BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. 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.

Whe a atile is pulished e post the pee eiees’ oets ad the authos’ esposes 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 hae ay uestios o BMJ Ope’s ope pee eie poess please eail [email protected]

http://bmjopen.bmj.com/ on September 30, 2021 by guest. Protected copyright. BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open

Comparative efficacy and acceptability of interventions for major depression in older persons: protocol for Bayesian network meta-analysis

ForJournal: peerBMJ Open review only Manuscript ID bmjopen-2017-019819

Article Type: Protocol

Date Submitted by the Author: 27-Sep-2017

Complete List of Authors: Liew, Tau Ming; Institute of Mental Health, Department of Geriatric Psychiatry

Primary Subject Mental health Heading:

Secondary Subject Heading: Geriatric medicine

major depression, older person, efficacy, acceptability, network meta- Keywords: analysis

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-2017-019819 on 21 January 2018. Downloaded from Page 1 of 21 BMJ Open

1 2 3 Comparative efficacy and acceptability of interventions for major depression in older 4 5 persons: protocol for Bayesian network meta-analysis 6 7 8 9 Tau Ming Liew1, 2 10 11 1Department of Geriatric Psychiatry, Institute of Mental Health, Singapore 12 13 2 14 Saw Swee Hock School of Public Health, National University of Singapore, Singapore 15 16 For peer review only 17 18 19 20 Correspondence to 21 22 Tau Ming Liew; 23 24 [email protected] 25 26 27 Department of Geriatric Psychiatry, Institute of Mental Health, 28 29 10 Buangkok View, Singapore 539747. 30 31 http://bmjopen.bmj.com/ 32 33 Keywords: major depression; older person; efficacy; acceptability; network meta-analysis 34 35 36 37 Number of words (Abstract): 287 38 39 on September 30, 2021 by guest. Protected copyright. 40 Number of words (main text): 2,009 41 42 Number of references: 26 43 44 Number of tables or figures: 1 45 46 47 48 49 50 51 52 53 54 55 56 57 58 1 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 2 of 21

1 2 3 ABSTRACT 4 5 6 7 Introduction: Major depression is a leading cause of disability, and has been associated with 8 9 adverse effects in older persons. While many pharmacological and non-pharmacological 10 11 interventions have been shown to be effective to address major depression in older persons, 12 13 14 there has not been a meta-analysis that consolidates all the available interventions and 15 16 compare the relativeFor benefits peer of these review available interventions. only In this study, we aim to 17 18 conduct a systematic review and network meta-analysis to compare the efficacy and 19 20 acceptability of all the known pharmacological and non-pharmacological interventions for 21 22 major depression in older persons. 23 24 Methods and analysis: We will search PubMed, Embase, PsycINFO, CINAHL, Web of 25 26 27 Science, Scopus, Cochrane Central Register of Controlled Trials and references of other 28 29 review articles for articles related to the keywords of ”randomized trial‘, ”major depression‘ 30 31 and ”older persons‘. Two reviewers will independently select the eligible articles. For each http://bmjopen.bmj.com/ 32 33 included article, the two reviewers will independently extract the data and assess the risk of 34 35 bias using the Cochrane risk of bias tool. Bayesian network meta-analyses will be conducted 36 37 to pool the efficacy (based on standardized mean difference of depression score) and all- 38 39 on September 30, 2021 by guest. Protected copyright. 40 cause attrition across all the included studies. The ranking probabilities for all interventions 41 42 will be estimated and the hierarchy of each interventions will be summarized as surface under 43 44 the cumulative ranking curve (SUCRA). Meta-regression and sub-group analyses will also 45 46 be performed to evaluate the effect of study-level covariates. The quality of the evidence will 47 48 be assessed using the Grading of Recommendations Assessment, Development and 49 50 Evaluation (GRADE) approach. 51 52 53 Ethics and dissemination: The results will be disseminated through conference 54 55 presentations and peer-reviewed publications. They will provide the consolidated evidence to 56 57 58 2 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 3 of 21 BMJ Open

1 2 3 inform clinicians on the best choice of intervention to address major depression in older 4 5 persons. 6 7 Trial registration number: International Prospective Register for Systematic Reviews 8 9 (PROSPERO) temporary registration number 75756 (submitted on 30th August 2017). 10 11 12 13 14 15 16 For peer review only 17 18 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 3 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 4 of 21

1 2 3 STRENGTHS AND LIMITATIONS OF THIS STUDY 4 5 6 7 • This systematic review and meta-analysis will provide a comprehensive summary on the 8 9 efficacy and acceptability of all available interventions for major depression in older 10 11 12 persons. 13 14 • The results will provide the highest level of evidence to inform clinicians on the best 15 16 choice of treatment,For peer from among review the many available only pharmacological and non- 17 18 pharmacological interventions. 19 20 • This protocol has been developed in accordance with the Preferred Reporting Items for 21 22 23 Systematic Review and Meta-analysis Protocols (PRISMA-P) statement and has been 24 25 submitted for registration with PROSPERO. 26 27 • The overall quality of evidence will be assessed using the Grading of Recommendations 28 29 Assessment, Development and Evaluation (GRADE) approach. 30 http://bmjopen.bmj.com/ 31 • This systematic review will be limited to studies in English language. 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 4 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 5 of 21 BMJ Open

1 2 3 INTRODUCTION 4 5 6 7 8 Rationale 9 10 11 12 13 Major depression has been identified by the World Health Organization as one of the leading 14 15 cause of disability globally.1 2 In older persons, its prevalence rates rise with the increase in 16 For peer review only 17 medical comorbidities,3 with reported rates of up to 5% in community-dwelling older 18 19 persons,3-5 5 to 10% in primary care3 6 and as high as 37% after critical care hospitalizations.3 20 21 7 Major depression has a significant impact on the older populations and has been linked to 22 23 4 8 9 4 10 24 higher risk of suicide, myocardial infarction, stroke, all-cause mortality and increasing

25 4 26 health services utilization. 27 28 29 30 A wide range of interventions have been available to treat major depression in older persons. 31 http://bmjopen.bmj.com/ 32 These include pharmacological and non-pharmacological interventions such as 33 34 ,11 ,12 cognitive behavioural therapy,13 problem solving 35 36 14 15 16 37 therapy, family interventions and physical exercise. Some of these interventions also

38 11 39 have had recent meta-analyses confirming their efficacy when compared to control groups. on September 30, 2021 by guest. Protected copyright. 40 13 14 16 41 However, none of the meta-analyses had provided comparisons among all the 42 43 pharmacological and non-pharmacological interventions to demonstrate the relative benefits 44 45 of each intervention. It is unknown whether all the interventions have comparable efficacy 46 47 and are equally suitable for older persons with major depression. 48 49 50 51 52 Objectives 53 54 55 56 57 58 5 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 6 of 21

1 2 3 In this study, we aim to conduct a systematic review and network meta-analysis to compare 4 5 the efficacy and acceptability of all the available pharmacological and non-pharmacological 6 7 interventions for major depression in older persons. The use of network meta-analysis allows 8 9 us to pool the evidence on various interventions and rank their benefits relative to each 10 11 other.17 It also allows us to conduct indirect comparison of the different interventions, even 12 13 14 when there is no direct evidence in the literature to allow head-to-head comparisons. 15 16 For peer review only 17 18 19 20 METHODS AND ANALYSIS 21 22 23 24 This protocol is developed in accordance with the Preferred Reporting Items for Systematic 25 26 18 19 27 Review and Meta-analysis (PRISMA) statement. It has also been submitted to the 28 29 International Prospective Register of Systematic Reviews (PROSPERO) for registration 30 31 (temporary registration number 75756, submitted on 30th August 2017). http://bmjopen.bmj.com/ 32 33 34 35 Eligibility criteria 36 37

38 39 on September 30, 2021 by guest. Protected copyright. 40 Participants and settings 41 42 43 44 We will include studies which recruit participants who are: (1) 60 years old and above; (2) 45 46 diagnosed with major depression based on formal criteria by the Diagnostic and Statistical 47 48 Manual of Mental Disorders (DSM) or International Classification of Diseases (ICD); and (3) 49 50 having a current episode of major depression (that is, the participants are currently 51 52 53 symptomatic and not in remission). 54 55 56 57 58 6 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 7 of 21 BMJ Open

1 2 3 We will exclude studies which recruit participants with treatment-resistant depression, 4 5 subthreshold depression, bipolar depression or psychotic depression. We will also exclude 6 7 participants who have major depression but are currently asymptomatic or in remission. 8 9 10 11 Interventions 12 13 14 15 16 We will includeFor studies peer which report review on pharmacologi onlycal interventions (such as 17 18 antidepressants, antipsychotics or other class of ) or non-pharmacological 19 20 interventions (such as cognitive behavioural therapy, problem solving therapy, 21 22 psychodynamic therapy or physical exercise). We will also include studies which report on 23 24 combinations of any of these pharmacological and non-pharmacological interventions. 25 26 27 28 29 Comparators 30 31 http://bmjopen.bmj.com/ 32 33 We will accept control conditions such as placebo intervention, waiting-list, treatment as 34 35 usual, as well as no intervention. We will also include studies with active comparators such 36 37 as those which compare between two different interventions within the same studies. 38 39 on September 30, 2021 by guest. Protected copyright. 40 41 42 Outcomes 43 44 45 46 We will only include a study if it reports at least one of the following outcome measures: (1) 47 48 depression score at the immediate post-intervention period; (2) proportion of participants in 49 50 each study arm with at least 50% improvement in depression score following intervention 51 52 53 (response rate); (3) Clinical Global ImpressionœImprovement scale (CGI-I); or (4) all-cause 54 55 attrition in each study arm at the immediate post-intervention period. 56 57 58 7 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 8 of 21

1 2 3 4 5 Study designs 6 7 8 9 We will only include randomized controlled trials (RCTs). The following study designs will 10 11 be excluded: qualitative studies, observational studies, non-randomized trials, reviews, meta- 12 13 14 analyses, case reports, case series, ecological studies, conference proceedings, letters, 15 16 comments and policyFor papers. peer review only 17 18 19 20 Language and time frame 21 22 23 24 We will only include studies which are reported in the English language. Apart from that, we 25 26 27 do not impose any time restriction to the publication year of the studies. 28 29 30 31 Information sources and search strategy http://bmjopen.bmj.com/ 32 33 34 35 We will search PubMed, Embase, PsycINFO, CINAHL, Web of Science, Scopus and 36 37 Cochrane Central Register of Controlled Trials for original articles related to the keywords of 38 39 on September 30, 2021 by guest. Protected copyright. 40 ”randomized trial‘, ”major depression‘ and ”older persons‘. Our search strategy for PubMed 41 42 is shown in Box 1. Similar search strategies will be used for the other databases. 43 44 Additionally, we will also hand-search the references of review articles related to the topic to 45 46 retrieve relevant articles which are not captured through our search of the electronic 47 48 databases. 49 50

51 52 53 Box 1. Search strategy for PubMed (MeSH, Medical Subject Headings) 54 55 1. "Randomized Controlled Trial" [Publication Type] OR 56 57 58 8 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 9 of 21 BMJ Open

1 2 3 2. "Randomized Controlled Trials as Topic"[Mesh] OR 4 5 3. "Random Allocation"[Mesh] OR 6 7 4. (random*[title/abstract] AND trial*[title/abstract])) 8 9 5. 1 or 2 or 3 or 4 10 11 12 6. "Depressive Disorder, Major/drug therapy"[Mesh] OR 13 14 7. "Depressive Disorder, Major/therapy"[Mesh] OR 15 16 8. (depress*[title]For AND major[title])peer ) review only 17 18 9. 6 or 7 or 8 19 20 10. elder*[title] OR 21 22 11. older[title] AND 23 24 25 12. (person*[title] OR people[title] OR adult*[title])) OR 26

27 13. 11 and 12 28 29 14. (late[title] AND life[title]) OR 30 31 15. geriatric[title]) http://bmjopen.bmj.com/ 32 33 16. 10 or 13 or 14 or 15 34 35 17. 5 and 9 and 16 36 37 38 39 on September 30, 2021 by guest. Protected copyright. 40 Study selection 41 42 43 44 All potential articles will be retrieved and organized in a data management software (Endnote 45 46 software, Thomson Reuters). After removing duplicate records, two reviewers will 47 48 independently screen through the titles and abstracts to retain eligible articles. The first 10% 49 50 of these titles and abstracts will be subjected to a calibration exercise between the two 51 52 53 reviewers to ensure mutual agreement. 54 55 56 57 58 9 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 10 of 21

1 2 3 After completing the screening phase, articles that are deemed as relevant by at least one of 4 5 the reviewers will be subjected to full-text review. The two reviewers will independently 6 7 confirm the eligibility of these articles based on the full texts. The first 10% of these full 8 9 texts will again undergo a calibration exercise by the two reviewers. After the full-text review, 10 11 the included articles will be used for qualitative synthesis. The chance-corrected agreement 12 13 14 between the two reviewers will be assessed using Cohen‘s Kappa (κ). 15 16 For peer review only 17 18 At any point during study selection, the reasons for excluding specific articles will be 19 20 recorded. Moreover, any disagreements between the two reviewers will be resolved by 21 22 discussion with a third reviewer. 23 24 25 26 27 Data extraction 28 29 30 31 Data from the selected studies will be extracted by two reviewers independently, and http://bmjopen.bmj.com/ 32 33 disagreements between the reviewers will be resolved by discussion with a third reviewer. 34 35 The extracted data will include the following information: 36 37 38 1. Study identification (first author, year of publication, geographic location) 39 on September 30, 2021 by guest. Protected copyright. 40 2. Study characteristics (study setting, study design, inclusion criteria, diagnostic criteria 41 42 of major depression, sample size, study duration) 43 44 3. Participant characteristics (age, gender, education, number of comorbidities, Mini 45 46 Mental State Examination score, baseline depression score, depression scale, duration 47 48 of the current episode of major depression) 49 50 51 4. Characteristics of intervention and comparator (description, depression score, all- 52 53 cause attrition) 54 55 56 57 58 10 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 11 of 21 BMJ Open

1 2 3 The original authors of the RCTs will be contacted when the required data are not available in 4 5 the published article. 6 7 8 9 Assessment of risk of bias 10 11

12 13 14 The risk of bias for each study will be assessed independently by two reviewers using the 15 20 16 Cochrane risk ofFor bias tool, peer focusing onreview the key criteria only of random sequence generation, 17 18 allocation concealment, blinding of outcome assessment, completeness of outcome data and 19 20 selective outcome reporting. Each criterion will be assigned a high, low or unclear risk of 21 22 bias. Any disagreements between the two reviewers will be resolved by discussion with a 23 24 third reviewer. 25 26 27 28 29 Outcome measures 30 31 http://bmjopen.bmj.com/ 32 33 Our primary outcomes are the efficacy and the acceptability of interventions. The efficacy 34 35 will be based on difference in the depression scores between the intervention and comparator 36 37 at the immediate post-intervention period, computed as standardized mean difference (SMD) 38 39 on September 30, 2021 by guest. Protected copyright. 40 for each RCT. The acceptability will be assessed by the relative risk (RR) of all-cause 41 42 attrition at the immediate post-intervention period. This will be based on information 43 44 extracted from each RCT, by subtracting those who were still available for data collection at 45 46 the immediate post-intervention period from those who were randomized at the start of the 47 48 RCT. Additionally, we will include a secondary outcome based on the RR of response rate at 49 50 the immediate post-intervention period. We define response rate as the proportion of 51 52 53 participants who have at least 50% improvement in depression score, or score much or very 54 55 much improved on the Clinical Global ImpressionœImprovement scale (CGI-I<3). 56 57 58 11 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 12 of 21

