The PREPARE Trial Guide for Planning Clinical Research to Avoid R

BJSM Online First, published on September 7, 2017 as 10.1136/bjsports-2017-097527 Education reviews Br J Sports Med: first published as 10.1136/bjsports-2017-097527 on 7 September 2017. Downloaded from

It is the aim of the PREPARE Trial guide Preparing for what the reporting to provide information to assist in the preparation and planning of clinical trial checklists will not tell you: the PREPARE research to enhance research quality and reduce avoidable research waste. The main Trial guide for planning clinical research focus is the randomised controlled trial (hereafter ‘trial’), but the guide is appli- to avoid research waste cable to other study types. To ease inter- pretation, practical examples are given 3 1,2,3 4 1,5 from the paper by Kise and colleagues. Thomas Bandholm, Robin Christensen, Kristian Thorborg, Where examples are not provided, it is 6 4,7 Shaun Treweek, Marius Henriksen the intention that the Kise paper—along with this guide—is to be used for, for example, journal club discussions of trial concepts. This narrative guide will need for this waste, but we believe that at least Introduction to be updated over time in order to keep some trials are discontinued because For the early career researcher or PhD its relevance. Hence, we encourage all of poor trial planning. Similarly, many student about to have a leading role in a readers who have constructive feedback to completed and published trials suffer clinical trial, it would be a tremendous use the PubMed Commons forum (http:// help to have a ‘How to’ guide, consisting from fundamental flaws or poor conduct www.ncbi.nlm.nih.gov/pubmedcom- of advice and key literature to help the due to lack of preparation and careful mons/) to share opinions and information researcher move successfully along the planning. Some flaws in the conception or (eg, published key literature that postdates initial trial pathway from research ques- planning phase prior to the finalised trial the PREPARE guide publication or links tion to protocol completion. While the protocol and registration are difficult to that are no longer functioning) relating ideal trial team should include experienced rectify once the trial starts. For example, to the guide, as this will then be linked to trialists, who participate in the prepara- patients may decline participation if the the guide’s PubMed entry. In this way, we tory research meetings, this support is study involves delaying or withholding will be able to take these comments into not always available. The PREPARE Trial treatment, or if the experimental condi- account in future updates. guide is intended to assist the early career tion is deemed unfeasible by patients. This researcher in this situation, but not replace can be avoided, for example, by involving an experienced trial team and regular patients in the planning of the trial or by The trial pathway research meetings. The guide may also conducting pilot or feasibility studies prior As an outline of the trial pathway, we serve as preparation for research meetings, to the full-scale trial. In many cases, the recommend using the list of headline trial or part of curricula for courses on research planning of a trial is as least as important processes that was recently published as 4 methodology, to be used by experienced as running or analysing it. part of a report from the first Trial Forge PhD supervisor, trialist or lecturer. The PREPARE Trial guide provides platform (www.trialforge.org) meeting Evidence from randomised controlled information to assist in the preparation (please see below for more information on 1 2 trials (RCTs) is often wasted (for a and planning of clinical trial research. the Trial Forge platform). The list provides http://bjsm.bmj.com/ collection of papers relating to this issue, The guide aims to reduce the number of the trial nomenclature to facilitate trial please see http://researchwaste.net/). circumstances in a trial where students discussions. Many trials are not published, and hence and supervisors need to compensate for the results are not reported, which is situations not considered during trial Research question, hypothesis indeed a waste. There are many reasons planning. The trial registration policy of and objective the International Committee of Medical Most trials start with a clinical problem, 1 Department of Occupational and Physical Therapy, Journal Editors (ICMJE: http://www. uncertainty of treatment approach or frus- on September 30, 2021 by guest. Protected copyright. Physical Medicine and Rehabilitation Research – icmje.org/) recommends that all medical Copenhagen (PMR-C), Amager-Hvidovre Hospital, tration with clinical outcomes from routine journal editors require registration of 5 University of Copenhagen, Copenhagen, Denmark practice. The ability of a trial designed 2Department of Orthopedic Surgery, Amager-Hvidovre clinical trials in a public trials registry at to address a clinical problem to become Hospital, University of Copenhagen, Copenhagen, or before the time of first patient enrol- a clinical ‘game changer’ depends on the Denmark 3 ment as a condition of consideration for ability of the researchers and clinicians Clinical Research Centre, Amager-Hvidovre Hospital, publication. Therefore, key decisions University of Copenhagen, Copenhagen, Denmark to turn the clinical problem into a good 4The Parker Institute, Copenhagen University Hospital relating to the methodological quality of research question that is relevant to the Bispebjerg-Frederiksberg, Copenhagen, Denmark the reported trial need to be made prior stakeholders: patients, clinicians, decision 5 Department of Orthopedic Surgery, Sports Orthopedic to protocol finalisation and trial registra- makers and others. Getting the research Research Centre – Copenhagen (SORC-C), Amager- tion. After this time, most trial changes Hvidovre Hospital, University of Copenhagen, question right at this stage is imperative, Copenhagen, Denmark have to be reported transparently in the as the question will guide the trial design, 6Health Services Research Unit, University of Aberdeen, trial report, and may ultimately limit the methods and analytic strategies.5 Aberdeen, UK study’s conclusions. For example, if the 7 To help frame your research question, we Department of Physical and Occupational Therapy, choice of primary outcome for the trial suggest you use the PICOT approach5 in Copenhagen University Hospital Bispebjerg- 6 Frederiksberg, Copenhagen, Denmark was rushed due to deadlines and has to conjunction with the FINER criteria. The be changed after trial commencement, PICOT approach requires that you specify Correspondence to Dr Thomas Bandholm, the sample size may not be reassessed or Copenhagen University Hospital, Hvidovre, Clinical the target Population (P), the Intervention Research Centre, Copenhagen DK-2650, Denmark; justified for the new primary outcome, of interest (I), the Comparator intervention thomas .quaade.bandholm@ regionh.dk limiting the study’s conclusions. (C), key Outcomes (O) and the Time frame

Bandholm T, et al. Br J Sports Med Month 2017 Vol 0 No 0 1 Copyright Article author (or their employer) 2017. Produced by BMJ Publishing Group Ltd under licence. Education reviews Br J Sports Med: first published as 10.1136/bjsports-2017-097527 on 7 September 2017. Downloaded from

