 year after the last patient was included and after completion of

466 all planned assessments)" Allocation concealment Low risk See above (selection bias) Blinding (performance bias Low risk Quote: "Double-blind randomised placebo-controlled and detection bias) study". Intervention was tablets supplied by the patients/carers pharmacy Blinding (performance bias Low risk Quote: "Double-blind randomised placebo-controlled and detection bias) study". Intervention was tablets supplied by the outcome assessment pharmacy Incomplete outcome data High risk 32% missing at 12 months with no clear reasons (attrition bias) given by study group Selective reporting High risk Adverse events and xerostomia data not fully (reporting bias) reported Other bias Low risk No other sources of bias are apparent

Büntzel 1998

Methods Location: Germany Number of centres: 1 Date of recruitment: not stated Participants Inclusion criteria: stage III or IV carcinoma of the head and neck, aged 16 to 80 and no evidence of systemic infection or liver or renal impairment. Tumour resected or excised before adjuvant RT Exclusion criteria: not reported Age: amifostine: median 61 (range 40-77); control: median 58 (range 38-75) Gender: amifostine 13 M, 1 F; control 12 M, 2 F Cancer type: tumour location (amifostine/control): larynx = 3/1, hypopharynx = 4/3, mesopharynx = 3/7, nose = 2/1, mouth = 2/2 Radiotherapy: 2 Gy fractions, 5 days a week for 6 weeks; maximum dose of 60 Gy (encompassing 75% of the major salivary glands) Chemotherapy: 20 min IV infusion of carboplatin (70 mg/m² days 1 to 5 and 21 to 25 of treatment) Number randomised: 28 (14 amifostine, 14 control) Number evaluated: 28 Interventions Amifostine versus no intervention Amifostine: (500 mg) 15 min IV before carboplatin (days 1 to 5 and days 21 to 25). Followed by antiemetic regimen to control nausea/vomiting Control: nothing Use of supportive drugs reported: (amifostine/control): G-CSF: 2/7; GM-CSF: 0/7; antibiotics: 4/10 Outcomes  Xerostomia: incidence and severity using WHO grading (0 to 4 scale - we report grade 2 and above)  Salivary flow rates: not reported  Adverse effects: hypotension  Survival data: not reported  Other oral symptoms: dysgeusia (taste disturbance), dysphagia (difficulty in swallowing), mucositis (WHO)  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: economic evaluation

Timing of assessment: xerostomia at end of RT and 1 year; other oral

467 symptoms at end of RT Funding US Bioscience who produce Ethyol-amifostine Trial Not registered registration Sample size Not reported calculation presented Notes Additional data presented but included extra 11 patients in amifostine group who were not entered in the study (not included in analyses)

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear risk Insufficient information generation (selection bias) Allocation concealment Unclear risk Insufficient information (selection bias) Blinding (performance bias High risk Amifostine versus no intervention and detection bias) patients/carers Blinding (performance bias High risk Not blinded and xerostomia is a subjective and detection bias) outcome outcome assessment Incomplete outcome data Low risk All participants included in analysis (attrition bias) Selective reporting Low risk Xerostomia and adverse events reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Büntzel 2010

Methods Location: Germany Number of centres: 6 Date of study: 2001 to 2007 Participants Inclusion criteria: SCCHN with deficiency in selenium and if radiation field included 75% of the major salivary glands Exclusion criteria: none reported Age: median 63.2 range 38.7-83.0 Gender (M:F): selenium 16:6; control 15:2 Cancer type: head and neck cancer Radiotherapy: 1.8 to 2.0 Gy to primary tumour and lymphatic neck during daily radiation treatment; to total dose 60-72 Gy Chemotherapy: unclear Number randomised: 40: 22 selenium, 18 control Number evaluated: 39: 22 selenium, 17 control Interventions Selenium versus no intervention Selenium: 500 µg sodium selenite, 2 days before RT, 500 µg selenite and radiation days (300 µg if official holiday). Administrated as oral fluid 1 hour before RT Control: no intervention

468 Outcomes  Xerostomia: RTOG grade for xerostomia  Salivary flow rates: not reported  Adverse effects: serious adverse events reported  Survival data: not reported  Other oral symptoms: mucositis RTOG, dysgeusia (taste disturbance RTOG), dysphagia (difficulty in swallowing RTOG)  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: 1, 2, 3, 4, 5, 6, 7 weeks from start of RT and 6 weeks after RT Funding Externally funded by Arzneimittel, Germany Trial Unclear registration Sample size Not reported calculation presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear risk Method not given, stated "randomised" generation (selection bias) Allocation concealment Unclear risk Allocation after consent obtained (selection bias) Blinding (performance bias High risk Participants receive selenium oral fluid prior to and detection bias) radiotherapy or not. Blinding not possible patients/carers Blinding (performance bias High risk Participants receive selenium oral fluid prior to and detection bias) radiotherapy or not, and their subjective outcome assessment assessment of xerostomia is included Incomplete outcome data Low risk Initial study requiring 60 patients per arm stopped (attrition bias) early due to slow accrual

113 screened. 93 selenium deficient. 40 consented. 1 withdrawal, 39 reported

Selenium concentrations reported in other article elsewhere Selective reporting High risk Xerostomia but no standard deviations. Total (reporting bias) adverse events reported but not per person Other bias Low risk No other sources of bias are apparent

Duncan 2005

Methods Location: Canada Number of centres: multicentre (unclear how many) Date of enrolment: September 1997 to September 1999

469 Participants Inclusion criteria: SCCHN, non-metastatic disease Exclusion criteria: none reported Age: lozenge median 59.7; placebo median 57.3 Gender (M:F): lozenge (48:18), placebo (52:15) Cancer type: oral cavity, oropharynx, hypopharynx, nasopharynx, larynx Radiotherapy: conventional radical or postoperative radiotherapy to a dose of 50 Gy or greater delivered in once daily fractions (1.8 to 2.4 Gy) Chemotherapy: not mentioned, probably none Number randomised: 138 (69 per group) Number evaluated: 133 (lozenge 66; placebo 67) Interventions Antimicrobial lozenge versus placebo Antimicrobial lozenge: BCoC, bacitracin, 6 mg; clotrimazole, 10 mg; gentamicin, 4 mg. Unclear how frequently taken or for how long Placebo: not described - assumed similar Outcomes  Xerostomia: item on trial specific checklist - 'Did you have mouth dryness (1-4 scale)?' and NCIC CTG ECTC physician-rated (using patient diary)  Salivary flow rates: not reported  Adverse effects: not reported  Survival data: not reported  Other oral symptoms: mucositis using OMAS (primary outcome), mouth pain, chewing, numbness, mouth opening, burning mouth  Other oral signs: not reported  Quality of life: 2 tools - European Organisation for Research and Treatment of Cancer Quality of Life questionnaire (EORT QLQ-C30), trial specific checklist  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: 2, 4, 6 during RT; 8-9, 12-14, 24 weeks on study Funding The National Cancer Institute of Canada, Clinical Trials Group Trial Unclear registration Sample size No calculation presented Notes Primarily study to prevent mucositis

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear risk Quote: "randomly assigned" generation (selection bias) Allocation concealment Unclear risk Not reported (selection bias) Blinding (performance bias Low risk Quote: "double-blind controlled trial". Placebo and detection bias) tablets given patients/carers Blinding (performance bias Low risk Quote: "double-blind controlled trial". Placebo and detection bias) tablets given outcome assessment Incomplete outcome data Low risk Compliance with quality of life forms reported to

470 (attrition bias) be 93.3% but reasons for dropouts not reported. Similar low rates of attrition per group Selective reporting High risk Xerostomia reported, adverse events not reported (reporting bias) for lozenge Other bias Low risk No other sources of bias are apparent

Fisher 2003

Methods Location: USA Number of centres: unclear Date of randomisation: March 1998 to February 2000 Participants Inclusion criteria: oral and oropharyngeal squamous cell carcinoma, Karnofsky Performance Score ≥ 60, no prior radiotherapy to the head and neck, planned irradiation of the oral cavity or oropharynx in which at least 50% of the major salivary glands are to receive > 50 Gy Exclusion criteria: salivary gland malignancy; use of cholinergic, anticholinergic, and tricyclic drugs; and patients with uncontrolled asthma, acute iritis, or narrow-angle glaucoma Age: pilocarpine 60.8 years; placebo 59.3 years Gender (M:F): pilocarpine 93:28, placebo 92:32 Cancer type: oral cavity 52; nasopharynx 3; oropharynx 104; hypopharynx 11; other 13; unknown 18 (evenly distributed across groups) Radiotherapy: 60-70 Gy with 50% of volume of major salivary glands receiving 50 Gy Chemotherapy: not stated Number randomised: 249; 3 ineligible, all from pilocarpine arm (121 pilocarpine, 125 placebo) Number evaluated: 166 end of RT (pilocarpine 89, placebo 77); 166 at 3 months (pilocarpine 85, placebo 81); 137 at 6 months (pilocarpine 68, placebo 69) Interventions Pilocarpine versus placebo Pilocarpine: 5 mg tablets 4 times daily starting 3 days before RT and continuing for 3 months Placebo: 5 mg tablets 4 times daily starting 3 days before RT and continuing for 3 months. 3 months after RT the placebo group were permitted to cross over to pilocarpine Outcomes  Xerostomia: not reported  Salivary flow rates: salivary gland scintigraphy (stimulated and unstimulated)  Adverse effects: drug toxicities reported  Survival data: not reported  Other oral symptoms: RTOG acute mucositis, mouth pain, dysgeusia (taste disturbance), dysmasesia (difficulty in chewing), dysphagia (difficulty in swallowing), dysphonia (difficulty in speaking)  Other oral signs: not reported  Quality of life: University of Washington QoL scale  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: pretreatment, end of RT, 3 months after end of RT, 6 months after end of RT Funding National Cancer Institute and MGI Pharma Inc.

471 Trial clinicaltrials.gov/show/NCT00003139ID - 11 registration Protocol available Sample size Not reported calculation presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear risk Insufficient information generation (selection bias) Allocation concealment Unclear risk Insufficient information (selection bias) Blinding (performance bias Low risk Pilocarpine versus placebo and detection bias) patients/carers Blinding (performance bias Low risk Pilocarpine versus placebo and detection bias) outcome assessment Incomplete outcome data Unclear risk 213/249 available for analysis. Dropouts very high (attrition bias) for salivary flow (33% end of RT and 3 months, 45% at 6 months) Selective reporting High risk Xerostomia not reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Gornitsky 2004

Methods Location: Canada Number of centres: 1 Date of randomisation: March 1998 to September 2001 Participants Inclusion criteria: scheduled to receive external beam radiotherapy, using a bilateral radiation technique encompassing ≥ 2/3 of all major and minor salivary glands for a minimum of 5000 cGy (200 cGy per day) for 5-7 weeks Exclusion criteria: clinically significant cardiovascular disease, chronic obstructive pulmonary disease, biliary tract disease, uncontrolled asthma, acute iritis, narrow angle glaucoma, participants who are pregnant or nursing. Hypersensitivity to pilocarpine, participants on tricyclic antidepressants, antihistamines with anticholinergic effects, beta blockers, or pilocarpine for ophthalmic indications were excluded Age (mean): pilocarpine 58 years; placebo 61 years Gender (M:F): pilocarpine 26:3, placebo 24:5 Cancer type: oral cavity 14; pharynx 13; tonsil 11; glottis 3; larynx 11; sinus 2; neck 1; unknown 1 (evenly distributed across groups) Radiotherapy: Mean dose = 64.7 Gy (pilocarpine group), 63.7 Gy (placebo group) Chemotherapy: pilocarpine 13 (45%); placebo 9 (32%) Number randomised: 58 Number evaluated: 58 (22 dropped out but ITT was used and missing data were calculated)

472 Interventions Pilocarpine versus placebo Phase 1 Pilocarpine: 5 mg tablets 5 times daily, half an hour before meals, before radiotherapy, and prior to sleep during the period of radiotherapy Placebo: identical tablets 5 times daily, half an hour before meals, before radiotherapy, and prior to sleep during the period of radiotherapy Phase 2 All received pilocarpine (5 mg) 4 times daily half an hour before meals and prior to sleep for 5 weeks Outcomes  Xerostomia: subjective assessment of xerostomia: VAS (rated 0-100)  Salivary flow rates: whole saliva secretion (unstimulated and stimulated) using the SAXON test  Adverse effects: not reported (data provided by author)  Survival data: not reported  Other oral symptoms: oral discomfort, difficulty with eating, dysphonia (difficulty in speaking), mucosal pain or burning (VAS, rated 0-100)  Other oral signs: not reported  Quality of life: global quality of life, sleeping problems (VAS, rated 0-100)  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: prior to RT, end of RT, 5 weeks after end of RT Funding Pharmacia Canada Trial Not registered registration Sample size Not reported calculation presented Notes Phase 2 data not included in the review

Additional data provided by author

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Comment from author: "block of four using a generation (selection bias) random number table... allocation sequence prepared by pharmacy of Jewish General Hospital" Allocation concealment Low risk Third party randomisation; coded bottles (selection bias) Blinding (performance bias Low risk Bottles only distinguished by number allocated by and detection bias) pharmacy. Investigators, treating physicians and patients/carers patients blinded Blinding (performance bias Low risk Subjective outcomes self reported (patients and detection bias) unaware of treatment group) outcome assessment Incomplete outcome data Low risk Analysis carried out on ITT basis. 38% dropout (attrition bias) rate. 58 randomised, 22 dropped out (9 pilocarpine, 13 placebo) Selective reporting High risk Adverse events not reported (reporting bias)

473 Other bias Low risk No other sources of bias are apparent

Grötz 2001

Methods Location: Germany Number of centres: 1 Date of randomisation: not stated Participants Inclusion criteria: scheduled to receive adjuvant or sole radiotherapy for head and neck cancer to a scheduled dose of 60 Gy. Cranial border of the field above the chin-mastoid line so salivary glands are located in the core irradiation field Exclusion criteria: salivary gland disorders Age: mean age = 55 years Gender: 22 M, 1 F Cancer type: head and neck Radiotherapy: total dose = 60 Gy Chemotherapy: unclear Number randomised: 48 Number evaluated: 23 Interventions Coumarin + troxerutin versus placebo Venalot Depot (coumarin 15 mg and troxerutin 90 mg) tablet: 2 tablets 3 times daily. Start 1 week before RT and 4 weeks after end of RT Control: placebo Outcomes  Xerostomia: not reported, only as part of total RTOG  Salivary flow rates: stimulated and unstimulated using sialoscintigraphy (sialometry abandoned as primary marker as not successfully collected). Acute radiation side effects RTOG score but for all organs  Adverse effects: reddened skin, nausea  Survival data: locoregional control  Other oral symptoms: not reported  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: 4 weeks after RT Funding Not stated Trial Unclear registration Sample size Not reported calculation presented Notes Unable to use data

Risk of bias table

474 Blinding (performance bias Low risk Venalot Depot versus placebo and detection bias) patients/carers Blinding (performance bias Low risk Salivary flow rates objective outcome and detection bias) outcome assessment Incomplete outcome data High risk 48 randomised, 25 dropped out. Dropouts per (attrition bias) group not specified Selective reporting High risk Xerostomia not reported. Data for total RTOG (reporting bias) score presented but no break down by condition or organ Other bias Low risk No other sources of bias are apparent

Haddad 2002

Methods Location: Iran Number of centres: 1 Date of recruitment: 1998-2000 Participants Inclusion criteria: 18-70 year old patients, irradiated to the head and neck, both parotid glands in the radiation fields (minimum 40 Gy). No previous history of irradiation in this region Exclusion criteria: asthma, chronic obstructive pulmonary disease, narrow- angle glaucoma, biliary or renal lithiasis and hypertensive, heart or psychiatric disorders requiring medical treatment Age: mean across groups = 43 years (range 18 to 70 years) Gender (M:F): across groups 36:24 Cancer type: primary site of tumour. Pilocarpine group: nasopharynx (n = 17), neck adenopathy (n = 1). Placebo group: maxilla (n = 2), nasopharynx (n = 13), tongue (n = 1), tonsil (n = 5) Radiotherapy: standard fractionation (1.8 to 2 Gy per day, 5 days a week) and cobalt-60 systems; mean parotid dose 58 Gy (pilocarpine 59 Gy; placebo 57 Gy) (range 45 to 70 Gy) Chemotherapy: none Number randomised: 60 Number evaluated: 39 (18 pilocarpine, 21 placebo) Interventions Pilocarpine versus placebo Pilocarpine hydrochloride: 5 mg 3 times daily for 3 months starting from the beginning of RT Placebo: 5 mg 3 times daily for 3 months starting from the beginning of RT Outcomes  Xerostomia: subjective evaluation score for xerostomia using 6 questions evaluated using VAS (0-100 mm). Objective grading of xerostomia according to the Late Effects of Normal Tissues Subjective, Objective, Management and Analytic (LENT SOMA) scale  Salivary flow rates: not reported  Adverse effects: lacrimation (excess tears, crying), nausea  Survival data: overall survival  Other oral symptoms: not reported  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: 6 months postRT

475 Funding Source of funding: Tehran University of Medical Sciences' research grant Trial Not registered registration Sample size calculation Not reported presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "...randomisation was performed at the start generation (selection bias) of radiotherapy by the sealed envelope method" Comment: although not clear the randomisation was probably done well as the pharmacy was involved in making and distributing the tables Allocation concealment Low risk Sealed envelopes, pharmacy involvement (selection bias) Blinding (performance bias Low risk Capsules only distinguished by a number recorded and detection bias) by the drug manufacturer. Investigators, treating patients/carers physicians and patients blinded Blinding (performance bias Low risk Capsules only distinguished by a number recorded and detection bias) by the drug manufacturer. Investigators, treating outcome assessment physicians and patients blinded Incomplete outcome data High risk 60 enrolled. 13/31 (42%) dropouts in pilocarpine (attrition bias) group; 8/29 (28%) dropouts in placebo group Selective reporting Low risk Xerostomia and adverse events reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Haddad 2009

