BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from
BMJ Open is committed to open peer review. As part of this commitment we make the peer review history of every article we publish publicly available.
When an article is published we post the peer reviewers’ comments and the authors’ responses online. We also post the versions of the paper that were used during peer review. These are the versions that the peer review comments apply to.
The versions of the paper that follow are the versions that were submitted during the peer review process. They are not the versions of record or the final published versions. They should not be cited or distributed as the published version of this manuscript.
BMJ Open is an open access journal and the full, final, typeset and author-corrected version of record of the manuscript is available on our site with no access controls, subscription charges or pay- per-view fees (http://bmjopen.bmj.com).
If you have any questions on BMJ Open’s open peer review process please email [email protected]
http://bmjopen.bmj.com/ on September 30, 2021 by guest. Protected copyright. BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from BMJ Open
Comparative efficacy of interventions for reducing injection and sexual risk behaviours to prevent HIV in drug users: protocol for Bayesian network meta-analysis
ForJournal: peerBMJ Open review only Manuscript ID bmjopen-2018-022811
Article Type: Protocol
Date Submitted by the Author: 07-Mar-2018
Complete List of Authors: lang, junjie; Wannan Medical College jin, lairun; Wannan Medical College Yao, Yingshui; Wannan Medical College,
EPIDEMIOLOGY, HIV & AIDS < INFECTIOUS DISEASES, Protocols & Keywords: guidelines < HEALTH SERVICES ADMINISTRATION & MANAGEMENT
http://bmjopen.bmj.com/
on September 30, 2021 by guest. Protected copyright.
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from Page 1 of 7 BMJ Open
1 2 3 4 Comparative efficacy of interventions for reducing injection 5 6 and sexual risk behaviours to prevent HIV in drug users: 7 8 9 protocol for Bayesian network meta-analysis 10 11 Junjie Lang Lairun Jin Yingshui Yao 12 13 14 School of Public Health, Wannan Medical College, Wuhu 241002, Anhui, China 15 16 Correspondence: YingshuiFor Yao, peer School of Publicreview Health, Wannan only Medical College, No. 22 Road 17 18 19 Wenchangxi, Yijiang district, Wuhu 241002, Anhui, China 20 21 E-mail: [email protected] 22 23 ABSTRACT 24 25 Introduction 26 Drug users are more vulnerable to AIDS than the general population. While several 27 interventions are effective for addressing HIV in drug users, no meta analysis has yet been 28 performed to compare interventions and determine the relative benefits of each. We intend to conduct 29 30 a Bayesian network meta analysis to compare all available interventions for reducing injection and 31 risky sexual behaviour for prevention of HIV in drug users. http://bmjopen.bmj.com/ 32 Methods and analysis 33 34 Studies will be retrieved by searching the following databases: Medline, Embase, PsycINFO, 35 Cochrane Central Register of Controlled Trials. Selection and abstraction of data will occur 36 simultaneously. Primary outcome measures will be injection risk behaviour and HIV risk behaviour. 37 HIV seroconversion, confirmed using an antibody test, will be the secondary outcome. Bayesian 38 39 network meta analyses will be conducted using the Markov Chains Monte Carlo method. The on September 30, 2021 by guest. Protected copyright. 40 Cochrane revised tool, Risk of Bias, will be used to assess the risk of bias. Grading of 41 Recommendations Assessment, Development, and Evaluation will be used to assess evidence quality. 42 43 Ethics and dissemination 44 The results of this study will be disseminated at professional conferences and via publications in 45 peer reviewed journals. This study will not include any confidential personal data or data on human 46 trials; therefore, ethical approval is not required. 47 48 PROSPERO registration number: CRD42018086999. 49 50 STRENGTHS AND LIMITATIONS OF THIS STUDY 51 1.This meta analysis will make a comprehensive comparison of interventions for reducing 52 53 injection and sexual risk behaviours to prevent HIV in drug users. 54 2. This article can be used as a reference for implementing relevant intervention measures. 55 3. This protocol is written in strict accordance with the Preferred Reporting Items for 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from BMJ Open Page 2 of 7
1 2 3 Systematic Reviews and Meta Analyses Protocols (PRISMA P). 4 4. The Grading of Recommendations Assessment, Development and Evaluation (GRADE) will 5 6 be used to evaluate the quality of evidence. 7 5. This meta analysis will be limited to studies which are published in English language and 8 have been peer reviewed. 9 INTRODUCTION 10 11 Injecting drug users are known to be at higher risk of HIV infection than the general population. 12 Data from the United Nations Office on Drugs and Crime (UNODC) indicate that the number of 13 people who inject drugs worldwide is approximately 12.7 million.1 The 2012 UNODC/WHO/The 14 Joint United Nations Programme on HIV/AIDS (UNAIDS)/World Bank global estimate of the number 15 16 of people who inject drugs and are living with HIV was 1.7 million (range: 0.9–4.8 million), For peer review only 2 17 corresponding to an average prevalence of HIV among people who inject drugs of 13.1%. 18 Furthermore, based on data published by UNAIDS, injecting drugs users accounted for 51% of people 19 20 with HIV infections in eastern Europe and central Asia, and 13% of new HIV infections in Asia and the 3 21 Pacific, in 2014. 22 HIV is a major contributor to the disease burden attributable to drug use globally.4 Effective 23 interventions are necessary to address HIV in drug users. There is a comprehensive package of nine 24 25 interventions, endorsed by UNAIDS, UNODC, and WHO, for the prevention, treatment, and care of 26 HIV in injecting drug users (IDUs), which includes: needle and syringe programmes (NSPs); opioid 27 substitution therapy (OST); antiretroviral therapy; and targeted information, education, and 28 communication (IEC) (among other measures).1 29 30 There have been several systematic reviews and meta analyses of HIV interventions in people 31 who inject drugs.5 9 These studies have confirmed the efficacy of interventions such as NSPs,5, 9 http://bmjopen.bmj.com/ 32 psychosocial interventions, and IEC;6, 7 8 however, none of the meta analyses evaluated the effects of 33 34 all of these interventions, or compared the relative benefits of each; therefore, information regarding 35 whether distinct types of intervention have comparable efficacy and are equally appropriate for 36 different drug users is lacking. 37 38 39 Objectives on September 30, 2021 by guest. Protected copyright. 40 In this study, we aim to compare the efficacy of all available interventions for reducing injection 41 and sexual risk behaviours to prevent HIV in drug users. A network meta analysis can combine direct 42 43 and indirect evidence to provide more precise and accurate (thus both internally and externally valid) 10 44 effect estimates. Moreover, based on effective statistical inference methods, it allows ranking of 45 investigated interventions to determine which among them is the most and least effective.11 46 47 48 METHODS AND ANALYSIS 49 This protocol follows the Preferred Reporting Items for Systematic Reviews and Meta Analyses 50 Protocols (PRISMA P).12 It has also been registered in the International Prospective Register of 51 Systematic Reviews (trial registration number: CRD42018086999). 52 53 54 Eligibility criteria 55 Types of participant 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from Page 3 of 7 BMJ Open