Feasibility or pilot studies Feasibility or pilot studies may be worth considering prior to a full-scale trial. They are smaller studies that precede a full- scale trial in order to assess the feasibility of important aspects of a full-scale trial (eg, inclusion rate or intervention accept- ability). It could be valuable to assess the feasibility of randomising patients to exercise therapy or arthroscopic surgery before starting the full-scale trial. In this way, it will become clear if patients actually accept being randomised to either of the two treatments. Again, this is an example of an issue where patient involvement in the design phase may help qualify the trial. In many cases, data from a pilot or feasibility study will help to attract Figure 1 The PICOT criteria to frame the research question (top) and an example of a research funding for the full-scale trial. This will objective using the PICOT approach (bottom). Modified with permission from Farrugia et al.7 help to demonstrate the chances of success and how worthwhile an investment the project is for funders. (T) over which the outcomes are assessed5 meniscal tear (when you believe that there Pilot or feasibility studies have received (figure 1). This approach will help you is), as this enables statistical testing. After increasing attention in recent years— formulate a research question that can be you have formulated your null hypothesis, evidenced by the birth of the scientific rephrased in the reader’s own words.5 This you can formulate an alternate hypoth- journal ‘Pilot and Feasibility Studies’ for approach will also help you consider design esis stating the nature or direction of any the publication of stand-alone pilot or details like initial piloting, relevant primary difference, for example, exercise therapy 9 outcome/endpoint and choice of compar- is superior to arthroscopic surgery.7 If feasibility studies. A Consolidated Stan- ator. If, for example, you plan to run a trial your subsequent trial findings are statisti- dards of Reporting Trials (CONSORT) extension for reporting of randomised investigating the effectiveness of exercise cally significant—and of a clinically rele- 10 therapy to treat a degenerative meniscal vant magnitude—you can reject the null pilot and feasibility studies also exists, tear, you will be guided to consider if a gold hypothesis and accept the alternate. When and is freely available from the Enhancing standard outcome measure exists for this the hypothesis has been formulated, you the Quality and Transparency of Health condition, and if there is an established gold can proceed to the primary research objec- Research (EQUATOR) website (http:// standard treatment or usual care interven- tive, which defines the specific aims of www.equator-network.org/). If you tion to be used as the comparator (control) your trial. The objective could be formu- consider undertaking a stand-alone pilot intervention. You will also be guided to lated as illustrated in figure 1. or feasibility study, we suggest you use http://bjsm.bmj.com/ consider if your question relates to whether Please see the following papers for the Standard Protocol Items: Recommen- the exercise therapy intervention is ‘better more information on how to develop your dations for Interventional Trials (SPIRIT) than’, ‘as good as’ or ‘no worse than’ the research question, hypothesis and objec- checklist (please see ‘Writing the trial established gold standard. These topics are tive using the PICOT approach and the protocol’ below) and register the study explained in more detail in the following. FINER criteria: refs 5 7 8. (please see ‘Trial registration’ below). The FINER criteria (Feasible, Interesting,

Novel, Ethical and Relevant) will help you on September 30, 2021 by guest. Protected copyright. define the desirable properties of your research question once formulated, and give some suggestions concerning how to achieve each attribute6 (figure 2). This includes having stakeholder input from, for example, patients already at this time to help qualify the research question and ensure its clinical relevance. Patient involvement is explained in more detail in the following. Once you have developed a good research question, it is time to develop the research hypothesis. The hypothesis will be based on your research question in a form that allows for statistical testing.7 The hypothesis should be formulated as a ‘null’ hypothesis, for example, no difference between exercise therapy and arthroscopic Figure 2 The FINER criteria for a good research question. Modified with permission partial meniscectomy for degenerative from Farrugia et al.7

2 Bandholm T, et al. Br J Sports Med Month 2017 Vol 0 No 0 Education reviews Br J Sports Med: first published as 10.1136/bjsports-2017-097527 on 7 September 2017. Downloaded from