1 2 3 4 5 Statistical analysis 6 7 8 9 We will conduct the network meta-analyses within a Bayesian framework using the Markov 10 11 Chains Monte Carlo method. Bayesian analysis provides probabilistic distributions of our 12 13 14 estimates-of-interest through large number of simulations, and hence produces results which 15 16 have more intuitiveFor interpretations. peer Forreview example, Bayesian only analysis generates the 95% 17 18 credible interval which can be accurately interpreted as the range containing 95% of the 19 20 estimates (based on the simulations). In the Bayesian analysis, we will run four Markov 21 22 chains simultaneously with different arbitrarily chosen initial values and with non- 23 24 informative priors. Each chain will have at least 10,000 simulations and at least the first 25 26 27 2,500 simulations will be discarded as burn-in. Convergence of the simulations will be 28 29 assessed with the trace plots, kernel density plots and Gelman-Rubin-Brooks plots. 30 31 http://bmjopen.bmj.com/ 32 33 We will employ both fixed-effects and random-effects models in the Bayesian analyses, and 34 35 will choose the final models based on the deviance information criterion (DIC). While there 36 37 is no rule-of-thumb on what constitute significant improvements in DIC, we can take 38 on September 30, 2021 by guest. Protected copyright. 39 21 40 reference from the guideline commonly used in the analogous Akaike Information Criteria: 41 42 values which are lesser by at least 10 points indicate significantly better model-fit and 43 44 parsimony. Hence, results from the random-effects model will be used if the random-effects 45 46 model has DIC which is smaller by at least 10 points compared to the fixed-effect model. We 47 48 will also compare the complexity of model between the fixed-effects and random-effects 49 50 models using pD (an indicator which has higher value when a model is more complex), with 51 52 53 preference for models which are more parsimonious (less complex). The global 54 2 55 heterogeneity will be assessed with I statistic. A common heterogeneity parameter will be 56 57 58 12 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 13 of 21 BMJ Open

1 2 3 assumed in the random-effects model. Inconsistency between direct and indirect sources of 4 5 evidence will be statistically assessed, by calculating the difference between direct and 6 7 indirect estimates in each closed loop in the network.22 8 9 10 11 We will estimate the ranking probabilities for all interventions and show the results 12 13 14 graphically in the form of rankograms and cumulative ranking probability plots. The 15 16 hierarchy of interventionsFor peerwill be summarized review as surface under only the cumulative ranking curve 17 18 (SUCRA) and presented in a scatterplot. SUCRAs have possible values ranging from 0% to 19 20 100%, with higher values indicating better efficacy or acceptability. Publication bias will be 21 22 assessed with comparison-adjusted funnel plot.23 24 23 24

25 26 27 We will conduct meta-regression analyses to determine whether the results of our network 28 29 meta-analyses will be affected by the following study-level covariates: sample size, study 30 31 duration, inclusion criteria, study setting, study design and risk of bias. A covariate is http://bmjopen.bmj.com/ 32 33 considered as a significant moderator if the 95% credible interval of its beta coefficient in 34 35 meta-regression does not include the value of zero. If a significant moderator is found, 36 37 further subgroup analyses will then be conducted to assess the effect of this moderator. 38 39 on September 30, 2021 by guest. Protected copyright. 40 41 42 The network meta-analyses will be conducted using JAGS (version 4.2.0), through the 43 44 GeMTC package of R (version 3.3.1). The —Network Graphs“ package in Stata statistical 45 46 software (version 14.0) will also used to produce some of the figures in this study, such as the 47 48 network plots, rankograms, cumulative ranking probability plots and comparison-adjusted 49 50 funnel plots. 23 25 51 52 53 54 55 Assessment of quality of evidence 56 57 58 13 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 14 of 21

1 2 3 4 5 We will use the Grading of Recommendations Assessment, Development and Evaluation 6 7 (GRADE) approach to report the quality of evidence on efficacy and acceptability of 8 9 interventions for major depression in older persons. Based on five key domains 10 11 (methodology quality, directness of evidence, heterogeneity, precision of effect estimates and 12 13 14 risk of publication bias), we will classify the quality of evidence in one of four levels œ high, 15 26 16 moderate, low andFor very low. peer review only 17 18 19 20 21 22 ETHICS AND DISSEMINATION 23 24

25 26 27 This systematic review will provide the consolidated evidence to inform clinicians on the best 28 29 choice of intervention, from among the many available options, to address major depression 30 31 in older persons. This systematic review will be reported in accordance with the http://bmjopen.bmj.com/ 32 33 recommendations of PRISMA statement.18 19 The results will be disseminated through 34 35 conference presentations and publications in peer-reviewed journal. 36 37

38 39 on September 30, 2021 by guest. Protected copyright. 40 41 42 FUNDING 43 44 45 46 TML was supported by a research fellowship under the Singapore Ministry of Health‘s 47 48 National Medical Research Council (Grant number: NMRC/Fellowship/0030/2016). The 49 50 funding source had no involvement in any part of the project. 51 52 53 54 55 56 57 58 14 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 15 of 21 BMJ Open

1 2 3 COMPETING INTERESTS 4 5 6 7 None declared. 8 9 10 11 12 13 14 15 16 For peer review only 17 18 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 15 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 16 of 21

1 2 3 REFERENCES 4 5 6 7 8 1. World Health Organization. The global burden of disease: 2004 update. Switzerland: 9 10 11 World Health Organization. 12 13 2. World Health Organization. Global Health Estimates 2015: Disease burden by Cause, Age, 14 15 Sex, by Country and by Region, 2000-2015. Geneva, 2016. 16 For peer review only 17 3. Taylor WD. Clinical practice. Depression in the elderly. The New England journal of 18 19 medicine 2014;371(13):1228-36. doi: 10.1056/NEJMcp1402180 [published Online 20 21 First: 2014/09/25] 22 23 24 4. Blazer DG. Depression in late life: review and commentary. The journals of gerontology 25 26 Series A, Biological sciences and medical sciences 2003;58(3):249-65. [published 27 28 Online First: 2003/03/14] 29 30 5. Volkert J, Schulz H, Härter M, et al. The prevalence of mental disorders in older people in 31 http://bmjopen.bmj.com/ 32 Western countries œ a meta-analysis. Ageing Research Reviews 2013;12(1):339-53. 33 34 doi: http://dx.doi.org/10.1016/j.arr.2012.09.004 35 36 37 6. Lyness JM, Caine ED, King DA, et al. Psychiatric Disorders in Older Primary Care 38 39 Patients. Journal of general internal medicine 1999;14(4):249-54. doi: on September 30, 2021 by guest. Protected copyright. 40 41 10.1046/j.1525-1497.1999.00326.x 42 43 7. Jackson JC, Pandharipande PP, Girard TD, et al. Depression, Posttraumatic Stress Disorder, 44 45 and Functional Disability in Survivors of Critical Illness: results from the BRAIN 46 47 ICU (Bringing to light the Risk Factors And Incidence of Neuropsychological 48 49 50 dysfunction in ICU survivors) Investigation: A Longitudinal Cohort Study. The lancet 51 52 Respiratory medicine 2014;2(5):369-79. doi: 10.1016/S2213-2600(14)70051-7 53 54 55 56 57 58 16 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 17 of 21 BMJ Open

1 2 3 8. Gan Y, Gong Y, Tong X, et al. Depression and the risk of coronary heart disease: a meta- 4 5 analysis of prospective cohort studies. BMC psychiatry 2014;14:371. doi: 6 7 10.1186/s12888-014-0371-z [published Online First: 2014/12/30] 8 9 9. Pan A, Sun Q, Okereke OI, et al. Depression and risk of stroke morbidity and mortality: a 10 11 meta-analysis and systematic review. Jama 2011;306(11):1241-9. doi: 12 13 14 10.1001/jama.2011.1282 [published Online First: 2011/09/22] 15 16 10. Cuijpers P,For Vogelzangs peer N, Twisk review J, et al. Comprehensive only meta-analysis of excess 17 18 mortality in depression in the general community versus patients with specific 19 20 illnesses. The American journal of psychiatry 2014;171(4):453-62. doi: 21 22 10.1176/appi.ajp.2013.13030325 [published Online First: 2014/01/18] 23 24 11. Tham A, Jonsson U, Andersson G, et al. Efficacy and tolerability of antidepressants in 25 26 27 people aged 65 years or older with major depressive disorder œ A systematic review 28 29 and a meta-analysis. Journal of affective disorders 2016;205:1-12. doi: 30 31 http://dx.doi.org/10.1016/j.jad.2016.06.013 http://bmjopen.bmj.com/ 32 33 12. Katila H, Mezhebovsky I, Mulroy A, et al. Randomized, double-blind study of the 34 35 efficacy and tolerability of extended release fumarate (quetiapine XR) 36 37 monotherapy in elderly patients with major depressive disorder. The American 38 39 on September 30, 2021 by guest. Protected copyright. 40 journal of geriatric psychiatry : official journal of the American Association for 41 42 Geriatric Psychiatry 2013;21(8):769-84. doi: 10.1016/j.jagp.2013.01.010 [published 43 44 Online First: 2013/04/10] 45 46 13. Gould RL, Coulson MC, Howard RJ. Cognitive behavioral therapy for depression in 47 48 older people: a meta-analysis and meta-regression of randomized controlled trials. 49 50 Journal of the American Geriatrics Society 2012;60(10):1817-30. doi: 51 52 53 10.1111/j.1532-5415.2012.04166.x [published Online First: 2012/09/26] 54 55 56 57 58 17 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 18 of 21

1 2 3 14. Kirkham JG, Choi N, Seitz DP. Meta-analysis of problem solving therapy for the 4 5 treatment of major depressive disorder in older adults. International journal of 6 7 geriatric psychiatry 2016;31(5):526-35. doi: 10.1002/gps.4358 [published Online 8 9 First: 2015/10/06] 10 11 15. Stahl ST, Rodakowski J, Saghafi EM, et al. Systematic review of dyadic and familyU 12 13 14 oriented interventions for lateUlife depression. International journal of geriatric 15 16 psychiatryFor 2016;31(9):963-73. peer doi: review 10.1002/gps.4434 only 17 18 16. Schuch FB, Vancampfort D, Rosenbaum S, et al. Exercise for depression in older adults: 19 20 a meta-analysis of randomized controlled trials adjusting for publication bias. Revista 21 22 brasileira de psiquiatria (Sao Paulo, Brazil : 1999) 2016;38(3):247-54. doi: 23 24 10.1590/1516-4446-2016-1915 [published Online First: 2016/09/10] 25 26 27 17. Tonin FS, Rotta I, Mendes AM, et al. Network meta-analysis: a technique to gather 28 29 evidence from direct and indirect comparisons. Pharmacy practice 2017;15(1):943. 30 31 doi: 10.18549/PharmPract.2017.01.943 [published Online First: 2017/05/16] http://bmjopen.bmj.com/ 32 33 18. Moher D, Shamseer L, Clarke M, et al. Preferred reporting items for systematic review 34 35 and meta-analysis protocols (PRISMA-P) 2015 statement. Systematic reviews 36 37 2015;4:1. doi: 10.1186/2046-4053-4-1 [published Online First: 2015/01/03] 38 39 on September 30, 2021 by guest. Protected copyright. 40 19. Shamseer L, Moher D, Clarke M, et al. Preferred reporting items for systematic review 41 42 and meta-analysis protocols (PRISMA-P) 2015: elaboration and explanation. BMJ 43 44 (Clinical research ed) 2015;349:g7647. doi: 10.1136/bmj.g7647 [published Online 45 46 First: 2015/01/04] 47 48 20. Higgins JPT, Green S. Cochrane Handbook for Systematic Reviews of Interventions: 49 50 Wiley 2011. 51 52 53 21. Burnham KP, Anderson DR. Model Selection and Multimodel Inference: A Practical 54 55 Information-Theoretic Approach: Springer New York 2003. 56 57 58 18 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 19 of 21 BMJ Open

1 2 3 22. Dias S, Welton NJ, Caldwell DM, et al. Checking consistency in mixed treatment 4 5 comparison meta-analysis. Statistics in medicine 2010;29(7-8):932-44. doi: 6 7 10.1002/sim.3767 [published Online First: 2010/03/10] 8 9 23. Chaimani A, Higgins JP, Mavridis D, et al. Graphical tools for network meta-analysis in 10 11 STATA. PloS one 2013;8(10):e76654. doi: 10.1371/journal.pone.0076654 [published 12 13 14 Online First: 2013/10/08] 15 16 24. Peters JL, SuttonFor AJ, Jonespeer DR, et al.review Assessing publication only bias in meta-analyses in the 17 18 presence of between-study heterogeneity. Journal of the Royal Statistical Society: 19 20 Series A (Statistics in Society) 2010;173(3):575-91. doi: 10.1111/j.1467- 21 22 985X.2009.00629.x 23 24 25. Chaimani A, Salanti G. Visualizing assumptions and results in network meta-analysis: 25 26 27 The network graphs package. Stata Journal 2015;15(4):905-50. 28 29 26. Schunemann H, Brozek J, Guyatt G, et al. GRADE handbook for grading quality of 30 31 evidence and strength of recommendation 2013. Available from: http://bmjopen.bmj.com/ 32 33 http://gdt.guidelinedevelopment.org/app/ (assessed 22 Nov 2016). 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 19 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from

BMJ Open Page 20 of 21

1 2 3 4 5 PRISMA-P (Preferred Reporting Items for Systematic review and Meta-Analysis Protocols) 2015 checklist: recommended items to 6 address in a systematic review protocol* 7 8 Section and topic Item Checklist item Page (Line) 9 No 10 ADMINISTRATIVE INFORMATION 11 Title: 12 Identification 1a Identify the Forreport as a protoco peerl of a systematic reviewreview only 1 (3) 13 Update 1b If the protocol is for an update of a previous systematic review, identify as such Not applicable 14 15 Registration 2 If registered, provide the name of the registry (such as PROSPERO) and registration number 3 (7) 16 Authors: http://bmjopen.bmj.com/ 17 Contact 3a Provide name, institutional affiliation, email address of all protocol authors; provide physical mailing address 1 (21) 18 of corresponding author 19 Contributions 3b Describe contributions of protocol authors and identify the guarantor of the review Not applicable 20 Amendments 4 If the protocol represents an amendment of a previously completed or published protocol, identify as such and Not applicable 21 list changes; otherwise, state plan for documenting important protocol amendments 22 Support: 23 Sources 5a Indicate sources of financial or other support for the review 14 (41) 24 Sponsor 5b Provide name for the review funder and/or sponsor on September 30, 2021 by guest. Protected copyright. Not applicable 25 Role of sponsor 5c Describe roles of funder(s), sponsor(s), and/or institution(s), if any, in developing the protocol 14 (47) 26 or funder 27 28 INTRODUCTION 29 Rationale 6 Describe the rationale for the review in the context of what is already known 5 (8) 30 Objectives 7 Provide an explicit statement of the question(s) the review will address with reference to participants, 5 (53) 31 interventions, comparators, and outcomes (PICO) 32 METHODS 33 34 Eligibility criteria 8 Specify the study characteristics (such as PICO, study design, setting, time frame) and report characteristics 6 (36) (such as years considered, language, publication status) to be used as criteria for eligibility for the review 35 36 Information sources 9 Describe all intended information sources (such as electronic databases, contact with study authors, trial 8 (32) registers or other grey literature sources) with planned dates of coverage 37 38 Search strategy 10 Present draft of search strategy to be used for at least one electronic database, including planned limits, such 8 (54) that it could be repeated 39 40 41 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from