1 2 3 People who inject opiates, cocaine, cannabis, and amphetamines (including ‘ecstasy’). People 4 who primarily misuse alcohol will be excluded from our study. 5 6 7 Interventions 8 We will include studies with interventions which are defined by WHO, UNODC, and UNAIDS, 9 including:1 10 11 needle and syringe programmes (NSPs); 12 opioid substitution therapy (OST) and other evidence based drug dependence treatment 13 programmes; 14 HIV testing and counselling (HTC); 15 16 antiretroviralFor therapy peer (ART); review only 17 prevention and treatment of sexually transmitted infections; 18 condom programmes; 19 20 targeted information, education, and communication (IEC) for people who inject drugs. 21 22 Comparators 23 Placebo controlled or no intervention. Studies which compare two different interventions within 24 25 the same investigation will also be accepted. 26 27 Outcomes 28 We will only accept a study if it contains at least one outcome measure of injection risk 29 30 behaviour, sexual risk behaviour, or HIV seroconversion. 31 http://bmjopen.bmj.com/ 32 Study designs and publication types 33 34 We will only include randomised controlled trials and publications which have been 35 peer reviewed. 36 37 Language and time frame 38 39 We only intend to include studies which are published in English. We will not place any time on September 30, 2021 by guest. Protected copyright. 40 restriction on the publication year. The search will be performed in May 2018. 41 42 43 Information sources and search strategy 44 We will search the following databases: Medline, Embase, PsycINFO, and the Cochrane Central 45 Register of Controlled Trials. The search strategy shown below was adapted from a previous 46 13 review, and improved by conferring with experts in a related field. The search strategies for other 47 48 databases will be adjusted according to their specific requirements. We will also carry out manual 49 searches of the reference lists of other review articles on related subjects, to retrieve additional 50 studies not identified by our original search. The following search terms will be used: 51 1. “drug users” OR “drug use” OR “drug abuse” OR “drug abuser” OR “drug abusers OR 52 53 “drug addict*” OR “substance abuse” OR “substance dependence” OR “drug 54 dependence” OR “drug dependency” OR “IDU” OR “IDUs” OR “injecting drug” OR 55 “intravenous drug” OR “intravenous substance” OR “injecting substance” OR exp 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from BMJ Open Page 4 of 7
1 2 3 substance abuse, intravenous/ 4 2. “HIV” OR “AIDS” OR “acquired immunodeficiency syndrome” OR “Acquired 5 6 Immunodeficiency Syndrome Virus” OR “AIDS Virus” OR “AIDS Viruses” OR 7 “Immunologic Deficiency Syndrome, Acquired” OR “Acquired Immune Deficiency 8 Syndrome” OR exp HIV/ OR exp HIV Infections/ 9 3. #1 AND #2 10 11 4. *Randomised Controlled Trial/OR (Randomised Controlled Trial).pt OR *Random 12 Allocation/. 13 5. (Randomised OR randomised OR (random* adj (assigned OR allocated OR assignment 14 OR allocation))). ab,ti. 15 16 6. #4 OR #5For peer review only 17 7. #3 AND #6 18 19 20 Study selection 21 We will import the search results into EndNote (data management software). After removing 22 duplicate articles, the first two authors will independently read the titles and abstracts to select 23 eligible articles according to the inclusion criteria. Then we will obtain the full texts of all articles 24 25 which appear to meet the inclusion criteria or where there is any uncertainty. The first two authors 26 will conduct full text reviews alone to confirm the eligibility of these articles. Cohen’s Kappa (κ) 27 will be used to measure the chance corrected agreement between the two authors. Any discrepancies 28 will be resolved by discussion with a third author and the reasons for excluding articles will be 29 30 recorded. 31 http://bmjopen.bmj.com/ 32 Data collection process 33 34 The first two authors will independently abstract the following information from the articles 35 collected as described above: 36 1. Study characteristics (first author, journal, year, country, sample size, etc.) 37 2. Participant characteristics (age, sex, manner of drug use, type of drug, the incidence rate of 38 39 injection risk behaviours or sexual risk behaviours at baseline, etc.) on September 30, 2021 by guest. Protected copyright. 40 3. Intervention characteristics (type, treatment dose, and duration, etc.) 41 4. Control characteristics 42 43 5. Any disagreements will be resolved by discussion with the third author and we will contact 44 the original authors of studies to resolve any uncertainties if necessary. 45 46 Outcome measures 47 48 Our primary outcome measures will be injection risk behaviours and HIV risk behaviours. We 49 will review all the acquired full texts to check the relative scale used by each study, then select the 50 authoritative scale, for example the HIV Risk Taking Behaviour Scale,14, 15 to assess the above 51 indicators. HIV seroconversion confirmed by an antibody test will be the secondary outcome, if 52 53 available. 54 55 Risk of bias in individual studies 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from Page 5 of 7 BMJ Open
1 2 3 We will use the Cochrane revised tool, Risk of Bias (RoB V.2.0),16 to determine the risk of bias. 4 The RoB tool contains five key domains: (1) randomisation process; (2) deviations from intended 5 6 interventions; (3) missing outcome data; (4) measurement of the outcome; and (5) selective reporting. 7 The risk of bias will be classified into three types for each domain: high, low, or some concerns. 8 Subsequently, we will arrive at an overall risk of bias, based on judgements from the five domains. The 9 first two authors will perform all assessments independently of each other. Any disagreements will be 10 11 resolved by discussion with a third author. 12 13 Data synthesis 14 We will use Stata software (13.0; Stata Corporation, College Station, Texas, USA) to conduct a 15 16 traditional pairwise Formeta analysis. peer If more thanreview five studies are included,only we will use the random 17 effects model to combine the data. Otherwise, we will use a fixed effect model, because the random 18 effects model may be imprecise in this situation.17 Dichotomous data will be evaluated using the risk 19 20 ratio (RR) with 95% confidence interval (CI), while continuous outcomes will be expressed as 2 21 standardized mean differences and 95% CI. Heterogeneity will be quantified using the I statistic. If 22 I2 > 50%, which indicates the presence of substantial heterogeneity,18 we will consider subgrouping 23 the intervention by study setting (receiving formal treatment or not), gender specific (single gender 24 25 or mixed gender), HIV testing (reported or not), methodological quality of the study, or geographical 26 area. 27 When pairwise meta analysis is completed, we will perform a network meta analysis using 28 WinBUGS 1.43 software. The Markov Chains Monte Carlo method will be used for Bayesian 29 30 analysis. When we run the WinBUGS program, we will set it to perform 100 000 simulations, and 31 the first 10 000 simulations will be discarded as burn in. Convergence of the model will be assessed http://bmjopen.bmj.com/ 32 by trace and Gelman Rubin Brooks plots.19 33 34 We will use both random and fixed effects models for the network analysis. Then we will 35 select the appropriate model on the basis of the deviance information criterion (DIC); the model with 36 the lower DIC will be preferred (a difference > 3 will be considered significant).20 The node splitting 37 method will be used to statistically assess the consistency between direct and indirect evidence.21 38 39 We will examine the assumptions of transitivity and similarity on account of clinical and on September 30, 2021 by guest. Protected copyright. 40 methodological characteristics; notably, there is no universal statistical method to analyse these effect 41 modifiers.22 We plan to investigate similarity based on factors including participant characteristics, 42 43 experimental design, study quality, and risk of bias, among others. 44 Furthermore, publication bias will be assessed using comparison adjusted funnel plots. All the 45 figures, including forest plots for each intervention, network plots, and comparison adjusted funnel 46 plots, will be produced using the ‘Network Graphs’ package in STATA. 47 48 49 Confidence in cumulative evidence 50 We will evaluate the quality of evidence for all outcomes according to the Grading of 51 Recommendations Assessment, Development and Evaluation (GRADE). Based on the domains of 52 53 methodology quality, consistency, directness, precision effect estimates, and publication bias, we will 54 rank the overall strength of evidence as high, moderate, low, or very low.23 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from BMJ Open Page 6 of 7
1 2 3 ETHICS AND DISSEMINATION 4 This Bayesian network meta analysis will include no confidential personal data and no data on 5 6 human trials. Consequently, ethical approval is not required. The procedures used for this study will 7 be reported in accordance with the Preferred Reporting Items for Systematic Reviews and 8 Meta Analysis (PRISMA) extension statement for network meta analyses of healthcare 9 interventions.24 The final results will be disseminated at professional conferences and through 10 11 publications in peer reviewed journals. 12 13 FOOTNOTES 14
15 16 Contributors: JJL contributedFor topeer the conception review and design of this only study protocol and is the guarantor 17 of the protocol. YSY developed the search strategy. LRJ will provide advice on data analysis and 18 presentation of study results. All authors approved the final version of the manuscript. 19 20 21 Funding: This research received no specific grant from any funding agency in the public, commercial 22 or not for profit sectors. 23 24 25 Disclaimer: None. 26 27 Competing interests: None declared. 28