The following papers provide more The following web resources provide objective. The two-group parallel design is detailed information on pilot and feasi- more detailed information on the evidence the most common type, in which an inter- bility studies: refs 9 11 12. base for trial decisions: www.trialforge. vention is measured against a comparator org, www.precis-2.org. (eg, placebo), or two active interventions Writing the trial protocol are compared. This design is often used, as the statistical inference (the process of The SPIRIT checklist as the backbone of Consult the reporting checklist when drawing conclusions about populations the overall work frame planning the trial or scientific truths from a data sample) is The trial protocol is a detailed descrip- Reporting checklists are evidence-based clearer when associated with a contrast tion of what you intend to do for the next recommendations on minimum reporting between only two groups. couple of years, so it is a piece of work standards and are mandatory require- The framework of your trial is also that you will want to get right. ments for trial reports in many scientific important to consider. This relates to Besides having experienced trialists journals. It is therefore nice to know in your research question. Are you aiming on your trial team, we suggest you use a the planning phase what the minimum to assess if an intervention is better than combination of the SPIRIT checklist for reporting requirement is, when you finally (superiority trial), is at least not worse trial protocols,13 the CONSORT checklist write the trial report. The SPIRIT checklist than (non-inferiority trial) or whether it for reporting trials14 and the introduc- is intended mainly for preparing protocols is equally effective as (equivalence trial) a tion-methods-results-and-discussion struc- for RCTs. It is intended to facilitate the comparator? These are important consid- ture for a scientific paper15 as the overall subsequent reporting of the trial using the erations, particularly for the sample size work frame. Consider using the SPIRIT CONSORT reporting checklist. However, estimation and for the conclusion that checklist as the backbone. We suggest you different types of RCTs exist, for which can be drawn. The trial designed to show also consult the International Council specific CONSORT extensions have been superiority is the most common, and for Harmonisation of Technical Require- created (for an introduction to superiority, if there is a statistically significant and ments for Pharmaceuticals for Human equivalence and non-inferiority trials, clinically important difference between Use (ICH) efficacy guideline concerning please see refs 19 and 20). So we suggest groups (in the hypothesised direction) the design, conduct, safety and reporting you consult the relevant CONSORT you can conclude that the intervention of clinical trials. Although the ICH guide- extension, which can be found at the of interest is superior to the comparator. line is mandatory for drug trials, it holds EQUATOR network’s website (http:// However, in many cases, trials do not excellent information in the planning of www.equator-network.org/). confirm the prespecified hypothesis of a non-drug trial. The guideline is freely The following web resources provide superiority because the results turn out available for download from the ICH more detailed information on reporting statistically non-significant. Hence, you website (http://www.ich.org/home.html). checklists: http://www.consort-statement. cannot conclude better performance or Likewise, the SPIRIT and CONSORT org/, http://www.equator-network.org/. superiority of the intervention in relation checklists are freely available for down- to the comparator. Does that then mean load from the EQUATOR network’s Study background and choice of that the intervention and comparator are website (www.equator-network.org). The comparator equally good? The answer is ‘No’. The checklists come with explanatory papers When you argue in your protocol the absence of evidence of superiority is not that elaborate in-depth on the specific http://bjsm.bmj.com/ need for undertaking the trial, the SPIRIT evidence of equivalence or non-inferiority. checklist items and give you examples of checklist requires you to summarise rele- There will always be some small differ- good reporting.16 17 We have added infor- vant studies and so does the CONSORT ence in the estimated effects that should mation below concerning some of the 14 reporting checklist. Both checklists be confirmed in a new trial with a (new) items in the checklists, which we think strongly suggest that an up-to-date prespecified hypothesis of equivalence may be useful to consult in addition to the systematic review of relevant studies be and an appropriate sample size estima- explanatory papers. summarised and cited in both the protocol 13 14 tion, which makes it possible to confirm 19 and paper. By doing so already in small differences as true. In contrast, on September 30, 2021 by guest. Protected copyright. The Trial Forge platform (www. the planning phase of your trial, you will if your non-inferiority trial successfully trialforge.org) provides an evidence reduce the risk of unnecessary research 21 documents that two interventions are not base for trial decisions duplication and waste. It is also likely to inferior to each other, it is possible to test The Trial Forge platform aims to increase help you choose your study comparator for superiority of one of the interventions the evidence base for trial decision making and indicate how likely your intervention on the same data, that is, a new trial may and efficiency.4 The initiative will collate is to work. not be necessary.19 or guide to existing high-quality evidence The following web resource provides You should also consider whether your on key trial processes, such as how to more detailed information on system- research question and the results you are tailor recruitment strategies to particular atic reviews with regard to fully util- expecting are confirmatory or explor- contexts or how to effectively train trial ising previous research and making atory.23 This has implications for the staff. You will have to decide on these your research evidence-based: www. nature and strength of the conclusion you issues and many more before the start of ebrnetwork.org; and the following paper can draw. Confirmatory research tests the trial, and you will need to describe provides more information on how predefined hypotheses, usually derived them in your trial protocol. to select a control intervention for a from a theory or the results of previous A key Trial Forge tool to support trial trial: ref22. studies that can be used to draw firm planning is PRECIS-2,18 which is a graph- and often meaningful conclusions. This ical tool to help you match design deci- Trial design is because the researcher knows exactly sions to the needs of the people you expect The design of your trial relates very closely what to look for and is confident in where to use the trial results. to your research question, hypothesis and to look for it. In contrast, exploratory

Bandholm T, et al. Br J Sports Med Month 2017 Vol 0 No 0 3 Education reviews Br J Sports Med: first published as 10.1136/bjsports-2017-097527 on 7 September 2017. Downloaded from research is less specific because the very important. Without a complete and researchers, in which we list issues that researcher may not know what to look for thorough description, trial personnel will we feel are important for researchers or where to look and typically generate have difficulty in delivering the interven- to consider carefully before they start hypotheses post hoc rather than test a tion during the trial. When the trial report including participants into a clinical trial. predefined hypothesis. Thus, exploratory becomes available, a poor description will The overall aim of these considerations research has fewer predefined hypotheses, mean that others may find the interven- is to increase research quality to benefit if any, due to a lack of theory or previous tion difficult to implement in clinical work both patients and society. Are there issues research. Exploratory research therefore or replicate for research. Recently, the that you – as a trial participant – feel are entails an increased risk of finding false Template for Intervention Description and important for researchers to consider associations, making the conclusions Replication (TIDieR) checklist was devel- when they plan a clinical trial?’ drawn correspondingly weak. A friendly oped to facilitate a complete and thorough To summarise the feedback, the patients note of caution should be made in associa- generic description of interventions.31 stated that the following was important tion with exploratory research. HARKing The TIDieR checklist is an extension of to consider: (1) what patients may gain (Hypothesising After the Results are the CONSORT checklist and the SPIRIT from the trial findings; (2) what the time Known)24 is a natural part of exploratory guideline when dealing with check- requirements are for participants, prefer- research, in which you subsequently test list items of intervention descriptions. ably with an illustration; (3) if participa- the HARKed hypothesis as a predefined Complex interventions (such as rehabilitation is compatible with normal working hypothesis in a new confirmatory trial. tion or exercise interventions) are particu- hours; (4) what is the plan for dissemi- You should be cautious not to report your larly prone to incomplete reporting.31 32 As nating the trial results to the participants; exploratory research as confirmatory. suggested previously, it is recommended to (5) that the researchers are friendly and A further important categorisation is consult this reporting checklist during the helpful; and (6) that the researchers are the difference between a pragmatic and planning phase of the trial, so that you will familiar with the condition. This infor- an explanatory trial. In the latter, many be able to make protocol choices accord- mation would not have been part of the conditions surrounding the trial will be ingly. As the TIDieR checklist is generic, PREPARE Trial guide had we not inter- controlled in order to better understand it is possible that you will need to supple- viewed trial participants. the outcome(s). In contrast, the prag- ment the intervention description with The following papers provide more matic trial aims at describing the effects research area-specific information. As detailed information on patient involve- of the intervention under conditions that with the CONSORT checklist, the TIDieR ment in research refs 33 34 reflect real-life or clinical practice as much checklist also comes with an informative as possible. The pragmatic trial design explanatory paper.31 Bias is necessary in certain analyses, such as The following web resources provide Bias is something that you do not want cost-effectiveness analyses. more detailed information on the in your trial but can be challenging to The following papers provide more TIDieR checklist: www.consort-state- avoid. It refers to systematic error in detailed information on different types of ment.org, www.equator-network.org. results or inferences, that is, a deviation trial designs: refs 19 25 26. from the truth. Biases can vary in magnitude: some are small and trivial compared Patient involvement