Methods Location: USA Number of centres: 4 Date of enrolment: May 2003 to April 2006 Participants Inclusion criteria: with stage III or IV, previously untreated, locally advanced, SCCHN. Primary tumour types allowed: oropharynx, hyperpharynx, oral cavity, larynx, unknown primary Exclusion criteria: grade > 2 peripheral neuropathy other serious comorbid illness, involuntary weight loss of > 20% of body weight in 3 months preceding study Age: amifostine mean 55; control 57 Gender: amifostine: 27 M, 2 F; control: 23 M, 6 F Cancer type: (amifostine/control) oropharynx = 18/17, oral cavity = 5/6, larynx = 3/5, unknown primary = 2/0, other = 1/1 Neck dissection: amifostine 48%; control 38%; no details reported Radiotherapy: concomitant boost radiation, 72 Gy in 42 fractions over 6 weeks. Use of IMRT not allowed Chemotherapy: 4 weekly doses of carboplatin/paclitaxel. Induction chemotherapy was used in 29 of 58 patients overall with docetaxel, cisplatin, and 5-fluorouracil Number randomised: 58 (29 per group) Number evaluated: unclear for xerostomia

476 Interventions Amifostine versus no intervention Subcutaneous daily amifostine at dose of 500 mg 30-60 min before daily RT (before morning dose only, when schedule moved to twice daily radiotherapy at day 19). Average number of amifostine doses was 25 (median 28 doses). Amifostine withheld for skin toxicity Outcomes  Xerostomia: Common Terminology Criteria for Adverse Events including xerostomia reported but not by group  Salivary flow rates: saliva collection with and without citric acid simulation  Adverse effects: not reported  Survival data: overall survival, progression-free survival, local control  Other oral symptoms: Common Terminology Criteria for Adverse Events for mucositis, swallowing measured  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: xerostomia and mucositis assessed weekly throughout RT, then every 4 weeks after RT; salivary flow rate assessed at 12, 24 and 52 weeks after RT; dysphagia (swallowing) assessed at 8, 12, 24 and 52 weeks after RT; survival - median follow-up 34 months after RT, minimum 26 months Funding Medimmune Oncology Trial Not registered registration Sample size Yes calculation presented Notes Quote: "Study stopped before completion of planned accrual because IMRT was becoming de facto standard technique in treating head and neck cancer"

Study focuses on survival

Not able to use data - contacted authors for data 19 February 2016

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "The randomisation process was centralised generation (selection bias) and managed through the Dana-Farber Cancer Institute protocol office" Comment: linked to Harvard University probably done well Allocation concealment Low risk Quote: "The randomisation process was centralised (selection bias) and managed through the Dana-Farber Cancer Institute protocol office" Blinding (performance bias High risk No intervention group as comparator - not blinded and detection bias) patients/carers Blinding (performance bias High risk Subjective assessment of xerostomia and detection bias) outcome assessment Incomplete outcome data Unclear risk Unclear how many participants dropped out

477 (attrition bias) Selective reporting High risk Badly reported xerostomia and no adverse events (reporting bias) Other bias Low risk No other sources of bias are apparent

Han 2010

Methods Location: China Number of centres: 2 Date of conduct: 1 October 2007 to 31 July 2009 Participants Inclusion criteria: quote: "First-visit patients; diagnosed as mid/moderate to advanced/terminal nasopharyngeal squamous carcinoma through pathological and radiographic examinations; Karnosfsky score ≥ 60; expected survival period > 6 months; without severe complications (e.g. hypertension, coronary heart disease, diabetes, history of mental illness)" Exclusion criteria: see above Age: Jinlong: mean 46.3 (SD 7.4), median 53; control: mean 47.4 (SD 6.8), median 52 Gender: Jinlong: 33 M, 16 F; control: 34 M, 14 F Cancer type: nasopharyngeal squamous carcinoma Radiotherapy: dose 60 to 76 Gy, 2 Gy per day, 5 times a week Chemotherapy: "concurrent chemoradiotherapy" (no further details) Number randomised: 97 (Jinlong: 49, control: 48) Number evaluated: 95 (Jinlong: 48, control: 47) Interventions Jinlong capsules versus no intervention 4 tablets once, 3 tablets every day Duration: 3 months Follow-up: 12 weeks after treatment Quote: "Jinlong capsule is a modern 'fresh medicine preparation' made of fresh gecko and fresh long-noded pit vipers, using cryogenic modern biochemical extracting and separation techniques. It maintained to the greatest degree the activity of effective ingredients of organisms, and reasonable compatibility among the ingredients. Basic research has shown that Jinlong can directly damage cancer cells by blocking the mitosis and proliferation of cancer cells, fix the p21 small protein molecule, restore the regulation of cancer cells, and turn cancer cells to normal cells..." Outcomes  Xerostomia: quote: "observe the patients for toxic and side effects during and after radiotherapy, assess the toxic and side effects according to RTOG's criteria"  Salivary flow rates: not reported  Adverse effects: leukopenia, nausea, vomiting, 1 participant had dizziness and blood pressure drop, 1 participant had skin rash  Survival data: not reported  Other oral symptoms: mucositis  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported Funding Not reported; conflicts of interest: not reported Trial Not registered registration Sample size No calculation

478 presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear risk Quote: "randomly divided" generation (selection bias) Allocation concealment Unclear risk Insufficient information (selection bias) Blinding (performance bias High risk Jinlong versus no intervention and detection bias) patients/carers Blinding (performance bias High risk Jinlong versus no intervention and detection bias) outcome assessment Incomplete outcome data Low risk Quotes: "1 patient quit because of myocardial (attrition bias) infarction (Tx Group)", "1 patient quit because of mucosa toxicity (control group)" Selective reporting Low risk Xerostomia and adverse events reported (reporting bias) Other bias Low risk No other sources of bias are apparent

He 2004

Methods Location: China Number of centres: 1 Date of conduct: not stated Participants Inclusion criteria: aged 20-70 years; Karnofsky Performance Score > 70; Hb 90 to 150/L; blood pressure 12-20/8-15 kPa; normal kidney and liver function; no severe infection such as septicaemia; no heart disease; no medical history of low blood pressure, no other cancer and no history of radiotherapy Exclusion criteria: see above Age: aged 20 to 70 (no further details) Gender: not reported Cancer type: amifostine: nasopharyngeal squamous cell carcinoma stage 1 = 1, stage 2 = 7, stage 3 = 8 and stage 4 = 1. Control: nasopharyngeal squamous cell carcinoma stage 1 = 1, stage 2 = 5, stage 3 = 1 and stage 4 = 1 Radiotherapy: conventional with nasopharyngeal tumour dose (65-74 Gy) Chemotherapy: none Number randomised: 32 (amifostine: 17; control: 15) Number evaluated: 32 (amifostine: 17; control: 15) - 1 participant left amifostine group due to GI tract side effect but analysis states 17 in this group (possible ITT analysis) Interventions Amifostine versus no intervention Amifostine (200 mg/m²), diluted with 'water for injection' at the concentration of 50 mg/mL, IV 15-30 min before RT Control: nothing Outcomes  Xerostomia: "mucositis and xerostomia according to RTOG's criteria" (0-4 scale; we report grade 2 and above)  Salivary flow rates: "method used to measure the amount of saliva: put a 0.2 g cotton ball under patient's tongue, after 3 minutes, use electronic balance to

479 measure its weight", reported as decrease in saliva/change score (unstimulated)  Adverse effects: GI tract reaction/side effects (nausea and vomiting)  Survival data: not reported  Other oral symptoms: mucositis (RTOG criteria)  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: 3, 5 and 7 weeks after start of RT Funding Not reported; conflicts of interest: not reported Trial Not registered registration Sample size No calculation presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear risk Quote: "randomized into" generation (selection bias) Comment: no further details given Allocation concealment Unclear risk Insufficient information (selection bias) Blinding (performance bias High risk Amifostine versus no intervention and detection bias) patients/carers Blinding (performance bias High risk Not possible due to no intervention group and and detection bias) subjective assessment of xerostomia outcome assessment Incomplete outcome data Low risk 1/32 participants dropped out, however, appears (attrition bias) to be included in analysis Selective reporting Low risk Xerostomia and adverse events reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Henke 2011

Methods Location: Australia, Canada and Europe Number of centres: 38 hospitals Date of conduct: January 2005 to August 2007 Participants Inclusion criteria: more than 18 years old; resected for pathohistologically documented high-risk stage 2 to 4B SCC of the oral cavity, oropharynx, hypopharynx, or larynx; ECOG score of 0 to 2; at least 2 of 9 areas of the oral or oropharyngeal mucosa due to receive at least 50 Gy RT Exclusion criteria: tumours of the lips, paranasal sinuses, salivary glands, or unknown primary site; metastatic disease; history of chronic pancreatitis or acute pancreatitis within the last year; prior RT to the head and neck region or prior

480 chemotherapy; previous treatment on this study or with other KGFs Age: palifermin: mean 56 (SD 8); placebo: mean 57 (SD 9) Gender: palifermin: 78 M, 14 F; placebo: 75 M, 19 F Cancer type: head and neck (oropharynx, oral cavity, larynx, hypopharynx, other) Radiotherapy: standard fractionation of once daily 2 Gy fractions, 5 days per week; total 60 Gy (for R0 resection) over 6 weeks, or 66 Gy (for R1 resection) over 7 weeks, both with allowable range of ± 15% Chemotherapy: cisplatin (100 mg/m²) IV after appropriate hydration on days 1 and 22 (for R0 resection), or days 1, 22 and 43 (for R1 resection) Number randomised: 186 (palifermin 92; placebo 94) Number evaluated: 186 (palifermin 92; placebo 94) Interventions Palifermin versus placebo Palifermin: (120 µg/kg) 3 days prior to start of, and then once per week during radiochemotherapy, i.e. 7 doses for those with R0 resection, 8 doses for those with R1 resection (total dose = 840 or 960 µg/kg respectively) Placebo: same schedule with placebo Outcomes  Xerostomia: incidence of grade ≥ 2 xerostomia (Common Terminology Criteria for Adverse Events (CTCAE) v 3.0, assessed at months 4, 6, 8, 10, 12, reported only at month 4  Salivary flow rates: not reported  Adverse effects: assessed weekly during study treatment  Survival data: overall and progression-free survival, incidence of disease recurrence and death  Other oral symptoms: incidence of dysphagia (difficulty in swallowing), OMWQ-HN 0 (no soreness) to 4 (extreme soreness) scale for mouth and throat soreness assessed weekly and reported as mean score  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported Funding Quote: "This study was supported by Amgen" (Amgen also named as sponsor on trials registry - pharmaceutical industry) Trial NCT00131638 (clinicaltrials.gov/ct2/show/NCT00131638) registration Sample size Yes calculation presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "Random assignment was made by a generation (selection bias) centralized interactive voice response system" Comment: large multicentre trial using high-tech randomisation method - likely to be done properly Allocation concealment Low risk Quote: "Random assignment was made by a (selection bias) centralized interactive voice response system" Comment: large multicentre trial using high-tech randomisation method - likely to be done properly Blinding (performance bias Low risk Quote: "placebo-controlled, double-blind study" and detection bias) Comment: blinding feasible

481 patients/carers Blinding (performance bias Low risk Quote: "placebo-controlled, double-blind study" and detection bias) Comment: blinding feasible outcome assessment Incomplete outcome data Low risk All cases accounted for. ITT analysis (participants (attrition bias) having no assessment assumed to have event) Selective reporting Low risk All outcomes reported. Low risk for xerostomia at (reporting bias) the time point used in meta-analysis (4 months) - however, it should be noted that xerostomia was measured up to 12 months but data not reported Other bias Low risk No other sources of bias are apparent

Hu 2005

Methods Location: China Number of centres: 1 Date of conduct: January 2002 to June 2004 Participants Inclusion criteria: head and neck patients confirmed by pathological examination Exclusion criteria: not reported Age (years): treatment: mean 51 (SD 19); control: mean 49 (SD 18) Gender: treatment: 36 M 34 F; control: 38 M, 32 F Cancer type: treatment: nasopharyngeal (52), tonsil (11) and tongue (7); cancer stage: I = 6, II = 20, III = 28 and IV = 16. Control: nasopharyngeal (51), tonsil (11) and tongue (8); cancer stage: I = 6, II = 19, III = 29 and IV = 16 Radiotherapy: overall dose: 70 Gy for nasopharyngeal carcinoma, 55-70 Gy for carcinoma of tonsil and tongue Chemotherapy: none Number randomised: 140 (treatment 70, control 70) Number evaluated: 140 (treatment 70, control 70) Interventions Shenqi Fanghou recipe versus no intervention Shenqi Fanghou recipe: dangshen (30 g), astragalus root (30 g), tuckahoe (30 g), Chinese yam (30 g), hedyotic diffusa (30 g), barbated skullcup herb (30 g), pueraria root (30 g), fragrant solomonseal rhizome (10 g), glossy privet fruit (10 g), stiff silkorm (10 g), grassleaf sweetflag rhizome (10 g), atractylodes macrocephala (10 g), semen coicis (50 g), dried tangerine peel (6 g), paris root (20 g), figwort root (15 g), common anemarrhena rhizome (15 g), gambir plant (15 g), scorpion (5 g), radix notoginseng (5 g), radix glycyrrhizae (5 g) Dosage: solution of 400 ml (200 ml in the morning and 200 ml in the afternoon), starting the first day of RT for 35 to 38 days Control group: nothing Follow-up: end of RT Outcomes  Xerostomia: subjective assessment of dry mouth: 1) mild: can eat dry cooked rice, 2) moderate: have difficulty in eating dry cooked rice, or 3) severe: cannot eat dry cooked rice  Salivary flow rates: not reported  Adverse effects: none  Survival data: survival after a follow-up of more than 1 year  Other oral symptoms: oropharyngeal mucosa reaction, difficulty in mouth opening  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported

482  Cost data: not reported Funding Source of funding: government (The Bureau of Science and Technology of Shenzhen City); conflicts of interest: not reported Trial Not registered registration Sample size No calculation presented

Risk of bias table

Comment: insufficient information Blinding (performance bias High risk Shenqi Fanghon recipe versus no intervention and detection bias) patients/carers Blinding (performance bias High risk Not possible due to no intervention group and and detection bias) subjective assessment of xerostomia outcome assessment

Incomplete outcome data Low risk No dropouts (attrition bias) Selective reporting Low risk Xerostomia and adverse events reported (quote: (reporting bias) "no adverse events") Other bias Low risk No other sources of bias are apparent

Jaguar 2015

Methods Location: Brazil Number of centres: 1 Date of conduct: January 2010 to March 2012 Participants Inclusion criteria: primary oral, oropharynx, or nasopharynx carcinomas (clinical stage ≥ II) scheduled to undergo 3-D radiotherapy (RTC3D) or IMRT, ≤ 75 years of age Exclusion criteria: hypersensitivity to bethanechol, hypotension, hyperthyroidism, peptic ulcer disease, epilepsy, angina, parkinsonism, and patients using tricyclic antidepressants, and antihistamines Age: bethanechol: mean 55.9 (range 21 to 75); placebo: mean 55.8 (range 28 to 75) Gender: bethanechol: 37 M, 11 F; placebo: 39 M, 10 F Cancer type: oral cavity, oropharynx, nasopharynx Radiotherapy: once-daily mega voltage (6 MV), given at 18 to 2.12 Gy per fraction, 5 days per week (duration unclear) for 7 weeks Chemotherapy: bethanechol 73%; placebo 71% (type of CT not reported) Number randomised: 97 (bethanechol 48, placebo 49)

483 Number evaluated: 84 (bethanechol 42, placebo 42) Interventions Bethanechol versus placebo Both groups: 1 tablet (25 mg) taken twice a day from beginning of RT and continued until 1 month after end of treatment (median 19 weeks) Outcomes  Xerostomia: observer-based grade and scored according to the subjective measures of Eisbruch (grade 0 to 3) - reported as grade 2 and above  Salivary flow rates: whole unstimulated and stimulated saliva flows collected over 5 min each and reported in ml/min (reported by RT-type subgroups - we combined the subgroups but numbers were not reported so we used the number randomised from table 1), also scintigraphy undertaken  Adverse effects: bethanechol toxicities using National Cancer Institute Common Terminology Criteria for Adverse Events – NCI CTCAE, v 3.0  Survival data: not reported  Other oral symptoms: not reported  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: xerostomia assessed weekly to 3 months postRT; saliva flow assessed during RT (range 30 to 35 Gy) and 2 months postRT Funding FAPESP (an independent public foundation) and CAPES (an organization of the Brazilian federal government under the Ministry of Education) Conflict of interest statement does not indicate whether there is conflict or not; quote: "All authors disclose any financial and personal relationships with other people or organizations" Trial Not registered registration Sample size Yes (reported in supplementary data online) calculation presented Notes Supplementary data online dx.doi.org/10.1016/j.radonc.2015.03.017

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "Using the Epi-Info® software version 6.04b, generation (selection bias) eight lists with a randomized sequence for patient allocation were generated, because a separate list was needed for each of the 8 strata defined by the 3 dichotomous stratification factors (randomization codes with block-size of eight)" Allocation concealment Unclear risk Authors do not state who randomised the participants (selection bias) and whether it was in a concealed manner Blinding (performance bias Low risk Quote: "a placebo was manipulated identical in and detection bias) color, shape and weight. Both bethanechol and patients/carers placebo therapies were coded as A and B. The clinician, patients as well as the statistician were unaware of the trial groups" Blinding (performance bias Low risk Placebo trial and detection bias)

484 outcome assessment Incomplete outcome data Low risk 13 out of 97 dropped out with reasons for dropouts (attrition bias) clearly stated by study group, but equal per group and similar reasons Selective reporting Low risk Xerostomia and adverse events reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Jellema 2006