29 30 Provenance and peer review: Not commissioned; externally peer reviewed. 31 http://bmjopen.bmj.com/ 32 33 REFERENCES 34 35 36 1 World Health Organization, United Nations Office on Drugs and Crime, and Joint United Nations Programme on HIV/AIDS. 37 WHO, UNODC, UNAIDS technical guide for countries to set targets for universal access to HIV prevention,treatment and care 38 39 for injecting drug users. Geneva: World Health Organization; 2012 on September 30, 2021 by guest. Protected copyright. 40 http://apps.who.int/iris/bitstream/10665/77969/1/9789241504379_eng.pdf?ua=1(accessed 24 Dec 2017). 41 2 United Nations Office on Drugs and Crime. World drug report 2014. http://www.unodc.org/wdr2014/(accessed 25 Dec 2017) 42 3 UNAIDS. GLOBAL AIDS UPDATE 2016. http://www.who.int/hiv/pub/arv/global aids update 2016 pub/en/(accessed 25 Dec 43 44 2017). 45 4 Degenhardt L, Hall W, Lynskey M, et al. Illicit drug use. In: Ezzati M, Lopez AD, Rodgers A, et al, eds. Comparative 46 quantification of health risks: global and regional burden of disease attributable to selected major risk factors. 2nd edn. Geneva: 47 48 World Health Organization, 2004: 1109–1176. 49 5 Jones L, Pickering L, Sumnall H, et al. Optimal provision of needle and syringe programmes for injecting drug users: A 50 systematic review. Int J Drug Policy 2010;21:335 42. 51 6 Meader N, Li R, Des Jarlais DC, et al. Psychosocial interventions for reducing injection and sexual risk behaviour for 52 53 preventing HIV in drug users. Cochrane Database Syst Rev 2010:CD007192. 54 7 Semaan S, Des Jarlais DC, Sogolow E, et al. A meta analysis of the effect of HIV prevention interventions on the sex behaviors 55 of drug users in the United States. J Acquir Immune Defic Syndr 2002;30 Suppl 1:S73 93. 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from Page 7 of 7 BMJ Open
1 2 3 8 Prendergast ML, Urada D, Podus D. Meta analysis of HIV risk reduction interventions within drug abuse treatment programs. 4 J Consult Clin Psychol 2001;69:389 405. 5 9 Fernandes RM, Cary M, Duarte G, et al. Effectiveness of needle and syringe Programmes in people who inject drugs An 6 7 overview of systematic reviews. BMC Public Health 2017;17:309. 8 10 Biondi Zoccai G, Abbate A, Benedetto U, et al. Network meta analysis for evidence synthesis: what is it and why is it posed to 9 dominate cardiovascular decision making. Int J Cardiol 2015;182:309 14. 10 11 11 Greco T, Landoni G, Biondi Zoccai G, et al. A Bayesian network meta analysis for binary outcome: how to do it. Stat Methods 12 Med Res 2016;25:1757 73. 13 12 Shamseer L, Moher D, Clarke M, et al. Preferred reporting items for systematic review and meta analysis protocols 14 (PRISMA P) 2015: elaboration and explanation. BMJ 2015;350:g7647. 15 16 13 Mathers BM, DegenhardtFor L, Ali H,peer et al. HIV prevention, review treatment, and care servicesonly for people who inject drugs: a systematic 17 review of global, regional, and national coverage. Lancet 2010;375:1014 28. 18 14 Darke S, Hall W, Heather N, et al. The reliability and validity of a scale to measure HIV risk taking behaviour among 19 intravenous drug users. AIDS 1991;5:181 5. 20 21 15 Ward J, Darke S, Hall W. The HIV Risk taking behaviour scale (HRBS) manual. 1990. 22 16 Higgins JPT, Sterne JAC, Savović J, et al. A revised tool for assessing risk of bias in randomized trials. In: Chandler J, 23 McKenzie J, BoutronI, eds. Cochrane methods cochrane database of systematic reviews,2016. 24 25 17 Higgins JP, Thompson SG, Spiegelhalter DJ. A re evaluation of random effects meta analysis. J R Stat Soc Ser A Stat Soc 26 2009;172:137 59. 27 18 Higgins JP, Thompson SG, Deeks JJ, et al. Measuring inconsistency in meta analyses. BMJ 2003;327:557 60. 28 19 Gelman A, Rubin DB. Inference from Iterative Simulation Using Multiple Sequences. Statistical Science 1992;7:457 72. 29 30 20 Spiegelhalter DJ, Best NG, Carlin BP, et al. Bayesian measures of model complexity and fit. Journal of the Royal Statistical 31 Society: Series B (Statistical Methodology) 2002;64:583 639. http://bmjopen.bmj.com/ 32 21 Dias S, Welton NJ, Caldwell DM, et al. Checking consistency in mixed treatment comparison meta‐analysis. Stat Med 33 34 2010;29:932 44. 35 22 Kim H, Gurrin L, Ademi Z, et al. Overview of methods for comparing the efficacies of drugs in the absence of head to head 36 clinical trial data. Br J Clin Pharmacol 2014;77:116 21. 37 23 Schunemann H, Brozek J, Guyatt G, et al. GRADE handbook for grading quality of evidence and strength of recommendation. 38 39 2013http:// gdt. guidelinedevelopment. org/ app/ (assessed 22 Dec 2017). on September 30, 2021 by guest. Protected copyright. 40 24 Hutton B, Salanti G, Caldwell DM, et al. The PRISMA extension statement for reporting of systematic reviews incorporating 41 network meta analyses of health care interventions: checklist and explanations. Ann Intern Med 2015;162:777 84. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from BMJ Open
Comparative efficacy of interventions for reducing injection and sexual risk behaviours to prevent HIV in injection drug users: protocol for Bayesian network meta-analysis ForJournal: peerBMJ Open review only Manuscript ID bmjopen-2018-022811.R1
Article Type: Protocol
Date Submitted by the 08-Sep-2018 Author:
Complete List of Authors: lang, junjie; Wannan Medical College, Wannan Medical College jin, lairun; Wannan Medical College Yao, Yingshui; Wannan Medical College,
Primary Subject Epidemiology Heading:
Secondary Subject Heading: HIV/AIDS
EPIDEMIOLOGY, HIV & AIDS < INFECTIOUS DISEASES, Protocols & Keywords: guidelines < HEALTH SERVICES ADMINISTRATION & MANAGEMENT http://bmjopen.bmj.com/
on September 30, 2021 by guest. Protected copyright.
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from Page 1 of 8 BMJ Open
1 2 3 4 1 Comparative efficacy of interventions for reducing injection 5 6 2 and sexual risk behaviours to prevent HIV in injection drug 7 8 9 3 users: protocol for Bayesian network meta-analysis 10 11 4 Junjie Lang Lairun Jin Yingshui Yao 12 13 14 5 School of Public Health, Wannan Medical College, Wuhu 241002, Anhui, China 15 16 6 Correspondence: YingshuiFor Yao, peer School of Publicreview Health, Wannan only Medical College, No. 22 Road 17 18 19 7 Wenchangxi, Yijiang district, Wuhu 241002, Anhui, China 20 21 E-mail: [email protected] 22 8 23 9 ABSTRACT 24 25 10 Introduction 26 11 Drug users are more vulnerable to AIDS than the general population. While several 27 12 interventions are effective for addressing HIV in injection drug users, no meta analysis has yet been 28 performed to compare interventions and determine the relative benefits of each. We intend to conduct 29 13 30 14 a Bayesian network meta analysis to compare all available interventions for reducing injection and 31 15 risky sexual behaviours for prevention of HIV in injection drug users. http://bmjopen.bmj.com/ 32 16 Methods and analysis 33 34 17 Studies will be retrieved by searching the following databases: Medline, Embase, PsycINFO, 35 18 Cochrane Central Register of Controlled Trials. The search will be performed between May and July 36 19 2018. Two authors will extract data independently. Primary outcome measures will be injection risk 37 behaviour and HIV risk behaviour. HIV seroconversion, confirmed using an antibody test, will be the 38 20 39 21 secondary outcome. Bayesian network meta analyses will be conducted using the Markov Chains on September 30, 2021 by guest. Protected copyright. 40 22 Monte Carlo method. The Cochrane revised tool, Risk of Bias, will be used to assess the risk of bias. 41 23 Grading of Recommendations Assessment, Development, and Evaluation will be used to assess 42 43 24 evidence quality. 44 25 Ethics and dissemination 45 26 The results of this study will be disseminated at professional conferences and via publications in 46 27 peer reviewed journals. This study will not include any confidential personal data or data on human 47 48 28 trials; therefore, ethical approval is not required. 49 29 PROSPERO registration number: CRD42018086999. 50 30 51 STRENGTHS AND LIMITATIONS OF THIS STUDY 52 31 53 32 Strengths: 54 33 1.This meta analysis will make a comprehensive comparison of interventions for reducing 55 34 injection and sexual risk behaviours to prevent HIV in drug users. 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from BMJ Open Page 2 of 8