Eligibility criteria and recruitment with the observed effect, and some are so http://bjsm.bmj.com/ Collaboration between patients and Specific research areas may have detailed substantial that an observed effect may be healthcare professionals (including recommendations for aspects of the entirely due to bias. It is usually impos- researchers) seems to be an appropriate protocol dealing with recruitment and sible to know to what extent biases have means to capture patient perspective33 enrolment (eg, diagnostic criteria). We affected the results of a particular study, and helps prevent a potential mismatch suggest you do a systematic search within although there is good empirical evidence between patient preferences and the scien- your specific research area. In general, that certain flaws in the design, conduct tific focus.34 There are many reasons why care is required when selecting eligibility and analysis of randomised clinical trials it is important to consider patient involve- on September 30, 2021 by guest. Protected copyright. criteria as many trials exclude partici- lead to bias.35–37 We suggest you plan your ment at the planning phase. To highlight pants who are potential candidates for the efforts to reduce bias, as if your trial was one, the relevance of your research will treatment under evaluation, which raises to be included in a systematic review, and likely increase. For example, patients will questions about external validity.27 28 accordingly assessed for risk of bias using be able to provide you with valuable infor- Regarding recruitment, Trial Forge (http:// the Cochrane Collaboration’s risk of bias mation on the symptoms they consider www.trialforge.org/) is currently preparing tool.38 most disabling, which can help you choose a Trinity Package, which is a comprehen- The following papers provide more the primary outcome for your trial. When sive summary of what is known about trial detailed information on identifying and confronted with an outline of the time recruitment.29 30 avoiding bias in research: refs 38 39. requirements for an intervention or the The following paper and web resource trial measurements schedule, patients provide more detailed information on will also be able to give you an idea, as to Sequence generation and allocation methods to improve recruitment to whether this intervention is feasible in a concealment trials: ref 29 and http://www.trialforge. real-life setting. Sequence generation and allocation org/. In order to collaborate with patients concealment together cover a domain when writing the PREPARE Trial guide, called selection bias.38 It refers to system- Interventions we asked the following question to two atic differences between baseline charac- Because the evaluation of interventions is patients who had just participated in a teristics of the groups that are compared. often the primary objective in a trial, the clinical trial: ‘We – as researchers – are Successful randomisation prevents quality of the intervention description is preparing a ‘how-to’ guide for other selection bias in allocating participants

4 Bandholm T, et al. Br J Sports Med Month 2017 Vol 0 No 0 Education reviews Br J Sports Med: first published as 10.1136/bjsports-2017-097527 on 7 September 2017. Downloaded from to interventions. Its success depends outcome set exists. Core outcome sets nl/). The COSMIN initiative has devel- on several interrelated processes. You represent the minimum that should be oped a critical appraisal tool (a checklist) must specify a rule for allocating partic- measured and reported in a clinical trial of containing standards for evaluating the ipants to interventions, based on some a specific patient population. A good place methodological quality of studies on the chance (random) process. We call this to start is the COMET (Core Outcome measurement properties of health-related sequence generation, the sequence being, Measures in Effectiveness Trials) Initiative outcome measures. When a possible rele- for example, a list of random 0s and and database website (http://www.comet- vant outcome measure has been identi- 1s, by which participants are assigned initiative.org/). The COMET Initiative fied, it is a good idea to pilot the outcome to the groups of the trial. Protecting or database is a repository of core outcome measure first. Piloting on your outcome concealing this sequence until partici- sets and can serve as an inspiration. It measure is just as important as piloting on pants have in fact been allocated to groups can be tempting to add a large number your intervention or inclusion procedures. is referred to as allocation (sequence) of secondary, supportive outcomes, but In order to master any test or outcome concealment. This is done so that involved please remember that the primary objec- measure, actual experience in performing trial personnel or participants are kept tive and sample size estimation pertain to the measuring/testing in the correct way, unaware of the upcoming allocation.40 the primary outcome only. Adding a high and knowing how your outcome measure For example, the person who recruits number of secondary outcomes will likely reacts in different situations, will often give trial participants cannot introduce bias by make the trial less efficient.45 you great advantage. Unnecessary ‘noise’ unintentionally recruiting patients with The outcome measure needs to from your measurement can hopefully be the perceived greatest needs every time measure what it is intended to (accuracy) avoided or reduced so that true between- he or she knows that the next participant do so in a precise manner (reliability), group differences can be detected. will be allocated to the active treatment. If and thereby measure when relevant The following papers provide more this bias was present, the treatment effect changes have occurred in the partici- detailed information on validity, reliability would most likely be overestimated.41 pants. When deciding on primary and and responsiveness of health-related The following papers provide more secondary outcomes, it is important to outcome measures: refs 47 48. detailed information on the types consider whether the outcome is either and methods of generating allocation a clinically meaningful outcome that sequences, as well as the principles of measures directly how patients feel, func- Sample size estimation stratified randomisation for randomised tion or survive, or a surrogate measure. Estimating the number of participants to be trials: refs 42 43 Surrogate measures do not measure these included in your trial can be troublesome parameters in a direct way, although they if you do not understand the fundamental may provide a causal pathway with a principles. Consulting a biostatistician is Blinding clinically meaningful outcome. Surrogate highly recommended, but it is a good idea Blinding to minimise the risk of bias is a measures can, however, be associated to be able to do the ground work yourself. cornerstone of intervention evaluation.44 with several problems; an interven- The sample size estimation should be made It is an attempt to hide the allocation tion that produces a positive surrogate for your defined primary outcome to esti- sequence from involved trial personnel response can have no or even a harmful mate how many participants you have to or participants following allocation to effect on the clinical outcome, and at include for the trial to be large enough to