Methods Location: the Netherlands Number of centres: 1 Date of recruitment: August 1999 to August 2003 Participants Inclusion criteria: stage III/IVB squamous cell carcinoma of the oral cavity, oropharynx, hypopharynx and/or larynx or lymph node metastases in the head and neck area from an unknown primary. Treatment with bilateral primary or postoperative radiotherapy with curative intent. 75% of the parotid gland volume expected to receive a radiation dose of at least 40 Gy. Minimal life expectancy of 12 months and a WHO performance score from 0 to 2. Good understanding of the Dutch language Exclusion criteria: distant metastases (M1), previously irradiated patients, patients treated in combination with induction or concurrent chemotherapy, and patients with tumours that originated in the salivary glands. Pregnant patients, those participating in another investigational trial or in poor general health or psychological conditions. Patients who had severe cardiovascular disease, poor renal function or sustained hypotension not secondary to antihypertensive medication Age: mean age = 55 (24 to 73) Gender: AMI-3: 20 M, 10 F; AMI-5: 22 M, 8 F; control: 18 M, 13 F Cancer type: head and neck at various sites and stages and lymph node classifications Submandibular gland removal: participants were stratified by this factor but numbers of participants affected are not reported Radiotherapy: megavolt equipment using isocentre techniques after 3- dimensional planning. 2 opposing lateral fields with an anterior field to cover the lower jugular and supraclavicular lymph node areas. All received 46 Gy to treated areas, boost doses varied from 56 Gy (in patients who had negative surgical margins) to 63.5 Gy (in patients who had lymph node metastasis with extranodal spread or positive margins). Patients treated primarily with radiotherapy received 70 Gy to macroscopic tumour Chemotherapy: none Number randomised: 91 (AMI-3: 30; AMI-5: 30; control: 31) Number evaluated: 71 (xerostomia at 12 months) (AMI-3: 22; AMI-5: 27; control: 22) Interventions 3 arms: Amifostine 1 versus amifostine 2 versus no intervention Group 1: amifostine 3 times weekly 200 mg/m² administered IV over 3 to 5 minutes 15 to 30 minutes before irradiation Group 2: amifostine 5 times weekly 200 mg/m² administered IV over 3 to 5 minutes 15 to 30 minutes before irradiation Control: nothing Outcomes  Xerostomia: late and acute radiation-induced xerostomia at grade 2 and above (0 to 4 scale - RTOG/EORTC Late Radiation Morbidity Scoring); patient-rated xerostomia and sticky saliva using QLQ-H&N35 (1 to 4 scale converted linearly to a 0 to 100 mm scale where higher scores = worse

485 symptoms) - not used  Salivary flow rates: not reported  Adverse effects: vomiting (emesis), nausea, hypotension, allergic reaction  Survival data: locoregional tumour control and overall survival  Other oral symptoms: not reported  Other oral signs: not reported  Quality of life: QoL-C30 version 3.0, the EORCT Core Questionnaire with supplemental head and neck specific module (QLQ-H&N35)  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: xerostomia and QoL assessed at end of RT and 6, 12, 18 and 24 months after RT; survival assessed to 60 months but reported in text at 2 years Funding Source of funding: not stated. Amifostine provided by Schering Plough Trial Not registered registration Sample size Reported calculation presented Notes Have only reported locoregional tumour control and overall survival data narratively as it did not seem sensible to combine the 2 amifostine arms due to differing results. Numbers per group unclear for xerostomia at end of RT and therefore not able to use

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Random assignment performed at university generation (selection bias) medical centre using a permuted block design Allocation concealment Unclear risk Not reported (selection bias) Blinding (performance bias High risk Blinding not possible and detection bias) patients/carers Blinding (performance bias High risk Blinding not mentioned. Xerosomia is subjective and detection bias) measure outcome assessment Incomplete outcome data High risk 22% dropouts at 12 months and difference in (attrition bias) attrition between groups (i.e. no amifostine = 29%, amifostine3 = 27%, amifostine5 = 10%) Selective reporting High risk Quality of life was measured but not reported (only (reporting bias) quote: "No significant differences") Other bias Low risk No other sources of bias are apparent

Jham 2007

Methods Location: Brazil Number of centres: 1 Date of enrolment: October 2004 to July 2005

486 Participants Inclusion criteria: adults with biopsy-proven malignant neoplasm of the head and neck who received external beam RT Exclusion criteria: conditions which may introduce adverse reaction to bethanechol: tricyclic antidepressants, antihistamines, betablockers, hypersensitivity Age: bethanechol: mean 57 (SD 15); control: 55 (SD 13) Gender: bethanechol: 17 M 5 F; control: 16 M 5 F Cancer type: malignant neoplasm of head and neck Radiotherapy: external beam RT, encompassing 1 or more salivary glands, minimum 45 Gy Chemotherapy: bethanechol 23%; control 48% (type of CT not reported) Number randomised: 43 (bethanechol 22; control 21) Number evaluated: range over outcomes (and time points). Xerostomia VAS at 08 to 40 weeks after RT: 30 (bethanechol 13; control 17) Interventions Bethanechol versus artificial saliva Bethanechol: 25 mg 3 times daily (6 am, 2 pm, 10 pm) administered with RT and used until end of RT Control: artificial saliva (OralBalance) - schedule not reported Outcomes  Xerostomia: subjective VAS scale (length not mentioned - not used), asking about dry mouth (yes/no)  Salivary flow rates: whole resting saliva and whole stimulated saliva collected over 5 minutes and reported in ml/min  Adverse effects: lacrimation, nervousness, frequent urination, sweating, warm face, cramps, diarrhoea, nausea  Survival data: death  Other oral: not reported  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: xerostomia and saliva flow assessed during RT (between 15th and 19th session), at end of RT and at least 2 months after RT (ranging from 8 to 40 weeks after) Funding CAPES (an organization of the Brazilian federal government under the Ministry of Education) gave financial support, Apsel Laboratories provided bethanechol, and Laclede provided artificial saliva Trial Not registered registration Sample size No calculation presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "Using the Epi-info software version 6.04b, 6 generation (selection bias) lists with randomized sequence for patient allocation were generated (random codes with block-size of 8). Prior to allocation patients were stratified by RT treatment and age" Allocation concealment Unclear risk Not reported

487 (selection bias) Blinding (performance bias High risk Quote: "...for obvious reasons it was not possible for and detection bias) the study to be double-blinded" patients/carers Blinding (performance bias High risk Quote: "...for obvious reasons it was not possible for and detection bias) the study to be double-blinded" outcome assessment Incomplete outcome data High risk Varies over outcomes. For xerostomia (VAS): 30% (attrition bias) dropped out or died (bethanechol 41%; control 19%). Differential dropout and (apart from death) reasons for dropouts unclear Selective reporting Low risk Xerostomia and adverse events reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Lajtman 2000

Methods Location: Croatia Number of centres: unclear Date of conduct: unclear Participants Inclusion criteria: patients scheduled to receive external beam radiation therapy to the major salivary glands completely or partially included in the field Exclusion criteria: significant cardiovascular, pulmonary, hepatic or pancreatic disorders or gastroduodenal ulcers Age: not reported Gender: not reported Cancer type: not reported Radiotherapy: weekly external beam radiation therapy for 4 to 8 weeks, no further details Chemotherapy: not stated Number randomised: unclear Number evaluated: 48 Interventions Pilocarpine versus placebo Pilocarpine: 5 mg capsules 4 times daily starting the day before RT and continuing for 3 months Placebo: 5 mg capsules 4 times daily starting the day before RT and continuing for 3 months Outcomes  Xerostomia: standardised questionnaire (subjective assessment, administered by clinician)  Salivary flow rates: stimulated salivary flow rate (parotid saliva by Carlson- Crittenden cup; submandibular/sublingual saliva by standardised suction device)  Adverse effects: not reported  Survival data: not reported  Other oral symptoms: not reported  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: 3 months (end of drug treatment), 6 months and 12 months Funding Unclear

488 Trial Unclear registration Sample size calculation No presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear risk Insufficient information generation (selection bias) Allocation concealment Unclear risk Insufficient information (selection bias) Blinding (performance bias Low risk Double-blind; pilocarpine versus placebo and detection bias) patients/carers Blinding (performance bias Low risk Double-blind; pilocarpine versus placebo and detection bias) outcome assessment Incomplete outcome data Unclear risk Unclear of number randomised to each group, (attrition bias) therefore number of dropouts unclear Selective reporting High risk Adverse effects not fully reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Lanzós 2010

Methods Location: Spain Number of centres: 1 Date of enrolment: May 2004 to May 2007 Participants Inclusion criteria: between 18 and 75 years of age. At least 10 teeth present in mouth. Willing to consent Exclusion criteria: presence of mucosal pathology, pregnant or undergoing orthodontic therapy Age: mouthwash: mean age 49.4 years (SD 15.4); control: mean age 54.3 years (SD 16.1) Gender (M:F): mouthwash 15:3, control 17:1 Cancer type: head and neck Radiotherapy: 50 to 80 Gy over 5 weeks Chemotherapy: probably none Number randomised: 36 (18 per group) Number evaluated: 16 at 4 weeks for stimulated saliva (mouthwash 9, control 7) Interventions Antiseptic mouthrinse versus placebo Mouthwash: CHX 0.12% and 0.05% cetylpyridinium by oral rinse 15 ml twice daily (morning and night). From start of RT for 28 days Placebo: control without active ingredient Outcomes  Xerostomia: not assessed  Salivary flow rates: stimulated saliva (ml/min), pH saliva (0/1/2)  Adverse effects: none reported  Survival data: not reported

489  Other oral symptoms: hiposialosis (drooling), mucositis, plaque, gingivitis, caries  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: 14, 28 days after RT started (i.e. no time points of interest) Funding Source of funding unclear; suspect pharmaceutical industry sponsored by intervention manufacture Perio-Aid Tratamiento Trial Unclear registration Sample size calculation No presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Computer generated list assigned by random generation (selection bias) number Allocation concealment Low risk Allocated after inclusion corresponding to (selection bias) numerically coded mouthrinse. Code only broken at end of study Blinding (performance bias Low risk List and numbered bottles provided by promoter. and detection bias) Participants and researchers blinded patients/carers Blinding (performance bias Low risk 1 single assessor blinded to allocation and detection bias) outcome assessment Incomplete outcome data High risk 56% attrition (antiseptic 50%, placebo 61%) for (attrition bias) outcome of interest (salivary flow rates) at 4 weeks Selective reporting High risk Xerostomia not reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Le 2011

Methods Location: North America and Europe Number of centres: 46 hospitals Date of conduct: August 2005 to September 2007 Participants Inclusion criteria: newly diagnosed unresected stage III to IV SCC of oral cavity, oropharynx, nasopharynx, hypopharynx or larynx, planned RT dose of more than 50 Gy to 2 subsites or oral cavity and oropharynx Exclusion criteria: evidence of secondary malignancy Age: mean 55.5 (SD 8.5) Gender: 159 M, 29 F Cancer type: SCC of oral cavity, oropharynx, nasopharynx, hypopharynx or larynx

490 Radiotherapy: mean 68 Gy in both arms for 43 days Chemotherapy: cisplatin 100 mg/m² IV infusion on days 1, 22, and 43 of RT Number randomised: 188 (94 per group) Number evaluated: 188, 185 adverse events Interventions Palifermin versus placebo Palifermin administered IV at 180 µg/Kg over a period of 30 to 60 seconds, in 8 weekly doses 3 days. Bolus injection before radiotherapy, then at weekend Placebo: matching as above (1.2 ml of sterile water +) Follow-up: median follow-up 25.9 months palifermin, 25.0 placebo Outcomes  Xerostomia: incidence of grade ≥ 2 xerostomia (Common Terminology Criteria for Adverse Events (CTCAE) v 3.0 Dry Mouth/Xerostomia scale - info from trials registry), assessed at months 4, 6, 8, 10, 12, reported only at month 4  Salivary flow rates: not reported  Adverse effects: assessed by Common Terminology Criteria for Adverse Events: nausea, constipation, decreased weight, vomiting, anaemia, leukopenia, fatigue, dehydration  Survival data: overall tumour response, time to locoregional tumour failure, incidence of secondary primary tumours, overall and progression-free survival  Other oral symptoms: dysphagia (difficulty in swallowing), OMWQ-HN 0 (no soreness) to 4 (extreme soreness) scale for mouth and throat soreness (assessed twice weekly by trained evaluators during radiochemotherapy)  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported Funding Externally funded by Amgen GSK Trial NCT00101582 (clinicaltrials.gov/show/NCT00101582) registration Sample size calculation Reported presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Centralised randomisation system. 1:1 allocation generation (selection bias) ratio Allocation concealment Low risk Centralised randomisation for all sites, probably (selection bias) allocation concealment Blinding (performance bias Low risk Quote: "placebo-controlled, double-blind study" and detection bias) Comment: blinding feasible patients/carers Blinding (performance bias Low risk Quote: "placebo-controlled, double-blind study" and detection bias) Comment: blinding feasible outcome assessment Incomplete outcome data Low risk All cases accounted for. ITT analysis (patients (attrition bias) having no assessment assumed to have event) Selective reporting Low risk All outcomes reported. Low risk for xerostomia at (reporting bias) the time point used in meta-analysis (4 months) -

491 however, it should be noted that xerostomia was measured up to 12 months but data not reported Other bias Low risk No other sources of bias are apparent

Lin 2014

Methods Location: Taiwan, Republic of China Number of centres: 1 Date of enrolment: January 2003 to November 2004 Participants Inclusion criteria: histological evidence of carcinoma of head and neck, to receive RT, life expectancy ≥ 3 months, ECOG status ≤ 2. Criteria such as white blood cells, platelets, haemoglobin had to be within certain parameters Exclusion criteria: prior RT, presence of oral lesions, severe organ failure, brain metastasis Age: TWBXM: mean 51 (SD 15); placebo: 54 (SD 16) Gender: TWBXM: 29 M, 9 F; placebo: 32 M 3 F Cancer type: treatment group: head and neck cancer stage 0 = 4, I = 4, II = 10, III = 4, IVA = 12, IVB = 4. Control group: head and neck cancer stage 0 = 3, I = 5, II = 8, III = 4, IVA = 10, IVB = 5 Radiotherapy: TWBXM: mean dose 6944.9 cGy; placebo: mean dose 7098.4 cGy Chemotherapy: not mentioned probably not given Number randomised: 73 (TWBXM 38; placebo 35) Number evaluated: 71 (TWBXM 38; placebo 33) Interventions Traditional Chinese medicine (TWBXM) versus placebo Tianwang Buxin Mini-pills (TWBXM) - 13 herbs listed in paper Placebo made of starch and designed to taste and look similar to intervention 3 g orally 3 times daily starting from initiation of RT until 1 month after RT completion Follow-up: end of RT (end of RT for dichotomous data, 1 month postRT for continuous data) Outcomes  Xerostomia: RTOG grades by evaluating physician - none, slight, moderate mouth dryness  Salivary flow rates: not reported  Adverse effects: EORTC QLQ-C30 - skin, nausea, vomiting, leukopenia, fatigue, pain, dyspnoea, insomnia, appetite loss, constipation, diarrhoea, body weight loss  Survival data: not reported  Other oral symptoms: EORTC QLQ-C30 - pain, swallowing, speech problems, mouth opening, teeth, mucositis, oral mucosa, loss of taste  Other oral signs: not reported  Quality of life: EORTC QLQ-C30 and head and neck specific QLQ-H & N35  Patient satisfaction: not reported  Cost data: not reported Funding Government. Committee on Chinese Medicine and Pharmacy, Department of Health, Executive Yuan, Taiwan (grants CCMP92-RD-011 and CCMP93-RD- 008). Quote: "All authors declare that there are no conflicts of interest" Trial Not registered registration Sample size No calculation presented

492 Notes Problematic data - emailed corresponding author 22 February 2016

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "...randomized to study medicine according generation (selection bias) to a computer-generated randomization schedule..." Allocation concealment Unclear risk Not reported (selection bias) Blinding (performance bias Low risk Quote: "All patients, the study nurses and doctors and detection bias) were blinded to the group of the treatment group" patients/carers Blinding (performance bias Low risk Quote: "All patients, the study nurses and doctors and detection bias) were blinded to the group of the treatment group" outcome assessment Incomplete outcome data Low risk The dropout rate was 3% to 7% (data reported (attrition bias) inconsistently in the report) at the end of RT, which is the preferred time point for this review. However, there were 29 (40%) dropouts at the time point of 1 month postRT Selective reporting Low risk Xerostomia and adverse events reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Lozada-Nur 1998

Methods Location: USA Number of centres: 1 Date of conduct: not stated Participants Inclusion criteria: not reported Exclusion criteria: not reported Age: pilocarpine: mean 51.5 years (range 29 to 76); placebo: 54.8 years (range 47 to 68) Gender (M:F): pilocarpine 9:2, placebo 9:2 Cancer type: nasopharyngeal cancer Radiotherapy: total dose 60 to 70 Gy Chemotherapy: some Number randomised: 22 (11 per group) Number evaluated: 11 per group for incidence of dry mouth. There are discrepancies over the numbers Interventions Pilocarpine versus placebo Pilocarpine: 5 mg tablets 3 times daily - 4 times daily, 2 weeks before RT and concurrently with RT Placebo Outcomes  Xerostomia: questionnaire  Salivary flow rates: resting salivary flow  Adverse effects: sweating, lacrimation (excess tears, crying), rhinorrhoea (watery discharge from the nose), diarrhoea, nausea, rhinitis, blurred vision, constipation, neuropathy  Survival data: not reported  Other oral symptoms: mucositis, dysphagia (difficulty in swallowing),

493 dysgeusia (taste disturbance), pain  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: prior to radiation, weekly during treatment, end of RT, 3 months after RT Funding Supported by a grant from MGI Pharma Inc. Trial Not registered or published registration Sample size No calculation presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Contact with authors confirmed that the random generation (selection bias) allocation was conducted by pharmaceutical company, producing coded containers Allocation concealment Low risk Quote: "Allocation sequence was provided in a (selection bias) logo-type format and kept by our dental assistant (in a locked cabinet)" Comment: further description of this from study authors implies allocation concealment Blinding (performance bias Low risk Quote: "double-blind" and detection bias) patients/carers Blinding (performance bias Low risk Quote: "double-blind" and detection bias) outcome assessment Incomplete outcome data Unclear risk 2 dropped out because of adverse events (severe (attrition bias) nausea and vomiting, and tumour growth). Unclear reasons for all dropouts for xerostomia outcome (6) Selective reporting Low risk Xerostomia and adverse events reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Patni 2004

Methods Location: India Number of centres: 1 Date of conduct: not reported Participants Inclusion criteria: locally advanced histologically proven squamous cell carcinoma of head and neck region. 75% or more of each parotid gland was included in the radiation portal. Age above 18 years. Expected survival > 12 months. Karnofsky Performance Status > 60. Normal haemogram, renal and liver functions, normal calcium levels Exclusion criteria: prior treatment for malignancy, associated hypotension or distant metastases