Page 21 of 21 BMJ Open

1 2 3 4 5 Study records: 6 Data 11a Describe the mechanism(s) that will be used to manage records and data throughout the review 9 (45) 7 management 8 Selection 11b State the process that will be used for selecting studies (such as two independent reviewers) through each 10 (3) 9 process phase of the review (that is, screening, eligibility and inclusion in metaanalysis) 10 Data collection 11c Describe planned method of extracting data from reports (such as piloting forms, done independently, in 10 (32) 11 process duplicate), any processes for obtaining and confirming data from investigators 12 Data items 12 List and defineFor all variables peerfor which data will bereview sought (such as PICO items, fundingonly sources), any pre 10 (37) 13 planned data assumptions and simplifications 14 Outcomes and 13 List and define all outcomes for which data will be sought, including prioritization of main and additional 11 (30) 15 prioritization outcomes, with rationale 16 Risk of bias in 14 Describe anticipated methods for assessing risk of bias of individual studies, including whether this will behttp://bmjopen.bmj.com/ 11 (10) 17 individual studies done at the outcome or study level, or both; state how this information will be used in data synthesis 18 Data synthesis 15a Describe criteria under which study data will be quantitatively synthesised Not applicable 19 15b If data are appropriate for quantitative synthesis, describe planned summary measures, methods of handling 12 (5) 20 data and methods of combining data from studies, including any planned exploration of consistency (such as 21 I2, Kendall’s τ) 22 15c Describe any proposed additional analyses (such as sensitivity or subgroup analyses, metaregression) 13 (27) 23 15d If quantitative synthesis is not appropriate, describe the type of summary planned Not applicable 24 Metabias(es) 16 Specify any planned assessment of metabias(es) (such as publication bias across studies, selective reporting on September 30, 2021 by guest. Protected copyright. 13 (22) 25 within studies) 26 Confidence in 17 Describe how the strength of the body of evidence will be assessed (such as GRADE) 14 (3) 27 cumulative evidence 28 29 * It is strongly recommended that this checklist be read in conjunction with the PRISMA-P Explanation and Elaboration (cite when available) for important 30 clarification on the items. Amendments to a review protocol should be tracked and dated. The copyright for PRISMA-P (including checklist) is held by the 31 PRISMA-P Group and is distributed under a Creative Commons Attribution Licence 4.0. 32 33 From: Shamseer L, Moher D, Clarke M, Ghersi D, Liberati A, Petticrew M, Shekelle P, Stewart L, PRISMA-P Group. Preferred reporting items for systematic review and 34 meta-analysis protocols (PRISMA-P) 2015: elaboration and explanation. BMJ. 2015 Jan 2;349(jan02 1):g7647. 35 36 37 38 39 40 41 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open

Comparative efficacy and acceptability of interventions for major depression in older persons: protocol for Bayesian network meta-analysis

ForJournal: peerBMJ Open review only Manuscript ID bmjopen-2017-019819.R1

Article Type: Protocol

Date Submitted by the Author: 28-Oct-2017

Complete List of Authors: Liew, Tau Ming; Institute of Mental Health, Department of Geriatric Psychiatry Lee, Cia Sin; SingHealth Polyclinics, Sengkang Polyclinic

Primary Subject Mental health Heading:

Secondary Subject Heading: Geriatric medicine

major depression, older person, efficacy, acceptability, network meta- Keywords: analysis 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-2017-019819 on 21 January 2018. Downloaded from Page 1 of 25 BMJ Open

1 2 3 1 Comparative efficacy and acceptability of interventions for major depression in older 4 5 2 persons: protocol for Bayesian network meta-analysis 6 7 3 8 9 4 Tau Ming Liew1, 2, Cia Sin Lee3 10 11 5 1Department of Geriatric Psychiatry, Institute of Mental Health, Singapore 12 13 6 2 14 Saw Swee Hock School of Public Health, National University of Singapore 15 3 16 7 SingHealth Polyclinics,For Singapore peer review only 17 18 8 19 20 9 21 22 10 Correspondence to 23 24 11 Tau Ming Liew; 25 26 12 27 [email protected] 28 29 13 Department of Geriatric Psychiatry, Institute of Mental Health, 30 31 14 10 Buangkok View, Singapore 539747. http://bmjopen.bmj.com/ 32 33 15 34 35 16 Keywords: major depression; older person; efficacy; acceptability; network metaanalysis 36 37 17 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 Number of words (Abstract): 290 41 42 19 Number of words (main text): 2,946 43 44 20 Number of references: 36 45 46 21 Number of tables or figures: 1 47 48 22 49 50 51 23 52 53 54 55 56 57 58 1 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 2 of 25

1 2 3 1 ABSTRACT 4 5 2 6 7 3 Introduction: Major depression is a leading cause of disability, and has been associated with 8 9 4 adverse effects in older persons. While many pharmacological and nonpharmacological 10 11 5 interventions have been shown to be effective to address major depression in older persons, 12 13 6 14 there has not been a metaanalysis that consolidates all the available interventions and 15 16 7 compare the relativeFor benefits peer of these review available interventions. only In this study, we aim to 17 18 8 conduct a systematic review and network metaanalysis to compare the efficacy and 19 20 9 acceptability of all the known pharmacological and nonpharmacological interventions for 21 22 10 major depression in older persons. 23 24 11 Methods and analysis: We will search MEDLINE, Embase, PsycINFO, Cumulative Index 25 26 12 27 to Nursing and Allied Health, Cochrane Central Register of Controlled Trials and references 28 29 13 of other review articles for articles related to the keywords of ‘randomized trial’, ‘major 30 31 14 depression’, ‘older persons’ and ‘treatments’. Two reviewers will independently select the http://bmjopen.bmj.com/ 32 33 15 eligible articles. For each included article, the two reviewers will independently extract the 34 35 16 data and assess the risk of bias using the Cochrane revised tool for Risk of Bias. Bayesian 36 37 17 network metaanalyses will be conducted to pool the depression scores (based on 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 standardized mean difference) and the allcause discontinuation across all included studies. 41 42 19 The ranking probabilities for all interventions will be estimated and the hierarchy of each 43 44 20 interventions will be summarized as surface under the cumulative ranking curve (SUCRA). 45 46 21 Metaregression and subgroup analyses will also be performed to evaluate the effect of 47 48 22 studylevel covariates. The quality of the evidence will be assessed using the Grading of 49 50 23 Recommendations Assessment, Development and Evaluation (GRADE) approach. 51 52 24 53 Ethics and dissemination: The results will be disseminated through conference 54 55 25 presentations and peerreviewed publications. They will provide the consolidated evidence to 56 57 58 2 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 3 of 25 BMJ Open

1 2 3 1 inform clinicians on the best choice of intervention to address major depression in older 4 5 2 persons. 6 7 3 Trial registration number: International Prospective Register for Systematic Reviews 8 9 4 (PROSPERO) number CRD42017075756. 10 11 5 12 13 14 6 15 16 7 For peer review only 17 18 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 3 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 4 of 25

1 2 3 1 STRENGTHS AND LIMITATIONS OF THIS STUDY 4 5 2 6 7 3 • This systematic review and metaanalysis will provide a comprehensive summary on the 8 9 4 efficacy and acceptability of all available interventions for major depression in older 10 11 5 12 persons. 13 14 6 • The results will provide the highest level of evidence to inform clinicians on the best 15 16 7 choice of treatment,For peer from among review the many available only pharmacological and non 17 18 8 pharmacological interventions. 19 20 9 • This protocol has been developed in accordance with the Preferred Reporting Items for 21 22 10 23 Systematic Review and Metaanalysis Protocols (PRISMAP) statement and has been 24 11 25 registered with PROSPERO. 26 27 12 • The overall quality of evidence will be assessed using the Grading of Recommendations 28 29 13 Assessment, Development and Evaluation (GRADE) approach. 30 http://bmjopen.bmj.com/ 31 14 • This systematic review will be limited to studies which are reported in English language 32 33 15 and have been peerreviewed. 34 35 16 36 37 38 17 39 on September 30, 2021 by guest. Protected copyright. 40 18 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 4 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 5 of 25 BMJ Open

1 2 3 1 INTRODUCTION 4 5 2 6 7 8 3 Rationale 9 10 4 11 12 13 5 Major depression has been identified by the World Health Organization as one of the leading 14 15 6 cause of disability globally.1 2 In older persons, its prevalence rates rise with the increase in 16 For peer review only 17 7 medical comorbidities,3 with reported rates of up to 5% in communitydwelling older 18 19 8 persons,35 5 to 10% in primary care3 6 and as high as 37% after critical care hospitalizations.3 20 21 9 7 Major depression has a significant impact on the older populations and has been linked to 22 23 10 4 8 9 4 10 24 higher risk of suicide, myocardial infarction, stroke, allcause mortality and increasing

25 4 26 11 health services utilization. 27 28 12 29 30 13 Many of the interventions for major depression in older persons have had recent meta 31 http://bmjopen.bmj.com/ 32 14 analyses confirming their efficacy when compared to control groups. These include 33 34 15 antidepressants,1114 cognitive behavioural therapy,15 problem solving therapy,16 35 36 16 1719 37 psychological interventions in general, and the various forms of nonpharmacological

38 2022 39 17 interventions. However, none of the metaanalyses had compared all the on September 30, 2021 by guest. Protected copyright. 40 41 18 pharmacological and nonpharmacological interventions together to demonstrate the relative 42 43 19 benefits of each intervention. It is unknown whether the different types of pharmacological 44 45 20 and nonpharmacological interventions have comparable efficacy and are equally suitable for 46 47 21 older persons with major depression. 48 49 22 50 51 52 23 Objectives 53 54 24 55 56 57 58 5 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 6 of 25

1 2 3 1 In this study, we aim to conduct a systematic review and network metaanalysis to compare 4 5 2 the efficacy and acceptability of all the available pharmacological and nonpharmacological 6 7 3 interventions for major depression in older persons. The use of network metaanalysis allows 8 9 4 us to pool the evidence on various interventions and rank their benefits relative to each 10 11 5 other.23 It also allows us to conduct indirect comparison of the different interventions, even 12 13 6 14 when there is no direct evidence in the literature to allow headtohead comparisons. 15 16 7 For peer review only 17 18 8 19 20 9 METHODS AND ANALYSIS 21 22 10 23 24 11 This protocol is developed in accordance with the Preferred Reporting Items for Systematic 25 26 12 24 25 27 Review and Metaanalysis (PRISMA) statement. It has also been registered with the 28 29 13 International Prospective Register of Systematic Reviews (PROSPERO) (registration number 30 31 14 CRD42017075756). http://bmjopen.bmj.com/ 32 33 15 34 35 16 Eligibility criteria 36 37 17 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 Participants and settings 41 42 19 43 44 20 We will include studies which recruited participants who were: 45 46 21 • 60 years old and above; 47 48 22 • diagnosed with major depression based on formal criteria by the Diagnostic and 49 50 23 51 Statistical Manual of Mental Disorders (DSM) or International Classification of Diseases 52 53 24 (ICD); and 54 55 56 57 58 6 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 7 of 25 BMJ Open

1 2 3 1 • having a current episode of major depression (that is, the participants were symptomatic 4 5 2 and not in remission at the point of recruitment; and the intervention was not intended 6 7 3 primarily for the prevention of future relapses). 8 9 4 10 11 5 12 We will exclude studies which recruited participants with treatmentresistant depression, 13 14 6 subthreshold depression, bipolar depression, depression in dementia or psychotic depression. 15 16 7 We will not includeFor maintenance peer studies review for major depression only as such studies primarily 17 18 8 focused on the prevention of relapses in participants who had been asymptomatic or in 19 20 9 remission at the point of recruitment. 21 22 10 23 24 11 25 Interventions 26 27 12 28 29 13 We will include studies with pharmacological interventions, including but not limited to: 30 31 14 • Antidepressants such as , , or ; http://bmjopen.bmj.com/ 32 33 15 • Antipsychotics such as , quetiapine, or ; 34 35 16 36 • Moodstabilizers such as valproate, , or gabapentin. 37 38 17 39 on September 30, 2021 by guest. Protected copyright. 40 18 We will include studies with nonpharmacological interventions, including but not limited to: 41 42 19 • Psychological interventions such as cognitive behavioural therapy, interpersonal therapy, 43 44 20 problem solving therapy, psychodynamic therapy or family interventions; 45 46 21 • Procedural interventions such as electroconvulsive therapy, transcranial magnetic 47 48 22 49 stimulation, transcranial directcurrent stimulation or bright light therapy. 50 51 23 52 53 24 We will also include studies which reported on combinations of any of these pharmacological 54 55 25 and nonpharmacological interventions. 56 57 58 7 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 8 of 25

1 2 3 1 4 5 2 We will exclude studies which focused primarily on health service models of care but were 6 7 3 not related to any modality of intervention, such as studies which evaluated the effectiveness 8 9 4 of home treatment, training of general practitioners, multidisciplinary approach or stepped 10 11 5 care approach. 12 13 6 14 15 16 7 Comparators For peer review only 17 18 8 19 20 9 We will accept control conditions such as placebo intervention, waitinglist, treatment as 21 22 10 usual, as well as no intervention. We will also include studies with active comparators such 23 24 11 as those which compare between two different interventions within the same studies. 25 26 12 27 28 29 13 Outcomes 30 31 14 http://bmjopen.bmj.com/ 32 33 15 We will only include a study if it reports the depression scores or the allcause 34 35 16 discontinuation in each study arm following intervention. 36 37 17 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 Study designs and publication types 41 42 19 43 44 20 We will only include randomized controlled trials (RCTs) which aimed to demonstrate the 45 46 21 superiority of a treatment to another (also known as superiority trials), and will not include 47 48 22 equivalence or noninferiority trials. The following study designs or publication types will 49 50 23 also be excluded: qualitative studies, observational studies, metaanalyses, case reports, case 51 52 24 53 series, ecological studies and policy papers. We intend to include only higherquality 54 55 56 57 58 8 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 9 of 25 BMJ Open

1 2 3 1 evidence and hence will exclude nonrandomized trials and publications which were not peer 4 5 2 reviewed (such as conference proceedings, letters and comments). 6 7 3 8 9 4 Language and time frame 10 11 5 12 13 6 14 We will only include studies which are reported in the English language. Apart from that, we 15 16 7 do not impose For any time peer restriction to review the publication year only of the studies. The search of 17 18 8 databases will be conducted in January 2018. 19 20 9 21 22 10 Information sources and search strategy 23 24 11 25 26 12 27 We will search MEDLINE, Embase, PsycINFO, Cumulative Index to Nursing and Allied 28 29 13 Health (CINAHL) and Cochrane Central Register of Controlled Trials (CENTRAL) for 30 31 14 original articles related to the keywords of ‘randomized trial’, ‘major depression’, ‘older http://bmjopen.bmj.com/ 32 33 15 persons’ and ‘treatments’. Our search strategy for MEDLINE is shown in Box 1. Similar 34 35 16 search strategies will be used for the other databases. Additionally, we will also handsearch 36 37 17 the references of review articles related to the topic to retrieve relevant articles which are not 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 captured through our search of the electronic databases. We will examine the full text of the 41 42 19 relevant articles and include the respective articles if they meet our eligibility criteria. 43 44 20 45 Box 1. Search strategy for MEDLINE (via Ovid interface) 46

47 1. *Therapeutics/ OR *Drug Therapy/ OR *Psychotropic Drugs/ OR *Antidepressive Agents/ OR 48 * Agents/ OR *Antimanic Agents/ OR */ OR *Psychotherapy/ 49 OR *Electroconvulsive Therapy/ OR *Transcranial Magnetic Stimulation/ OR *Transcranial 50 Direct Current Stimulation/ OR *Phototherapy/ 51 2. (* OR selective inhibitor OR SSRI OR citalopram OR 52 53 OR OR sertraline OR OR OR (serotonin adj2 54 epinephrine adj reuptake adj inhibitor) OR SNRI OR venlafaxine OR OR 55 OR OR OR OR noradrenergic and specific 56 serotonergic antidepressant OR NaSSA OR mirtazapine OR TCA OR OR amersergide 57 58 9 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 10 of 25

1 2 3 OR OR OR OR OR chlorpoxiten OR 4 OR clorimipramine OR OR OR dibenzipin OR 5 dothiepin OR OR OR OR OR metapramine OR 6 OR OR OR OR OR 7 8 OR OR OR OR ).ab,ti 9 3. (antipsychotic* OR OR OR benperidol OR OR 10 flupenthixol OR OR OR OR sulpiride OR 11 OR OR OR OR fluspirilene OR 12 methotrimeprazine OR risperidone OR OR quetiapine OR olanzapine OR 13 14 OR amisulpiride OR aripiprazole OR OR OR ).ab,ti 15 4. ((mood adj stabili*) OR (antimanic adj (agent* OR drug*)) OR anticonvuls* OR anti convuls* 16 OR carbamazepineFor OR peer ethosuximide review OR gabapentin OR only lacosamide OR lamotrigine OR 17 levetiracetam OR lithium OR OR phenobarbital OR phenytoin OR pregabalin 18 OR rufinamide OR tiagabine OR OR valproic acid OR valproate OR verapamil OR 19 vigabatrin OR zonisamide).ab,ti 20