trial arms. It can be difficult to achieve in http://bjsm.bmj.com/ worst lead to the implementation of provide a reliable answer in terms of the some types of trials, such as trials evalu- harmful treatments.46 primary hypothesis and outcome. If for ating non-pharmacological interventions. Finally, you should consider the some reason the sample size is estimated So, when you plan your trial, it is rele- fact that you will need to report on all on some other basis, we suggest you make vant to know of the different types of prespecified outcomes in your trial, and this clear and provide a justification for blinding that exist to blind participants, report transparently if you want to report this approach.49 intervention providers or outcome asses- new outcomes that were added after the In order to estimate and subsequently sors. These include, for example, hypoth- trial began. We suggest you carefully report your sample size sufficiently so on September 30, 2021 by guest. Protected copyright. esis-blinding, sham procedures and similar plan and report your trial outcomes, as that others can replicate it, you will need attention-control interventions. You will if your trial intent and report were to to specify several items, depending on be asked to describe the type of blinding be investigated by the COMPare trials your trial design. For a classic superiority (if used) when you eventually report your project (http://compare-trials.org/). The trial with an intervention and control trial, in accordance with the CONSORT COMPare team systematically checked (comparator) group, these include the checklist. every trial published in the top 5 medical primary outcome (including whether it is The following papers provide more journals between October 2015 and a binary or continuous outcome); the test detailed information on blinding methods January 2016, with the purpose of statistic (for example t-test or χ2 statis- and how to report them: refs 42 44 searching for misreported findings and tics); the minimal clinically important outcome switching. This team’s effort difference in response between groups, Outcomes revealed a large degree of inconsistency that is, the smallest effect of your inter- Choosing the most relevant primary in outcome reporting. vention worth detecting; the probability outcome for your trial needs careful With respect to the issue of selecting of erroneously rejecting the null hypoth- consideration. The primary outcome outcome measures of good methodolog- esis or, in other words, the probability should reflect the needs of the end user ical quality for a specific patient popula- of a false-positive conclusion (‘α’, the (eg, patient) and have sound psychometric tion, a fine resource to guide you is the type I error, commonly set at 0.05); and properties (ie, be reliable and valid). Patient Consensus-based Standards for the Selec- the probability of erroneously failing to involvement in the planning process can tion of Health Measurement Instruments reject the null hypothesis or, in other assist in this process, especially if no core (COSMIN) website (http://www.cosmin. words, the probability of a false-negative

Bandholm T, et al. Br J Sports Med Month 2017 Vol 0 No 0 5 Education reviews Br J Sports Med: first published as 10.1136/bjsports-2017-097527 on 7 September 2017. Downloaded from conclusion (‘β’, the type II error, treatment effects that are to be estimated When you eventually get to report the commonly set to 0.20 (statistical power in order to satisfy the primary objec- statistical analyses in the trial report, of 80)). Hence, an α and β of 0.05 and tive of the trial. The statistical methods descriptive or summary statistics are 0.20, respectively, means that you desire used to accomplish these tasks should be very important.14 Because the statis- less than 5% risk of a type I error and described for the primary (and preferably tical analysis plan holds information on less than 20% risk of a type II error for the key secondary) outcome variables, and how to report descriptive statistics, it is your trial. the underlying statistical model should be important to plan this before the trial The following papers provide more made clear. starts. The CONSORT explanation and detailed information on the principles of We recommend that you clearly distin- elaboration paper is a great resource for sample size estimations and how to report guish between the primary analysis this.17 It presents examples of tables of them: Refs50–52. of the primary outcome variable and results as well as, for example, informa- supporting secondary analyses of the tion on how and when to use SD, 95% CIs primary or secondary outcome variables. and p values. Please note that CONSORT Statistical analysis plan For example, you may have decided that emphasises contrasts between groups, As stated above for the sample size esti- your primary analysis for the primary also known as effect sizes. It is therefore mation, consulting a biostatistician is outcome variable is to be on the full much more important for you to plan highly recommended when you write analysis data set (intention to treat-pop- a clear presentation of between-group your statistical analysis plan, but it is a ulation), using multiple imputations changes (contrasts) than within-group good idea to be able to do the ground for participants with missing data and changes. work yourself. An important part of the without analysis adjustments. You may The following papers provide more trial protocol is the plan for the statis- then want to supplement this primary detailed information on how to develop tical analyses that you will run on your analysis, which the reader will trust and write your statistical analysis data. It can be provided in detail in the the most, with a supportive secondary plan: refs 17 49 53 55 57–59. trial protocol, but it can also be in the analysis on the per-protocol data set, form of a separate document named in which participants are excluded if ‘Statistical analysis plan’ that gives more they do not meet a predefined minimal Trial registration details than that expected of a trial treatment exposure. If in this secondary A trial registry is a web-based resource protocol. Whichever format is chosen, analysis of the primary outcome you also for patients, relatives, healthcare profes- the main thing is that you have to plan intend to use the baseline data to adjust sionals and the public that provides access in detail, the statistical analyses for your the treatment estimates for potential to information on public and privately trial data, before the trial starts, or at the baseline differences, this can then also be supported trials. Essentially, the registry latest make minor last minute changes specified. information is a summary of your trial before breaking the trial blind. You also Different approaches exist by which protocol that becomes publicly accessible. have the possibility of publishing your you can create a full analysis data set for At the same time, the registry assigns a statistical analysis plan, either as part of intention-to-treat-analyses. The prin- unique identifier (trial number) to the a protocol or as a stand-alone document. ciple implies that the analysis includes trial, which can be used to track the trial Many journals now publish trial proto- all randomised participants. Compliance and trial-related information. More and cols and some publish statistical analysis with this principle would necessitate more journals require that clinical trials http://bjsm.bmj.com/ plans as stand-alone documents. complete follow-up of all randomised have been preregistered, that is, registered As a frame for developing and writing participants so that no outcome data in a trial registry before enrolment of your statistical analysis plan, we suggest were missing. This is difficult to achieve, the first participant. This is also the trial you use the ICH-E9 guideline.49 The as some trial participants are almost registration policy of the ICMJE (please guideline is freely available for down- always lost to follow-up because they see the ICMJE website for more details: load from the ICH website (http://www. move, become sick or do not show up http://www.icmje.org/). We suggest you ich.org/home.html). In addition, the as scheduled. The term ‘Full analysis set’ use a copy/paste approach for trial regis- on September 30, 2021 by guest. Protected copyright. Statistical Analyses and Methods in the is used to describe the analysis data set, tration activities, so that the trial informa- Published Literature (SAMPL) guide- which is as complete as possible, and as tion is identical between the trial protocol lines53 will also be good to consult in the close as possible, to the intention-to-treat and the registry. If you are in doubt as to planning phase. The SAMPL guideline is population. To create a full analysis set, whether your trial is a clinical trial and freely available for download from the you will have to replace missing values needs registration, we suggest you use the EQUATOR network´s homepage (http:// with substituted values—a process called yes/no decision tree in the Hinman et al www.equator-network.org/). Finally, a imputation. Imputation techniques range paper,60 or even better, register the trial comprehensive guidance document for from carrying forward the last observa- regardless. The AllTrials campaign calls statistical analysis plan content is currently tion (data point) to the use of complex for registration of all types of trials (http:// being developed.54 mathematical models.55 The main thing www.alltrials.net/). The statistical analysis plan for your trial is that you plan and describe how you Some trial registries have adopted an should include all the principal features intend to handle missing data in your element to its platform to collect data- of the proposed confirmatory analysis trial, and that you justify the approach sharing plans, which is also an item in the for the primary outcome variable(s) and chosen. The same can be said if you plan SPIRIT checklist.13 It is now the recom- the way you intend to handle anticipated to run subgroup (ie, stratified) analyses mendation of the ICMJE that authors analysis problems, such as missing data. or adjust for the influence of covari- choose a registry that includes a data- For a confirmatory analysis, the statis- ates, as these procedures will most likely sharing plan element.61 Although it may tical analysis plan should specify the impact your sample size and increase the not become reality,62 the ICMJE recently hypotheses that are to be tested and the risk of bias.56 proposed that trial report authors share