494 Age: not reported Gender: 65% M Cancer type: head and neck Radiotherapy: external radiation therapy with gamma rays to a dose of 66 to 72 Gy with conventional fractionation Chemotherapy: 40 mg/m² cisplatin weekly Number randomised: 170 (85 per group) Number evaluated: 170 (85 per group) Interventions Amifostine versus no intervention Amifostine: 250 mg IV over 3 minutes for 4 days a week from day 1 of radiotherapy until completion of treatment Control: nothing Outcomes  Xerostomia: acute and late xerostomia grade 2 and above (RTOG 0-4 scale)  Salivary flow rates: parotid scintigraphy (no data)  Adverse effects: not reported  Survival data: disease-free survival at 24 months and tumour response  Other oral symptoms: mucositis (RTOG)  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessments: xerostomia at 3 and 12 months after RT; parotid scintigraphy at 3, 6, 12 and 24 months after RT; survival at 24 months Funding Not reported Trial Not registered or published registration Sample size No calculation presented Notes Abstract with additional information provided by study authors

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear risk Randomised but unclear method generation (selection bias) Allocation concealment Unclear risk Not mentioned (selection bias) Blinding (performance bias High risk Not possible due to 'no treatment' group and detection bias) patients/carers Blinding (performance bias High risk Not possible due to 'no treatment' group. and detection bias) Xerostomia is subjective outcome outcome assessment Incomplete outcome data Low risk No dropouts (attrition bias) Selective reporting High risk No information on adverse events (reporting bias) Other bias Low risk No other sources of bias are apparent

495

Peng 2006

Methods Location: China Number of centres: 1 Date of conduct: October 2003 to October 2005 Participants Inclusion criteria: quote: "1) H&N SC patients diagnosed with pathological examinations, who cannot endure surgery or cannot be treated with radical resection; 2) first-visit patients who have not received cancer treatment, with no tumour metastasis; 3) WBC > 4.0*109/L, platelet count > 100.0*109/L, Hb > 10 g/L, normal function of heart, lungs, liver and kidneys; 4) Karnofsky score ≥ 60; 5) patients gave informed consent" Exclusion criteria: see above Age at baseline (years): amifostine: median 58; control: median 57 Gender: amifostine: 12 M, 6 F; control: 12 M, 6 F Cancer type: head and neck squamous carcinoma (amifostine/control): oral = 4/3, nasopharyngeal = 6/6, oropharyngeal = 5/5, hypopharyngeal = 1/1, laryngeal = 1/1, paranasalsinus = 1/1, glottis = 0/1 Radiotherapy: quote: "Primary site of tumour and cervical lymph nodes: 74.4 Gy overall, 1.2 Gy per time, 2 times per day (> 6 hours between these 2 times), for 4 to 5 weeks; separate routine fractionated RT for the neck area: 50 Gy overall, 25 times, 5 weeks" Chemotherapy: quote: "Continued IV infusion of 5 FU 750 mg/m² using pump, 24 hours per day for 3 days. On the 5th day after 5 FU use, intravenous infusion of cisplatin 50 mg/m² with 250 or 500 ml saline (2 to 4 hours). IV infusion of docetaxel 75 mg/m² with 250 ml saline (< 1.5 hours). Chemotherapy performed in rounds: 1st round – during RT, 2nd round – begin in the 5th week after RT, 3rd round – begin in the 9th week after RT" Number randomised: 37 (amifostine 18, control 19) Number evaluated: 36 (amifostine 18, control 18) Interventions Amifostine versus no intervention Amifostine: quote: "400 mg amifostine each time, intravenous infusion 15 minutes before RT and chemotherapy (finish in 5 to 7 minutes)" Control: no intervention other than the same RT and chemotherapy Outcomes  Xerostomia: only results for 'acute' and 'chronic' dry mouth reported; time and standards for assessments not described  Salivary flow rates: not reported  Adverse effects: hypotension, nausea, vomiting, dizziness, fatigue, hiccup, sneezing, facial flush  Survival data: not reported  Other oral symptoms: mucositis  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: not reported (therefore unable to use data) Funding Not reported; conflicts of interest: not reported Trial Not registered registration Sample size No calculation presented

496 Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear risk Quote: "patients were randomized to" generation (selection bias) Allocation concealment Unclear risk No details given (selection bias) Blinding (performance bias High risk Amifostine versus no intervention. Blinding not and detection bias) possible patients/carers Blinding (performance bias High risk Blinding not mentioned. Xerostomia is and detection bias) subjective measure outcome assessment Incomplete outcome data Unclear risk Quote: "1 patient in the amifostine group quitted (attrition bias) due to financial reasons" Selective reporting Unclear risk Xerostomia and adverse events reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Pimentel 2014

Methods Location: Brazil Number of centres: 1 Date of conduct: not reported Participants Inclusion criteria: newly diagnosed head and neck cancer beginning treatment with RT Exclusion criteria: previous RT, concomitant chemotherapy, cardiopathy, hypertension, diabetes, allergy to pilocarpine, Sjögren syndrome, salivary gland tumours, chronic lung disease, glaucoma, peptic ulcer, taking betablockers or drugs that could alter salivary flow Age: mean 60 years (not given by group) Gender (M:F): 8:3 Cancer type: oral (n = 1); oropharynx (n = 3); mouth floor and tongue (n = 2); larynx (n = 4); pharynx (n = 1) Radiotherapy: 35 to 50 Gy, with daily doses about 2 Gy Chemotherapy: none Number randomised: unclear whether 29 or 11 (see attrition bias) Number evaluated: 11 (pilocarpine 5, placebo 6) Interventions Pilocarpine versus placebo Pilocarpine: 5 mg 3 times daily for duration of RT Placebo: saline solution 3 times daily for duration of RT Outcomes  Xerostomia: patient-reported feeling of dry mouth  Salivary flow rates: unstimulated (USF) and stimulated saliva (SSF) (ml/min)  Adverse effects: reported narratively  Survival data: locoregional control: not reported  Other oral symptoms: oral mucositis, ulcers  Other oral signs: difficulty in eating  Quality of life: only eating  Patient satisfaction: not reported

497  Cost data: not reported

Timing of assessment: weekly during RT (4 weeks) Funding The National Research Council (CNPq) Trial Not registered registration Sample size No calculation presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Quote: "...dispensing pharmacy held custody of the generation (selection bias) samples, separating those from the group who took pilocarpine solution from those which took the placebo. All patients were assigned a number corresponding to the medicine bottle. Researchers were not granted access to that information prior to the end of the survey" Allocation concealment Low risk Quote: "...dispensing pharmacy held custody of the (selection bias) samples, separating those from the group who took pilocarpine solution from those which took the placebo. All patients were assigned a number corresponding to the medicine bottle. Researchers were not granted access to that information prior to the end of the survey" Blinding (performance bias Low risk Double-blind - see above and detection bias) patients/carers Blinding (performance bias Low risk Double-blind - see above and detection bias) outcome assessment Incomplete outcome data High risk Quote: "We pre-selected 29 patients; however, the (attrition bias) careful selection of the population was directly reflected in the number of enrolled patients and in the end, only 11 were included in the survey. We consider that this low number is a result not only of the exclusion stemming from previously established eligibility criteria but also from the breach of protocol" Selective reporting High risk Poorly reported data for xerostomia and no adverse (reporting bias) events Other bias Low risk No other sources of bias are apparent

Reshma 2012

Methods Location: India Number of centres: 1 Date of conduct: not stated Participants Inclusion criteria: carcinoma of the head and neck stage III and IV, aged

498 between 30 and 70, to receive RT, normal haematology, biochemistry and Karnofsky Performance Index > 70% Exclusion criteria: poor general condition, associated co-morbidities, psychiatric conditions Age: 30-70 Gender: unclear Cancer type: SCC of head and neck Radiotherapy: 60 Gy on cobolt 60 for 30 days over 6 weeks Chemotherapy: not reported but probably none Number randomised: 20 or 40 unclear Number evaluated: no apparent dropouts but unclear how many started Interventions Tulasi (Ocimum Sanctum) versus placebo Tulasi (Ocimum Sanctum) Placebo: vitamin B complex 2 capsules of 250 mg orally half an hour prior to RT *Healthy control group also included but not used (not randomised) Follow-up: 29 days of RT Outcomes  Xerostomia: unclear, quote: "patients were assessed at the end of every week for grade of mucositis, skin reaction and salivary status"  Salivary flow rates: not reported  Adverse effects: not reported  Survival data: not reported  Other oral symptoms: not reported  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported Funding Not reported Trial Not registered registration Sample size No calculation presented Notes No useable data - emailed corresponding author 27 January 2016

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear risk Quote: "Patients were randomized into 2 arms" generation (selection bias) Comment: insufficient information on method of random sequence generation Allocation concealment Unclear risk Not reported (selection bias) Blinding (performance bias Low risk Placebo-controlled and detection bias) patients/carers Blinding (performance bias Low risk Placebo-controlled and detection bias) outcome assessment Incomplete outcome data Low risk No dropouts (attrition bias)

499 Selective reporting High risk No adverse event data reported. Poorly reported (reporting bias) xerostomia Other bias Low risk No other sources of bias are apparent

Rode 1999

Methods Location: Slovenia Number of centres: 1 Dates and duration of recruitment period: not stated Participants Inclusion criteria: irradiated for head and neck cancer and salivary glands included in the irradiation field Exclusion criteria: not reported Age: aged 32 to 72, median 62 years Gender (M:F): 60:9 Cancer type: oral cavity (n = 14), oropharynx (n = 33), hypopharynx (n = 8), larynx (n = 11), other (n = 3) (evenly distributed across groups) Radiotherapy: 5 Gy per day, 5 days a week. Irradiated volume reduced twice during irradiation treatment: at 40 Gy for shielding the spinal cord and at 60 Gy for treating the area of original disease, up to 70 Gy. Postoperative patients received 50 Gy - 56 Gy with parallel opposed portals only. 44 patients had postoperative RT and 25 were treated by RT alone Mean irradiation dose (Gy) delivered to area of salivary glands Chemotherapy: not stated Number randomised: 69 (A: 9, B: 30, C: 30) Number evaluated: 69 (A: 9, B: 30, C: 30) Interventions Pilocarpine (postRT) + Biperiden (during RT) versus no intervention Group Aª: pilocarpine during RT and 6 weeks after. Pilocarpine hydrochloride (5 mg) perorally 3 times daily administered 1 hour before irradiation Group B: Biperiden during RT and pilocarpine after RT group. Biperidin chloride (2 mg tablets) 1 and a half hours before irradiation, and pilocarpine hydrochloride (5 mg 3 times daily) for 6 weeks after RT Group C: no intervention Outcomes  Xerostomia: not reported  Salivary flow rates: mean quantity of non-stimulated saliva secretion (ml/min)  Adverse effects: not reported  Survival data: not reported  Other oral symptoms: mucositis, swallowing WHO criteria  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: end of RT, 3 months, 6 months and 12 months after end of RT Funding Source of funding: none Trial Not registered registration Sample size No calculation presented Notes ªGroup A data not used: randomisation to Group A was stopped after the first 9 patients for ethical reasons - 3 months after RT total cessation of saliva secretion

500 was observed in all except 1 patient

Follow-up: 12 months

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Information provided by author: "Sequence centrally generation (selection bias) generated" Allocation concealment Low risk Author included the following in an email (selection bias) "randomization with permuted blocs, participants were allocated to the treatment groups randomly. Allocation sequence was generated centrally, treating physician (radiologist) enrolled patient and participants were assigned to the groups by specialist in dental medicine"

Comment: centralised random allocation and was probably adequately concealed Blinding (performance bias High risk No blinding undertaken and detection bias) patients/carers Blinding (performance bias Low risk Salivary flow objective measure and detection bias) outcome assessment Incomplete outcome data Low risk All randomised patients included in analysis (attrition bias) Selective reporting High risk Xerostomia and adverse events not reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Sangthawan 2001

Methods Location: Thailand Number of centres: 1 Date of recruitment: January 1998 to January 1999 Participants Inclusion criteria: histologically documented squamous cell carcinoma of head and neck who would receive definite or postoperative radiation Exclusion criteria: significant uncontrolled cardiac, pulmonary, renal or occular disease or required tricyclic antidepressants or antihistamine with anticholinergic effects, betablocker, pilocarpine for ophthalmic indications or chemotherapy Age: pilocarpine: 57 years; placebo: 58 years Gender (M:F): 49:11 Cancer type: oropharynx (n = 27); nasopharynx (n = 14); others (n = 19) Radiotherapy: Cobalt-60 or 6 MV photon machine. Standard arrangement - opposing lateral portals, loaded 1:1 and/or anterior low neck field. Both parotids treated to a dose of at least 50 Gy with an equal daily dose of 1.8-2.0 Gy Chemotherapy: none Number randomised: 60 (30 per group) Number evaluated: 47 (25 pilocarpine; 22 placebo) Interventions Pilocarpine versus placebo

501 Pilocarpine jelly: self administered 5 mg 3 times daily at meal times for duration of RT (7 weeks) Placebo: self administered 3 times daily at meal times for duration of RT (7 weeks) Follow-up: 6 months after RT Outcomes  Xerostomia: subjective evaluation scores for xerostomia questionnaire (100 mm VAS for each of 5 questions)  Salivary flow rates: not reported  Adverse effects: reported ("non-specific symptoms such as nausea, vomiting, dizziness, urinary frequency, palpitation, sweating  and tearing")  Survival data: not reported  Other oral symptoms: not reported  Other oral signs: disability to oral intake, amount of meals, use of analgesics  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: 6 months after RT Funding Faculty of Medicine, Prince of Songkla University Trial Not registered registration Sample size No calculation presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Information provided by author: "random number generation (selection bias) table" Allocation concealment Low risk Information provided by author: "clinician not (selection bias) participating in study generated allocation sequence. Treatment codes concealed in sealed envelopes. Treatment coding not disclosed to investigator or patient" Blinding (performance bias Low risk Quote: "identically appearing placebo... patients and and detection bias) investigators were unaware of which treatment was patients/carers administered" Blinding (performance bias Low risk Information provided by author: "treatment code and detection bias) was disclosed to the investigator only after outcome assessment completion of the analysis of the results of the study" Incomplete outcome data Unclear risk 22% dropout rate (attrition bias) Selective reporting High risk Poor reporting of xerostomia without SD (reporting bias) Other bias Low risk No other sources of bias are apparent

Vacha 2003

502 Methods Location: Germany Number of centres: 1 Date of conduct: October 1996 to February 1999 Participants Inclusion criteria: advanced tumours of larynx, oro or hypopharynx, all received surgery and were to receive RT plus CT, pathologically confirmed cancer, between 18 and 70 years, performance status WHO 0-II, adequate bone marrow and liver and renal function Exclusion criteria: severe internal medical disorders, hyper or hypotension, history of RT or CT, women with inadequate contraception, secondary malignancies, recurrent tumours Age: amifostine: mean 53.5; control: mean 55.1 Gender: amifostine: M 21, F 4; control: M 19, F 6 Cancer type: amifostine: SSC (24) and adenocarcinoma (1); control: SSC (24) and lymphoepithelial cancer (1) Neck dissection involving removal of submandibular glands: radical bilateral (amifostine 32%, control 40%); radical unilateral + selective contralateral (amifostine 16%, control 12%); radical unilateral (amifostine 36%, control 16%) Radiotherapy: conventionally fractionated RT (5 x 2 Gy/week). Total dose 60 Gy for completely resected tumours, 70 Gy incomplete resected tumours Chemotherapy: individually planned. 70 mg/m² carboplatin on treatment days 1 to 5 and 29 to 33 just before RT session (all participants had this) Number randomised: 56 (not reported by group) Number evaluated: 50 (25 per group); 41 (amifostine 19, control 22) for xerostomia Interventions Amifostine versus no intervention 250 mg amifostine given intravenously as short infusion over 10 to 15 minutes Outcomes  Xerostomia: RTOG (not useable)  Salivary flow rates: not reported  Adverse effects: skin toxicity, loss of hair  Survival data: not reported  Other oral symptoms: oral mucositis  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: weekly during RT until week 6 (during RT; no usable data) Funding Not reported Trial Not registered registration Sample size No calculation presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear risk Quote: "...were randomized to receive..." generation (selection bias) Allocation concealment Unclear risk Not reported (selection bias)

503 Blinding (performance bias High risk No treatment control group and detection bias) patients/carers Blinding (performance bias High risk Xerostomia is a subjective assessment by the patient and detection bias) outcome assessment Incomplete outcome data High risk Reasons for dropouts by study group unclear. 6 (attrition bias) dropouts altogether: allergic skin reaction (1) after 5th application of amifostine, patient refusal (3), second malignancy (1), surgical complications (1). 27% attrition overall for xerostomia - participants with larynx cancer excluded from analysis Selective reporting High risk Xerostomia selectively and poorly reported (only (reporting bias) significant data reported) Other bias Low risk No other sources of bias are apparent

Valdez 1993

Methods Location: USA Number of centres: multisite unclear how many Date of conduct and duration of recruitment period: not stated Participants Inclusion criteria: scheduled to receive external beam radiation therapy to the major salivary glands, completely or partially Exclusion criteria: significant cardiovascular, pulmonary, hepatic, or pancreatic disorders or gastroduodenal ulcers. Women with childbearing potential were required to have a pregnancy test with negative results before entry and to use contraception during the study Age: pilocarpine: 22 to 65 years, mean 42.6 years; placebo: 21 to 56 years, mean 40.2 years Gender (M:F): 6:4 Cancer type: mix of SCC, mucoepidermoid carcinoma, Hodgkin disease and malignant lymphoma Radiotherapy: mean dose 41.9 Gy Chemotherapy: not stated Number randomised: 10 (5 per group) Number evaluated: 9 (pilocarpine 5; placebo 4) Interventions Pilocarpine versus placebo Pilocarpine: 5 mg capsules 4 times daily for 3 months starting the day before RT Placebo: 5 mg capsules 4 times daily for 3 months starting the day before RT All participants received a rigorous preventative oral hygiene regimen including topical fluoride applications Outcomes  Xerostomia: subjective assessment using questionnaire  Salivary flow rates: stimulated salivary flow rates (μl/min)  Adverse effects: not reported  Survival data: quote: "all tumour responded favourably and all were in complete remission for the remainder of the study"  Other oral symptoms: not reported  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: 3, 4, 5, 6 and 12 months from start of RT