21 5. (psychotherap* OR therap* OR cognitive behavio* therapy OR cognitive therapy OR behavio* 22 therapy OR interpersonal therapy OR interpersonal therapy OR problem solving therapy OR 23 problemsolving therapy OR (family adj (therapy OR intervention)) OR bibliotherapy OR 24 mindful* OR (group adj (therapy OR intervention)) OR psychodynamic OR psychoanalytic OR 25 emotionfocused OR emotion focused OR reminiscen* OR life review OR lifereview).ab,ti 26

27 6. (electroconvulsive therapy OR electroconvulsive therapy OR Transcranial Magnetic 28 Stimulation OR Transcranial Direct Current Stimulation OR light therapy).ab,ti 29 7. #1 OR #2 OR #3 OR #4 OR #5 OR #6 30 8. *Depressive Disorder, Major/ OR (major adj (depressive OR depression)).ab,ti http://bmjopen.bmj.com/ 31 9. *Aged/ OR *"Aged, 80 and over"/ OR (elder* OR (older adj (person* OR people OR adult*)) 32 33 OR (late adj life) OR geriatric).ab,ti 34 10. *Randomized Controlled Trial/ OR (Randomized Controlled Trial).pt OR *Random 35 Allocation/ 36 11. (singleblind* OR doubleblind* OR trebleblind* OR tripleblind*).ab,ti 37 12. (single* OR doubl* OR trebl* OR tripl*) adj5 blind*).ab,ti 38 13. (random*).ab,ti on September 30, 2021 by guest. Protected copyright. 39 40 14. (randomized OR randomised OR (random* adj (assigned OR allocated OR assignment OR 41 allocation))).ab,ti 42 15. #10 OR ((#11 OR #12) AND #13) OR #14 43 16. #7 AND #8 AND #9 AND #15 44 1 45 46 47 2 Study selection 48 49 3 50 51 4 All potential articles will be retrieved and organized in a data management software (Endnote 52 53 5 software, Thomson Reuters). After removing duplicate records, two reviewers will 54 55 6 independently screen through the titles and abstracts to retain eligible articles. The first 10% 56 57 58 10 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 11 of 25 BMJ Open

1 2 3 1 of these titles and abstracts will be subjected to a calibration exercise between the two 4 5 2 reviewers to ensure mutual agreement. 6 7 3 8 9 4 After completing the screening phase, articles that are deemed as relevant by at least one of 10 11 5 the reviewers will be subjected to fulltext review. The two reviewers will independently 12 13 6 14 confirm the eligibility of these articles based on the full texts. The first 10% of these full 15 16 7 texts will again undergoFor a calibrationpeer exercise review by the two reviewers. only After the fulltext review, 17 18 8 the included articles will be used for qualitative synthesis. The chancecorrected agreement 19 20 9 between the two reviewers will be assessed using Cohen’s Kappa (κ). 21 22 10 23 24 11 25 At any point during study selection, the reasons for excluding specific articles will be 26 27 12 recorded. Moreover, any disagreements between the two reviewers will be resolved by 28 29 13 discussion with a third reviewer. 30 31 14 http://bmjopen.bmj.com/ 32 33 15 Data extraction 34 35 16 36 37 17 38 Data from the selected studies will be extracted by two reviewers independently, and 39 on September 30, 2021 by guest. Protected copyright. 40 18 disagreements between the reviewers will be resolved by discussion with a third reviewer. 41 42 19 The extracted data will include the following information: 43 44 20 1. Study identification (first author, year of publication, geographic location) 45 46 21 2. Study characteristics (study setting, study design, inclusion criteria, diagnostic criteria 47 48 22 of major depression, sample size) 49 50 23 51 3. Participant characteristics (age, gender, education, number of comorbidities, Mini 52 53 24 Mental State Examination score, baseline depression score, depression scale, duration 54 55 25 of the current episode of major depression) 56 57 58 11 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 12 of 25

1 2 3 1 4. Characteristics of intervention and comparator (description, treatment dose/intensity, 4 5 2 treatment duration, depression score, allcause discontinuation) 6 7 3 8 9 4 The original authors of the RCTs will be contacted when the required data are not available in 10 11 5 the published article. 12 13 14 6 15 16 7 Assessment of riskFor of bias peer review only 17 18 8 19 20 9 The risk of bias for each study will be assessed independently by two reviewers using the 21 22 10 Cochrane revised tool for Risk of Bias (RoB 2.0),26 focusing on biases related to five key 23 24 11 domains: randomization process, deviations from intended interventions, missing outcome 25 26 12 27 data, measurement of the outcome and selection of the reported result. Each domain will 28 29 13 receive a judgement on the risk of bias (high, low or some concerns) and an overall risk of 30 31 14 bias will be assigned based on the judgements from the five domains. Any disagreements http://bmjopen.bmj.com/ 32 33 15 between the two reviewers will be resolved by discussion with a third reviewer. 34 35 16 36 37 17 Outcome measures 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 41 42 19 Our primary outcomes are the efficacy and the acceptability of interventions. The efficacy 43 44 20 will be based on the difference in depression scores between the intervention and comparator 45 46 21 upon the completion of intervention (we will give preference to the primary timepoint 47 48 22 predefined in the original study), computed as standardized mean difference (SMD) for each 49 50 23 RCT. The acceptability will be assessed by the relative risk (RR) of allcause discontinuation 51 52 24 53 of the intervention. Each intervention will only be grouped by its generic name for 54 55 25 pharmacological interventions (such as mirtazapine, citalopram, quetiapine, valproate or 56 57 58 12 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 13 of 25 BMJ Open

1 2 3 1 lithium) or by its known modality for nonpharmacological interventions (such as cognitive 4 5 2 behavioural therapy, problem solving therapy, or transcranial magnetic stimulation). We will 6 7 3 not categorize the interventions further in our analyses of the outcome measures. In the event 8 9 4 that the active arm of a RCT involves combinations of interventions, it will be reported as the 10 11 5 respective combinations (such as citalopram–cognitive behavioural therapy combination, or 12 13 6 14 mirtazapine–quetiapine–problem solving therapy combination). 15 16 7 For peer review only 17 18 8 Statistical analysis 19 20 9 21 22 10 We will first conduct pairwise metaanalysis with the randomeffects model (DerSimonian 23 24 11 and Laird method)27 provided there are at least two included studies for each pairwise 25 26 12 2 27 comparison. We will use the I statistic and the Q test to assess heterogeneity in each 28 2 28 29 13 pairwise metaanalysis. In the presence of substantial heterogeneity (I >50%) in a 30 31 14 particular intervention, we will consider subgrouping the intervention by its dose/intensity http://bmjopen.bmj.com/ 32 33 15 and duration, and use the subgroups of that intervention in the subsequent network meta 34 35 16 analyses. 36 37 17 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 We will then conduct the network metaanalyses within a Bayesian framework using the 41 42 19 Markov Chains Monte Carlo method. Bayesian analysis provides probabilistic distributions 43 44 20 of our estimatesofinterest through large number of simulations, and hence produces results 45 46 21 which have more intuitive interpretations. For example, Bayesian analysis generates the 95% 47 48 22 credible interval which can be accurately interpreted as the range containing 95% of the 49 50 23 estimates (based on the simulations). In the Bayesian analysis, we will run four Markov 51 52 24 53 chains simultaneously with different arbitrarily chosen initial values and with non 54 55 25 informative priors. Each chain will have at least 10,000 simulations and at least the first 56 57 58 13 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 14 of 25

1 2 3 1 2,500 simulations will be discarded as burnin. Convergence of the simulations will be 4 5 2 assessed with the trace plots, kernel density plots and GelmanRubinBrooks plots. 6 7 3 8 9 4 We will employ both fixedeffects and randomeffects models in the Bayesian analyses, and 10 11 5 will choose the final models based on the deviance information criterion (DIC). While there 12 13 6 14 is no ruleofthumb on what constitute significant improvements in DIC, we can take 15 29 16 7 reference from theFor guideline peer commonly reviewused in the analogous only Akaike Information Criteria: 17 18 8 values which are lesser by at least 10 points indicate significantly better modelfit and 19 20 9 parsimony. Hence, results from the randomeffects model will be used if the randomeffects 21 22 10 model has DIC which is smaller by at least 10 points compared to the fixedeffect model. We 23 24 11 will also compare the complexity of model between the fixedeffects and randomeffects 25 26 12 27 models using pD (an indicator which has higher value when a model is more complex), with 28 29 13 preference for models which are more parsimonious (less complex). The global 30 31 14 heterogeneity will be assessed with I2 statistic. A common heterogeneity parameter will be http://bmjopen.bmj.com/ 32 33 15 assumed in the randomeffects model. Inconsistency between direct and indirect sources of 34 35 16 evidence will be statistically assessed using the nodesplitting method,30 31 which generates a 36 37 17 pvalue for the difference between direct and indirect estimates in each closedloop in the 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 network (pvalues of <0.05 indicates the presence of inconsistency between direct and 41 42 19 indirect estimates in a particular closedloop). 43 44 20 45 46 21 We will estimate the ranking probabilities for all interventions and show the results 47 48 22 graphically in the form of rankograms and cumulative ranking probability plots. The 49 50 23 hierarchy of interventions will be summarized as surface under the cumulative ranking curve 51 52 24 53 (SUCRA) and presented in a scatterplot. SUCRAs have possible values ranging from 0% to 54 55 56 57 58 14 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 15 of 25 BMJ Open

1 2 3 1 100%, with higher values indicating better efficacy or acceptability. Publication bias will be 4 5 2 assessed with comparisonadjusted funnel plot.32 33 6 7 3 8 9 4 We will conduct metaregression analyses to determine whether the results of our network 10 11 5 metaanalyses will be affected by the following studylevel covariates: sample size, study 12 13 6 14 duration, inclusion criteria, study setting, study design and risk of bias. A covariate is 15 16 7 considered as a For significant peer moderator ifreview the 95% credible only interval of its beta coefficient in 17 18 8 metaregression does not include the value of zero. If a significant moderator is found, 19 20 9 further subgroup analyses will then be conducted to assess the effect of this moderator. 21 22 10 23 24 11 The pairwise metaanalyses will be conducted with STATA (version 14). The network meta 25 26 12 27 analyses will be conducted using JAGS (version 4.2.0), through the GeMTC package of R 28 29 13 (version 3.3.1). The “Network Graphs” package in Stata statistical software (version 14.0) 30 31 14 will also used to produce some of the figures in this study, such as the network plots, http://bmjopen.bmj.com/ 32 33 15 rankograms, cumulative ranking probability plots and comparisonadjusted funnel plots. 32 34 34 35 16 36 37 17 Assessment of quality of evidence 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 41 42 19 We will use the Grading of Recommendations Assessment, Development and Evaluation 43 44 20 (GRADE) approach to report the quality of evidence on efficacy and acceptability of 45 46 21 interventions for major depression in older persons. Based on five key domains 47 48 22 (methodology quality, directness of evidence, heterogeneity, precision of effect estimates and 49 50 23 risk of publication bias), we will classify the quality of evidence in one of four levels – high, 51 52 24 35 53 moderate, low and very low. 54 55 25 56 57 58 15 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 16 of 25

1 2 3 1 LIMITATIONS 4 5 2 6 7 3 Several limitations of this study should be noted. First, there can possibly be heterogeneity in 8 9 4 the dose/intensity and the duration of each intervention, which may limit the interpretation of 10 11 5 the metaanalysis. To address this potential limitation, we will first conduct pairwise meta 12 13 6 2 14 analyses to evaluate the amount of heterogeneity using the I statistic and the Q test. In the 15 2 28 16 7 presence of substantialFor heterogeneity peer (I >50%)review in a particular only intervention, we will consider 17 18 8 subgrouping the intervention by its dose/intensity and duration, and use the more 19 20 9 homogeneous subgroups of that intervention in the subsequent network metaanalyses. In the 21 22 10 network metaanalyses, we will also evaluate for inconsistency between direct and indirect 23 24 11 estimates using nodesplitting method, and evaluate for heterogeneity using metaregression 25 26 12 27 and subgroup analyses. Second, we will exclude nonEnglish and nonpeer reviewed 28 29 13 publications (such as conference proceedings and letters), which may potentially raise the 30 31 14 concern of publication bias. The exclusion of nonpeer reviewed publications is related to http://bmjopen.bmj.com/ 32 33 15 our intention of including only higherquality evidence. Regardless, we will monitor the 34 35 16 impact of such decision and any possible publication bias using comparisonadjusted funnel 36 37 17 plot. Third, we will use allcause discontinuation as a crude composite measure of treatment 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 acceptability. Allcause discontinuation was chosen (instead of discontinuation due to 41 42 19 specific reasons) because this information is more readily available in almost all RCTs, 43 44 20 especially among nonpharmacological RCTs where it can be more challenging to clearly 45 46 21 attribute the cause of discontinuation to specific reasons such as adverse effects. Hence, the 47 48 22 use of allcause discontinuation will allow us to compare the acceptability of both 49 50 23 pharmacological and nonpharmacological interventions within the same model in network 51 52 24 53 metaanalysis. 54 55 25 56 57 58 16 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 17 of 25 BMJ Open

1 2 3 1 ETHICS AND DISSEMINATION 4 5 2 6 7 3 This systematic review will provide the consolidated evidence to inform clinicians on the best 8 9 4 choice of intervention, from among the many available options, to address major depression 10 11 5 in older persons. This systematic review will be reported in accordance with the 12 13 6 36 14 recommendations of PRISMA statement for network metaanalyses. It is expected to be 15 16 7 completed by JanuaryFor 2020,peer and the review results will be disseminatedonly through conference 17 18 8 presentations and publications in peerreviewed journal. 19 20 9 21 22 10 CONTRIBUTORS 23 24 11 25 26 12 27 TML conceived the idea for this systematic review, developed the initial methodology, wrote 28 29 13 the first draft and act as the guarantor of the protocol. CSL provided critical feedback on the 30 31 14 search strategy, methodology and manuscript. All authors approved the final version of the http://bmjopen.bmj.com/ 32 33 15 manuscript. 34 35 16 36 37 17 FUNDING 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 41 42 19 TML was supported by research grants under the Singapore Ministry of Health’s National 43 44 20 Medical Research Council (Grant No.: NMRC/Fellowship/0030/2016 and 45 46 21 NMRC/CSSSP/0014/2017). The funding source had no involvement in any part of the 47 48 22 project. 49 50 23 51 52 24 53 54 55 25 COMPETING INTERESTS 56 57 58 17 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 18 of 25

1 2 3 1 4 5 2 None declared. 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 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 18 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 19 of 25 BMJ Open

1 2 3 REFERENCES 4 5 6 7 8 1. World Health Organization. The global burden of disease: 2004 update. Switzerland: 9 10 11 World Health Organization. 12 13 2. World Health Organization. Global Health Estimates 2015: Disease burden by Cause, Age, 14 15 Sex, by Country and by Region, 20002015. Geneva, 2016. 16 For peer review only 17 3. Taylor WD. Clinical practice. Depression in the elderly. The New England journal of 18 19 medicine 2014;371(13):122836. doi: 10.1056/NEJMcp1402180 [published Online 20 21 First: 2014/09/25] 22 23 24 4. Blazer DG. Depression in late life: review and commentary. The journals of gerontology 25 26 Series A, Biological sciences and medical sciences 2003;58(3):24965. [published 27 28 Online First: 2003/03/14] 29 30 5. Volkert J, Schulz H, Härter M, et al. The prevalence of mental disorders in older people in 31 http://bmjopen.bmj.com/ 32 Western countries – a metaanalysis. Ageing Research Reviews 2013;12(1):33953. 33 34 doi: http://dx.doi.org/10.1016/j.arr.2012.09.004 35 36 37 6. Lyness JM, Caine ED, King DA, et al. Psychiatric Disorders in Older Primary Care 38 39 Patients. Journal of general internal medicine 1999;14(4):24954. doi: on September 30, 2021 by guest. Protected copyright. 40 41 10.1046/j.15251497.1999.00326.x 42 43 7. Jackson JC, Pandharipande PP, Girard TD, et al. Depression, Posttraumatic Stress Disorder, 44 45 and Functional Disability in Survivors of Critical Illness: results from the BRAIN 46 47 ICU (Bringing to light the Risk Factors And Incidence of Neuropsychological 48 49 50 dysfunction in ICU survivors) Investigation: A Longitudinal Cohort Study. The lancet 51 52 Respiratory medicine 2014;2(5):36979. doi: 10.1016/S22132600(14)700517 53 54 55 56 57 58 19 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 20 of 25