6 Bandholm T, et al. Br J Sports Med Month 2017 Vol 0 No 0 Education reviews Br J Sports Med: first published as 10.1136/bjsports-2017-097527 on 7 September 2017. Downloaded from

Funding This research was supported in part by a donation from the Oak Foundation to The Parker Institute, Bispebjerg – Frederiksberg Hospital, Copenhagen (RC and MH). Competing interests RC reports non-financial support from board membership, grants from consultancy (AbbVie, Amgen, Axellus A/S, Bristol- Myers Squibb, Cambridge Weight Plan, Celgene, Eli Lilly, Hospira, MSD, Norpharma, Novartis, Orkla Health, Pfizer, Roche, Sobi, Takeda), personal fees from employment (Research Unit for Musculoskeletal Function and Physiotherapy, Institute of Sports Science and Clinical Biomechanics, University of Southern Denmark), non-financial support from expert testimony, grants from grants/grants pending (Axellus A/S, AbbVie, Cambridge Weight Plan, Janssen, MSD, Mundipharma, Novartis and Roche), grants from payment for lectures including service on speakers bureaus (Abbott, Amgen, Axellus, Bayer HealthCare Pharmaceuticals, Biogen Idec, Bristol-Myers Squibb, Cambridge Weight Plan, Ipsen, Janssen, Laboratoires Expanscience, MSD, Mundipharma, Norpharma, Novartis, Pfizer, Figure 3 Summary of key issues to consider when you prepare your trial. COMET, Core Outcome Roche, Rottapharm-Madaus, Sobi and Wyeth), grants Measures in Effectiveness Trials; COSMIN, Consensus-based Standards for the Selection of Health from payment for manuscript preparation (Axellus, Measurement Instruments; PICOT, Population, Intervention, Comparison group, Outcome of Bristol-Myers Squibb, and Cambridge Weight Plan, interest, Time; SPIRIT, Standard Protocol Items: Recommendations for Interventional Trials; TIDieR, Aleris-Hamlet (via Norpharma)), non-financial support from patents (planned, pending or issued), Template for Intervention Description and Replication. non-financial support from royalties, grants from payment for development of educational presentations (Bristol-Myers Squibb, MSD, Pfizer), non-financial the de-identified individual patient data the trial report or at a permanent online support from stock/stock options, grants from travel/ accommodations/meeting expenses unrelated to underlying the results presented in the repository. activities listed (Abbott, AbbVie, Axellus, Bristol-Myers article no later than 6 months after publi- The following papers provide more Squibb, Cambridge Weight Plan, Celgene, Laboratoires cation to enhance trial transparency.61 detailed information on publication of Expanscience, Norpharma, Novartis, Pfizer, Roche, Nevertheless, it is important for you to trial protocols: refs 65 66. Rottapharm-Madaus and Wyeth), non-financial support develop and register a data-sharing plan, from other (err on the side of full disclosure), outside the submitted work; RC is involved in many healthcare which, at the same time, complies with Summary initiatives and research that could benefit from national regulations. The preparatory steps in the trial pathway wide uptake of this publication (including Cochrane, The following web resource and papers are very important for a good trial design OMERACT, IDEOM, RADS and the GRADE Working provide more detailed information on Group). The Musculoskeletal Statistics Unit, The Parker and may increase the chance that your Institute, is grateful for the financial support received registration and reporting of trial results: research is not wasted at later stages. The

from public and private foundations, companies and http://bjsm.bmj.com/ http://www.alltrials.net/ and refs 61 63 64. main point of the PREPARE Trial guide private individuals over the years. The Parker Institute is to help you to prepare your research is supported by a core grant from the Oak Foundation; Publishing the trial protocol thoroughly, based on what you will the Oak Foundation is a group of philanthropic As preregistration of your trial with a organisations that, since its establishment in 1983, face at later steps in the trial pathway. has given grants to not-for-profit organisations around trial registry commonly holds a summary Figure 3 and online supplementary table the world. ST reports grants from Chief Scientist Office of the trial protocol only, it is possible 1 summarise the key issues that will help during the conduct of the study. The Health Services to publish the full protocol in a scientific you make informed trial decisions by care- Research Unit, University of Aberdeen, receives core journal. The general idea is that once a fully considering them in the preparatory funding from the Chief Scientist Office of the Scottish on September 30, 2021 by guest. Protected copyright. Government Health Directorates. MH reports grants trial protocol becomes public, there will phase. be people waiting for the results, which from the Oak Foundation during the conduct of the study. KT and TB report no conflicts of interest. in turn reduces the chance of non-publi- Acknowledgements We thank Mette Merete cation of the results.65 It also means that Pedersen, Postdoc, PhD, and Rasmus Skov Husted, Patient consent Obtained. readers can check for differences between Research Assistant, MSc, Clinical Research Centre, Provenance and peer review Not commissioned; Hvidovre Hospital, Copenhagen, for valuable comments externally peer reviewed. what was planned and what is reported to an early version of the manuscript, and Mikkel B in the final trial report. Publication of Klausen, PhD student, Metropolitan University College, Data sharing statement The manuscript does not the trial protocol also provides an oppor- Copenhagen, for practical help with patient interviews. contain any data. tunity to publish details that may not fit We also thank Jeanette Kirk, Postdoc, PhD, Clinical © Article author(s) (or their employer(s) unless into the final study report. Many journals Research Centre, Hvidovre Hospital, Copenhagen and otherwise stated in the text of the article) 2017. All Rasmus S Husted, PhD student, Clinical Research have word count and appendix restric- rights reserved. No commercial use is permitted unless Centre, Hvidovre Hospital, Copenhagen for providing otherwise expressly granted. tions, and if your intervention or other valuable stakeholder input and comments. Finally, we aspects of your trial methodology needs thank former trial participants Nadia W Houman and ►► Additional material is published online only. To view please visit the journal online (http://dx.doi.org/ elaboration, a published trial protocol Benny Frederiksen for coproducing the PREPARE Trial guide by providing valuable patient-specific input. 10.1136/ bjsports-2017- 097527). serves as an excellent reference for these details. If you choose not to publish your Contributors TB conceived the idea for the guide and drafted the first manuscript. TB, RC, KT, ST and MH all trial protocol as a stand-alone paper, we provided intellectual input and contributed to drafting suggest you consider making it publicly revisions of the manuscript. TB, RC, KT, ST and MH all available as supplementary material to read and approved the final version to be published.