504 Funding National Institute of Dental and Craniofacial Research (NIDCR). Randomisation sequence generation and allocation were undertaken by US National Institutes of Health (NIH) pharmaceutical service Trial Not registered registration Sample size No calculation presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Sequence generation not explicit but conducted by generation (selection bias) NIH Pharmaceutical Development Service Allocation concealment Low risk Conducted by NIH Pharmaceutical Development (selection bias) Service; intervention dispensed in coded bottles Blinding (performance bias Low risk Pilocarpine versus placebo and detection bias) patients/carers Blinding (performance bias Low risk Participants are blinded and providing saliva and and detection bias) questionnaire data outcome assessment Incomplete outcome data Low risk 1/5 participants in placebo group dropped out. No (attrition bias) dropouts in pilocarpine group Selective reporting High risk No adverse event data reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Veerasarn 2006

Methods Location: Thailand Number of centres: 5 Date of recruitment: February 1999 to September 2001 Participants Inclusion criteria: histological proven squamous cell carcinoma of head and neck region; ECOG performance statue 0-2; adequate bone marrow, liver and renal functions; age 18-70 years; no prior definite/radical surgery, chemotherapy, radiotherapy or biological response modifier; no evidence of distant metastasis; life expectancy ≥ 12 months; able to comply with a follow-up schedule; weight loss ≤ 10% in previous 3 months Exclusion criteria: concomitant malignant disease in other parts of the body; active uncontrolled infection; pregnant or lactating women; medical or psychiatric illness that compromise the patient's ability to complete the study; concomitant use of chemotherapy Age: amifostine: mean 55 (23-70); control: mean 52 (23 to 69) Gender: amifostine: 24 M, 8 F; control: 27 M, 8 F Cancer type: oral cavity, oropharynx, nasopharynx, larynx, hypopharynx Radiotherapy: standard fractionation (2 Gy, 5 days a week). Duration: 5 to 8 weeks. Definite RT = 70 Gy. Postoperative RT = 50 Gy. Amifostine group: definite RT = 15; postoperative RT = 17. Control group: definite RT = 18; postoperative RT = 17 Chemotherapy: none

505 Number randomised: 67 (amifostine 32, control 35) Number evaluated: 62 (amifostine 32, control 30) Interventions Amifostine versus no intervention Amifostine: (Ethyol) (200 mg/m²) IV (3 to 5 min), 30 min before RT. 5 consecutive days a week for 5 to 7 weeks (during RT) Control: nothing Outcomes  Xerostomia: 1) questionnaire (6 questions) (RTOG/EORTC acute and late radiation morbidity scoring criteria) for xerostomia: dryness of mouth, oral comfort, quality of sleep, ability to speak, ability to chew and swallow and ability to wear dentures (average score 0 to 10: 0 = normal); 2) RTOG 0 to 4 scale - grade 2 and above  Salivary flow rates: unstimulated and stimulated whole saliva collection (mg/5 min) and scintigraphy  Adverse effects: nausea, vomiting, hypotension  Survival data: disease-free survival  Other oral symptoms: mucositis (RTOG grade 2-3)  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessments: weekly during RT (for 6 weeks), then at end of RT and at 1, 2, 3, 6, 12, 18 and 24 months after RT; survival at 24 months after RT Funding Source of funding: unclear Trial Not registered or published registration Sample size No calculation presented Notes Information on randomisation, and numbers and SDs from correspondence

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Low risk Random number table generation (selection bias) Allocation concealment Low risk Central allocation (selection bias) Blinding (performance bias High risk Amifostine versus no intervention and detection bias) patients/carers Blinding (performance bias High risk Xerostomia is subjective measure and detection bias) outcome assessment Incomplete outcome data High risk Quote: "We excluded 5 cases in the control as they (attrition bias) did not have salivary gland function, or had severe salivary gland impairment"

Comment: we are assuming that there were no other dropouts (assessment made on 32 in amifostine, 30 in control)

506 Selective reporting Low risk Xerostomia and adverse events reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Wang 1998

Methods Location: China Number of centres: 1 Date of recruitment: May 1996 to October 1997 Participants Inclusion criteria: patients with head and neck cancer, pathology confirmed, aged 20 to 70 and radiation fields including the main salivary glands, total dose > 60 Gy Exclusion criteria: severe systemic disease or history of chronic diseases of the salivary glands Age (years): treatment: mean 46.8 (range 22 to 68); placebo: mean 45.5 (range 24 to 70) Gender: treatment: 20 M, 4 F; placebo: 21 M, 5 F Cancer type: treatment: nasopharyngeal = 17, throat = 4, tonsil = 3; placebo: nasopharyngeal = 18, throat = 4, tonsil = 3, hard palate = 1 Radiotherapy: 6 MV x-ray with accelerator linear, SL-75 produced by Philips. 2 Gy, 5 times per week for 6 to 7 weeks (total dose 60-70 Gy) Chemotherapy: none Number randomised: 50 (treatment 24, placebo 26) Number evaluated: 50 (treatment 24, placebo 26) Interventions Chinese medicine versus placebo (Dobell's solution) Chinese medicine and Dobell's solution (20 ml 3 times daily) rinsing and spray inhalation for the duration of the RT starting from the beginning of RT Chinese medicine: formulation of fragrant solomonseal rhizome (30 g), dwarf lilyturf tuber root (20 g), peach seed (24 g), dendrobium stem (30 g), wolfberry fruit (30 g), rehmannia dried root (40 g), prepared rehmannia root (40 g), American ginseng (30 g), safflower (20 g), chuanxiong rhizome (20 g). Mixture broken into a powder, soaked in 1000 ml of water, 40 degrees Centigrade for 6 hours, ultrasound oscillation for 10 minutes, centrifugation and the liquid part is used for clinical use Placebo: Dobell's solution Follow-up: end of RT Outcomes  Xerostomia: subjective evaluation score (VAS) for xerostomia, before and during the RT (at 10, 20, 30, 40, 50 and 60 Gy)  Salivary flow rates: stimulated salivary flow rates, in morning around 9 am (at least 1 hour after breakfast) patients rinsed with water, then chewed gum and saliva collected after 5 minutes, before and during the RT (at 20, 40 and 60 Gy)  Adverse effects: not reported  Survival data: not reported  Other oral symptoms: not reported  Other oral signs: not reported  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported Funding Not reported; conflicts of interest: not reported Trial Not registered registration

507 Sample size No calculation presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear risk Quote: "randomly divided" generation (selection bias) Comment: no further details given Allocation concealment Unclear risk Insufficient information (selection bias) Blinding (performance bias Low risk Placebo used and detection bias) patients/carers Blinding (performance bias Low risk Placebo used and detection bias) outcome assessment Incomplete outcome data Low risk No dropouts (attrition bias) Selective reporting High risk Adverse effects not reported. Data for xerostomia (reporting bias) not usable (reported as a graph with no SDs) Other bias Low risk No other sources of bias are apparent

Warde 2002

Methods Location: Canada Number of centres: 1 Date and duration of recruitment period: not stated Participants Inclusion criteria: squamous cell head and neck cancer, scheduled to receive RT with the inclusion of > 50% of both parotid glands in the radiation fields to doses > 50 Gy. Primary treatment and postoperative RT participants Exclusion criteria: previous RT or chemotherapy or pre-existing xerostomia from other causes. Medical contraindication to pilocarpine Age: pilocarpine: mean 56.2 years; placebo: mean 57.8 years Gender (M:F): 94:36 Cancer type: SSC of head and neck Radiotherapy: 60-70 Gy in 2 Gy daily fractions (68 participants), 60-64 Gy in 40 fractions during 4 weeks using twice daily treatments (33 participants), 50 Gy in 25 daily fractions (7 participants), 60 Gy in 25 daily fractions (8 participants) and 51 Gy in 20 daily fractions (7 participants) Chemotherapy: not stated Number randomised: 130 (65 per group) Number evaluated: for xerostomia: 92 (pilocarpine 48, placebo 44) at 3 months postRT; 87 (pilocarpine 46, placebo 41) at 6 months postRT Interventions Pilocarpine versus placebo Pilocarpine: 5 mg tablets 3 times daily starting day 1 of RT and continued until 1 month after completion of RT Placebo: tablets 3 times daily starting day 1 of RT and continued until 1 month after completion of RT Outcomes  Xerostomia: VAS (patient-completed) assessing patient's perception of

508 dryness of their mouth (7 questions). Scores from 0 to 100, low scores = most difficulty  Salivary flow rates: not reported  Adverse effects: excessive sweating, acute toxicity of therapy (RTOG)  Survival data: not reported  Other oral symptoms: not reported  Other oral signs: feeding tube inserted  Quality of life: patients' quality of life (McMaster University Head and Neck Questionnaire (HNRQ)). Score 1-7, lower score = poorer quality of life  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: 1, 3 and 6 months after end of RT Funding Unrestricted educational grant from Pharmacia Canada Trial Not registered registration Sample size Yes calculation presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear risk Insufficient information generation (selection bias) Allocation concealment Unclear risk Insufficient information (selection bias) Blinding (performance bias Low risk Pilocarpine versus placebo, "double-blind" and detection bias) patients/carers Blinding (performance bias Low risk Self reported and detection bias) outcome assessment Incomplete outcome data High risk 130 participants randomised. 19/65 dropouts in (attrition bias) pilocarpine group; 24/65 dropouts in placebo group. 8 in pilocarpine and 4 in placebo dropped out for toxicity, otherwise reasons unclear Selective reporting Low risk Xerostomia and adverse events reported (reporting bias) Other bias Low risk No other sources of bias are apparent

Watanabe 2010

Methods Location: Japan Number of centres: 1 Date of conduct: January and October 2009 Participants Inclusion criteria: head and neck cancer scheduled for RT or RT + CT Exclusion criteria: not reported Age: polaprezinc: 67.4 (range 53 to 78); control: 62.7 (range 35 to 86) Gender (M:F): polaprezinc 13:3, control 11:4

509 Cancer type: head and neck, mainly pharyngeal and laryngeal Radiotherapy (mean dose and duration): polaprezinc: 51 Gy (range 30-70), 37 days (range 21-55); control: 58 Gy (range 36-70), 45 days (range 28-79) Chemotherapy: some but unclear what, concomitantly carried out for 56% of polaprezinc and 80% of control Number randomised: 31 (polaprezinc 16, control 15) Number evaluated: 31 (polaprezinc 16, control 15) Interventions Polaprezinc versus azulene oral rinse Polaprezinc granules (0.5 g) were dissolved in 20 ml of 5% sodium alginate Azulene oral rinse prepared by pouring 7 drops of 4% solution into 100 ml water Both groups administered via oral rinse. Rinse for 3 minutes, 4 times daily, from start to end RT. After rinsing, polaprezinc swallowed but azulene spat out. From start to end of RT Outcomes  Xerostomia: CTCAE 0-3 grade  Salivary flow rates: not reported  Adverse effects: not reported  Survival data: tumour response by RECIST criteria for specific group of patients  Other oral symptoms: pain, dysgeusia (taste disturbance), mucositis  Other oral signs: amount of meals  Quality of life: not reported  Patient satisfaction: not reported  Cost data: not reported

Timing of assessment: over RT period (maximum score) Funding Source of funding unclear Trial Unclear registration Sample size No calculation presented

Risk of bias table

Authors' Bias Support for judgement judgement Random sequence Unclear risk Quote: "...were randomly assigned". No details generation (selection bias) given Allocation concealment Unclear risk No details given (selection bias) Blinding (performance bias High risk None. Polaprezinc swallowed after rinsing, while and detection bias) azulene is spat out. Quote: "open trial" patients/carers Blinding (performance bias High risk Xerostomia subjective measurement, and and detection bias) different interventions outcome assessment Incomplete outcome data Low risk None (attrition bias) Selective reporting High risk No adverse events reported (reporting bias) Other bias Low risk No other sources of bias are apparent

510 Footnotes

CHX = chlorhexidine; CRT = chemoradiotherapy; CT = chemotherapy; CTCAE = Common Terminology Criteria for Adverse Events; ECOG = Eastern Cooperative Oncology Group score; EORTC = European Organisation for Research and Treatment of Cancer; F = female; G-CSF = granulocyte-colony stimulating factor; GI = gastrointestinal; GM-CSF = granulocyte- macrophage colony-stimulating factor; Gy = gray; HNSCC = head and neck squamous cell carcinoma; IMRT = intensity-modulated radiation therapy; ITT = intention-to-treat; IV = intravenous; KGFs = keratinocyte growth factors; M = male; min = minute; MV = megavolt; NCI CTC = National Cancer Institute Common Toxicity Criteria; NCIC CTG ECTC = National Cancer Institute of Canada Clinical Trials Group Expanded Common Toxicity Criteria; OMAS = Oral Mucositis Assessment Scale; QoL = quality of life; RECIST = Response Evaluation Criteria In Solid Tumours; RT = radiotherapy; RTOG = Radiation Therapy Oncology Group; SC = subcutaneous; SCC = squamous cell carcinoma; SCCHN = squamous cell carcinoma of the head and neck; SD = standard deviation; TNM = tumour, node and metastasis; VAS = visual analogue scale; WHO = World Health Organization; wt/vol = weight/volume.

Characteristics of excluded studies

Anné 2002 (249)

Reason for exclusion Not an RCT - compares a subcutaneous amifostine group with the intravenous amifostine and no-treatment arms from another study (Brizel 2000)

Bagga 2007 (250)

Reason for exclusion Abstract; insufficient information

Bakowski 1978 (251)

Reason for exclusion CCT

Belcaro 2008 (252)

Reason for exclusion Not head and neck cancer patients

Bohuslavizki 1998 (170)

Reason for exclusion Radioactive iodine not radiotherapy

Borg 2007 (253)

Reason for exclusion Abstract; insufficient information

Bourhis 2000 (254)

Reason for exclusion Salivary gland dysfunction was not a reported outcome

Braaksma 2002 (255)

511

Reason for exclusion Data unavailable for the outcomes of interest to this review

Braaksma 2005 (256)

Reason for exclusion Salivary gland dysfunction was not a reported outcome

Chambers 2005 (257)

Reason for exclusion Abstract; insufficient information

Demiroz 2012 (258)

Reason for exclusion Not an RCT

Fallahi 2013 (259)

Reason for exclusion Radioactive iodine not radiotherapy

Fan 2011 (260)

Reason for exclusion Unclear whether this is an RCT or not. Study authors contacted February 2016 but no reply received

Franzén 1995 (261)

Reason for exclusion Salivary gland dysfunction was not a reported outcome

Fuertes 2004 (177)

Reason for exclusion Quasi-randomised trial (case history number)

Goyal 2007 (262)

Reason for exclusion Abstract; insufficient information

Gu 2014 (263)

Reason for exclusion Abstract; insufficient information

Johnson 2002 (264)

Reason for exclusion Not an RCT

Karacetin 2004 (265)

Reason for exclusion Quasi-randomised trial

Koukourakis 2000 (266)

512 Reason for exclusion No formal assessment of salivary gland dysfunction; data not presented for head and neck cancer patients

Kumarchandra 2010 (267)

Reason for exclusion Abstract; insufficient information

Manoor 2014 (268)

Reason for exclusion Abstract; insufficient information

Mateos 2001 (269)

Reason for exclusion Quasi-randomised trial (alternate assignment)

Mitine 2000 (270)

Reason for exclusion Abstract; insufficient information

Mix 2013 (271)

Reason for exclusion Abstract; insufficient information

Nicolatou-Galitis 2003 (272)

Reason for exclusion Not an RCT

Norberg-Spaak 1996 (273)

Reason for exclusion Abstract; insufficient information

Norberg-Spaak 1997 (274)

Reason for exclusion Abstract; insufficient information

Nyárády 2006 (275)

Reason for exclusion Quasi-randomised trial (alternate assignment)

Park 2012 (276)

Reason for exclusion Abstract; insufficient information

Park 2012a (277)

Reason for exclusion Abstract; insufficient information

Peters 1999 (278)

Reason for exclusion Quasi-randomised trial (day of birth)

513

Qian 2003 (279)

Reason for exclusion Dissertation; unable to obtain a copy

Resubal 2011 (280)

Reason for exclusion Abstract; insufficient information

Rieger 2012 (281)

Reason for exclusion Control group not relevant for this review

Rischin 2010 (282)

Reason for exclusion Tirapazamine is cancer treatment drug not a radiation protector

Rudat 2005 (283)

Reason for exclusion Abstract; insufficient information

Schönekäs 1999 (284)

Reason for exclusion Not an RCT

Sharma 2012 (285)

Reason for exclusion Intervention not defined as a pharmacological agent

Strnad 1997 (286)

Reason for exclusion Abstract; insufficient information

Su 2006 (133)

Reason for exclusion Salivary gland dysfunction measured as an adverse event following administration of G-CSF

Takahashi 1986 (169)

Reason for exclusion Not an RCT

Thorstad 2003 (287)

Reason for exclusion Not an RCT

Uchiyama 2005 (288)

Reason for exclusion Unclear whether this is an RCT or not

Zale 1993 (289)

514

Reason for exclusion Abstract; insufficient information

Zimmerman 1997 (290)

Reason for exclusion Not an RCT

Footnotes

CCT = controlled clinical trial; G-CSF = granulocyte-colony stimulating factor; RCT = randomised controlled trial.

Characteristics of studies awaiting classification

Yu 2009 (291)

Methods RCT Participants Nasopharynx cancer patients with poorly differentiated squamous carcinoma, at clinical phase III to VI, aged 25 to 69 years, first session of treatment, radiation field included > 75% of the area of salivary glands Interventions Amifostine versus no intervention Outcomes Xerostomia; salivary flow rates; adverse effects Notes Method/scale used to assess xerostomia unclear and further information required from study authors; the results do not have the potential to change any conclusions of the review

Characteristics of ongoing studies

NCT02430298

Study name Topical and oral melatonin for preventing concurrent radiochemotherapy induced oral mucositis and xerostomia in head and neck cancer patients Methods Parallel, double-blind RCT Participants Head and neck cancer adult patients Interventions Topical and oral melatonin versus placebo Outcomes Xerostomia; oral mucositis; quality of life Starting date July 2013 Contact Nutjaree Pratheepawanit Johns, Khon Kaen University information Notes clinicaltrials.gov/show/NCT02430298

Footnotes

RCT = randomised controlled trial.