1 2 3 1 2. This article can be used as a reference for implementing relevant intervention measures. 4 2 3. This protocol is written in strict accordance with the Preferred Reporting Items for 5 6 3 Systematic Reviews and Meta Analyses Protocols (PRISMA P). 7 4 4. The Grading of Recommendations Assessment, Development and Evaluation (GRADE) will 8 5 be used to evaluate the quality of evidence. 9 6 Limitations: 10 11 7 1. This meta analysis will be limited to studies which are published in English language and 12 8 have been peer reviewed. 13 9 2. Given this meta analysis will only include randomised controlled trials (RCTs), there is a 14 possibility that the study participant population will not be representative of the overall population. 15 10 16 11 INTRODUCTIONFor peer review only 17 12 Injecting drug users are known to be at higher risk of HIV infection than the general population. 18 13 Data from the United Nations Office on Drugs and Crime (UNODC) indicate that the number of 19 1 20 14 people who inject drugs worldwide is approximately 12.7 million. The 2018 UNODC/WHO/The 21 15 Joint United Nations Programme on HIV/AIDS (UNAIDS)/World Bank global estimate of the number 22 16 of people who inject drugs and are living with HIV was 1.7 million (range: 0.9–4.8 million), 23 2 17 corresponding to an average prevalence of HIV among people who inject drugs of 13.1%. 24 25 18 Furthermore, based on data published by UNAIDS, injecting drugs users accounted for 51% of people 26 19 with HIV infections in eastern Europe and central Asia, and 13% of new HIV infections in Asia and the 27 20 Pacific, in 2014.3 28 HIV is a major contributor to the disease burden attributable to drug use globally.4 Effective 29 21 30 22 interventions are necessary to address HIV in drug users. There is a comprehensive package of nine 31 23 interventions, endorsed by UNAIDS, UNODC, and WHO, for the prevention, treatment, and care of http://bmjopen.bmj.com/ 32 24 HIV in injecting drug users (IDUs), which includes: needle and syringe programmes (NSPs); opioid 33 34 25 substitution therapy (OST); antiretroviral therapy; and targeted information, education, and 1 35 26 communication (IEC) (among other measures). 36 27 There have been several systematic reviews and meta analyses of HIV interventions in people 37 28 who inject drugs.5 9 These studies have confirmed the efficacy of interventions such as NSPs,5, 9 38 6, 7 8 39 29 psychosocial interventions, and IEC; however, none of the meta analyses evaluated the effects of on September 30, 2021 by guest. Protected copyright. 40 30 all of these interventions, or compared the relative benefits of each; therefore, information regarding 41 31 whether distinct types of intervention have comparable efficacy and are equally appropriate for 42 43 32 different populations of injection drug users are lacking. 44 33 45 34 Objectives 46 35 In this study, we aim to compare the efficacy of all available interventions for reducing injection 47 48 36 and sexual risk behaviours to prevent HIV in drug users. A network meta analysis can combine direct 49 37 and indirect evidence to provide more precise and accurate (thus both internally and externally valid) 50 38 effect estimates.10 Moreover, based on effective statistical inference methods, it allows ranking of 51 investigated interventions to determine which among them is the most and least effective.11 52 39 53 40 54 41 METHODS AND ANALYSIS 55 42 This protocol follows the Preferred Reporting Items for Systematic Reviews and Meta Analyses 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from Page 3 of 8 BMJ Open
1 2 3 1 Protocols (PRISMA P).12 It has also been registered in the International Prospective Register of 4 2 Systematic Reviews (trial registration number: CRD42018086999). 5 6 3 7 4 Eligibility criteria 8 5 Types of participant 9 6 People who inject opiates, cocaine, cannabis, and amphetamines (including ‘ecstasy’) will be 10 11 7 included. People who primarily misuse alcohol will be excluded. 12 8 13 9 Interventions 14 Interventions which are defined by WHO, UNODC, and UNAIDS, including:1 15 10 16 11 needle andFor syringe peer programmes review (NSPs); only 17 12 opioid substitution therapy (OST) and other evidence based drug dependence treatment 18 13 programmes; 19 20 14 HIV testing and counselling (HTC); 21 15 antiretroviral therapy (ART); 22 16 prevention and treatment of sexually transmitted infections; 23 17 condom programmes; 24 25 18 targeted information, education, and communication (IEC) for people who inject drugs. 26 19 27 20 Comparators 28 Placebo controlled or no intervention. Studies which compare two different interventions within 29 21 30 22 the same investigation will also be accepted. 31 23 http://bmjopen.bmj.com/ 32 24 Outcomes 33 34 25 Injection risk behaviour, sexual risk behaviour, or HIV seroconversion. 35 26 36 27 Study designs and publication types 37 Randomised controlled trials and peer reviewed publications. 38 28 39 29 Setting on September 30, 2021 by guest. Protected copyright. 40 30 Setting 41 31 There will be no restrictions by type of clinical setting, and authors will include studies at all 42 43 32 levels of healthcare setting. 44 33 Language and time frame 45 34 We only intend to include studies which are published in English and indexed from 1980 to 46 35 May 2018. 47 48 36 49 37 Information sources and search strategy 50 38 We will search the following databases: Medline, Embase, PsycINFO, and the Cochrane Central 51 Register of Controlled Trials. The search will be performed between May and July 2018.The search 52 39 13 53 40 strategy shown below was adapted from a previous review, and improved by conferring with 54 41 experts in a related field. The search strategies for other databases will be adjusted according to their 55 42 specific requirements. We will also carry out manual searches of the reference lists of other review 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from BMJ Open Page 4 of 8
1 2 3 1 articles on related subjects, to retrieve additional studies not identified by our original search. The 4 2 following search terms will be used: 5 6 3 1. “drug users” OR “drug use” OR “drug abuse” OR “drug abuser” OR “drug abusers OR 7 4 “drug addict*” OR “substance abuse” OR “substance dependence” OR “drug 8 5 dependence” OR “drug dependency” OR “IDU” OR “IDUs” OR “injecting drug” OR 9 6 “intravenous drug” OR “intravenous substance” OR “injecting substance” OR exp 10 11 7 substance abuse, intravenous/ 12 8 2. “HIV” OR “AIDS” OR “acquired immunodeficiency syndrome” OR “Acquired 13 9 Immunodeficiency Syndrome Virus” OR “AIDS Virus” OR “AIDS Viruses” OR 14 “Immunologic Deficiency Syndrome, Acquired” OR “Acquired Immune Deficiency 15 10 16 11 Syndrome”For OR exp peer HIV/ OR exp review HIV Infections/ only 17 12 3. #1 AND #2 18 13 4. *Randomised Controlled Trial/OR (Randomised Controlled Trial).pt OR *Random 19 20 14 Allocation/. 21 15 5. (Randomised OR randomised OR (random* adj (assigned OR allocated OR assignment 22 16 OR allocation))). ab,ti. 23 17 6. #4 OR #5 24 25 18 7. #3 AND #6 26 19 27 20 Study selection 28 We will import the search results into EndNote (data management software). After removing 29 21 30 22 duplicate articles, the first two authors will independently read the titles and abstracts to select 31 23 eligible articles according to the inclusion criteria. Then we will obtain the full texts of all articles http://bmjopen.bmj.com/ 32 24 which appear to meet the inclusion criteria or where there is any uncertainty. The first two authors 33 34 25 will conduct full text reviews alone to confirm the eligibility of these articles. Cohen’s Kappa (κ) 35 26 (calculated by R software 3.44) will be used to measure the chance corrected agreement between the 36 27 two authors. Any discrepancies will be resolved by discussion with a third author and the reasons for 37 excluding articles will be recorded. 38 28 39 29 on September 30, 2021 by guest. Protected copyright. 40 30 Data collection process 41 31 The first two authors will independently use Excel 2016 software to abstract the following 42 43 32 information from the articles collected as described above: 44 33 1. Study characteristics (first author, journal, year, country, sample size, etc.) 45 34 2. Participant characteristics (age, sex, manner of drug use, type of drug, the incidence rate of 46 35 injection risk behaviours or sexual risk behaviours at baseline, etc.) 47 48 36 3. Intervention characteristics (type, treatment dose, and duration, etc.) 49 37 4. Control characteristics 50 38 Any disagreements will be resolved by discussion with the third author and we will contact the 51 original authors of studies to resolve any uncertainties if necessary. 52 39 53 40 54 41 Outcome measures 55 42 Our primary outcome measures will be injection risk behaviours and HIV risk behaviours. HIV 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from Page 5 of 8 BMJ Open