Bandholm T, et al. Br J Sports Med Month 2017 Vol 0 No 0 7 Education reviews Br J Sports Med: first published as 10.1136/bjsports-2017-097527 on 7 September 2017. Downloaded from

To cite Bandholm T, Christensen R, Thorborg K, et al. 21 Lund H, Juhl C, Christensen R. Systematic reviews and 44 Boutron I, Guittet L, Estellat C, et al. Reporting Br J Sports Med Published Online First: [please include research waste. The Lancet 2016;387:123–4. methods of blinding in randomized trials assessing Day Month Year]. doi:10.1136/bjsports-2017-097527 22 Mann H, Djulbegovic B. Choosing a control nonpharmacological treatments. PLoS Med intervention for a randomised clinical trial. BMC Med Accepted 9 August 2017 2007;4:e61. Res Methodol 2003;3:7. 45 Getz KA, Stergiopoulos S, Marlborough M, et al. Br J Sports Med 2017;0:1–8. 23 Christensen R, Langberg H. Statistical principles for Quantifying the magnitude and cost of collecting doi:10.1136/bjsports-2017-097527 prospective study protocols:: design, analysis, and extraneous protocol data. Am J Ther 2015;22:117–24. reporting. Int J Sports Phys Ther 2012;7:504. 46 la Cour JL, Brok J, Gøtzsche PC. Inconsistent reporting 24 Kerr NL. HARKing: hypothesizing after the results are of surrogate outcomes in randomised clinical trials: References known. Pers Soc Psychol Rev 1998;2:196–217. cohort study. BMJ 2010;341:c3653. 1 Ioannidis JP. Clinical trials: what a waste. BMJ 25 Ford I, Norrie J, TrialsP. N Engl J Med 47 Mokkink LB, Terwee CB, Knol DL, et al. The COSMIN 2014;349:g7089. 2016;375:454–63. checklist for evaluating the methodological quality of 2 Chapman SJ, Shelton B, Mahmood H, et al. 26 Kaji AH, Lewis RJ. Noninferiority Trials: Is a new studies on measurement properties: a clarification of Discontinuation and non-publication of surgical treatment almost as effective as another? JAMA its content. BMC Med Res Methodol 2010;10:22. randomised controlled trials: observational study. BMJ 2015;313:2371–2. 48 Terwee CB, Mokkink LB, Knol DL, et al. Rating 2014;349:g6870. 27 Saunders C, Byrne CD, Guthrie B, et al. External the methodological quality in systematic reviews 3 Kise NJ, Risberg MA, Stensrud S, et al. Exercise validity of randomized controlled trials of glycaemic of studies on measurement properties: a scoring therapy versus arthroscopic partial meniscectomy for control and vascular disease: how representative are system for the COSMIN checklist. Qual Life Res degenerative meniscal tear in middle aged patients: participants? Diabet Med 2013;30:300–8. 2012;21:651–7. randomised controlled trial with two year follow-up. 28 Treweek S, Dryden R, McCowan C, et al. Do 49 Statistical Principles for Clinical Trials : ICH. http:// BMJ 2016;354:i3740. participants in adjuvant breast cancer trials reflect www.ich. org/products/ guidelines/efficacy/ efficacy- 4 Treweek S, Altman DG, Bower P, et al. Making the breast cancer patient population? Eur J Cancer single/article/ statistical-principles- for-clinical- trials.html randomised trials more efficient: report of the first 2015;51:907–14. (accessed 22 Aug 2016). meeting to discuss the Trial Forge platform. Trials 29 Treweek S, Lockhart P, Pitkethly M, et al. Methods to 50 Charles P, Giraudeau B, Dechartres A, et al. Reporting 2015;16:261. improve recruitment to randomised controlled trials: of sample size calculation in randomised controlled 5 Thabane L, Thomas T, Ye C, et al. Posing the research cochrane systematic review and meta-analysis. BMJ trials: review. BMJ 2009;338:b1732. question: not so simple. Canadian Journal of Open 2013;3:e002360. 51 Schulz KF, Grimes DA. Sample size calculations in Anesthesia 2009;56:71–9. 30 Gardner HR, Fraser C, MacLennan G, et al. A protocol randomised trials: mandatory and mystical. Lancet 6 Hulley SB, Cummings SR, Browner WS, et al. Designing for a systematic review of non-randomised evaluations 2005;365:1348–53. Clinical Research. Fourth edition. Philadelphia: LWW, of strategies to improve participant recruitment to 52 Cook JA, Hislop J, Altman DG, et al. Specifying 2013. randomised controlled trials. Syst Rev 2016;5:131. the target difference in the primary outcome for a 7 Farrugia P, Petrisor BA, Farrokhyar F, et al. 31 Hoffmann TC, Glasziou PP, Boutron I, et al. Better randomised controlled trial: guidance for researchers. Practical tips for surgical research: research reporting of interventions: template for intervention Trials 2015;16:12. questions, hypotheses and objectives. Can J Surg description and replication (TIDieR) checklist and 53 Lang TA, Altman DG. Basic statistical reporting 2010;53:278–81. guide. BMJ 2014;348:g1687–g1687–g1687. for articles published in biomedical journals: 8 Rios LP, Ye C, Thabane L. Association between framing 32 Abell B, Glasziou P, Hoffmann T. Reporting the “Statistical Analyses and Methods in the Published of the research question using the PICOT format and and replicating trials of exercise-based cardiac Literature” or the SAMPL Guidelines. Int J Nurs Stud reporting quality of randomized controlled trials. BMC rehabilitation: do we know what the researchers 2015;52:5–9. Med Res Methodol 2010;10:11. actually did? Circ Cardiovasc Qual Outcomes 54 Krishan A, Stocken D, Lewis S, et al. Development of 9 Lancaster GA. Pilot and feasibility studies come of 2015;8:187–94. guidance for statistical analysis plans (SAPs) for clinical age!. Pilot Feasibility Stud 2015;1:1. 33 Crowe S, Fenton M, Hall M, et al. Patients’, clinicians’ trials. Trials 2015;16:O41. 10 Eldridge SM, Chan CL, Campbell MJ, et al. CONSORT and the research communities’ priorities for treatment 55 Vickers AJ, Altman DG. Statistics notes: missing 2010 statement: extension to randomised pilot and research: there is an important mismatch. Res Involv outcomes in randomised trials. BMJ 2013;346:f3438. feasibility trials. BMJ 2016;355:i5239. Engagem 2015:1. 56 Sun X, Briel M, Busse JW, et al. Credibility of claims 11 Abbott JH. The distinction between randomized clinical 34 de Wit MP, Berlo SE, Aanerud GJ, et al. European of subgroup effects in randomised controlled trials: trials (RCTs) and preliminary feasibility and pilot league against rheumatism recommendations for http://bjsm.bmj.com/ systematic review. BMJ 2012;344:e1553. studies: what they are and are not. J Orthop Sports the inclusion of patient representatives in scientific 57 Vickers AJ, Altman DG. Statistics notes: analysing Phys Ther 2014;44:555–8. projects. Ann Rheum Dis 2011;70:722–6. controlled trials with baseline and follow up 12 Eldridge S, Bond C, Campbell M, et al. Defining 35 Savović J, Jones HE, Altman DG, et al. Influence of measurements. BMJ 2001;323:1123–4. feasibility and pilot studies in preparation for reported study design characteristics on intervention 58 Schulz KF, Grimes DA. Sample size slippages in randomised controlled trials: using consensus methods effect estimates from randomized, controlled trials. randomised trials: exclusions and the lost and and validation to develop a conceptual framework. Ann Intern Med 2012;157:429–38. wayward. Lancet 2002;359:781–5. Trials 2015;16:O87. 36 Schulz KF, Chalmers I, Hayes RJ, et al. Empirical 59 Burke JF, Sussman JB, Kent DM, et al. Three simple 13 Chan AW, Tetzlaff JM, Altman DG, et al. SPIRIT 2013 evidence of bias. Dimensions of methodological quality