1 2 3 8. Gan Y, Gong Y, Tong X, et al. Depression and the risk of coronary heart disease: a meta 4 5 analysis of prospective cohort studies. BMC psychiatry 2014;14:371. doi: 6 7 10.1186/s128880140371z [published Online First: 2014/12/30] 8 9 9. Pan A, Sun Q, Okereke OI, et al. Depression and risk of stroke morbidity and mortality: a 10 11 metaanalysis and systematic review. Jama 2011;306(11):12419. doi: 12 13 14 10.1001/jama.2011.1282 [published Online First: 2011/09/22] 15 16 10. Cuijpers P,For Vogelzangs peer N, Twisk review J, et al. Comprehensive only metaanalysis of excess 17 18 mortality in depression in the general community versus patients with specific 19 20 illnesses. The American journal of psychiatry 2014;171(4):45362. doi: 21 22 10.1176/appi.ajp.2013.13030325 [published Online First: 2014/01/18] 23 24 11. Tham A, Jonsson U, Andersson G, et al. Efficacy and tolerability of antidepressants in 25 26 27 people aged 65 years or older with major depressive disorder – A systematic review 28 29 and a metaanalysis. Journal of affective disorders 2016;205:112. doi: 30 31 http://dx.doi.org/10.1016/j.jad.2016.06.013 http://bmjopen.bmj.com/ 32 33 12. Nelson JC, Delucchi K, Schneider LS. Efficacy of second generation antidepressants in 34 35 latelife depression: a metaanalysis of the evidence. The American journal of 36 37 geriatric psychiatry : official journal of the American Association for Geriatric 38 39 on September 30, 2021 by guest. Protected copyright. 40 Psychiatry 2008;16(7):55867. doi: 10.1097/JGP.0b013e3181693288 [published 41 42 Online First: 2008/07/02] 43 44 13. Thorlund K, Druyts E, Wu P, et al. Comparative efficacy and safety of selective serotonin 45 46 reuptake inhibitors and serotonin reuptake inhibitors in older adults: a 47 48 network metaanalysis. Journal of the American Geriatrics Society 2015;63(5):10029. 49 50 doi: 10.1111/jgs.13395 [published Online First: 2015/05/07] 51 52 53 14. Tedeschini E, Levkovitz Y, Iovieno N, et al. Efficacy of antidepressants for latelife 54 55 depression: a metaanalysis and metaregression of placebocontrolled randomized 56 57 58 20 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 21 of 25 BMJ Open

1 2 3 trials. The Journal of clinical psychiatry 2011;72(12):16608. doi: 4 5 10.4088/JCP.10r06531 [published Online First: 2012/01/17] 6 7 15. Gould RL, Coulson MC, Howard RJ. Cognitive behavioral therapy for depression in 8 9 older people: a metaanalysis and metaregression of randomized controlled trials. 10 11 Journal of the American Geriatrics Society 2012;60(10):181730. doi: 12 13 14 10.1111/j.15325415.2012.04166.x [published Online First: 2012/09/26] 15 16 16. Kirkham JG,For Choi N,peer Seitz DP. review Metaanalysis of ponlyroblem solving therapy for the 17 18 treatment of major depressive disorder in older adults. International journal of 19 20 geriatric psychiatry 2016;31(5):52635. doi: 10.1002/gps.4358 [published Online 21 22 First: 2015/10/06] 23 24 17. Cuijpers P, van Straten A, Smit F. Psychological treatment of latelife depression: a meta 25 26 27 analysis of randomized controlled trials. International journal of geriatric psychiatry 28 29 2006;21(12):113949. doi: 10.1002/gps.1620 [published Online First: 2006/09/07] 30 31 18. Cuijpers P, Karyotaki E, Pot AM, et al. Managing depression in older age: psychological http://bmjopen.bmj.com/ 32 33 interventions. Maturitas 2014;79(2):16069. doi: 10.1016/j.maturitas.2014.05.027 34 35 19. Linde K, Sigterman K, Kriston L, et al. Effectiveness of Psychological Treatments for 36 37 Depressive Disorders in Primary Care: Systematic Review and MetaAnalysis. The 38 39 on September 30, 2021 by guest. Protected copyright. 40 Annals of Family Medicine 2015;13(1):5668. doi: 10.1370/afm.1719 41 42 20. Holvast F, Massoudi B, Oude Voshaar RC, et al. Nonpharmacological treatment for 43 44 depressed older patients in primary care: A systematic review and metaanalysis. PloS 45 46 one 2017;12(9):e0184666. doi: 10.1371/journal.pone.0184666 47 48 21. Apostolo J, BobrowiczCampos E, Rodrigues M, et al. The effectiveness of non 49 50 pharmacological interventions in older adults with depressive disorders: A systematic 51 52 53 review. International journal of nursing studies 2016;58:5970. doi: 54 55 10.1016/j.ijnurstu.2016.02.006 [published Online First: 2016/04/19] 56 57 58 21 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 22 of 25

1 2 3 22. Sukhato K, Lotrakul M, Dellow A, et al. Efficacy of homebased nonpharmacological 4 5 interventions for treating depression: a systematic review and network metaanalysis 6 7 of randomised controlled trials. BMJ open 2017;7(7) 8 9 23. Tonin FS, Rotta I, Mendes AM, et al. Network metaanalysis: a technique to gather 10 11 evidence from direct and indirect comparisons. Pharmacy practice 2017;15(1):943. 12 13 14 doi: 10.18549/PharmPract.2017.01.943 [published Online First: 2017/05/16] 15 16 24. Moher D, ShamseerFor L,peer Clarke M, etreview al. Preferred reporting only items for systematic review 17 18 and metaanalysis protocols (PRISMAP) 2015 statement. Systematic reviews 19 20 2015;4:1. doi: 10.1186/2046405341 [published Online First: 2015/01/03] 21 22 25. Shamseer L, Moher D, Clarke M, et al. Preferred reporting items for systematic review 23 24 and metaanalysis protocols (PRISMAP) 2015: elaboration and explanation. BMJ 25 26 27 (Clinical research ed) 2015;349:g7647. doi: 10.1136/bmj.g7647 [published Online 28 29 First: 2015/01/04] 30 31 26. Higgins JPT, Sterne JAC, Savović J, et al. A revised tool for assessing risk of bias in http://bmjopen.bmj.com/ 32 33 randomized trials. In: Chandler J, McKenzie J, Boutron I, et al., eds. Cochrane 34 35 Methods Cochrane Database of Systematic Reviews 2016, Issue 10 (Suppl 1). 36 37 27. DerSimonian R, Laird N. Metaanalysis in clinical trials. Controlled clinical trials 38 39 on September 30, 2021 by guest. Protected copyright. 40 1986;7(3):17788. [published Online First: 1986/09/01] 41 42 28. Higgins JP, Thompson SG, Deeks JJ, et al. Measuring inconsistency in metaanalyses. 43 44 BMJ (Clinical research ed) 2003;327(7414):55760. doi: 10.1136/bmj.327.7414.557 45 46 [published Online First: 2003/09/06] 47 48 29. Burnham KP, Anderson DR. Model Selection and Multimodel Inference: A Practical 49 50 InformationTheoretic Approach: Springer New York 2003. 51 52 53 54 55 56 57 58 22 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 23 of 25 BMJ Open

1 2 3 30. Dias S, Welton NJ, Caldwell DM, et al. Checking consistency in mixed treatment 4 5 comparison metaanalysis. Statistics in medicine 2010;29(78):93244. doi: 6 7 10.1002/sim.3767 [published Online First: 2010/03/10] 8 9 31. van Valkenhoef G, Dias S, Ades AE, et al. Automated generation of nodesplitting 10 11 models for assessment of inconsistency in network metaanalysis. Research synthesis 12 13 14 methods 2016;7(1):8093. doi: 10.1002/jrsm.1167 [published Online First: 15 16 2015/10/16]For peer review only 17 18 32. Chaimani A, Higgins JP, Mavridis D, et al. Graphical tools for network metaanalysis in 19 20 STATA. PloS one 2013;8(10):e76654. doi: 10.1371/journal.pone.0076654 [published 21 22 Online First: 2013/10/08] 23 24 33. Peters JL, Sutton AJ, Jones DR, et al. Assessing publication bias in metaanalyses in the 25 26 27 presence of betweenstudy heterogeneity. Journal of the Royal Statistical Society: 28 29 Series A (Statistics in Society) 2010;173(3):57591. doi: 10.1111/j.1467 30 31 985X.2009.00629.x http://bmjopen.bmj.com/ 32 33 34. Chaimani A, Salanti G. Visualizing assumptions and results in network metaanalysis: 34 35 The network graphs package. Stata Journal 2015;15(4):90550. 36 37 35. Schunemann H, Brozek J, Guyatt G, et al. GRADE handbook for grading quality of 38 39 on September 30, 2021 by guest. Protected copyright. 40 evidence and strength of recommendation 2013. Available from: 41 42 http://gdt.guidelinedevelopment.org/app/ (assessed 22 Nov 2016). 43 44 36. Hutton B, Salanti G, Caldwell DM, et al. The PRISMA extension statement for reporting 45 46 of systematic reviews incorporating network metaanalyses of health care 47 48 interventions: checklist and explanations. Annals of internal medicine 49 50 2015;162(11):77784. doi: 10.7326/m142385 [published Online First: 2015/06/02] 51 52 53 54 55 56 57 58 23 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from

BMJ Open Page 24 of 25

1 2 3 4 5 PRISMA-P (Preferred Reporting Items for Systematic review and Meta-Analysis Protocols) 2015 checklist: recommended items to 6 address in a systematic review protocol* 7 8 Section and topic Item Checklist item Page (Line) 9 No 10 ADMINISTRATIVE INFORMATION 11 Title: 12 Identification 1a Identify the Forreport as a protoco peerl of a systematic reviewreview only 1 (2) 13 Update 1b If the protocol is for an update of a previous systematic review, identify as such Not applicable 14 15 Registration 2 If registered, provide the name of the registry (such as PROSPERO) and registration number 3 (5) 16 Authors: http://bmjopen.bmj.com/ 17 Contact 3a Provide name, institutional affiliation, email address of all protocol authors; provide physical mailing address 1 (11) 18 of corresponding author 19 Contributions 3b Describe contributions of protocol authors and identify the guarantor of the review Not applicable 20 Amendments 4 If the protocol represents an amendment of a previously completed or published protocol, identify as such and Not applicable 21 list changes; otherwise, state plan for documenting important protocol amendments 22 Support: 23 Sources 5a Indicate sources of financial or other support for the review 17 (10) 24 Sponsor 5b Provide name for the review funder and/or sponsor on September 30, 2021 by guest. Protected copyright. Not applicable 25 Role of sponsor 5c Describe roles of funder(s), sponsor(s), and/or institution(s), if any, in developing the protocol 17 (12) 26 or funder 27 28 INTRODUCTION 29 Rationale 6 Describe the rationale for the review in the context of what is already known 5 (5) 30 Objectives 7 Provide an explicit statement of the question(s) the review will address with reference to participants, 5 (24) 31 interventions, comparators, and outcomes (PICO) 32 METHODS 33 34 Eligibility criteria 8 Specify the study characteristics (such as PICO, study design, setting, time frame) and report characteristics 6 (14) (such as years considered, language, publication status) to be used as criteria for eligibility for the review 35 36 Information sources 9 Describe all intended information sources (such as electronic databases, contact with study authors, trial 9 (8) registers or other grey literature sources) with planned dates of coverage 37 38 Search strategy 10 Present draft of search strategy to be used for at least one electronic database, including planned limits, such 9 (11) that it could be repeated 39 40 41 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from

Page 25 of 25 BMJ Open

1 2 3 4 5 Study records: 6 Data 11a Describe the mechanism(s) that will be used to manage records and data throughout the review 10 (4) 7 management 8 Selection 11b State the process that will be used for selecting studies (such as two independent reviewers) through each 10 (5) 9 process phase of the review (that is, screening, eligibility and inclusion in metaanalysis) 10 Data collection 11c Describe planned method of extracting data from reports (such as piloting forms, done independently, in 11 (14) 11 process duplicate), any processes for obtaining and confirming data from investigators 12 Data items 12 List and defineFor all variables peerfor which data will bereview sought (such as PICO items, fundingonly sources), any pre 11 (16) 13 planned data assumptions and simplifications 14 Outcomes and 13 List and define all outcomes for which data will be sought, including prioritization of main and additional 12 (16) 15 prioritization outcomes, with rationale 16 Risk of bias in 14 Describe anticipated methods for assessing risk of bias of individual studies, including whether this will behttp://bmjopen.bmj.com/ 12 (6) 17 individual studies done at the outcome or study level, or both; state how this information will be used in data synthesis 18 Data synthesis 15a Describe criteria under which study data will be quantitatively synthesised Not applicable 19 15b If data are appropriate for quantitative synthesis, describe planned summary measures, methods of handling 13 (7) 20 data and methods of combining data from studies, including any planned exploration of consistency (such as 21 I2, Kendall’s τ) 22 15c Describe any proposed additional analyses (such as sensitivity or subgroup analyses, metaregression) 13 (14) 23 15d If quantitative synthesis is not appropriate, describe the type of summary planned Not applicable 24 Metabias(es) 16 Specify any planned assessment of metabias(es) (such as publication bias across studies, selective reporting on September 30, 2021 by guest. Protected copyright. 14 (22) 25 within studies) 26 Confidence in 17 Describe how the strength of the body of evidence will be assessed (such as GRADE) 15 16) 27 cumulative evidence 28 29 * It is strongly recommended that this checklist be read in conjunction with the PRISMA-P Explanation and Elaboration (cite when available) for important 30 clarification on the items. Amendments to a review protocol should be tracked and dated. The copyright for PRISMA-P (including checklist) is held by the 31 PRISMA-P Group and is distributed under a Creative Commons Attribution Licence 4.0. 32 33 From: Shamseer L, Moher D, Clarke M, Ghersi D, Liberati A, Petticrew M, Shekelle P, Stewart L, PRISMA-P Group. Preferred reporting items for systematic review and 34 meta-analysis protocols (PRISMA-P) 2015: elaboration and explanation. BMJ. 2015 Jan 2;349(jan02 1):g7647. 35 36 37 38 39 40 41 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open

Comparative efficacy and acceptability of interventions for major depression in older persons: protocol for Bayesian network meta-analysis

ForJournal: peerBMJ Open review only Manuscript ID bmjopen-2017-019819.R2

Article Type: Protocol

Date Submitted by the Author: 23-Nov-2017

Complete List of Authors: Liew, Tau Ming; Institute of Mental Health, Department of Geriatric Psychiatry Lee, Cia Sin; SingHealth Polyclinics, Sengkang Polyclinic

Primary Subject Mental health Heading:

Secondary Subject Heading: Geriatric medicine

major depression, older person, efficacy, acceptability, network meta- Keywords: analysis 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-2017-019819 on 21 January 2018. Downloaded from Page 1 of 27 BMJ Open

1 2 3 1 Comparative efficacy and acceptability of interventions for major depression in older 4 5 2 persons: protocol for Bayesian network meta-analysis 6 7 3 8 9 4 Tau Ming Liew1, 2, Cia Sin Lee3 10 11 5 1Department of Geriatric Psychiatry, Institute of Mental Health, Singapore 12 13 6 2 14 Saw Swee Hock School of Public Health, National University of Singapore 15 3 16 7 SingHealth Polyclinics,For Singapore peer review only 17 18 8 19 20 9 21 22 10 Correspondence to 23 24 11 Tau Ming Liew; 25 26 12 27 [email protected] 28 29 13 Department of Geriatric Psychiatry, Institute of Mental Health, 30 31 14 10 Buangkok View, Singapore 539747. http://bmjopen.bmj.com/ 32 33 15 34 35 16 Keywords: major depression; older person; efficacy; acceptability; network metaanalysis 36 37 17 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 Number of words (Abstract): 290 41 42 19 Number of words (main text): 2,946 43 44 20 Number of references: 36 45 46 21 Number of tables or figures: 1 47 48 22 49 50 51 23 52 53 54 55 56 57 58 1 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 2 of 27