515 APPENDIX 12: Salivary gland dysfunction review - study data not included in analyses

1 Pilocarpine versus no treatment/placebo (other outcomes)

Outcome Study ID Time point Pilocarpine Control Results Comments Oral related symptoms (other than salivary gland dysfunction/xerostomia) Mean 38.7 Mean 56.7 Gornitsky (SD 31.6) (SD 26.7) Mean difference -18.00 (95% CI - Oral discomfort 2004 37.41 to 1.41), P = 0.07 n = 16 n = 20 Mean 57.5 Mean 37.3 Gornitsky (SD 26.8) (SD 27.5) Mean difference 20.20 (95% CI Speech difficulties 2004 1.93 to 38.47), P = 0.03 n = 16 n = 18 End of Mean 47.4 Mean 61.8 Gornitsky radiotherapy (SD 33.9) (SD 25.4) Mean difference -14.40 (95% CI - Eating difficulties 2004 35.38 to 6.58), P = 0.18 n = 15 n = 17 Mean 38.8 Mean 53.6 Gornitsky (SD 33.9) (SD 34.2) Mean difference -14.80 (95% CI - Mucosal pain 2004 37.07 to 7.47), P = 0.19 n = 17 n = 19 Pimentel RR 0.30 (95% CI 0.05 to 1.89), P = Oral complications 1/5 4/6 2014 0.20 Adverse events Abacioglu Random-effects meta-analysis of 5 Sweating 2/12 0/12 1997 studies: RR 2.98 (95% CI 1.43 to

516 Fisher 2003 18/118 5/114 6.22), P = 0.004

Gornitsky 3/28 1/28 2 2004 Heterogeneity: I = 0%, P = 0.52 Lozada-Nur 5/12 1/12 1998 Sangthawan 1/29 2/24 2001 Abacioglu RR 3.00 (95% CI 0.13 to 67.06), P Chilling 1/12 0/12 1997 = 0.49 Gornitsky 7/28 5/28 Random-effects meta-analysis of 3 2004 studies: RR 1.39 (95% CI 0.63 to Nausea Haddad 2002 3/18 3/21 3.05), P = 0.41

Lozada-Nur 2/12 1/12 2 1998 Heterogeneity: I = 0%, P = 0.93 Fisher 2003 13/118 10/114 Random-effects meta-analysis of 3 Gornitsky studies: RR 1.28 (95% CI 0.70 to 6/28 5/28 Vomiting 2004 2.35), P = 0.43

Lozada-Nur 2/12 1/12 2 1998 Heterogeneity: I = 0%, P = 0.92 Gornitsky RR 0.67 (95% CI 0.12 to 3.69), P = Headache 2/28 3/28 2004 0.64 Fisher 2003 3/118 0/114 Random-effects meta-analysis of 3

Excessive lacrimation Haddad 2002 1/18 0/21 studies: RR 2.54 (95% CI 0.70 to (tears) 9.17), P = 0.15 Sangthawan 4/25 2/22

517 2001 Heterogeneity: I2 = 0%, P = 0.71 Lozada-Nur RR 1.50 (95% CI 0.30 to 7.43), P = Dysphasia 3/12 2/12 1998 0.62 RR 1.45 (95% CI 0.25 to 8.51), P = Weakness Fisher 2003 3/118 2/114 0.68 Gornitsky Random-effects meta-analysis of 2 0/28 1/28 2004 studies: RR 1.02 (95% CI 0.11 to Nervous 9.33), P = 0.99 Lozada-Nur 1/12 0/12 1998 Heterogeneity: I2 = 0%, P = 0.33 Fisher 2003 2/118 5/114 Random-effects meta-analysis of 3 Lozada-Nur studies: RR 0.87 (95% CI 0.41 to 1/12 1/12 Rhinitis 1998 1.86), P = 0.72

Sangthawan 8/29 6/24 2 2001 Heterogeneity: I = 0%, P = 0.53 Lozada-Nur RR 3.00 (95% CI 0.13 to 67.06), P Blurred vision 1/12 0/12 1998 = 0.49 Fisher 2003 7/118 5/114 Random-effects meta-analysis of 2

studies: RR 0.87 (95% CI 0.43 to Urinary frequency Sangthawan 1.75), P = 0.70 6/25 8/22 2001 Heterogeneity: I2 = 0%, P = 0.32 Gornitsky 0/28 2/28 Random-effects meta-analysis of 2 Dizziness 2004 studies: RR 0.80 (95% CI 0.18 to Sangthawan 4/25 3/22

518 2001 3.45), P = 0.76

Heterogeneity: I2 = 13%, P = 0.28 Sangthawan RR 0.10 (95% CI 0.01 to 1.73), P = Palpitation 0/25 4/22 2001 0.11 RR 2.90 (95% CI 0.12 to 70.44), P Skin flushing Fisher 2003 1/118 0/114 = 0.51 RR 1.93 (95% CI 0.18 to 21.02), P Motor tremors Fisher 2003 2/118 1/114 = 0.59 Mean 37.3 Mean 49.6 Gornitsky End of (SD 36.4) (SD 36.9) Mean difference -12.30 (95% CI - Sleep problems 2004 radiotherapy 36.27 to 11.67), P = 0.31 n = 17 n = 19 RTOG (grade 3; mucous No statistically significant membrane, pharynx and Warde 2002 difference between larynx) treatment groups

Footnotes

CI = confidence interval; RR = risk ratio; RTOG = Radiation Therapy Oncology Group; SD = standard deviation.

2 Biperiden plus pilocarpine versus no treatment/placebo (other outcomes)

Outcome Study ID Time point Pilocarpine Control Results Comments Dysphagia (WHO grade 3+) Rode 1999 12 months after RT 1/30 4/30 RR 0.25 (95% CI 0.03 to 2.11), P = 0.20

Footnotes

519 CI = confidence interval; RR = risk ratio; RT = radiotherapy; WHO = World Health Organization.

3 Amifostine versus no treatment/placebo (other outcomes)

Outcome Study ID Time point Amifostine Control Results Comments Assessed at end "No significant Jellema of RT and 6, 12, Quality of life No data No data differences between the 3 2006 18 and 24 treatment arms" months after RT Antonadou Random-effects meta- 14/22 23/23 2002 analysis of 2 studies: RR Dysphagia 0.50 (95% CI 0.17 to (difficulty in End of RT 1.48); P = 0.21 Büntzel swallowing) (0-4 1/14 5/14 1998 2 scale): grade 3 and Heterogeneity: I = 40%, above P = 0.20 Antonadou RR 0.70 (95% CI 0.13 to By 8 weeks after RT, no participants had grade 3 4 weeks after RT 2/22 3/23 2002 3.78); P = 0.68 or above dysphagia Dysgeusia (taste disturbance) (0-4 Büntzel RR 0.24 (95% CI 0.10 to End of RT 3/14 14/14 scale): grade 2 and 1998 0.61); P = 0.003 above Cost data (mean per Büntzel USD patient supportive End of RT USD 4401 P = 0.02 1998 5873 care costs) Antonadou Random-effects meta- Vomiting 1/22 0/23 2002 analysis of 5 studies: RR

520 Brizel 2000 55/150 11/153 4.90 (95% CI 2.87 to

Buentzel 8.38); P < 0.00001 8/66 2/64 2006 Heterogeneity: I2 = 0%, P "1 patient left due to gastrointestinal tract = 0.96 reaction/side effect, all other patients completed the treatment"

He 2004 1/17 0/15 "At the beginning of treatment, nausea and

vomiting was obvious for amifostine group, but after treating with metoclopramide, there was no significant difference between 2 groups in gastrointestinal tract reaction/side effect" Jellema 10/60 0/31 2006 Data not reported in control group. Unknown if Peng 2006 10/18 this was due to 0 events Veerasarn Data not reported in control group. Unknown if 18/32 2006 this was due to 0 events Antonadou Random-effects meta- 3/22 0/23 2002 analysis of 3 studies: RR Brizel 2000 22/150 2/153 9.20 (95% CI 2.84 to 29.83); P = 0.0002 Hypotension Büntzel 2/14 0/14 2 1998 Heterogeneity: I = 0%, P = 0.88 Veerasarn Data not reported in control group. Unknown if 5/32 2006 this was due to 0 events

521 Brizel 2000 66/150 25/153

Buentzel 4/66 4/64 2006 Random-effects meta- "1 patient left due to gastrointestinal tract analysis of 4 studies: RR reaction/side effect, all other patients completed 2.60 (95% CI 1.81 to the treatment" 3.74); P < 0.00001 He 2004 1/17 0/15 "At the beginning of treatment, nausea and

2 vomiting was obvious for amifostine group, but Nausea Heterogeneity: I = 0%, P = 0.45 after treating with metoclopramide, there was no significant difference between 2 groups in gastrointestinal tract reaction/side effect" Jellema 23/60 3/31 2006 Data not reported in control group. Unknown if Peng 2006 10/18 this was due to 0 events Veerasarn Data not reported in control group. Unknown if 20/32 2006 this was due to 0 events Brizel 2000 8/150 0/153 Random-effects meta-

Buentzel analysis of 3 studies: RR 2/66 0/64 2006 7.51 (95% CI 1.40 to Allergic response 40.39); P = 0.02 Jellema 4/60 0/31 2 2006 Heterogeneity: I = 0%, P = 0.77 Asthenia (weakness Buentzel RR 2.91 (95% CI 0.31 to 3/66 1/64 or lack of energy) 2006 27.24); P = 0.35

522 Vacha Similar in both groups and increased Alopecia 2003 continuously during the treatment Vacha Similar in both groups and increased Skin toxicity 2003 continuously during the treatment "..dizziness, fatigue, hiccup, sneezing, facial Peng 2006 flush all in less than 5% of the patients" Hot flush Veerasarn Data not reported in control group. Unknown if 17/32 2006 this was due to 0 events Somnolence Veerasarn Data not reported in control group. Unknown if 18/32 (drowsiness) 2006 this was due to 0 events "..dizziness, fatigue, hiccup, sneezing, facial Peng 2006 flush all in less than 5% of the patients" Sneezing Veerasarn Data not reported in control group. Unknown if 13/32 2006 this was due to 0 events "..dizziness, fatigue, hiccup, sneezing, facial Peng 2006 flush all in less than 5% of the patients" Hiccup Veerasarn Data not reported in control group. Unknown if 10/32 2006 this was due to 0 events "...dizziness, fatigue, hiccup, sneezing, facial Dizziness Peng 2006 flush all in less than 5% of the patients" "..dizziness, fatigue, hiccup, sneezing, facial Fatigue Peng 2006 flush all in less than 5% of the patients" Footnotes

CI = confidence interval; RR = risk ratio; RT = radiotherapy; USD = US dollars.

523 4 Amifostine: comparison of different doses (other outcomes)

Amifostine 3 times Amifostine 5 times Outcome Study ID Time point Results Comments weekly weekly Quality of Jellema Assessed at end of RT and 6, 12, "No significant differences No data No data life 2006 18 and 24 months after RT between the 3 treatment arms" Jellema RR 0.64 (95% CI 0.33 to 1.25); P Nausea 9/30 14/30 2006 = 0.19 Jellema RR 0.25 (95% CI 0.06 to 1.08); P Vomiting 2/30 8/30 2006 = 0.06 Allergic Jellema RR 1.00 (95% CI 0.15 to 6.64); P 2/30 2/30 response 2006 = 1 Footnotes

CI = confidence interval; RR = risk ratio; RT = radiotherapy.

5 Amifostine: different routes of administration (other outcomes)

Outcome Study ID Time point Intravenous Subcutaneous Results Comments Nausea/vomiting Bardet 2011 29% 36% P = 0.267

Hypotension Bardet 2011 20% 8% P = 0.007

Skin rash Bardet 2011 10% 22% P = 0.012

Local pain at injection site Bardet 2011 0% 8% P = 0.001

Fever Bardet 2011 2% 0% P = 0.256

Asthenia (weakness or lack of energy) Bardet 2011 1% 6% P = 0.054

524 6 Chinese medicine (other outcomes)

Outcome Study Intervention Time point Study Control Results Comments ID 38.9 Dysphasia (difficulty in Lin Chinese medicine End of RT 50.2 (26.3); n = 35 (25.9); n P = 0.07 swallowing) 2014 (TWBXM) = 33 (score for EORTC- 26.7 Lin Chinese medicine 1 month H&N35 questionnaire: 30.2 (29.8); n = 23 (24.8); n P = 0.65 2014 (TWBXM) after RT mean (SD)) = 21 RR 0.96 Dysgeusia (taste Lin Chinese medicine (95% CI 0.79 disturbance) (0 to 3 End of RT 32/38 29/33 2014 (TWBXM) to 1.16); P = scale): grade 1 and above 0.13 28.6 Lin Chinese medicine Speech difficulty End of RT 36.3 (26.7); n = 35 (26.2); n P = 0.23 2014 (TWBXM) = 33 (mean (SD) score for 22.7 EORTC-H&N35 Lin Chinese medicine 1 month 27.4 (28.6); n = 23 (19.5); n P = 0.50 questionnaire) 2014 (TWBXM) after RT = 21 41.4 Difficulty in mouth Lin Chinese medicine End of RT 39.6 (28.2); n = 35 (27.7); n P = 0.79 opening 2014 (TWBXM) = 33 (mean (SD) score for 33.3 Lin Chinese medicine 1 month EORTC-H&N35 32.2 (31.5); n = 23 (24.1); n P = 0.88 2014 (TWBXM) after RT questionnaire) = 21 Hu Chinese medicine "During the RR 0.42 Difficulty in mouth 22/70 52/70 2005 (Shenqi Fanghou treatment" (95% CI 0.29

525 opening recipe) to 0.61); P < 0.001 (0 to 2 scale): grade 1 and above RR 1.01 Skin toxicity (0 to 3 Lin Chinese medicine (95% CI 0.88 End of RT 35/38 30/33 scale): grade 1 and above 2014 (TWBXM) to 1.17); P = 0.82 RR 0.84 Chinese medicine Skin toxicity (0 to 4 Hu "During the (95% CI 0.74 (Shenqi Fanghou 57/70 68/70 scale): grade 1 and above 2005 treatment" to 0.94); P = recipe) 0.002 Quote: "toxicities during and after treatment were assessed" Skin toxicity (prevalence Han Chinese medicine according to RTOG 46.82% 58.32% 2010 (Jinlong capsule) Comment: time point for standards) assessment unclear; minor discrepancies in presented data RR 2.61 Nausea/vomiting (0 to 3 Lin Chinese medicine (95% CI 0.93 End of RT 12/38 4/33 scale): grade 1 and above 2014 (TWBXM) to 7.30); P = 0.183 RR 0.29 Lin Chinese medicine (95% CI 0.03 Hoarseness End of RT 1/38 3/33 2014 (TWBXM) to 2.65); P = 0.26

526 42.4 Lin Chinese medicine Fatigue End of RT 43.2 (26.2); n = 35 (23.0); n P = 0.88 2014 (TWBXM) = 33 (mean (SD) score for 36.4 EORTC-C30 Lin Chinese medicine 1 month 31.2 (28.3); n = 23 (25.0); n P = 0.51 questionnaire) 2014 (TWBXM) after RT = 21 41.7 Lin Chinese medicine Pain End of RT 46.8 (23.2); n = 35 (27.4); n P = 0.40 2014 (TWBXM) = 33 (mean (SD) score for 40.9 EORTC-C30 Lin Chinese medicine 1 month 35.9 (27.0); n = 23 (29.9); n P = 0.54 questionnaire) 2014 (TWBXM) after RT = 21 42.4 Lin Chinese medicine Pain End of RT 55.4 (25.1); n = 35 (20.5); n P = 0.02 2014 (TWBXM) = 33 (mean (SD) score for 37.8 EORTC-H&N35 Lin Chinese medicine 1 month 31.6 (24.2); n = 23 (23.3); n P = 0.35 questionnaire) 2014 (TWBXM) after RT = 21 16.7 Lin Chinese medicine Dyspnea End of RT 17.1 (23.1); n = 35 (20.7); n P = 0.93 2014 (TWBXM) = 33 (mean (SD) score for 13.6 EORTC-C30 Lin Chinese medicine 1 month 20.5 (21.2); n = 23 (22.2); n P = 0.28 questionnaire) 2014 (TWBXM) after RT = 21 31.2 Insomnia Lin Chinese medicine End of RT 40.5 (25.0); n = 35 (25.3); n P = 0.13 2014 (TWBXM) = 33

527 (mean (SD) score for 31.8 Lin Chinese medicine 1 month EORTC-C30 30.8 (24.8); n = 23 (28.1); n P = 0.28 2014 (TWBXM) after RT questionnaire) = 21 45.8 Lin Chinese medicine Appetite loss End of RT 45.0 (30.7); n = 35 (29.0); n P = 0.91 2014 (TWBXM) = 33 (mean (SD) score for 34.9 EORTC-C30 Lin Chinese medicine 1 month 28.2 (26.1); n = 23 (30.0); n P = 0.42 questionnaire) 2014 (TWBXM) after RT = 21 29.2 Lin Chinese medicine Constipation End of RT 37.8 (27.4); n = 35 (20.3); n P = 0.15 2014 (TWBXM) = 33 (mean (SD) score for 25.7 EORTC-C30 Lin Chinese medicine 1 month 29.5 (30.3); n = 23 (20.4); n P = 0.63 questionnaire) 2014 (TWBXM) after RT = 21 6.2 Lin Chinese medicine Diarrhoea End of RT 9.0 (15.0); n = 35 (13.2); n P = 0.42 2014 (TWBXM) = 33 (mean (SD) score for 6.1 EORTC-C30 Lin Chinese medicine 1 month 9.0 (15.1); n = 23 (16.7); n P = 0.53 questionnaire) 2014 (TWBXM) after RT = 21 Leukopenia, nausea, vomiting, 1 participant had Han Chinese medicine Not Adverse effects dizziness and blood pressure 2010 (Jinlong capsule) reported drop, 1 participant had skin rash Hu Chinese medicine "During the Not Adverse effects No adverse event 2005 (Shenqi Fanghou treatment" reported

528 recipe)

Footnotes

CI = confidence interval; EORTC = European Organisation for Research and Treatment of Cancer; H&N = head and neck; RR = risk ratio; RT = radiotherapy; RTOG = Radiation Therapy Oncology Group; SD = standard deviation; TWBXM = Tianwang Buxin Mini-pills.