rules to ensure reasonably credible subgroup analyses. on September 30, 2021 by guest. Protected copyright. statement: defining standard protocol items for clinical associated with estimates of treatment effects in trials. Ann Intern Med 2013;158:200–7. controlled trials. JAMA 1995;273:408–12. BMJ 2015;351:h5651. 14 Schulz KF, Altman DG, Moher D. CONSORT 2010 37 Wood L, Egger M, Gluud LL, et al. Empirical evidence 60 Hinman RS, Buchbinder R, Craik RL, et al. Is this a statement: updated guidelines for reporting of bias in treatment effect estimates in controlled clinical trial? and should it be registered? Phys Ther parallel group randomised trials. PLoS Med trials with different interventions and outcomes: meta- 2015;95:810–4. 2010;2010:e1000251. epidemiological study. BMJ 2008;336:601–5. 61 Taichman DB, Backus J, Baethge C, et al. Sharing 15 Sollaci LB, Pereira MG. The introduction, methods, 38 Higgins JP, Altman DG, Gøtzsche PC, et al. The clinical trial data: a proposal from the international results, and discussion (IMRAD) structure: a fifty-year Cochrane Collaboration’s tool for assessing risk of bias committee of medical journal editors. PLoS Med survey. J Med Libr Assoc 2004;92:364–71. in randomised trials. BMJ 2011;343:d5928. 2016;13:e1001950. 16 Chan AW, Tetzlaff JM, Gøtzsche PC, et al. SPIRIT 2013 39 Pannucci CJ, Wilkins EG. Identifying and avoiding bias 62 Horton R. Offline: Data sharing-why editors may have explanation and elaboration: guidance for protocols of in research. Plast Reconstr Surg 2010;126:619–25. got it wrong. Lancet 2016;388:1143. clinical trials. BMJ 2013;346:e7586. 40 Viera AJ, Bangdiwala SI. Eliminating bias in 63 Emdin C, Odutayo A, Hsiao A, et al. Association of 17 Moher D, Hopewell S, Schulz KF, et al. CONSORT 2010 randomized controlled trials: importance of allocation cardiovascular trial registration with positive study explanation and elaboration: updated guidelines concealment and masking. Fam Med 2007;39:132–7. findings: epidemiological study of randomized trials for reporting parallel group randomised trials. BMJ 41 Kunz R, Vist G, Oxman AD. Randomisation to protect (ESORT). JAMA Intern Med 2015;175:304–7. 2010;340:c869. against selection bias in healthcare trials. Cochrane 64 Kaplan RM, Irvin VL. Likelihood of null effects of large 18 Loudon K, Treweek S, Sullivan F, et al. The PRECIS-2 Database Syst Rev 2007:MR000012. NHLBI clinical trials has increased over time. PLoS One tool: designing trials that are fit for purpose. BMJ 42 Schulz KF, Grimes DA. Allocation concealment in 2015;10:e0132382. 2015;350:h2147. randomised trials: defending against deciphering. 65 Ohtake PJ, Childs JD. Why publish study protocols? 19 Lesaffre E. Superiority, equivalence, and non-inferiority Lancet 2002;359:614–8. Phys Ther 2014;94:1208–9. trials. Bull NYU Hosp Jt Dis 2008;66:150–4. 43 Kernan WN, Viscoli CM, Makuch RW, et al. Stratified 66 Skogvoll E, Kramer-Johansen J. Publication of clinical 20 Schumi J, Wittes JT. Through the looking glass: randomization for clinical trials. J Clin Epidemiol trial protocols--what can we learn? Scand J Trauma understanding non-inferiority. Trials 2011;12:106. 1999;52:19–26. Resusc Emerg Med 2013;21:12.

8 Bandholm T, et al. Br J Sports Med Month 2017 Vol 0 No 0