1 2 3 1 seroconversion confirmed by an antibody test will be the secondary outcome, if available. The 4 2 efficacy will be based on the difference in injection risk behaviours and HIV risk behaviours between 5 6 3 the intervention and comparator on the completion of intervention. HIV risk behaviours include sex 7 4 times (vaginal or anal), frequency of condoms used when had sex, whether engaging in sex with 8 5 other partners concurrently. Injection risk behaviour will be defined as having shared syringes, 9 6 containers, filters or water to inject drugs in the previous month and backloading/frontloading. We 10 11 7 will review all the acquired full texts to check the relative scale used by each study, for example the 14, 15 12 8 HIV Risk Taking Behaviour Scale, to assess the above indicators. 13 9 14 10 Risk of bias in individual studies 15 16 16 11 We will use theFor Cochrane peer revised tool, Riskreview of Bias (RoB V.2.0),only to determine the risk of bias. 17 12 The RoB tool contains five key domains: (1) randomisation process; (2) deviations from intended 18 13 interventions; (3) missing outcome data; (4) measurement of the outcome; and (5) selective reporting. 19 20 14 The risk of bias will be classified into three types for each domain: high, low, or some concerns. 21 15 Subsequently, we will arrive at an overall risk of bias, based on judgements from the five domains. A 22 16 summary of risk of bias of all the domains will be provided for each trial. The first two authors will 23 17 perform all assessments independently of each other. Any disagreements will be resolved by 24 25 18 discussion with a third author. 26 19 27 20 Data synthesis 28 We will use Stata software (13.0; Stata Corporation, College Station, Texas, USA) to conduct a 29 21 30 22 traditional pairwise meta analysis. If more than five studies are included, we will use the random 31 23 effects model to combine the data. Otherwise, we will use a fixed effect model, because the random http://bmjopen.bmj.com/ 32 24 effects model may be imprecise in this situation.17 Dichotomous data will be evaluated using the risk 33 34 25 ratio (RR) with 95% confidence interval (CI), while continuous outcomes will be expressed as 2 35 26 standardized mean differences and 95% CI. Heterogeneity will be quantified using the I statistic. If 36 27 I2 > 50%, which indicates the presence of substantial heterogeneity,18 we will consider subgrouping 37 the intervention by study setting (receiving formal treatment or not), gender specific (single gender 38 28 39 29 or mixed gender), HIV testing (reported or not), methodological quality of the study, drug types or on September 30, 2021 by guest. Protected copyright. 40 30 geographical area. 41 31 When pairwise meta analysis is completed, we will perform a network meta analysis using 42 43 32 WinBUGS 1.43 software. The Markov Chains Monte Carlo method will be used for Bayesian 44 33 analysis. When we run the WinBUGS program, we will set it to perform 100 000 simulations, and 45 34 the first 10 000 simulations will be discarded as burn in. Convergence of the model will be assessed 46 19 35 by trace and Gelman Rubin Brooks plots. 47 48 36 We will use both random and fixed effects models for the network meta analysis. Then we will 49 37 select the appropriate model on the basis of the deviance information criterion (DIC); the model with 50 38 the lower DIC will be preferred (a difference > 3 will be considered significant).20 The node splitting 51 method will be used to statistically assess the consistency between direct and indirect evidence.21 52 39 53 40 We will examine the assumptions of transitivity (similarity in the distribution of potential effect 54 41 modifiers across the different pairwise comparisons) on account of clinical and methodological 55 42 characteristics; notably, there is no universal statistical method to analyse these effect modifiers.22 23 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from BMJ Open Page 6 of 8
1 2 3 1 We plan to investigate similarity based on factors including participant characteristics, experimental 4 2 design, study quality, and risk of bias, among others. 5 6 3 Furthermore, publication bias will be assessed using comparison adjusted funnel plots. All the 7 4 figures, including forest plots for each intervention, network plots, and comparison adjusted funnel 8 5 plots, will be produced using the ‘Network Graphs’ package in STATA. 9 6 We will performed sensitivity analysis to address whether the combined estimates of the 10 11 7 interventions are dominated by one or several studies, especially those with a high risk of bias. Then 12 8 we will exclude the trials to test the robustness of our study result. Second, we will test whether the 13 9 imputation of the missing values affects the result of the meta analysis. 14
15 10 16 11 Confidence in cumulativeFor evidence peer review only 17 12 We will evaluate the quality of evidence for all outcomes according to the Grading of 18 13 Recommendations Assessment, Development and Evaluation (GRADE). Based on the domains of 19 20 14 methodology quality, consistency, directness, precision effect estimates, and publication bias, we will 24 21 15 rank the overall strength of evidence as high, moderate, low, or very low. 22 16 Patient and public involvement 23 17 Patients will not be involved. 24 25 18 ETHICS AND DISSEMINATION 26 19 This Bayesian network meta analysis will include no confidential personal data and no data on 27 20 human trials. Consequently, ethical approval is not required. The procedures used for this study will 28 be reported in accordance with the Preferred Reporting Items for Systematic Reviews and 29 21 30 22 Meta Analysis (PRISMA) extension statement for network meta analyses of healthcare 31 23 interventions.25 The final results will be disseminated at professional conferences and through http://bmjopen.bmj.com/ 32 24 publications in peer reviewed journals. 33 34 25 35 26 FOOTNOTES 36 27 37 Contributors: JJL contributed to the conception and design of this study protocol and is the guarantor 38 28 39 29 of the protocol. YSY developed the search strategy. LRJ will provide advice on data analysis and on September 30, 2021 by guest. Protected copyright. 40 30 presentation of study results. All authors approved the final version of the manuscript. 41 31 42 43 32 Funding: This research received no specific grant from any funding agency in the public, commercial 44 33 or not for profit sectors. 45 34 46 35 Disclaimer: None. 47 48 36 49 37 Competing interests: None declared. 50 38 51 Provenance and peer review: Not commissioned; externally peer reviewed. 52 39 53 40 54 55 41 REFERENCES 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from Page 7 of 8 BMJ Open
1 2 3 1 4 2 1 World Health Organization, United Nations Office on Drugs and Crime, and Joint United Nations Programme on HIV/AIDS. 5 WHO, UNODC, UNAIDS technical guide for countries to set targets for universal access to HIV prevention,treatment and care 6 3 7 4 for injecting drug users. Geneva: World Health Organization; 2012 8 5 http://apps.who.int/iris/bitstream/10665/77969/1/9789241504379_eng.pdf?ua=1(accessed 24 Dec 2017). 9 6 2 United Nations Office on Drugs and Crime. World drug report 2018. 10 11 7 http://www.unodc.org/wdr2018/prelaunch/WDR18_Booklet_1_EXSUM.pdf (accessed 25 Aug 2018) 12 8 3 UNAIDS. GLOBAL AIDS UPDATE 2016. http://www.who.int/hiv/pub/arv/global aids update 2016 pub/en/(accessed 25 Dec 13 9 2017). 14 10 4 Degenhardt L, Hall W, Lynskey M, et al. Illicit drug use. In: Ezzati M, Lopez AD, Rodgers A, et al, eds. Comparative 15 16 11 quantification of healthFor risks: global peer and regional burdenreview of disease attributable only to selected major risk factors. 2nd edn. Geneva: 17 12 World Health Organization, 2004: 1109–1176. 18 13 5 Jones L, Pickering L, Sumnall H, et al. Optimal provision of needle and syringe programmes for injecting drug users: A 19 systematic review. Int J Drug Policy 2010;21:335 42. 20 14 21 15 6 Meader N, Li R, Des Jarlais DC, et al. Psychosocial interventions for reducing injection and sexual risk behaviour for 22 16 preventing HIV in drug users. Cochrane Database Syst Rev 2010:CD007192. 23 17 7 Semaan S, Des Jarlais DC, Sogolow E, et al. A meta analysis of the effect of HIV prevention interventions on the sex behaviors 24 25 18 of drug users in the United States. J Acquir Immune Defic Syndr 2002;30 Suppl 1:S73 93. 26 19 8 Prendergast ML, Urada D, Podus D. Meta analysis of HIV risk reduction interventions within drug abuse treatment programs. 27 20 J Consult Clin Psychol 2001;69:389 405. 28 9 Fernandes RM, Cary M, Duarte G, et al. Effectiveness of needle and syringe Programmes in people who inject drugs An 29 21 30 22 overview of systematic reviews. BMC Public Health 2017;17:309. 31 23 10 Biondi Zoccai G, Abbate A, Benedetto U, et al. Network meta analysis for evidence synthesis: what is it and why is it posed to http://bmjopen.bmj.com/ 32 24 dominate cardiovascular decision making. Int J Cardiol 2015;182:309 14. 33 34 25 11 Greco T, Landoni G, Biondi Zoccai G, et al. A Bayesian network meta analysis for binary outcome: how to do it. Stat Methods 35 26 Med Res 2016;25:1757 73. 36 27 12 Shamseer L, Moher D, Clarke M, et al. Preferred reporting items for systematic review and meta analysis protocols 37 28 (PRISMA P) 2015: elaboration and explanation. BMJ 2015;350:g7647. 38 39 29 13 Mathers BM, Degenhardt L, Ali H, et al. HIV prevention, treatment, and care services for people who inject drugs: a systematic on September 30, 2021 by guest. Protected copyright. 40 30 review of global, regional, and national coverage. Lancet 2010;375:1014 28. 41 31 14 Darke S, Hall W, Heather N, et al. The reliability and validity of a scale to measure HIV risk taking behaviour among 42 intravenous drug users. AIDS 1991;5:181 5. 43 32 44 33 15 Ward J, Darke S, Hall W. The HIV Risk taking behaviour scale (HRBS) manual. 1990. 45 34 16 Higgins JPT, Sterne JAC, Savović J, et al. A revised tool for assessing risk of bias in randomized trials. In: Chandler J, 46 35 McKenzie J, BoutronI, eds. Cochrane methods cochrane database of systematic reviews,2016. 47 48 36 17 Higgins JP, Thompson SG, Spiegelhalter DJ. A re evaluation of random effects meta analysis. J R Stat Soc Ser A Stat Soc 49 37 2009;172:137 59. 50 38 18 Higgins JP, Thompson SG, Deeks JJ, et al. Measuring inconsistency in meta analyses. BMJ 2003;327:557 60. 51 19 Gelman A, Rubin DB. Inference from Iterative Simulation Using Multiple Sequences. Statistical Science 1992;7:457 72. 52 39 53 40 20 Spiegelhalter DJ, Best NG, Carlin BP, et al. Bayesian measures of model complexity and fit. Journal of the Royal Statistical 54 41 Society: Series B (Statistical Methodology) 2002;64:583 639. 55 42 21 Dias S, Welton NJ, Caldwell DM, et al. Checking consistency in mixed treatment comparison meta‐ analysis. Stat Med 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from BMJ Open Page 8 of 8