1 2 3 1 ABSTRACT 4 5 2 6 7 3 Introduction: Major depression is a leading cause of disability, and has been associated with 8 9 4 adverse effects in older persons. While many pharmacological and nonpharmacological 10 11 5 interventions have been shown to be effective to address major depression in older persons, 12 13 6 14 there has not been a metaanalysis that consolidates all the available interventions and 15 16 7 compare the relativeFor benefits peer of these review available interventions. only In this study, we aim to 17 18 8 conduct a systematic review and network metaanalysis to compare the efficacy and 19 20 9 acceptability of all the known pharmacological and nonpharmacological interventions for 21 22 10 major depression in older persons. 23 24 11 Methods and analysis: We will search MEDLINE, Embase, PsycINFO, Cumulative Index 25 26 12 27 to Nursing and Allied Health, Cochrane Central Register of Controlled Trials and references 28 29 13 of other review articles for articles related to the keywords of ‘randomized trial’, ‘major 30 31 14 depression’, ‘older persons’ and ‘treatments’. Two reviewers will independently select the http://bmjopen.bmj.com/ 32 33 15 eligible articles. For each included article, the two reviewers will independently extract the 34 35 16 data and assess the risk of bias using the Cochrane revised tool for Risk of Bias. Bayesian 36 37 17 network metaanalyses will be conducted to pool the depression scores (based on 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 standardized mean difference) and the allcause discontinuation across all included studies. 41 42 19 The ranking probabilities for all interventions will be estimated and the hierarchy of each 43 44 20 interventions will be summarized as surface under the cumulative ranking curve (SUCRA). 45 46 21 Metaregression and subgroup analyses will also be performed to evaluate the effect of 47 48 22 studylevel covariates. The quality of the evidence will be assessed using the Grading of 49 50 23 Recommendations Assessment, Development and Evaluation (GRADE) approach. 51 52 24 53 Ethics and dissemination: The results will be disseminated through conference 54 55 25 presentations and peerreviewed publications. They will provide the consolidated evidence to 56 57 58 2 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 3 of 27 BMJ Open

1 2 3 1 inform clinicians on the best choice of intervention to address major depression in older 4 5 2 persons. 6 7 3 Trial registration number: International Prospective Register for Systematic Reviews 8 9 4 (PROSPERO) number CRD42017075756. 10 11 5 12 13 14 6 15 16 7 For peer review only 17 18 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 3 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 4 of 27

1 2 3 1 STRENGTHS AND LIMITATIONS OF THIS STUDY 4 5 2 6 7 3 • This systematic review and metaanalysis will provide a comprehensive summary on the 8 9 4 efficacy and acceptability of all available interventions for major depression in older 10 11 5 12 persons. 13 14 6 • The results will provide the highest level of evidence to inform clinicians on the best 15 16 7 choice of treatment,For peer from among review the many available only pharmacological and non 17 18 8 pharmacological interventions. 19 20 9 • This protocol has been developed in accordance with the Preferred Reporting Items for 21 22 10 23 Systematic Review and Metaanalysis Protocols (PRISMAP) statement and has been 24 11 25 registered with PROSPERO. 26 27 12 • The overall quality of evidence will be assessed using the Grading of Recommendations 28 29 13 Assessment, Development and Evaluation (GRADE) approach. 30 http://bmjopen.bmj.com/ 31 14 • This systematic review will be limited to studies which are reported in English language 32 33 15 and have been peerreviewed. 34 35 16 36 37 38 17 39 on September 30, 2021 by guest. Protected copyright. 40 18 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 4 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 5 of 27 BMJ Open

1 2 3 1 INTRODUCTION 4 5 2 6 7 8 3 Rationale 9 10 4 11 12 13 5 Major depression has been identified by the World Health Organization as one of the leading 14 15 6 cause of disability globally.1 2 In older persons, its prevalence rates rise with the increase in 16 For peer review only 17 7 medical comorbidities,3 with reported rates of up to 5% in communitydwelling older 18 19 8 persons,35 5 to 10% in primary care3 6 and as high as 37% after critical care hospitalizations.3 20 21 9 7 Major depression has a significant impact on the older populations and has been linked to 22 23 10 4 8 9 4 10 24 higher risk of suicide, myocardial infarction, stroke, allcause mortality and increasing

25 4 26 11 health services utilization. 27 28 12 29 30 13 Many of the interventions for major depression in older persons have had recent meta 31 http://bmjopen.bmj.com/ 32 14 analyses confirming their efficacy when compared to control groups. These include 33 34 15 antidepressants,1114 cognitive behavioural therapy,15 problem solving therapy,16 35 36 16 1719 37 psychological interventions in general, and the various forms of nonpharmacological

38 2022 39 17 interventions. However, none of the metaanalyses had compared all the on September 30, 2021 by guest. Protected copyright. 40 41 18 pharmacological and nonpharmacological interventions together to demonstrate the relative 42 43 19 benefits of each intervention. It is unknown whether the different types of pharmacological 44 45 20 and nonpharmacological interventions have comparable efficacy and are equally suitable for 46 47 21 older persons with major depression. 48 49 22 50 51 52 23 Objectives 53 54 24 55 56 57 58 5 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 6 of 27

1 2 3 1 In this study, we aim to conduct a systematic review and network metaanalysis to compare 4 5 2 the efficacy and acceptability of all the available pharmacological and nonpharmacological 6 7 3 interventions for major depression in older persons. The use of network metaanalysis allows 8 9 4 us to pool the evidence on various interventions and rank their benefits relative to each 10 11 5 other.23 It also allows us to conduct indirect comparison of the different interventions, even 12 13 6 14 when there is no direct evidence in the literature to allow headtohead comparisons. 15 16 7 For peer review only 17 18 8 19 20 9 METHODS AND ANALYSIS 21 22 10 23 24 11 This protocol is developed in accordance with the Preferred Reporting Items for Systematic 25 26 12 24 25 27 Review and Metaanalysis (PRISMA) statement. It has also been registered with the 28 29 13 International Prospective Register of Systematic Reviews (PROSPERO) (registration number 30 31 14 CRD42017075756). http://bmjopen.bmj.com/ 32 33 15 34 35 16 Eligibility criteria 36 37 17 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 Participants and settings 41 42 19 43 44 20 We will include studies which recruited participants who were: 45 46 21 • 60 years old and above; 47 48 22 • diagnosed with major depression based on formal criteria by the Diagnostic and 49 50 23 51 Statistical Manual of Mental Disorders (DSM) or International Classification of Diseases 52 53 24 (ICD); and 54 55 56 57 58 6 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 7 of 27 BMJ Open

1 2 3 1 • having a current episode of major depression (that is, the participants were symptomatic 4 5 2 and not in remission at the point of recruitment; and the intervention was not intended 6 7 3 primarily for the prevention of future relapses). 8 9 4 10 11 5 12 We will exclude studies which recruited participants with treatmentresistant depression, 13 14 6 subthreshold depression, bipolar depression, depression in dementia or psychotic depression. 15 16 7 We will not includeFor maintenance peer studies review for major depression only as such studies primarily 17 18 8 focused on the prevention of relapses in participants who had been asymptomatic or in 19 20 9 remission at the point of recruitment. 21 22 10 23 24 11 25 Interventions 26 27 12 28 26 27 29 13 We will include studies with pharmacological interventions, including but not limited to: 30 31 14 • Antidepressants such as citalopram, sertraline, venlafaxine or mirtazapine; http://bmjopen.bmj.com/ 32 33 15 • Antipsychotics such as risperidone, quetiapine, olanzapine or aripiprazole; 34 35 16 36 • Moodstabilizers such as valproate, carbamazepine, lithium or gabapentin. 37 38 17 39 on September 30, 2021 by guest. Protected copyright. 40 18 We will include studies with nonpharmacological interventions, including but not limited to: 41 42 19 2830 43 44 20 • Psychological interventions such as cognitive behavioural therapy, problem solving 45 46 21 therapy, interpersonal therapy, family interventions or psychodynamic therapy; 47 48 22 49 • Procedural interventions such as electroconvulsive therapy, transcranial magnetic 50 51 23 stimulation, transcranial directcurrent stimulation or bright light therapy. 52 53 24 54 55 56 57 58 7 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 8 of 27

1 2 3 1 We will also include studies which reported on combinations of any of these pharmacological 4 5 2 and nonpharmacological interventions. 6 7 3 8 9 4 We will exclude studies which focused primarily on health service models of care but were 10 11 5 not related to any modality of intervention, such as studies which evaluated the effectiveness 12 13 6 14 of home treatment, training of general practitioners, multidisciplinary approach or stepped 15 16 7 care approach. For peer review only 17 18 8 19 20 9 Comparators 21 22 10 23 24 11 We will accept control conditions such as placebo intervention, waitinglist, treatment as 25 26 12 27 usual, as well as no intervention. We will also include studies with active comparators such 28 29 13 as those which compare between two different interventions within the same studies. 30 31 14 http://bmjopen.bmj.com/ 32 33 15 Outcomes 34 35 16 36 37 17 We will only include a study if it reports the depression scores or the allcause 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 discontinuation in each study arm following intervention. 41 42 19 43 44 20 Study designs and publication types 45 46 21 47 48 22 We will only include randomized controlled trials (RCTs) which aimed to demonstrate the 49 50 23 superiority of a treatment to another (also known as superiority trials), and will not include 51 52 24 53 equivalence or noninferiority trials. The following study designs or publication types will 54 55 25 also be excluded: qualitative studies, observational studies, metaanalyses, case reports, case 56 57 58 8 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 9 of 27 BMJ Open

1 2 3 1 series, ecological studies and policy papers. We intend to include only higherquality 4 5 2 evidence and hence will exclude nonrandomized trials and publications which were not peer 6 7 3 reviewed (such as conference proceedings, letters and comments). 8 9 4 10 11 5 Language and time frame 12 13 6 14 15 16 7 We will only includeFor studies peer which are reportedreview in the English only language. Apart from that, we 17 18 8 do not impose any time restriction to the publication year of the studies. The search of 19 20 9 databases will be conducted in January 2018. 21 22 10 23 24 11 25 Information sources and search strategy 26 12 27 28 29 13 We will search MEDLINE, Embase, PsycINFO, Cumulative Index to Nursing and Allied 30 31 14 Health (CINAHL) and Cochrane Central Register of Controlled Trials (CENTRAL) for http://bmjopen.bmj.com/ 32 33 15 original articles related to the keywords of ‘randomized trial’, ‘major depression’, ‘older 34 35 16 persons’ and ‘treatments’. Our search strategy for MEDLINE is shown in Box 1. Similar 36 37 17 search strategies will be used for the other databases. Additionally, we will also handsearch 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 the references of review articles related to the topic to retrieve relevant articles which are not 41 42 19 captured through our search of the electronic databases. We will examine the full text of the 43 44 20 relevant articles and include the respective articles if they meet our eligibility criteria. 45 46 21 47 48 Box 1. Search strategy for MEDLINE (via Ovid interface) 49 1. *Therapeutics/ OR *Drug Therapy/ OR *Psychotropic Drugs/ OR *Antidepressive Agents/ OR 50 *Antipsychotic Agents/ OR *Antimanic Agents/ OR *Anticonvulsants/ OR *Psychotherapy/ 51 OR *Electroconvulsive Therapy/ OR *Transcranial Magnetic Stimulation/ OR *Transcranial 52 Direct Current Stimulation/ OR *Phototherapy/ 53 54 2. (antidepressant* OR “selective serotonin ” OR SSRI OR citalopram OR 55 fluoxetine OR paroxetine OR sertraline OR escitalopram OR fluvoxamine OR “serotonin and 56 epinephrine reuptake inhibitor” OR “serotonin epinephrine reuptake inhibitor” OR SNRI OR 57 58 9 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 10 of 27

1 2 3 venlafaxine OR desvenlafaxine OR duloxetine OR milnacipran OR reboxetine OR bupropion 4 OR “noradrenergic and specific serotonergic antidepressant” OR NaSSA OR mirtazapine OR 5 TCA OR tricyclic OR amersergide OR amineptine OR amitriptyline OR amoxapine OR 6 butriptyline OR chlorpoxiten OR clomipramine OR clorimipramine OR demexiptiline OR 7 8 desipramine OR dibenzipin OR dothiepin OR doxepin OR imipramine OR lofepramine OR 9 melitracen OR metapramine OR nortriptyline OR noxiptiline OR opipramol OR protriptyline 10 OR quinupramine OR trimipramine OR tianeptine OR trazodone OR nefazodone OR 11 agomelatine).ab,ti 12 3. (antipsychotic* OR haloperidol OR trifluoperazine OR benperidol OR chlorprothixene OR 13 14 flupenthixol OR clopenthixol OR chlorpromazine OR prochlorperazine OR sulpiride OR 15 periciazine OR perphenazine OR pimozide OR promazine OR fluspirilene OR 16 methotrimeprazineFor OR peer risperidone ORreview paliperidone OR only quetiapine OR olanzapine OR 17 amisulpride OR amisulpiride OR aripiprazole OR clozapine OR sertindole OR zotepine).ab,ti 18 4. ((mood adj stabili*) OR (antimanic adj (agent* OR drug*)) OR anticonvuls* OR anti convuls* 19 OR carbamazepine OR ethosuximide OR gabapentin OR lacosamide OR lamotrigine OR 20 21 levetiracetam OR lithium OR oxcarbazepine OR phenobarbital OR phenytoin OR pregabalin 22 OR rufinamide OR tiagabine OR topiramate OR valproic acid OR valproate OR verapamil OR 23 vigabatrin OR zonisamide).ab,ti 24 5. (psychotherap* OR therap* OR (cognitive adj behavio* adj therapy) OR “cognitive therapy” 25 OR behavio* adj therapy OR “problem solving therapy” OR “problemsolving therapy” OR 26 27 “interpersonal therapy” OR “interpersonal therapy” OR (family adj (therapy OR intervention)) 28 OR psychodynamic OR psychoanalytic OR bibliotherapy OR mindful* OR (group adj (therapy 29 OR intervention)) OR emotionfocused OR “emotion focused” OR reminiscen* OR “life 30 review” OR lifereview).ab,ti http://bmjopen.bmj.com/ 31 6. (“electroconvulsive therapy” OR “electroconvulsive therapy” OR “Transcranial Magnetic 32 33 Stimulation” OR “Transcranial Direct Current Stimulation” OR “light therapy”).ab,ti 34 7. #1 OR #2 OR #3 OR #4 OR #5 OR #6 35 8. *Depressive Disorder, Major/ OR (major adj (depressive OR depression)).ab,ti 36 9. *Aged/ OR *"Aged, 80 and over"/ OR (elder* OR (older adj (person* OR people OR adult*)) 37 OR (late adj life) OR geriatric).ab,ti 38 10. *Randomized Controlled Trial/ OR (Randomized Controlled Trial).pt OR *Random on September 30, 2021 by guest. Protected copyright. 39 40 Allocation/ 41 11. (randomized OR randomised OR (random* adj (assigned OR allocated OR assignment OR 42 allocation))).ab,ti 43 12. #10 OR #11 44 13. #7 AND #8 AND #9 AND #12 45 46 1 47 48 2 Study selection 49 50 3 51 52 4 All potential articles will be retrieved and organized in a data management software (Endnote 53 54 5 software, Thomson Reuters). After removing duplicate records, two reviewers will 55 56 57 58 10 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 11 of 27 BMJ Open