7 Palifermin versus placebo (other outcomes)

Outcome Study Time Palifermin Placebo Results ID point Oral related symptoms (other than salivary gland dysfunction/xerostomia) Le 2011 29/94 19/91 Random-effects meta-analysis of 3 studies: Brizel 3 months 61/64 31/32 RR 1.32 (95% CI 0.55 to 3.13); P = 0.54 Dysphagia 2008 postRT Henke 2 32/92 20/93 Heterogeneity: I = 94%, P < 0.00001 2011 n = 94, mean = n = 94, mean = Random-effects meta-analysis of 2 studies: Le 2011 1.66, SD = 0.73 1.86, SD = 0.65 mean difference -0.12 (95% CI -0.27 to 0.02); Mouth and throat soreness - 0 (no soreness) 3 months P = 0.10 to 4 (extreme soreness) OMWQ-HN scale Henke postRT n = 92, mean = n = 94, mean = 2011 1.52, SD = 0.69 1.57, SD = 0.63 Heterogeneity: I2 = 13%, P = 0.28 Adverse events Nausea Le 2011 47/94 42/91 Random-effects meta-analysis of 2 studies: Brizel 48/67 26/32 RR 0.96 (95% CI 0.77 to 1.19); P = 0.69 2008 Heterogeneity: I2 = 28%, P = 0.24

529 Fever Brizel 30/67 13/32 RR 1.10 (95% CI 0.67 to 1.81); P = 0.70 2008 Constipation Le 2011 31/94 24/91 Random-effects meta-analysis of 2 studies: Brizel 28/67 13/32 RR 1.15 (95% CI 0.82 to 1.60); P = 0.42 2008 Heterogeneity: I2 = 0%, P = 0.57 Diarrhoea Brizel 14/67 8/32 Random-effects meta-analysis of 2 studies: 2008 RR 1.28 (95% CI 0.49 to 3.36); P = 0.61 Henke 11/92 5/93 2 2011 Heterogeneity: I = 57%, P = 0.13 Insomnia Brizel 12/67 4/32 Random-effects meta-analysis of 2 studies: 2008 RR 0.77 (95% CI 0.23 to 2.55); P = 0.67 Henke 5/92 12/93 2 2011 Heterogeneity: I = 63%, P = 0.10 Dyspnea Brizel 9/67 1/32 RR 1.10 (95% CI 0.67 to 1.81); P = 0.70 2008 Cough Brizel 8/67 5/32 RR 0.76 (95% CI 0.27 to 2.15); P = 0.61 2008 Headache Brizel 8/67 2/32 Random-effects meta-analysis of 2 studies: 2008 RR 2.13 (95% CI 0.86 to 5.28); P = 0.10 Henke 9/92 4/93 2 2011 Heterogeneity: I = 0%, P = 0.86 Decreased weight Le 2011 29/94 27/91 Random-effects meta-analysis of 2 studies: Brizel 7/67 4/32 RR 1.01 (95% CI 0.67 to 1.52); P = 0.96 2008 Heterogeneity: I2 = 0%, P = 0.73 Dizziness Brizel 5/67 4/32 RR 0.60 (95% CI 0.17 to 2.07); P = 0.42

530 2008 Anxiety Brizel 4/67 5/32 RR 0.38 (95% CI 0.11 to 1.33); P = 0.13 2008 Hypomagnesemia Brizel 4/67 4/32 RR 0.48 (95% CI 0.13 to 1.79); P = 0.27 2008 Vomiting Le 2011 26/94 26/91 Random-effects meta-analysis of 2 studies: Brizel 33/67 16/32 RR 0.98 (95% CI 0.72 to 1.33); P = 0.89 2008 Heterogeneity: I2 = 0%, P = 0.96 Radiation skin injury Le 2011 25/94 13/91 RR 1.10 (95%CI 0.67 to 1.81); P = 0.70 Anaemia Le 2011 21/94 34/91 Random-effects meta-analysis of 2 studies: Brizel 10/67 3/32 RR 0.83 (95% CI 0.33 to 2.05); P = 0.68 2008 Heterogeneity: I2 = 54%, P = 0.14 Fatigue Le 2011 21/94 20/91 Random-effects meta-analysis of 3 studies: Henke RR 0.88 (95% CI 0.60 to 1.30); P = 0.52 7/92 14/93 2011 Heterogeneity: I2 = 2%, P = 0.36 Brizel 17/67 8/32 2008 Leukopenia Le 2011 21/94 12/91 Random-effects meta-analysis of 2 studies: RR 1.01 (95% CI 0.37 to 2.78); P = 0.98 Henke 12/92 20/93 2011 Heterogeneity: I2 = 79%, P = 0.03 Granulocytopenia Brizel 20/67 6/32 RR 1.59 (95% CI 0.47 to 5.39); P = 0.45 2008 Pharyngolaryngeal pain Le 2011 20/94 23/91 RR 0.84 (95% CI 0.50 to 1.42); P = 0.52

531 Hypokalemia Le 2011 19/94 8/91 RR 2.04 (95% CI 0.98 to 4.28); P = 0.06 Pyrexia Le 2011 16/94 19/91 RR 0.82 (95% CI 0.45 to 1.48); P = 0.50 Henke Mucosal inflammation 4/92 10/93 RR 0.40 (95% CI 0.13 to 1.24); P = 0.11 2011 Henke Asthenia 13/92 7/93 RR 1.88 (95% CI 0.78 to 4.49); P = 0.16 2011 Henke Abdominal pain 7/92 2/93 RR 3.54 (95% CI 0.75 to 16.58); P = 0.11 2011 Henke Back pain 6/92 1/93 RR 6.07 (95% CI 0.74 to 49.40); P = 0.09 2011 1/92 Henke Febrile neutropenia 0/93 RR 3.03 (95% CI 0.13 to 73.48); P = 0.50 2011 Considered "serious adverse event" Dehydration Le 2011 13/94 19/91 Random-effects meta-analysis of 3 studies: Henke RR 0.75 (95% CI 0.45 to 1.25); P = 0.27 6/92 13/93 2011 Heterogeneity: I2 = 30%, P = 0.24 Brizel 20/67 8/32 2008

Footnotes

CI = confidence interval; OMWQ-HN = Oral Mucositis Weekly Questionnaire - Head and Neck Cancer; RR = risk ratio; RT = radiotherapy; SD = standard deviation.

532 8 Bethanechol versus placebo (other outcomes)

Outcome Study ID Results Adverse Jaguar No statistical difference between the groups in bethanechol-related toxicity. Quote: "No patient experienced severe (grade 3) effects 2015 toxicity and no one dropped out of the study due to adverse effects"

9 Bethanechol versus artificial saliva (other outcomes)

Study Time Artificial Outcome Bethanechol Results Comments ID point saliva RR 6.70 (95% CI Lacrimation 3/22 0/21 0.37 to 122.29); P (watering eyes) = 0.2 RR 6.70 (95% CI Nervousness 3/22 0/21 0.37 to 122.29); P

= 0.2 RR 6.70 (95% CI Frequent 3/22 0/21 0.37 to 122.29); P urination Jham End of = 0.2 2007 RT "1 patient using bethanechol dropped out of the study due to RR 4.78 (95% CI excessive sweating (Grade 2 severity; National Cancer Institute Sweating 2/22 0/21 0.24 to 94.12); P = Common Terminology Criteria for Adverse Events – NCI CTCAE, v 0.3 3" RR 4.78 (95% CI Warm face 2/22 0/21 0.24 to 94.12); P =

0.3 Cramps 1/22 0/21 RR 2.87 (95% CI

533 0.12 to 66.75); P = 0.51 RR 2.87 (95% CI Diarrhoea 1/22 0/21 0.12 to 66.75); P =

0.51 RR 0.48 (95% CI Nausea 1/22 2/21 0.05 to 4.88); P =

0.53

Footnotes

CI = confidence interval; RR = risk ratio; RT = radiotherapy.

10 Selenium versus no intervention (other outcomes)

Outcome Study ID Time point Reported in text Loss of taste Büntzel 6 weeks after "Ageusia was milder in the selenium group. But the difference was not significant" 2010 end RT Dysphagia Büntzel 6 weeks after "The only significant difference was observed at week 7, when the selenium group had developed a mean value 2010 end RT of 1.533 versus 2.167 in the control group (P = 0.05)" Adverse Büntzel 6 weeks after "23 serious adverse events (SAEs) were seen in the selenium group, compared to 22 in the control group (P = events 2010 end RT 0.476). No statistically significant differences in toxicities were found using the 2-tailed Fisher's exact test"

Footnotes

RT = radiotherapy.

534 11 Antiseptic mouthrinse versus placebo (other outcomes)

Outcome Study ID Time point Antiseptic rinse Placebo Results Drooling Lanzós 4 weeks from Increased 6 Increased 3 RR 1.43 (95% CI 0.46 to 4.39); P = 0.53 2010 baseline No change or decreased No change or decreased 8 7

6/14 3/10 Adverse Lanzós "No relevant adverse events were reported in any events 2010 group"

Footnotes

CI = confidence interval; RR = risk ratio.

12 Antimicrobial lozenge versus placebo (other outcomes)

Outcome Study ID Time point Antimicrobial lozenge Placebo Results Mouth pain Duncan 2005 Worse over 6 months 32/66 32/62 RR 0.94 (95% CI 0.66 to 1.33); P=0.72 Sore/burning mouth Duncan 2005 Worse over 6 months 32/65 32/62 RR 0.95 (95% CI 0.68 to 1.35); P=0.79 Throat pain Duncan 2005 Worse over 6 months 29/66 36/65 RR 0.79 (95% CI 0.56 to 1.12); P=0.19 Dryness in mouth Duncan 2005 Worse over 6 months 55/66 46/65 RR 1.18 (95% CI 0.97 to 1.42); P=0.09 Nausea Duncan 2005 Worse over 6 months 27/66 14/65 RR 1.90 (95% CI 1.10 to 3.28); P=0.02 Diarrhoea Duncan 2005 Worse over 6 months 6/66 3/65 RR 1.97 (95% CI 0.51 to 7.54); P=0.32 Constipation Duncan 2005 Worse over 6 months 24/66 26/65 RR 0.91 (95% CI 0.59 to 1.42); P=0.67

535 Footnotes

CI = confidence interval; RR = risk ratio.

13 Polaprezinc versus azulene rinse (other outcomes)

Outcome Study ID Time point Polaprezinc Azulene rinse Results Pain > 2 (0-3 scale) Watanabe 2010 Over RT period 5/16 13/15 RR 0.36 (95% CI 0.17 to 0.77); P = 0.008 Taste disturbance > 2 (0-3 scale) Watanabe 2010 Over RT period 1/16 8/15 RR 0.12 (95% CI 0.02 to 0.83); P = 0.03 Disability of oral intake Watanabe 2010 Over RT period 2/16 6/15 RR 0.31 (95% CI 0.07 to 1.31); P = 0.11 Footnotes

CI = confidence interval; RR = risk ratio; RT = radiotherapy.

14 Venalot Depot (coumarin/troxerutin) versus placebo

Outcome Study ID Results Adverse events Grötz 2001 "No adverse events could be attributed to the experimental medication"

536 APPENDIX 13: Salivary gland dysfunction review - secondary outcome analyses

Analysis 1.8: Pilocarpine versus control; Outcome: Overall survival

Analysis 1.9: Pilocarpine versus control; Outcome: Quality of life (various scales) 537

Analysis 3.6: Amifostine versus control; Outcome: Overall survival at 12 to 24 months postradiotherapy

Study ID Time point Amifostine Control Comments Haddad Median follow-up 34 months after radiotherapy, "No differences noted" 2009 minimum 26 months 3 times weekly = 84% Jellema Reported narratively rather than as a risk ratio due to differing 24 months 70% 2006 results in the amifostine arms 5 times weekly = 58%

Analysis 3.7: Amifostine versus control; Outcome: Overall survival - narrative data

538

Analysis 3.8: Amifostine versus control; Outcome: Progression-free survival at 12 to 24 months postradiotherapy

Analysis 3.9: Amifostine versus control; Outcome: Progression-free survival 539 Study ID Time point Amifostine Control Comments Haddad 2009 Median follow-up 34 months after radiotherapy, minimum 26 months "No differences noted"

Analysis 3.10: Amifostine versus control; Outcome: Progression-free survival - narrative data

Analysis 3.11: Amifostine versus control; Outcome: Locoregional tumour control at 12 to 24 months postradiotherapy

540 Study ID Time point Amifostine Control Comments Haddad Median follow-up 34 months after radiotherapy, "No differences noted" 2009 minimum 26 months 3 times weekly = 67% Jellema Reported narratively rather than as a risk ratio due to differing 24 months 79% 2006 results in the amifostine arms 5 times weekly = 83% Patni 2004 24 month No data No data "Amifostine does not alter the response or the survival"

Analysis 3.12: Amifostine versus control; Outcome: Locoregional tumour control - narrative data

Analysis 3.13: Amifostine versus control; Outcome: Disease-free survival 541 Study ID Time point Amifostine Control Comments Patni 2004 24 months No data No data "Amifostine does not alter the response or the survival" Veerasarn 2006 24 months No data No data "There was no statistical difference in 2-year disease-free survival"

Analysis 3.14: Amifostine versus control; Outcome: Disease-free survival - narrative data

Analysis 3.15: Amifostine versus control; Outcome: Quality of life (Patient Benefit Questionnaire) 542

Study ID Time point Amifostine 3 times weekly Amifostine 5 times weekly Comments Jellema 2006 24 months 84% 58%

Analysis 4.2: Amifostine (comparison of doses); Outcome: Overall survival - narrative data

Study ID Time point Amifostine 3 times weekly Amifostine 5 times weekly Comments Jellema 2006 24 months 67% 83%

Analysis 4.3: Amifostine (comparison of doses); Outcome: Locoregional tumour control - narrative data

Analysis 5.2: Amifostine (intravenous versus subcutaneous); Outcome: Overall survival 543

Analysis 5.3: Amifostine (intravenous versus subcutaneous); Outcome: Locoregional tumour control

Analysis 6.4: Chinese medicine versus control; Outcome: Overall survival (12 months post-RT)

544

Analysis 6.5: Chinese medicine versus control; Outcome: Quality of life (EORTC-C30)

545

Analysis 7.2: Palifermin versus placebo; Outcome: Overall survival at 42 to 72 months

Analysis 7.3: Palifermin versus placebo; Outcome: Progression-free survival at 42 to 72 months

546

Analysis 9.4: Bethanechol versus artificial saliva; Outcome: Overall survival

Analysis 12.2: Antimicrobial lozenge versus placebo; Outcome: Quality of life 547 APPENDIX 14: Results of individual assessors’ appraisal of included guidelines

AGREE II items (211)

Guideline Assessor 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Overall Would you assessment recommend?