1 2 3 1 2010;29:932 44. 4 2 22 Kim H, Gurrin L, Ademi Z, et al. Overview of methods for comparing the efficacies of drugs in the absence of head to head 5 clinical trial data. Br J Clin Pharmacol 2014;77:116 21. 6 3 7 4 23 Jansen JP ,Naci H . Is network meta analysis as valid as standard pairwise meta analysis? It all depends on 8 5 the distribution of effect modifiers. BMC Med 2013;11:159 9 6 24 Schunemann H, Brozek J, Guyatt G, et al. GRADE handbook for grading quality of evidence and strength of recommendation. 10 11 7 2013http:// gdt. guidelinedevelopment. org/ app/ (assessed 22 Dec 2017). 12 8 25 Hutton B, Salanti G, Caldwell DM, et al. The PRISMA extension statement for reporting of systematic reviews incorporating 13 9 network meta analyses of health care interventions: checklist and explanations. Ann Intern Med 2015;162:777 84. 14 15 10 16 For peer review only 17 18 11 19 20 21 22 23 24 25 26 27 28 29 30 31 http://bmjopen.bmj.com/ 32 33 34 35 36 37 38 39 on September 30, 2021 by guest. Protected copyright. 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from BMJ Open
Comparative efficacy of interventions for reducing injection and sexual risk behaviours to prevent HIV in injection drug users: protocol for Bayesian network meta-analysis ForJournal: peerBMJ Open review only Manuscript ID bmjopen-2018-022811.R2
Article Type: Protocol
Date Submitted by the 12-Nov-2018 Author:
Complete List of Authors: lang, junjie; Wannan Medical College, Wannan Medical College jin, lairun; Wannan Medical College Yao, Yingshui; Wannan Medical College, school of public health; Wannan Medical College, Insititute of Chronic Disease Prevention and Control
Primary Subject Epidemiology Heading:
Secondary Subject Heading: HIV/AIDS
EPIDEMIOLOGY, HIV & AIDS < INFECTIOUS DISEASES, Protocols & http://bmjopen.bmj.com/ Keywords: guidelines < HEALTH SERVICES ADMINISTRATION & MANAGEMENT
on September 30, 2021 by guest. Protected copyright.
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 8 BMJ Open
1 BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from 2 3 4 1 Comparative efficacy of interventions for reducing injection and 5 6 7 2 sexual risk behaviours to prevent HIV in injection drug users: 8 9 10 3 protocol for Bayesian network meta-analysis 11 12 4 Junjie Lang Lairun Jin Yingshui Yao 13 14 15 5 School of Public Health, Wannan Medical College, Wuhu 241002, Anhui, China 16 17 18 6 Correspondence: YingshuiFor Yao, peer School of Publicreview Health, Insititute only of Chronic Disease Prevention 19 20 7 and Control, Wannan Medical College, No. 22 Road Wenchangxi, Yijiang district, Wuhu 241002, 21 22 23 8 Anhui, China 24 25 26 9 E-mail: [email protected] 27 28 10 ABSTRACT 29 11 Introduction 30 31 12 Drug users are more vulnerable to AIDS than the general population. While several 32 13 interventions are effective for addressing HIV in injection drug users, no meta-analysis has yet been 33 14 performed to compare interventions and determine the relative benefits of each. We intend to 34
35 15 conduct a Bayesian network meta-analysis to compare all available interventions evaluated by an http://bmjopen.bmj.com/ 36 16 RCT for reducing injection and risky sexual behaviours for prevention of HIV in injection drug 37 17 users. 38 39 18 Methods and analysis 40 19 Studies will be retrieved by searching the following databases: Medline, Embase, PsycINFO, 41 20 Cochrane Central Register of Controlled Trials. The search will be performed between May and July 42
43 21 2018 for literature published between 1980 and May 2018. Two authors will extract data on September 30, 2021 by guest. Protected copyright. 44 22 independently. Primary outcome measures will be injection risk behaviour and HIV risk behaviour. 45 46 23 HIV seroconversion, confirmed using an antibody test, will be the secondary outcome. Bayesian 47 24 network meta-analyses will be conducted using the Markov Chains Monte Carlo method. The 48 25 Cochrane revised tool, Risk of Bias, will be used to assess the risk of bias. Grading of 49 50 26 Recommendations Assessment, Development, and Evaluation will be used to assess evidence 51 27 quality. 52 28 Ethics and dissemination 53 54 29 The results of this study will be disseminated at professional conferences and via publications in 55 30 peer-reviewed journals. This study will not include any confidential personal data or data on human 56 31 trials; therefore, ethical approval is not required. 57 58 32 PROSPERO registration number: CRD42018086999. 59 33 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 2 of 8
1 BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from 2 3 1 STRENGTHS AND LIMITATIONS OF THIS STUDY 4 5 2 Strengths: 6 3 1.This meta-analysis will conduct a comprehensive comparison of interventions for reducing 7 injection and sexual risk behaviours to prevent HIV in injection drug users. 8 4 9 5 2. This article can give advice for implementing relevant interventions. 10 6 3. This protocol is written in strict accordance with the Preferred Reporting Items for 11 12 7 Systematic Reviews and Meta-Analyses Protocols (PRISMA-P). 13 8 4. This meta-analysis will be limited to studies which are published in English language and 14 9 have been peer reviewed. 15 16 10 5. Given this meta-analysis will only include randomised controlled trials (RCTs), there is a 17 11 possibility that the study participant population will not be representative of the overall population. 18 12 We only focus on personsFor who injectpeer drugs andreview interventions definedonly by WHO, UNODC, and 19 20 13 UNAIDS(nor all interventions). 21 14 INTRODUCTION 22 15 Injecting drug users are known to be at higher risk of HIV infection than the general population. 23 24 16 Data from the United Nations Office on Drugs and Crime (UNODC) indicate that the number of 25 17 people who inject drugs worldwide is approximately 12.7 million.1 The 2018 UNODC/WHO/The 26 Joint United Nations Programme on HIV/AIDS (UNAIDS)/World Bank global estimate of the 27 18 28 19 number of injecting drug users and are living with HIV was 1.7 million (range: 0.9–4.8 million), 29 20 corresponding to an average prevalence of HIV among injecting drug users of 13.1%.2 Furthermore, 30 31 21 based on data published by UNAIDS, injecting drugs users accounted for 51% of people with HIV 32 22 infections in eastern Europe and central Asia, and 13% of new HIV infections in Asia and the 33 23 Pacific, in 2014.3 34 4 35 24 HIV is a major contributor to the disease burden attributable to drug use globally. Effective http://bmjopen.bmj.com/ 36 25 interventions are necessary to address HIV in injection drug users. There is a comprehensive package 37 26 of nine interventions, endorsed by UNAIDS, UNODC, and WHO, for the prevention, treatment, and 38 39 27 care of HIV in injecting drug users (IDUs), which includes: needle and syringe programmes (NSPs); 40 28 opioid substitution therapy (OST); antiretroviral therapy; and targeted information, education, and 41 29 communication (IEC) (among other measures).1 42
43 30 There have been several systematic reviews and meta-analyses of HIV interventions in injecting on September 30, 2021 by guest. Protected copyright. 44 31 drug users.5-9 These studies have confirmed the efficacy of interventions such as NSPs,5, 9 45 6, 7 8 46 32 psychosocial interventions, and IEC; however, none of the meta-analyses evaluated the effects of 47 33 all of these interventions, or compared the relative benefits of each; therefore, information regarding 48 34 whether distinct types of intervention have comparable efficacy and are equally appropriate for 49 50 35 different populations of injection drug users are lacking. 51 36 52 37 Objectives 53 54 38 In this study, we aim to compare the efficacy of all available interventions for reducing injection 55 39 and sexual risk behaviours to prevent HIV in injection drug users. 56 40 57 58 41 METHODS AND ANALYSIS 59 42 A network meta-analysis can combine direct and indirect evidence to provide more precise and 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 8 BMJ Open