1 2 3 1 independently screen through the titles and abstracts to retain eligible articles. The first 10% 4 5 2 of these titles and abstracts will be subjected to a calibration exercise between the two 6 7 3 reviewers to ensure mutual agreement. 8 9 4 10 11 5 After completing the screening phase, articles that are deemed as relevant by at least one of 12 13 6 14 the reviewers will be subjected to fulltext review. The two reviewers will independently 15 16 7 confirm the eligibilityFor of thesepeer articles reviewbased on the full texts.only The first 10% of these full 17 18 8 texts will again undergo a calibration exercise by the two reviewers. After the fulltext review, 19 20 9 the included articles will be used for qualitative synthesis. The chancecorrected agreement 21 22 10 between the two reviewers will be assessed using Cohen’s Kappa (κ). 23 24 11 25 26 27 12 At any point during study selection, the reasons for excluding specific articles will be 28 29 13 recorded. Moreover, any disagreements between the two reviewers will be resolved by 30 31 14 discussion with a third reviewer. http://bmjopen.bmj.com/ 32 33 15 34 35 16 Data extraction 36 37 17 38 39 on September 30, 2021 by guest. Protected copyright. 40 18 Data from the selected studies will be extracted by two reviewers independently, and 41 42 19 disagreements between the reviewers will be resolved by discussion with a third reviewer. 43 44 20 The extracted data will include the following information: 45 46 21 1. Study identification (first author, year of publication, geographic location) 47 48 22 2. Study characteristics (study setting, study design, inclusion criteria, diagnostic criteria 49 50 23 51 of major depression, sample size) 52 53 54 55 56 57 58 11 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 12 of 27

1 2 3 1 3. Participant characteristics (age, gender, education, number of comorbidities, Mini 4 5 2 Mental State Examination score, baseline depression score, depression scale, duration 6 7 3 of the current episode of major depression) 8 9 4 4. Characteristics of intervention and comparator (description, treatment dose/intensity, 10 11 5 treatment duration, depression score, allcause discontinuation) 12 13 14 6 15 16 7 The original authorsFor of the peer RCTs will be review contacted when the only required data are not available in 17 18 8 the published article. 19 20 9 21 22 10 Assessment of risk of bias 23 24 11 25 26 12 27 The risk of bias for each study will be assessed independently by two reviewers using the 28 31 29 13 Cochrane revised tool for Risk of Bias (RoB 2.0), focusing on biases related to five key 30 31 14 domains: randomization process, deviations from intended interventions, missing outcome http://bmjopen.bmj.com/ 32 33 15 data, measurement of the outcome and selection of the reported result. Each domain will 34 35 16 receive a judgement on the risk of bias (high, low or some concerns) and an overall risk of 36 37 17 bias will be assigned based on the judgements from the five domains. Any disagreements 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 between the two reviewers will be resolved by discussion with a third reviewer. 41 42 19 43 44 20 Outcome measures 45 46 21 47 48 22 Our primary outcomes are the efficacy and the acceptability of interventions. The efficacy 49 50 23 will be based on the difference in depression scores between the intervention and comparator 51 52 24 53 upon the completion of intervention (we will give preference to the primary timepoint 54 55 25 predefined in the original study), computed as standardized mean difference (SMD) for each 56 57 58 12 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 13 of 27 BMJ Open

1 2 3 1 RCT. The acceptability will be assessed by the relative risk (RR) of allcause discontinuation 4 5 2 of the intervention. When the information is available, we will also capture a secondary 6 7 3 outcome of discontinuation due to adverse effects of interventions and evaluate the RR of 8 9 4 discontinuation due to adverse effects. Each intervention will only be grouped by its generic 10 11 5 name for pharmacological interventions (such as citalopram, risperidone, or valproate) or by 12 13 6 14 its known modality for nonpharmacological interventions (such as cognitive behavioural 15 16 7 therapy, problemFor solving therapy,peer or electroconvulsiv reviewe therapy). only We will not categorize the 17 18 8 interventions further in our analyses of the outcome measures. In the event that the active 19 20 9 arm of a RCT involves combinations of interventions, it will be reported as the respective 21 22 10 combinations (such as citalopram–cognitive behavioural therapy combination, or 23 24 11 risperidone–problem solving therapy combination). 25 26 12 27 28 29 13 Statistical analysis 30 31 14 http://bmjopen.bmj.com/ 32 33 15 We will first conduct pairwise metaanalysis provided there are at least two included studies 34 35 16 for each pairwise comparison. If there are at least five included studies, we will use the 36 37 17 random effects model (DerSimonian and Laird method)32 to pool the results because this 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 model does not assume homogeneity among the pooled studies. If there are less than five 41 33 34 42 19 included studies, the random effects model is imprecise in its estimations and we will 43 44 20 choose the fixed effect model (MantelHaenszel method)35 instead. We will use the I2 45 46 21 statistic and the Q test to assess heterogeneity in each pairwise metaanalysis. In the presence 47 48 22 of substantial heterogeneity (I2>50%)36 in a particular intervention, we will consider sub 49 50 23 grouping the intervention by its dose/intensity and duration, and use the subgroups of that 51 52 24 53 intervention in the subsequent network metaanalyses. 54 55 25 56 57 58 13 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 14 of 27

1 2 3 1 We will then conduct the network metaanalyses within a Bayesian framework using the 4 5 2 Markov Chains Monte Carlo method. Bayesian analysis provides probabilistic distributions 6 7 3 of our estimatesofinterest through large number of simulations, and hence produces results 8 9 4 which have more intuitive interpretations. For example, Bayesian analysis generates the 95% 10 11 5 credible interval which can be accurately interpreted as the range containing 95% of the 12 13 6 14 estimates (based on the simulations). In the Bayesian analysis, we will run four Markov 15 16 7 chains simultaneouslyFor withpeer different review arbitrarily chosen only initial values and with non 17 18 8 informative priors. Each chain will have at least 10,000 simulations and at least the first 19 20 9 2,500 simulations will be discarded as burnin. Convergence of the simulations will be 21 22 10 assessed with the trace plots, kernel density plots and GelmanRubinBrooks plots. 23 24 11 25 26 12 27 We will employ both fixedeffects and randomeffects models in the Bayesian analyses, and 28 29 13 will choose the final models based on the deviance information criterion (DIC). While there 30 31 14 is no ruleofthumb on what constitute significant improvements in DIC, we can take http://bmjopen.bmj.com/ 32 33 15 reference from the guideline commonly used in the analogous Akaike Information Criteria:37 34 35 16 values which are lesser by at least 10 points indicate significantly better modelfit and 36 37 17 parsimony. Hence, results from the randomeffects model will be used if the randomeffects 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 model has DIC which is smaller by at least 10 points compared to the fixedeffect model. We 41 42 19 will also compare the complexity of model between the fixedeffects and randomeffects 43 44 20 models using pD (an indicator which has higher value when a model is more complex), with 45 46 21 preference for models which are more parsimonious (less complex). The global 47 48 22 heterogeneity will be assessed with I2 statistic. A common heterogeneity parameter will be 49 50 23 assumed in the randomeffects model. Inconsistency between direct and indirect sources of 51 52 24 38 39 53 evidence will be statistically assessed using the nodesplitting method, which generates a 54 55 25 pvalue for the difference between direct and indirect estimates in each closedloop in the 56 57 58 14 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 15 of 27 BMJ Open

1 2 3 1 network (pvalues of <0.05 indicates the presence of inconsistency between direct and 4 5 2 indirect estimates in a particular closedloop). 6 7 3 8 9 4 We will estimate the ranking probabilities for all interventions and show the results 10 11 5 graphically in the form of rankograms and cumulative ranking probability plots. The 12 13 6 14 hierarchy of interventions will be summarized as surface under the cumulative ranking curve 15 16 7 (SUCRA) and presentedFor inpeer a scatterplot. review SUCRAs have possible only values ranging from 0% to 17 18 8 100%, with higher values indicating better efficacy or acceptability. Publication bias will be 19 20 9 assessed with comparisonadjusted funnel plot.40 41 21 22 10 23 24 11 We will conduct metaregression analyses to determine whether the results of our network 25 26 12 27 metaanalyses will be affected by the following studylevel covariates: sample size, study 28 29 13 duration, inclusion criteria, study setting, study design and risk of bias. A covariate is 30 31 14 considered as a significant moderator if the 95% credible interval of its beta coefficient in http://bmjopen.bmj.com/ 32 33 15 metaregression does not include the value of zero. If a significant moderator is found, 34 35 16 further subgroup analyses will then be conducted to assess the effect of this moderator. 36 37 17 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 The pairwise metaanalyses will be conducted with STATA (version 14). The network meta 41 42 19 analyses will be conducted using JAGS (version 4.2.0), through the GeMTC package of R 43 44 20 (version 3.3.1). The “Network Graphs” package in Stata statistical software (version 14.0) 45 46 21 will also used to produce some of the figures in this study, such as the network plots, 47 48 22 rankograms, cumulative ranking probability plots and comparisonadjusted funnel plots. 40 42 49 50 23 51 52 24 53 Assessment of quality of evidence 54 55 25 56 57 58 15 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 16 of 27

1 2 3 1 We will use the Grading of Recommendations Assessment, Development and Evaluation 4 5 2 (GRADE) approach to report the quality of evidence on efficacy and acceptability of 6 7 3 interventions for major depression in older persons. Based on five key domains 8 9 4 (methodology quality, directness of evidence, heterogeneity, precision of effect estimates and 10 11 5 risk of publication bias), we will classify the quality of evidence in one of four levels – high, 12 13 6 43 14 moderate, low and very low. 15 16 7 For peer review only 17 18 8 LIMITATIONS 19 20 9 21 22 10 Several limitations of this study should be noted. First, there can possibly be heterogeneity in 23 24 11 the dose/intensity and the duration of each intervention, which may limit the interpretation of 25 26 12 27 the metaanalysis. To address this potential limitation, we will first conduct pairwise meta 28 2 29 13 analyses to evaluate the amount of heterogeneity using the I statistic and the Q test. In the 30 31 14 presence of substantial heterogeneity (I2>50%)36 in a particular intervention, we will consider http://bmjopen.bmj.com/ 32 33 15 subgrouping the intervention by its dose/intensity and duration, and use the more 34 35 16 homogeneous subgroups of that intervention in the subsequent network metaanalyses. In the 36 37 17 network metaanalyses, we will also evaluate for inconsistency between direct and indirect 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 estimates using nodesplitting method, and evaluate for heterogeneity using metaregression 41 42 19 and subgroup analyses. Second, we will exclude nonEnglish and nonpeer reviewed 43 44 20 publications (such as conference proceedings and letters). The exclusion of nonpeer 45 46 21 reviewed publications is related to our intention of including only higherquality evidence. 47 48 22 Regardless, we will monitor the impact of such decision and any possible publication bias 49 50 23 using comparisonadjusted funnel plot. Third, we will use allcause discontinuation as a 51 52 24 53 crude composite measure of treatment acceptability. Allcause discontinuation was chosen 54 55 25 (instead of discontinuation due to specific reasons) because this information is more readily 56 57 58 16 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 17 of 27 BMJ Open

1 2 3 1 available in almost all RCTs, especially among nonpharmacological RCTs where it can be 4 5 2 more challenging to clearly attribute the cause of discontinuation to specific reasons such as 6 7 3 adverse effects. Hence, the use of allcause discontinuation will allow us to compare the 8 9 4 acceptability of both pharmacological and nonpharmacological interventions within the 10 11 5 same model in network metaanalysis. 12 13 6 14 15 16 7 ETHICS AND ForDISSEMINATION peer review only 17 18 8 19 20 9 This systematic review will provide the consolidated evidence to inform clinicians on the best 21 22 10 choice of intervention, from among the many available options, to address major depression 23 24 11 in older persons. This systematic review will be reported in accordance with the 25 26 12 44 27 recommendations of PRISMA statement for network metaanalyses. It is expected to be 28 29 13 completed by January 2020, and the results will be disseminated through conference 30 31 14 presentations and publications in peerreviewed journal. http://bmjopen.bmj.com/ 32 33 15 34 35 16 CONTRIBUTORS 36 37 17 38 39 on September 30, 2021 by guest. Protected copyright. 18 40 TML conceived the idea for this systematic review, developed the initial methodology, wrote 41 42 19 the first draft and act as the guarantor of the protocol. CSL provided critical feedback on the 43 44 20 search strategy, methodology and manuscript. All authors approved the final version of the 45 46 21 manuscript. 47 48 22 49 50 23 FUNDING 51 52 24 53 54 55 56 57 58 17 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 18 of 27

1 2 3 1 TML was supported by research grants under the Singapore Ministry of Health’s National 4 5 2 Medical Research Council (Grant No.: NMRC/Fellowship/0030/2016 and 6 7 3 NMRC/CSSSP/0014/2017). The funding source had no involvement in any part of the 8 9 4 project. 10 11 5 12 13 14 6 15 16 7 COMPETING ForINTERESTS peer review only 17 18 8 19 20 9 None declared. 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 18 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 19 of 27 BMJ Open

1 2 3 REFERENCES 4 5 6 7 8 1. World Health Organization. The global burden of disease: 2004 update. Switzerland: 9 10 11 World Health Organization. 12 13 2. World Health Organization. Global Health Estimates 2015: Disease burden by Cause, Age, 14 15 Sex, by Country and by Region, 20002015. Geneva, 2016. 16 For peer review only 17 3. Taylor WD. Clinical practice. Depression in the elderly. The New England journal of 18 19 medicine 2014;371(13):122836. doi: 10.1056/NEJMcp1402180 [published Online 20 21 First: 2014/09/25] 22 23 24 4. Blazer DG. Depression in late life: review and commentary. The journals of gerontology 25 26 Series A, Biological sciences and medical sciences 2003;58(3):24965. [published 27 28 Online First: 2003/03/14] 29 30 5. Volkert J, Schulz H, Härter M, et al. The prevalence of mental disorders in older people in 31 http://bmjopen.bmj.com/ 32 Western countries – a metaanalysis. Ageing Research Reviews 2013;12(1):33953. 33 34 doi: http://dx.doi.org/10.1016/j.arr.2012.09.004 35 36 37 6. Lyness JM, Caine ED, King DA, et al. Psychiatric Disorders in Older Primary Care 38 39 Patients. Journal of general internal medicine 1999;14(4):24954. doi: on September 30, 2021 by guest. Protected copyright. 40 41 10.1046/j.15251497.1999.00326.x 42 43 7. Jackson JC, Pandharipande PP, Girard TD, et al. Depression, Posttraumatic Stress Disorder, 44 45 and Functional Disability in Survivors of Critical Illness: results from the BRAIN 46 47 ICU (Bringing to light the Risk Factors And Incidence of Neuropsychological 48 49 50 dysfunction in ICU survivors) Investigation: A Longitudinal Cohort Study. The lancet 51 52 Respiratory medicine 2014;2(5):36979. doi: 10.1016/S22132600(14)700517 53 54 55 56 57 58 19 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 20 of 27

1 2 3 8. Gan Y, Gong Y, Tong X, et al. Depression and the risk of coronary heart disease: a meta 4 5 analysis of prospective cohort studies. BMC psychiatry 2014;14:371. doi: 6 7 10.1186/s128880140371z [published Online First: 2014/12/30] 8 9 9. Pan A, Sun Q, Okereke OI, et al. Depression and risk of stroke morbidity and mortality: a 10 11 metaanalysis and systematic review. Jama 2011;306(11):12419. doi: 12 13 14 10.1001/jama.2011.1282 [published Online First: 2011/09/22] 15 16 10. Cuijpers P,For Vogelzangs peer N, Twisk review J, et al. Comprehensive only metaanalysis of excess 17 18 mortality in depression in the general community versus patients with specific 19 20 illnesses. The American journal of psychiatry 2014;171(4):45362. doi: 21 22 10.1176/appi.ajp.2013.13030325 [published Online First: 2014/01/18] 23 24 11. Tham A, Jonsson U, Andersson G, et al. Efficacy and tolerability of antidepressants in 25 26 27 people aged 65 years or older with major depressive disorder – A systematic review 28 29 and a metaanalysis. Journal of affective disorders 2016;205:112. doi: 30 31 http://dx.doi.org/10.1016/j.jad.2016.06.013 http://bmjopen.bmj.com/ 32 33 12. Nelson JC, Delucchi K, Schneider LS. Efficacy of second generation antidepressants in 34 35 latelife depression: a metaanalysis of the evidence. The American journal of 36 37 geriatric psychiatry : official journal of the American Association for Geriatric 38 39 on September 30, 2021 by guest. Protected copyright. 40 Psychiatry 2008;16(7):55867. doi: 10.1097/JGP.0b013e3181693288 [published 41 42 Online First: 2008/07/02] 43 44 13. Thorlund K, Druyts E, Wu P, et al. Comparative efficacy and safety of selective serotonin 45 46 reuptake inhibitors and serotoninnorepinephrine reuptake inhibitors in older adults: a 47 48 network metaanalysis. Journal of the American Geriatrics Society 2015;63(5):10029. 49 50 doi: 10.1111/jgs.13395 [published Online First: 2015/05/07] 51 52 53 14. Tedeschini E, Levkovitz Y, Iovieno N, et al. Efficacy of antidepressants for latelife 54 55 depression: a metaanalysis and metaregression of placebocontrolled randomized 56 57 58 20 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 21 of 27 BMJ Open