Alberta 2017 PR 7 7 7 4 1 4 6 2 2 6 2 4 2 7 5 4 5 1 3 2 4 7 5 5 Y (+) AL 6 6 7 6 1 2 6 7 2 6 2 5 2 6 4 5 6 3 5 1 5 1 3 4 Y (+) HW 7 7 7 4 1 2 6 1 1 4 5 4 1 7 6 6 6 1 5 1 1 7 7 5 Y (+) AMG 7 7 7 4 1 1 5 1 1 6 4 5 1 7 5 5 6 1 5 1 1 7 7 5 Y (+) Bensinger PR 6 6 5 6 1 7 1 2 3 3 5 2 1 3 5 5 7 1 4 1 1 1 3 5 Y (+) 2008 AL 2 1 1 1 1 5 1 1 6 1 6 5 1 1 3 5 7 5 2 2 4 1 1 3 N HW 7 1 7 5 1 7 1 1 1 1 2 2 1 1 4 4 7 1 1 1 1 7 7 3 N AMG 7 1 7 4 1 7 1 1 1 1 3 3 1 1 3 3 7 1 1 1 1 7 7 3 N Buglione PR 5 5 4 5 1 3 4 4 3 4 3 4 5 1 4 3 4 1 1 2 1 1 5 4 Y (+) 2016 AL 5 2 2 6 1 1 6 5 2 6 6 7 6 1 3 5 6 1 1 2 1 1 6 4 Y (+) HW 7 1 7 5 1 4 6 5 4 3 4 5 4 1 5 4 7 1 1 1 1 7 7 4 Y (+) AMG 7 1 7 6 1 5 5 5 5 2 4 5 5 1 5 4 7 1 1 1 1 7 7 4 Y (+) Butterworth PR 3 3 3 3 1 3 1 1 1 1 2 1 1 1 1 1 3 1 1 1 1 1 1 1 N 2016 AL 3 2 4 1 1 5 1 1 1 1 2 2 1 1 2 2 5 1 1 1 1 1 1 2 N HW 5 1 7 1 1 3 1 1 1 1 1 4 1 1 3 2 7 1 1 1 1 1 1 1 N AMG 1 1 1 1 1 1 1 1 1 1 1 1 2 4 1 2 7 1 1 1 1 1 1 2 N Califano PR 6 5 5 5 1 5 1 1 1 2 1 4 1 1 4 2 6 1 1 1 1 7 3 4 Y (+) 2015 AL 4 2 4 6 1 6 1 1 3 1 4 5 1 1 6 6 6 1 1 1 3 6 7 3 N HW 7 4 6 5 1 4 1 1 1 1 1 4 1 1 5 4 7 1 1 1 1 7 5 2 N AMG 5 7 5 5 7 7 7 7 7 7 5 7 5 7 6 6 6 7 7 7 7 7 7 2 N Cohen 2016 PR 6 6 7 6 4 7 7 6 4 6 2 7 4 7 6 6 7 6 7 2 1 7 7 6 Y 548 AL 7 7 7 7 5 7 6 5 6 4 5 6 6 6 5 6 6 7 7 2 6 6 7 6 Y HW 7 7 7 7 5 7 7 7 7 7 6 7 7 7 6 6 7 7 7 1 5 7 7 6 Y AMG 5 5 6 7 5 7 7 6 3 3 4 2 7 7 5 5 7 6 6 6 1 7 7 5 Y (+) Davies 2010 PR 5 2 3 5 1 3 4 2 2 3 5 4 1 1 5 4 6 1 1 1 1 4 1 4 Y (+) AL 3 2 3 3 1 1 5 4 4 2 5 2 1 1 4 6 6 3 1 1 2 3 1 3 N HW 7 7 7 5 1 2 5 3 1 4 3 3 1 1 5 4 7 1 1 1 1 7 1 2 N AMG 7 6 6 5 1 2 5 1 5 2 4 5 5 5 1 5 4 7 1 1 1 7 1 4 Y (+) De Sanctis PR 5 5 4 5 1 3 4 4 3 4 3 4 5 1 4 3 4 1 1 2 1 1 5 4 Y (+) 2016 AL 5 2 2 6 1 1 6 5 6 6 6 6 6 1 3 5 5 1 1 2 6 1 6 4 Y (+) HW 7 7 7 7 1 4 4 4 1 5 4 4 1 1 5 5 7 1 1 1 1 7 6 3 N AMG 7 5 7 5 1 3 5 5 1 5 4 3 5 1 4 3 5 4 1 1 1 7 7 4 Y (+) Edmonds PR 6 6 4 4 1 4 7 2 3 2 2 4 1 1 4 4 5 2 1 1 1 4 3 4 Y (+) 2012 AL 2 3 3 2 1 5 4 3 3 1 2 6 1 1 6 4 6 2 2 1 4 2 6 3 N HW 7 7 7 3 1 4 1 1 1 1 4 6 1 1 5 4 6 1 1 1 1 1 1 2 N AMG 7 4 5 3 1 5 5 3 2 2 2 4 5 1 3 3 4 4 1 1 1 4 7 3 N Fogh 2014 PR 4 4 3 3 1 4 1 1 2 1 2 2 1 1 2 2 1 1 1 1 1 1 1 2 N AL 2 2 5 1 1 1 1 1 2 1 5 4 1 1 3 5 2 4 1 2 5 1 5 2 N HW 7 5 6 1 1 7 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 7 7 1 N AMG 7 2 3 1 1 5 1 1 1 1 1 1 5 1 2 3 1 1 1 1 1 1 1 1 N Hensley PR 4 4 4 6 4 6 6 7 6 6 5 4 4 4 7 4 7 2 1 3 1 1 3 5 Y (+) 2009 AL 4 2 2 7 5 5 6 6 6 6 6 7 7 1 6 6 7 4 1 7 2 1 6 5 Y (+) HW 7 6 7 7 7 4 6 7 7 7 6 7 1 1 7 5 5 1 1 1 1 7 7 5 Y AMG 7 5 5 7 6 5 7 7 5 6 6 6 7 2 7 6 7 1 1 5 1 7 7 6 Y Hutchinson PR 4 3 3 3 1 2 2 1 2 1 2 2 1 1 2 1 3 1 1 1 1 1 3 2 N 2014 AL 6 1 3 1 1 1 1 1 1 1 3 3 1 1 1 2 1 1 1 1 1 1 1 2 N HW 7 6 7 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 4 1 N AMG 7 5 7 1 1 1 1 1 1 1 1 3 4 1 2 1 1 1 1 1 1 4 6 1 N 549 Jensen 2010 PR 6 6 4 7 1 3 7 7 6 6 5 7 1 6 6 4 6 1 1 3 1 1 5 6 Y AL 5 6 5 5 1 1 6 6 7 6 6 5 1 3 4 6 6 3 2 5 3 1 5 4 Y (+) HW 7 7 7 6 1 1 5 5 1 5 4 5 1 1 3 3 4 1 1 1 1 1 6 1 N AMG 7 6 5 7 1 2 7 7 5 7 5 5 7 2 5 5 7 1 1 6 1 7 7 5 Y (+) Lacouture PR 6 5 4 4 1 3 4 3 3 4 2 3 1 1 3 3 5 1 1 1 1 7 3 4 Y (+) 2011 AL 5 3 5 4 1 1 3 3 5 2 6 5 1 1 5 6 6 1 6 1 5 7 6 4 Y (+) HW 7 7 5 6 1 1 3 5 3 3 2 1 1 1 2 2 7 1 1 1 1 7 7 3 N AMG 7 5 5 5 1 2 3 3 3 5 3 1 5 1 3 3 6 1 1 1 1 7 7 3 N Lalla 2014 PR 6 6 4 6 1 5 6 7 7 7 5 7 1 3 7 6 7 2 3 1 1 7 5 7 Y AL 5 3 5 3 1 1 4 6 6 7 2 5 1 2 3 6 7 5 6 4 1 7 7 4 Y (+) HW 7 7 7 6 1 2 7 6 4 6 2 1 1 1 7 7 7 1 1 1 1 7 7 4 Y (+) AMG 7 4 5 6 1 3 7 7 6 6 6 6 6 6 7 6 7 3 4 1 1 7 7 6 Y Mirabile PR 5 5 4 5 1 3 4 4 2 4 3 4 5 1 4 4 5 1 1 1 1 1 5 4 Y (+) 2016 AL 5 2 2 6 1 1 6 5 2 6 2 6 6 1 2 5 5 1 1 2 1 1 3 3 N HW 7 7 7 6 1 7 5 6 1 5 2 1 1 1 7 6 7 1 1 1 1 7 7 3 N AMG 7 4 6 5 1 3 5 6 5 3 3 6 5 1 2 2 5 1 1 2 1 7 7 3 N RCS_BSDO PR 7 5 4 2 1 3 2 1 1 1 1 4 1 2 5 4 5 1 1 1 1 1 1 3 N H 2012 AL 5 2 4 1 6 3 1 1 1 1 4 5 1 2 7 7 7 1 6 1 7 1 1 3 N HW 7 7 7 1 1 4 1 1 1 1 1 1 1 1 1 1 6 1 1 1 1 1 1 1 N AMG 7 5 6 2 1 2 1 1 1 3 1 1 4 4 4 4 2 1 1 1 1 1 1 2 N Sandherr PR 7 6 6 4 1 5 3 1 2 2 2 4 1 1 4 3 5 1 1 1 1 1 3 4 Y (+) 2015 AL 4 2 4 3 1 3 3 1 4 2 3 2 2 1 2 4 1 1 1 2 4 1 7 3 N HW 7 7 7 5 1 7 2 1 1 3 1 1 1 1 1 1 4 1 1 1 1 1 7 1 N AMG 7 7 6 3 1 2 3 1 1 3 2 5 1 1 3 3 1 1 1 1 1 1 7 2 N Snowden PR 4 4 4 4 1 4 1 1 1 3 2 3 4 4 3 3 3 1 1 1 1 4 1 3 N 2011 AL 3 3 3 3 5 3 1 1 2 5 5 5 5 5 2 3 5 2 1 2 1 2 2 3 N HW 7 7 7 1 1 7 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 N 550 AMG 7 7 7 3 7 3 3 1 4 5 5 5 5 1 6 4 7 1 1 5 1 1 1 4 Y (+) Tolentino PR 5 5 4 2 1 2 1 1 2 1 2 3 1 1 3 2 2 1 1 1 1 1 1 2 N 2011 AL 3 2 2 1 1 1 1 1 1 1 3 4 1 1 2 2 1 1 1 1 1 1 1 2 N HW 7 7 7 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 N AMG 7 5 5 1 1 2 1 1 1 1 1 1 1 1 2 2 1 1 1 1 1 1 1 1 N

Footnotes

PR: Philip Riley AL: Anne Littlewood HW: Helen Worthington AMG: Anne-Marie Glenny Y: yes Y (+): yes with modifications N: no

551 APPENDIX 15: Protocol for future research on the development of a core outcome set for oral mucositis prevention in adult cancer patients

Core outcome set for prevention of oral mucositis in cancer patients

Background Treatments for cancer, such as chemotherapy, radiotherapy of the head and neck, and targeted therapy, generally lead to toxic side effects in the oral cavity of the cancer patient (3). One such devastating side-effect is oral mucositis, which affects at least 75% of high risk patients (those receiving radiotherapy for head and neck cancers or high dose chemotherapy) (2). The process leading to oral mucositis is a complex and multifactorial one, culminating with ulceration of varying severity (8). This can then cause pain and difficulties with eating, swallowing and talking (2), which in turn can lead to the need for pain relief medication, nutritional support (e.g. a feeding tube), specialist oral hygiene care, treatment of the mucositis, and in some instances hospitalisation (4-6). All this negatively impacts on the cancer patient’s quality of life, but perhaps more seriously, the patient may need to alter or even halt their optimal cancer treatment programme, which can affect their chances of survival (3, 6, 7). Immunosuppressed patients have an increased risk of bacteraemia and sepsis (if whole bacteria invade the ulcers and cross into the underlying submucosa) which can be fatal, but at the very least require antibiotic medication and hospitalisation (2, 6, 7). The incremental costs associated with oral mucositis have been reported in one study to range, depending on the severity, from $1700 to $6000 per patient (10), and another study reported a median incremental cost of $18,515 per patient (61). Therefore, due to all these factors, the priority must be to prevent oral mucositis from occurring, and the research to support such endeavours must be of the highest quality.

In order to form clear and answerable research questions, researchers must consider the population, intervention, comparison, and outcome(s) that are relevant to their question (234). The outcome measures are the endpoint of the research and “clinical trials are only as credible as their endpoints”(235). It is unethical to waste limited healthcare resources by assessing outcomes that are not considered important by clinicians who are intended to use the research, or patients who are intended to benefit from the research. Also, it is well-established that if proper consideration is not given to outcomes, this gives rise to further problems. Firstly, inconsistency of outcome measures leads to heterogeneity and the inability to compare and summarise research when conducting systematic reviews (67). Secondly, there is the potential for outcome reporting bias, whereby results are selectively reported according to statistical significance/favourability (70). For example, one study reported that statistically significant outcomes were 2.2 to 4.7 times more likely to be fully reported than non- significant outcomes, and that 40% to 62% of studies had at least one primary outcome that was changed, introduced, or omitted (69). This can lead to a distorted view of the effects of an intervention in the literature, leading to an incorrect assertion that it is more beneficial than it actually is (68).

There is general agreement in the literature that the way forward in tackling the above problems is to develop disease/condition-specific core outcome sets to be used in the conduct and reporting of clinical trials (68). In an area where there is a known core outcome set, it would be expected that it would be included in any subsequent clinical trials as a minimum (67). The COMET (Core Outcome Measures in Effectiveness Trials) Initiative promotes the development of core outcome sets and is involved in 552 advancing the associated methodology. This project to develop a core outcome set for the prevention of oral mucositis in patients with cancer receiving treatment is registered on the COMET initiative’s free online database in order to avoid duplication of effort (www.comet-initiative.org).

The authors of the present study are aware of only one previous study in this field, which was published in 2002 and attempted to identify the most important outcomes to measure (71). The study brought together a panel of 11 people representing different stakeholders who discussed outcomes at a single meeting until consensus was reached. The authors of that study should be commended for their work and especially for including all relevant stakeholders (dentist, oncologists/haematologists, patients, statistician, health economist, oncology nurses, pharmaceutical industry representatives). However, there may be some limitations to the study which suggest that more rigorous and up-to-date work should be carried out: 1) the scope of the study was limited to bone marrow transplant patients, nurses and clinicians; 2) the sample size was small; 3) the panel was not international and therefore the study may lack external validity; 4) the authors state that they cannot be certain that the views of the panel are representative of their respective stakeholder groups; 5) the panel met face-to-face resulting in the loss of anonymity in their answers/opinions, and therefore some stakeholders may have been more coercive than others when trying to achieve consensus. It will be of interest to discover whether or not using alternative methods to the older study will produce the same results.

Aim The aim of this study is to produce an up-to-date core outcome set to be used in future clinical trials on the prevention of oral mucositis in patients of 16 years of age (young adults) and older receiving treatment for any type of cancer.

Objectives The objectives are:

 to carry out a systematic review of the literature in order to produce a list of outcomes previously used in clinical trials;  to identify which outcomes are most important to patients;  to identify which outcomes clinicians and other stakeholders believe to be the most important; and  to integrate the two sets of opinions into one core outcome set.

Methods Systematic review Two reviewers will extract outcomes independently and in duplicate from any new eligible studies identified from the updated electronic searches for the Cochrane systematic review ‘Interventions for preventing oral mucositis for patients with cancer receiving treatment’ (14). This will provide a list of outcomes that will be provided to the Delphi participants. As recommended by Kirkham et al, an outcome matrix will also be produced to show whether outcomes were fully, partially or not reported in each study (237). This will be reported in a separate paper exploring diversity of outcomes and possible outcome reporting bias. This will also include outcomes reported in the included studies of previous version of the Cochrane review in order to investigate any temporal relationship in outcome reporting. The Cochrane systematic review only includes randomised controlled trials (RCTs) as these are the gold standard study design when assessing the effectiveness of interventions (43). An additional search for

553 observational studies to identify long-term or rare outcomes will not be carried out as these outcomes are unlikely to be feasible for study in RCTs (292).

Identification of outcomes of importance to patients/carers Purposive sampling will be used to obtain a mixture of six to eight participants to take part in a focus group. Adults and young adults (16 years or older), with or without carers, who have experienced oral mucositis whilst undergoing treatment for cancer will be eligible to participate. They must also be able to understand verbal explanations and written information given in English, and to give consent to take part. The Christie cancer hospital in Manchester will recruit participants for this stage through advertising on a closed Facebook group and through a specialist nurse. Patient information leaflets will be provided to people expressing a wish to participate. This will be well in advance of the focus group taking place (greater than 24 hours), to allow them time to properly understand their participation in the research and make a decision on whether to take part. There will be contact information for the chief investigator so that they can get in touch to ask any questions they may have. Written consent would then be taken at the start of the focus group. The purpose of the focus group will be to elicit outcomes of importance to patients/carers using open questions. The focus group will be audio- recorded and transcribed so as not to miss any important information. As there is a risk of distress for participants when discussing such a sensitive topic, a distress protocol is presented in appendix 1.

Identification of outcomes of importance to all stakeholders Stakeholders of interest are: patients, carers, surgeons, oncologists, nurses, radiologists, dieticians, pharmacists, oral medicine specialists, dentists, health economists, researchers, pharmaceutical industry, and guideline developers. Multiple relevant groups will be contacted to ask their members to participate (e.g. Macmillan nurses, The Royal College of Radiologists, the Mucositis Study Group of the Multinational Association of Supportive Care in Cancer/International Society of Oral Oncology, etc). Patients involved in the focus group will also be given the option to participate in this phase. Although there are no formal statistical methods for estimating required sample sizes for such Delphi processes, a minimum of 50 participants will be aimed for.

A Delphi process (using SelectSurvey) will be used to preserve anonymity and avoid participants being influenced by other participants who may otherwise dominate a face- to-face meeting.

To minimise attrition, only people who respond to an invitation to participate, will be included in the Delphi process. The text suggested by Sinha et al will be used to emphasise the importance of remaining in the process until completion (293): “Thank you for agreeing to participate in our study. It is very important that you complete the questionnaires in each round. The reliability of the results could be compromised if people drop out of the study before it is completed, because they feel that the rest of the group does not share their opinions. If people drop out because they feel their opinions are in the minority, the final results will overestimate how much the sample of participants agreed on this topic.”

Participants will be sent the list of previously used outcomes (maximum number of 30) obtained from the systematic review, and also the list of outcomes resulting from the focus group, to avoid the risk of them missing important outcomes.

554 The numbers invited, numbers of respondents, and their stakeholder group at all stages will be recorded and clearly reported.

Delphi round 1 Participants will be asked for their job title/role (and asked to describe their clinical research experience). They will be provided with the list of outcomes from the systematic review (in alphabetical order to avoid any misconceptions that they have been provided to them in a preconceived order of importance) and the focus group results, and asked to score them. They will also be asked to add any other missing outcomes and score those too.

The Grading of Recommendations Assessment, Development and Evaluation (GRADE) scoring system will be used. This is on a 1 to 9 scale to maximise objectivity and minimise bias on what constitutes consensus (67). A score of 1 to 3 indicates limited importance, 4 to 6 important but not critical, and 7 to 9 critical.

Participants will be given 2 weeks to respond, with reminders sent after 1 week.

The number of participants scoring each outcome, the distribution of scores, and attrition will be analysed, summarised and a document drawn up.

Any drop-outs will not be included in subsequent rounds. Any outcomes already achieving consensus (see below) will not be included in subsequent rounds.

Delphi round 2 Participants will be provided with the summary of results from the previous round and asked to score the list of remaining outcomes again (i.e. those that have not achieved consensus).

Participants will be given 2 weeks to respond, with reminders sent after 1 week.

Results will be summarised again as for round 1.

Delphi round 3 If a third round is required (i.e. if consensus is not already achieved) participants will be provided with the summary of results from the previous round and asked to score any remaining outcomes.

Results will be summarised again as for previous rounds.

Consensus Consensus will be defined as suggested by Williamson et al (67): inclusion of an outcome in the core set if at least 70% of the participants scored it as 7 to 9, with less than 15% scoring it as 1 to 3; exclusion of an outcome in the core set if at least 70% of the participants scored it as 1 to 3, with less than 15% scoring it as 7 to 9; any other distribution of scores will be interpreted as lack of agreement for inclusion in the core outcome set, and therefore will be excluded.

Appendix 1 Regarding the focus group, participants will be upset/distressed by their condition/past condition (i.e. a life threatening illness) but we do not believe that our study/questions will cause further distress with it being a discussion about a side-effect of treatment. If

555 distress occurs, participants will be asked if they are ok and if they want to continue. Two adjacent meeting rooms will be booked so that one facilitator can take them next door to recover, if necessary. The other facilitator will pause the meeting and wait with the other participants. The distressed participant will be asked if they wish to be referred to counselling services provided by staff at The Christie or/and MacMillan. At all times participants will be free to leave, or even to sit and listen rather than take part.

Macmillan helpline: 0808 808 0000

The Christie (self-referral): 0161 446 3681/8004

556