1 BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from 2 3 1 accurate (thus both internally and externally valid) effect estimates.10 Moreover, based on effective 4 5 2 statistical inference methods, it allows ranking of investigated interventions to determine which among 6 3 them is the most and least effective.11 7 This protocol follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 8 4 9 5 Protocols (PRISMA-P).12 It has also been registered in the International Prospective Register of 10 6 Systematic Reviews (trial registration number: CRD42018086999). 11 12 7 13 8 Eligibility criteria for reports focused on 14 9 Types of participant 15 16 10 People who inject opiates, cocaine, cannabis, and amphetamines (including ‘ecstasy’) will be 17 11 included. People who primarily misuse alcohol will be excluded. 18 12 For peer review only 19 20 13 Interventions 21 14 Interventions which are defined by WHO, UNODC, and UNAIDS will be included:1 22 15 . needle and syringe programmes (NSPs); 23 24 16 . opioid substitution therapy (OST) and other evidence-based drug dependence treatment 25 17 programmes; 26 . HIV testing and counselling (HTC); 27 18 28 19 . antiretroviral therapy (ART); 29 20 . prevention and treatment of sexually transmitted infections; 30 31 21 . condom programmes; 32 22 . targeted information, education, and communication (IEC) for people who inject drugs. 33 23 34
35 24 Comparators http://bmjopen.bmj.com/ 36 25 Placebo-controlled or no intervention. Studies which compare two different interventions within 37 26 the same investigation will also be accepted. 38 39 27 40 28 Outcomes 41 29 Injection risk behaviour, sexual risk behaviour, or HIV seroconversion. 42
43 30 on September 30, 2021 by guest. Protected copyright. 44 31 Study designs and publication types 45 46 32 Randomised controlled trials and peer-reviewed publications. 47 33 Setting 48 34 Any healthcare setting. 49 50 35 Language and time frame 51 36 We will include studies which are published in English and published from 1980 to May 2018. 52 37 53 54 38 Information sources and search strategy 55 39 We will search the following databases: Medline, Embase, PsycINFO, and the Cochrane 56 40 Central Register of Controlled Trials. The search will be performed between May and July 2018.The 57 13 58 41 search strategy shown below was adapted from a previous review, and improved by conferring 59 42 with experts in a related field. The search strategies for other databases will be adjusted according to 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 8
1 BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from 2 3 1 their specific requirements. We will also carry out manual searches of the reference lists of other 4 5 2 review articles on related subjects, to retrieve additional studies not identified by our original search. 6 3 The following search terms will be used: 7 1. “drug users” OR “drug use” OR “drug abuse” OR “drug abuser” OR “drug abusers OR 8 4 9 5 “drug addict*” OR “substance abuse” OR “substance dependence” OR “drug 10 6 dependence” OR “drug dependency” OR “IDU” OR “IDUs” OR “injecting drug” OR 11 12 7 “intravenous drug” OR “intravenous substance” OR “injecting substance” OR exp 13 8 substance abuse, intravenous/ 14 9 2. “HIV” OR “AIDS” OR “acquired immunodeficiency syndrome” OR “Acquired 15 16 10 Immunodeficiency Syndrome Virus” OR “AIDS Virus” OR “AIDS Viruses” OR 17 11 “Immunologic Deficiency Syndrome, Acquired” OR “Acquired Immune Deficiency 18 12 Syndrome”For OR exp peerHIV/ OR exp review HIV Infections/ only 19 20 13 3. #1 AND #2 21 14 4. *Randomised Controlled Trial/OR (Randomised Controlled Trial).pt OR *Random 22 15 Allocation/. 23 24 16 5. (Randomised OR randomised OR (random* adj (assigned OR allocated OR assignment 25 17 OR allocation))). ab,ti. 26 6. #4 OR #5 27 18 28 19 7. #3 AND #6 29 20 30 31 21 Study selection 32 22 We will import the search results into EndNote (data management software). After removing 33 23 duplicate articles, the first two authors will independently read the titles and abstracts to select 34
35 24 eligible articles according to the inclusion criteria. Then we will obtain the full-texts of all articles http://bmjopen.bmj.com/ 36 25 which appear to meet the inclusion criteria or where there is any uncertainty. The first two authors 37 26 will conduct full-text reviews independently to confirm the eligibility of these articles. Cohen’s 38 39 27 Kappa (κ) (calculated by R software 3.44) will be used to measure the chance-corrected agreement 40 28 between the two authors. Any discrepancies will be resolved by discussion with a third author and 41 29 the reasons for excluding articles at full report will be recorded. 42
43 30 on September 30, 2021 by guest. Protected copyright. 44 31 Data collection process 45 46 32 The first two authors will independently use Excel 2016 software to abstract the following 47 33 information from the articles collected as described above: 48 34 1. Study characteristics (first author, journal, year, country, sample size, etc.) 49 50 35 2. Participant characteristics (age, sex, manner of drug use, type of drug, the incidence rate of 51 36 injection risk behaviours or sexual risk behaviours at baseline, etc.) 52 37 3. Intervention characteristics (type, treatment dose, and duration, etc.) 53 54 38 4. Control characteristics 55 39 Any disagreements will be resolved by discussion with the third author and we will contact the 56 40 original authors of studies to resolve any uncertainties if necessary. 57 58 41 59 42 Outcome measures 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 8 BMJ Open
1 BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from 2 3 1 Our primary outcome measures will be injection risk behaviours and HIV risk behaviours. HIV 4 5 2 seroconversion confirmed by an antibody test will be the secondary outcome, if available. The 6 3 efficacy will be based on the difference in injection risk behaviours and HIV risk behaviours 7 between the intervention and comparator on the completion of intervention. HIV risk behaviours 8 4 9 5 include sex types (vaginal or anal), frequency of condom protected sex and, whether engaging in sex 10 6 with other partners concurrently. Injection risk behaviour will be defined as having shared syringes, 11 12 7 containers, filters or water to inject drugs in the previous month and backloading/frontloading. We 13 8 will review all the acquired full-texts to check the relative scale used by each study, for example the 14 9 HIV Risk-Taking Behaviour Scale,14, 15 to assess the above indicators. 15 16 10 17 11 Risk of bias in individual studies 18 12 We will use the CochraneFor revisedpeer tool, Riskreview of Bias (RoB only V.2.0),16 to determine the risk of bias. 19 20 13 The RoB tool contains five key domains: (1) randomisation process; (2) deviations from intended 21 14 interventions; (3) missing outcome data; (4) measurement of the outcome; and (5) selective 22 15 reporting. The risk of bias will be classified into three types for each domain: high, low, or some 23 24 16 concerns. Subsequently, we will arrive at an overall risk of bias, based on judgements from the five 25 17 domains. A summary of risk of bias of all the domains will be provided for each trial. The first two 26 authors will perform all assessments independently of each other. Any disagreements will be 27 18 28 19 resolved by discussion with a third author. 29 20 30 31 21 Data synthesis 32 22 We will use Stata software (13.0; Stata Corporation, College Station, Texas, USA) to conduct a 33 23 traditional pairwise meta-analysis. If more than five studies are included, we will use the random 34
35 24 effects model to combine the data. Otherwise, we will use a fixed effect model, because the random http://bmjopen.bmj.com/ 36 25 effects model may be imprecise in this situation.17 Dichotomous data will be evaluated using the risk 37 26 ratio (RR) with 95% confidence interval (CI), while continuous outcomes will be expressed as 38 2 39 27 standardized mean differences and 95% CI. Heterogeneity will be quantified using the I statistic. If 40 28 I2 > 50%, which indicates the presence of substantial heterogeneity,18 we will consider subgrouping 41 29 the intervention by study setting (receiving formal treatment or not), gender-specific (single-gender 42
43 30 or mixed gender), HIV testing (reported or not), methodological quality of the study, drug types or on September 30, 2021 by guest. Protected copyright. 44 31 geographical area. 45 46 32 When pairwise meta-analysis is completed, we will perform a network meta-analysis using 47 33 WinBUGS 1.43 software. The Markov Chains Monte Carlo method will be used for Bayesian 48 34 analysis. When we run the WinBUGS program, we will set it to perform 100 000 simulations, and 49 50 35 the first 10 000 simulations will be discarded as burn-in. Convergence of the model will be assessed 51 36 by trace and Gelman-Rubin-Brooks plots.19 52 37 We will use both random- and fixed-effects models for the network meta-analysis. Then we will 53 54 38 select the appropriate model on the basis of the deviance information criterion (DIC); the model with 55 39 the lower DIC will be preferred (a difference > 3 will be considered significant).20 The node-splitting 56 40 method will be used to statistically assess the consistency between direct and indirect evidence.21 57 58 41 We will examine the assumptions of transitivity (similarity in the distribution of potential effect 59 42 modifiers across the different pairwise comparisons) on account of clinical and methodological 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 8