1 2 3 trials. The Journal of clinical psychiatry 2011;72(12):16608. doi: 4 5 10.4088/JCP.10r06531 [published Online First: 2012/01/17] 6 7 15. Gould RL, Coulson MC, Howard RJ. Cognitive behavioral therapy for depression in 8 9 older people: a metaanalysis and metaregression of randomized controlled trials. 10 11 Journal of the American Geriatrics Society 2012;60(10):181730. doi: 12 13 14 10.1111/j.15325415.2012.04166.x [published Online First: 2012/09/26] 15 16 16. Kirkham JG,For Choi N,peer Seitz DP. review Metaanalysis of ponlyroblem solving therapy for the 17 18 treatment of major depressive disorder in older adults. International journal of 19 20 geriatric psychiatry 2016;31(5):52635. doi: 10.1002/gps.4358 [published Online 21 22 First: 2015/10/06] 23 24 17. Cuijpers P, van Straten A, Smit F. Psychological treatment of latelife depression: a meta 25 26 27 analysis of randomized controlled trials. International journal of geriatric psychiatry 28 29 2006;21(12):113949. doi: 10.1002/gps.1620 [published Online First: 2006/09/07] 30 31 18. Cuijpers P, Karyotaki E, Pot AM, et al. Managing depression in older age: psychological http://bmjopen.bmj.com/ 32 33 interventions. Maturitas 2014;79(2):16069. doi: 10.1016/j.maturitas.2014.05.027 34 35 19. Linde K, Sigterman K, Kriston L, et al. Effectiveness of Psychological Treatments for 36 37 Depressive Disorders in Primary Care: Systematic Review and MetaAnalysis. The 38 39 on September 30, 2021 by guest. Protected copyright. 40 Annals of Family Medicine 2015;13(1):5668. doi: 10.1370/afm.1719 41 42 20. Holvast F, Massoudi B, Oude Voshaar RC, et al. Nonpharmacological treatment for 43 44 depressed older patients in primary care: A systematic review and metaanalysis. PloS 45 46 one 2017;12(9):e0184666. doi: 10.1371/journal.pone.0184666 47 48 21. Apostolo J, BobrowiczCampos E, Rodrigues M, et al. The effectiveness of non 49 50 pharmacological interventions in older adults with depressive disorders: A systematic 51 52 53 review. International journal of nursing studies 2016;58:5970. doi: 54 55 10.1016/j.ijnurstu.2016.02.006 [published Online First: 2016/04/19] 56 57 58 21 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 22 of 27

1 2 3 22. Sukhato K, Lotrakul M, Dellow A, et al. Efficacy of homebased nonpharmacological 4 5 interventions for treating depression: a systematic review and network metaanalysis 6 7 of randomised controlled trials. BMJ open 2017;7(7) 8 9 23. Tonin FS, Rotta I, Mendes AM, et al. Network metaanalysis: a technique to gather 10 11 evidence from direct and indirect comparisons. Pharmacy practice 2017;15(1):943. 12 13 14 doi: 10.18549/PharmPract.2017.01.943 [published Online First: 2017/05/16] 15 16 24. Moher D, ShamseerFor L,peer Clarke M, etreview al. Preferred reporting only items for systematic review 17 18 and metaanalysis protocols (PRISMAP) 2015 statement. Systematic reviews 19 20 2015;4:1. doi: 10.1186/2046405341 [published Online First: 2015/01/03] 21 22 25. Shamseer L, Moher D, Clarke M, et al. Preferred reporting items for systematic review 23 24 and metaanalysis protocols (PRISMAP) 2015: elaboration and explanation. BMJ 25 26 27 (Clinical research ed) 2015;349:g7647. doi: 10.1136/bmj.g7647 [published Online 28 29 First: 2015/01/04] 30 31 26. MacQueen GM, Frey BN, Ismail Z, et al. Canadian Network for Mood and Anxiety http://bmjopen.bmj.com/ 32 33 Treatments (CANMAT) 2016 Clinical Guidelines for the Management of Adults with 34 35 Major Depressive Disorder: Section 6. Special Populations: Youth, Women, and the 36 37 Elderly. Canadian journal of psychiatry Revue canadienne de psychiatrie 38 39 on September 30, 2021 by guest. Protected copyright. 40 2016;61(9):588603. doi: 10.1177/0706743716659276 [published Online First: 41 42 2016/08/04] 43 44 27. Kennedy SH, Lam RW, McIntyre RS, et al. Canadian Network for Mood and Anxiety 45 46 Treatments (CANMAT) 2016 Clinical Guidelines for the Management of Adults with 47 48 Major Depressive Disorder: Section 3. Pharmacological Treatments. Canadian 49 50 journal of psychiatry Revue canadienne de psychiatrie 2016;61(9):54060. doi: 51 52 53 10.1177/0706743716659417 [published Online First: 2016/08/04] 54 55 56 57 58 22 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 23 of 27 BMJ Open

1 2 3 28. Parikh SV, Quilty LC, Ravitz P, et al. Canadian Network for Mood and Anxiety 4 5 Treatments (CANMAT) 2016 Clinical Guidelines for the Management of Adults with 6 7 Major Depressive Disorder: Section 2. Psychological Treatments. Canadian journal 8 9 of psychiatry Revue canadienne de psychiatrie 2016;61(9):52439. doi: 10 11 10.1177/0706743716659418 [published Online First: 2016/08/04] 12 13 14 29. Milev RV, Giacobbe P, Kennedy SH, et al. Canadian Network for Mood and Anxiety 15 16 TreatmentsFor (CANMAT) peer 2016 Clinical review Guidelines for only the Management of Adults with 17 18 Major Depressive Disorder: Section 4. Neurostimulation Treatments. Canadian 19 20 journal of psychiatry Revue canadienne de psychiatrie 2016;61(9):56175. doi: 21 22 10.1177/0706743716660033 [published Online First: 2016/08/04] 23 24 30. Ravindran AV, Balneaves LG, Faulkner G, et al. Canadian Network for Mood and 25 26 27 Anxiety Treatments (CANMAT) 2016 Clinical Guidelines for the Management of 28 29 Adults with Major Depressive Disorder: Section 5. Complementary and Alternative 30 31 Medicine Treatments. Canadian journal of psychiatry Revue canadienne de http://bmjopen.bmj.com/ 32 33 psychiatrie 2016;61(9):57687. doi: 10.1177/0706743716660290 [published Online 34 35 First: 2016/08/04] 36 37 31. Higgins JPT, Sterne JAC, Savović J, et al. A revised tool for assessing risk of bias in 38 39 on September 30, 2021 by guest. Protected copyright. 40 randomized trials. In: Chandler J, McKenzie J, Boutron I, et al., eds. Cochrane 41 42 Methods Cochrane Database of Systematic Reviews 2016, Issue 10 (Suppl 1). 43 44 32. DerSimonian R, Laird N. Metaanalysis in clinical trials. Controlled clinical trials 45 46 1986;7(3):17788. [published Online First: 1986/09/01] 47 48 33. Higgins JP, Thompson SG, Spiegelhalter DJ. A reevaluation of randomeffects meta 49 50 analysis. Journal of the Royal Statistical Society Series A, (Statistics in Society) 51 52 53 2009;172(1):13759. doi: 10.1111/j.1467985X.2008.00552.x [published Online First: 54 55 2009/04/22] 56 57 58 23 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from BMJ Open Page 24 of 27

1 2 3 34. Guyatt G. Users' Guides to the Medical Literature: A Manual for EvidenceBased Clinical 4 5 Practice, 3E: McGrawHill Education 2014. 6 7 35. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective 8 9 studies of disease. Journal of the National Cancer Institute 1959;22(4):71948. 10 11 [published Online First: 1959/04/01] 12 13 14 36. Higgins JP, Thompson SG, Deeks JJ, et al. Measuring inconsistency in metaanalyses. 15 16 BMJ (ClinicalFor research peer ed) 2003;327(7414):55760. review only doi: 10.1136/bmj.327.7414.557 17 18 [published Online First: 2003/09/06] 19 20 37. Burnham KP, Anderson DR. Model Selection and Multimodel Inference: A Practical 21 22 InformationTheoretic Approach: Springer New York 2003. 23 24 38. Dias S, Welton NJ, Caldwell DM, et al. Checking consistency in mixed treatment 25 26 27 comparison metaanalysis. Statistics in medicine 2010;29(78):93244. doi: 28 29 10.1002/sim.3767 [published Online First: 2010/03/10] 30 31 39. van Valkenhoef G, Dias S, Ades AE, et al. Automated generation of nodesplitting http://bmjopen.bmj.com/ 32 33 models for assessment of inconsistency in network metaanalysis. Research synthesis 34 35 methods 2016;7(1):8093. doi: 10.1002/jrsm.1167 [published Online First: 36 37 2015/10/16] 38 39 on September 30, 2021 by guest. Protected copyright. 40 40. Chaimani A, Higgins JP, Mavridis D, et al. Graphical tools for network metaanalysis in 41 42 STATA. PloS one 2013;8(10):e76654. doi: 10.1371/journal.pone.0076654 [published 43 44 Online First: 2013/10/08] 45 46 41. Peters JL, Sutton AJ, Jones DR, et al. Assessing publication bias in metaanalyses in the 47 48 presence of betweenstudy heterogeneity. Journal of the Royal Statistical Society: 49 50 Series A (Statistics in Society) 2010;173(3):57591. doi: 10.1111/j.1467 51 52 53 985X.2009.00629.x 54 55 56 57 58 24 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from Page 25 of 27 BMJ Open

1 2 3 42. Chaimani A, Salanti G. Visualizing assumptions and results in network metaanalysis: 4 5 The network graphs package. Stata Journal 2015;15(4):90550. 6 7 43. Schunemann H, Brozek J, Guyatt G, et al. GRADE handbook for grading quality of 8 9 evidence and strength of recommendation 2013. Available from: 10 11 http://gdt.guidelinedevelopment.org/app/ (assessed 22 Nov 2016). 12 13 14 44. Hutton B, Salanti G, Caldwell DM, et al. The PRISMA extension statement for reporting 15 16 of systematicFor reviews peer incorporating review network only metaanalyses of health care 17 18 interventions: checklist and explanations. Annals of internal medicine 19 20 2015;162(11):77784. doi: 10.7326/m142385 [published Online First: 2015/06/02] 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 25 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from

BMJ Open Page 26 of 27

1 2 3 4 5 PRISMA-P (Preferred Reporting Items for Systematic review and Meta-Analysis Protocols) 2015 checklist: recommended items to 6 address in a systematic review protocol* 7 8 Section and topic Item Checklist item Page (Line) 9 No 10 ADMINISTRATIVE INFORMATION 11 Title: 12 Identification 1a Identify the Forreport as a protoco peerl of a systematic reviewreview only 1 (2) 13 Update 1b If the protocol is for an update of a previous systematic review, identify as such Not applicable 14 15 Registration 2 If registered, provide the name of the registry (such as PROSPERO) and registration number 3 (5) 16 Authors: http://bmjopen.bmj.com/ 17 Contact 3a Provide name, institutional affiliation, email address of all protocol authors; provide physical mailing address 1 (11) 18 of corresponding author 19 Contributions 3b Describe contributions of protocol authors and identify the guarantor of the review Not applicable 20 Amendments 4 If the protocol represents an amendment of a previously completed or published protocol, identify as such and Not applicable 21 list changes; otherwise, state plan for documenting important protocol amendments 22 Support: 23 Sources 5a Indicate sources of financial or other support for the review 17 (10) 24 Sponsor 5b Provide name for the review funder and/or sponsor on September 30, 2021 by guest. Protected copyright. Not applicable 25 Role of sponsor 5c Describe roles of funder(s), sponsor(s), and/or institution(s), if any, in developing the protocol 17 (12) 26 or funder 27 28 INTRODUCTION 29 Rationale 6 Describe the rationale for the review in the context of what is already known 5 (5) 30 Objectives 7 Provide an explicit statement of the question(s) the review will address with reference to participants, 5 (24) 31 interventions, comparators, and outcomes (PICO) 32 METHODS 33 34 Eligibility criteria 8 Specify the study characteristics (such as PICO, study design, setting, time frame) and report characteristics 6 (14) (such as years considered, language, publication status) to be used as criteria for eligibility for the review 35 36 Information sources 9 Describe all intended information sources (such as electronic databases, contact with study authors, trial 9 (8) registers or other grey literature sources) with planned dates of coverage 37 38 Search strategy 10 Present draft of search strategy to be used for at least one electronic database, including planned limits, such 9 (11) that it could be repeated 39 40 41 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 BMJ Open: first published as 10.1136/bmjopen-2017-019819 on 21 January 2018. Downloaded from

Page 27 of 27 BMJ Open

1 2 3 4 5 Study records: 6 Data 11a Describe the mechanism(s) that will be used to manage records and data throughout the review 10 (4) 7 management 8 Selection 11b State the process that will be used for selecting studies (such as two independent reviewers) through each 10 (5) 9 process phase of the review (that is, screening, eligibility and inclusion in metaanalysis) 10 Data collection 11c Describe planned method of extracting data from reports (such as piloting forms, done independently, in 11 (14) 11 process duplicate), any processes for obtaining and confirming data from investigators 12 Data items 12 List and defineFor all variables peerfor which data will bereview sought (such as PICO items, fundingonly sources), any pre 11 (16) 13 planned data assumptions and simplifications 14 Outcomes and 13 List and define all outcomes for which data will be sought, including prioritization of main and additional 12 (16) 15 prioritization outcomes, with rationale 16 Risk of bias in 14 Describe anticipated methods for assessing risk of bias of individual studies, including whether this will behttp://bmjopen.bmj.com/ 12 (6) 17 individual studies done at the outcome or study level, or both; state how this information will be used in data synthesis 18 Data synthesis 15a Describe criteria under which study data will be quantitatively synthesised Not applicable 19 15b If data are appropriate for quantitative synthesis, describe planned summary measures, methods of handling 13 (7) 20 data and methods of combining data from studies, including any planned exploration of consistency (such as 21 I2, Kendall’s τ) 22 15c Describe any proposed additional analyses (such as sensitivity or subgroup analyses, metaregression) 13 (14) 23 15d If quantitative synthesis is not appropriate, describe the type of summary planned Not applicable 24 Metabias(es) 16 Specify any planned assessment of metabias(es) (such as publication bias across studies, selective reporting on September 30, 2021 by guest. Protected copyright. 14 (22) 25 within studies) 26 Confidence in 17 Describe how the strength of the body of evidence will be assessed (such as GRADE) 15 16) 27 cumulative evidence 28 29 * It is strongly recommended that this checklist be read in conjunction with the PRISMA-P Explanation and Elaboration (cite when available) for important 30 clarification on the items. Amendments to a review protocol should be tracked and dated. The copyright for PRISMA-P (including checklist) is held by the 31 PRISMA-P Group and is distributed under a Creative Commons Attribution Licence 4.0. 32 33 From: Shamseer L, Moher D, Clarke M, Ghersi D, Liberati A, Petticrew M, Shekelle P, Stewart L, PRISMA-P Group. Preferred reporting items for systematic review and 34 meta-analysis protocols (PRISMA-P) 2015: elaboration and explanation. BMJ. 2015 Jan 2;349(jan02 1):g7647. 35 36 37 38 39 40 41 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48