1 BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from 2 3 1 characteristics; notably, there is no universal statistical method to analyse these effect modifiers.22-23 4 5 2 We plan to investigate similarity based on factors including participant characteristics, experimental 6 3 design, study quality, and risk of bias, among others. 7 Furthermore, publication bias will be assessed using comparison-adjusted funnel plots. All the 8 4 9 5 figures, including forest plots for each intervention, network plots, and comparison-adjusted funnel 10 6 plots, will be produced using the ‘Network Graphs’ package in STATA. 11 12 7 We will performed sensitivity analysis to address whether the combined estimates of the 13 8 interventions are dominated by one or several studies, especially those with a high risk of bias. Then 14 9 we will exclude the trials to test the robustness of our study result. Second, we will test whether the 15 16 10 imputation of the missing values affects the result of the meta-analysis. 17 11 18 12 Confidence in cumulativeFor evidence peer review only 19 20 13 We will evaluate the quality of evidence for all outcomes according to the Grading of 21 14 Recommendations Assessment, Development and Evaluation (GRADE). Based on the domains of 22 15 methodology quality, consistency, directness, precision effect estimates, and publication bias, we 23 24 24 16 will rank the overall strength of evidence as high, moderate, low, or very low. 25 17 26 Patient and public involvement 27 18 28 19 Patients will not be involved. 29 20 30 31 21 ETHICS AND DISSEMINATION 32 22 This Bayesian network meta-analysis will include no confidential personal data and no data on 33 23 human trials. Consequently, ethical approval is not required. The procedures used for this study will 34
35 24 be reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta- http://bmjopen.bmj.com/ 36 25 Analysis (PRISMA) extension statement for network meta-analyses of healthcare interventions.25 37 26 The final results will be disseminated at professional conferences and through publications in peer- 38 39 27 reviewed journals. 40 28 41 29 FOOTNOTES 42
43 30 on September 30, 2021 by guest. Protected copyright. 44 31 Contributors: JJL contributed to the conception and design of this study protocol and is the guarantor 45 46 32 of the protocol. YSY developed the search strategy. LRJ will provide advice on data analysis and 47 33 presentation of study results. All authors approved the final version of the manuscript. 48 34 49 50 35 Funding: Anhui Provincial Department of Education upgrade quality project - Master 51 36 Studio(2014msgzs151); Key Projects for Academic Support of Top-level Talents in 52 37 Universities(gxbjZD2016073) 53 54 38 55 39 Disclaimer: None. 56 40 57 58 41 Competing interests: None. 59 42 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 8 BMJ Open
1 BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from 2 3 1 Provenance and peer review: Not commissioned; externally peer reviewed. 4 5 2 6 7 3 REFERENCES 8 9 4 10 5 1 World Health Organization, United Nations Office on Drugs and Crime, and Joint United Nations Programme on HIV/AIDS. 11 WHO, UNODC, UNAIDS technical guide for countries to set targets for universal access to HIV prevention,treatment and care 12 6 13 7 for injecting drug users. Geneva: World Health Organization; 2012 14 8 http://apps.who.int/iris/bitstream/10665/77969/1/9789241504379_eng.pdf?ua=1(accessed 24 Dec 2017). 15 16 9 2 United Nations Office on Drugs and Crime. World drug report 2018. 17 10 http://www.unodc.org/wdr2018/prelaunch/WDR18_Booklet_1_EXSUM.pdf (accessed 25 Aug 2018) 18 11 3 UNAIDS. GLOBAL AIDSFor UPDATE peer 2016. http://www.who.int/hiv/pub/arv/global-aids-update-2016-pub/en/(accessed review only 25 Dec 19 20 12 2017). 21 13 4 Degenhardt L, Hall W, Lynskey M, et al. Illicit drug use. In: Ezzati M, Lopez AD, Rodgers A, et al, eds. Comparative 22 14 quantification of health risks: global and regional burden of disease attributable to selected major risk factors. 2nd edn. Geneva: 23 24 15 World Health Organization, 2004: 1109–1176. 25 16 5 Jones L, Pickering L, Sumnall H, et al. Optimal provision of needle and syringe programmes for injecting drug users: A 26 17 systematic review. Int J Drug Policy 2010;21:335-42. 27 28 18 6 Meader N, Li R, Des Jarlais DC, et al. Psychosocial interventions for reducing injection and sexual risk behaviour for 29 19 preventing HIV in drug users. Cochrane Database Syst Rev 2010:CD007192. 30 7 Semaan S, Des Jarlais DC, Sogolow E, et al. A meta-analysis of the effect of HIV prevention interventions on the sex 31 20 32 21 behaviors of drug users in the United States. J Acquir Immune Defic Syndr 2002;30 Suppl 1:S73-93. 33 22 8 Prendergast ML, Urada D, Podus D. Meta-analysis of HIV risk-reduction interventions within drug abuse treatment programs. 34
35 23 J Consult Clin Psychol 2001;69:389-405. http://bmjopen.bmj.com/ 36 24 9 Fernandes RM, Cary M, Duarte G, et al. Effectiveness of needle and syringe Programmes in people who inject drugs - An 37 25 overview of systematic reviews. BMC Public Health 2017;17:309. 38 39 26 10 Biondi-Zoccai G, Abbate A, Benedetto U, et al. Network meta-analysis for evidence synthesis: what is it and why is it posed to 40 27 dominate cardiovascular decision making. Int J Cardiol 2015;182:309-14. 41 28 11 Greco T, Landoni G, Biondi-Zoccai G, et al. A Bayesian network meta-analysis for binary outcome: how to do it. Stat Methods 42
43 29 Med Res 2016;25:1757-73. on September 30, 2021 by guest. Protected copyright. 44 30 12 Shamseer L, Moher D, Clarke M, et al. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA- 45 31 P) 2015: elaboration and explanation. BMJ 2015;350:g7647. 46 47 32 13 Mathers BM, Degenhardt L, Ali H, et al. HIV prevention, treatment, and care services for people who inject drugs: a 48 33 systematic review of global, regional, and national coverage. Lancet 2010;375:1014-28. 49 14 Darke S, Hall W, Heather N, et al. The reliability and validity of a scale to measure HIV risk-taking behaviour among 50 34 51 35 intravenous drug users. AIDS 1991;5:181-5. 52 36 15 Ward J, Darke S, Hall W. The HIV Risk-taking behaviour scale (HRBS) manual. 1990. 53 54 37 16 Higgins JPT, Sterne JAC, Savović J, et al. A revised tool for assessing risk of bias in randomized trials. In: Chandler J, 55 38 McKenzie J, BoutronI, eds. Cochrane methods cochrane database of systematic reviews,2016. 56 39 17 Higgins JP, Thompson SG, Spiegelhalter DJ. A re-evaluation of random-effects meta-analysis. J R Stat Soc Ser A Stat Soc 57 58 40 2009;172:137-59. 59 41 18 Higgins JP, Thompson SG, Deeks JJ, et al. Measuring inconsistency in meta-analyses. BMJ 2003;327:557-60. 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 8
1 BMJ Open: first published as 10.1136/bmjopen-2018-022811 on 28 January 2019. Downloaded from 2 3 1 19 Gelman A, Rubin DB. Inference from Iterative Simulation Using Multiple Sequences. Statistical Science 1992;7:457-72. 4 5 2 20 Spiegelhalter DJ, Best NG, Carlin BP, et al. Bayesian measures of model complexity and fit. Journal of the Royal Statistical 6 3 Society: Series B (Statistical Methodology) 2002;64:583-639. 7 4 21 Dias S, Welton NJ, Caldwell DM, et al. Checking consistency in mixed treatment comparison meta‐analysis. Stat Med 8 9 5 2010;29:932-44. 10 6 22 Kim H, Gurrin L, Ademi Z, et al. Overview of methods for comparing the efficacies of drugs in the absence of head-to-head 11 clinical trial data. Br J Clin Pharmacol 2014;77:116-21. 12 7 13 8 23 Jansen JP ,Naci H . Is network meta-analysis as valid as standard pairwise meta-analysis? It all depends on 14 9 the distribution of effect modifiers. BMC Med 2013;11:159 15 16 10 24 Schunemann H, Brozek J, Guyatt G, et al. GRADE handbook for grading quality of evidence and strength of recommendation. 17 11 2013http:// gdt. guidelinedevelopment. org/ app/ (assessed 22 Dec 2017). 18 12 25 Hutton B, Salanti G, CaldwellFor DM, peeret al. The PRISMA review extension statement only for reporting of systematic reviews incorporating 19 20 13 network meta-analyses of health care interventions: checklist and explanations. Ann Intern Med 2015;162:777-84. 21 22 14 23 24 25 15 26 27 28 29 30 31 32 33 34
35 http://bmjopen.bmj.com/ 36 37 38 39 40 41 42
43 on September 30, 2021 by guest. Protected copyright. 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml