Design

This study is a systematic review and IPD meta-analysis registered in the PROSPERO International prospective register of systematic reviews (PROSPERO: CRD42020219891). Additional important protocol amendments will be updated in this register. This paper accords with the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) statement.^{Reference Shamseer, Moher, Clarke, Ghersi, Liberati and Petticrew11} The 1 September 2019 was the funding date for this study, after which work on the protocol started. Data extraction will start after PROSPERO registration. The study is expected to be completed 31 March 2023. This protocol builds on and extends prior work by our group.^{Reference Driessen, Abbass, Barber, Connolly Gibbons, Dekker and Fokkema12,Reference Weitz, Kleiboer, van Straten, Hollon and Cuijpers13}

Eligibility criteria

We will include (a) randomised clinical trials (b) comparing antidepressant medication and IPT (c) in the acute-phase treatment of (d) adults with depression.

We will include studies in which antidepressant medication and IPT are directly compared among participants randomly assigned to these treatments. We will consider any type of standard oral antidepressant medication within the therapeutic dose range (such as, selective serotonin reuptake inhibitors (SSRIs), selective serotonin–noradrenaline reuptake inhibitors (SNRIs), tricyclic antidepressants (TCAs) and monoamine oxidase inhibitors (MAOIs)).

An intervention will be considered to be an IPT when it is a psychotherapy based on the manuals developed by Klerman & Weissman for IPT or for the briefer version called interpersonal counselling.^{Reference Klerman, Weissman, Rounsaville and Chevron14–Reference Weissman, Hankerson, Scorza, Olfson, Verdeli and Shea18} We will include IPT in any delivery format (such as individual, group, face to face, telephone or video-conferencing), as long as a clinician delivers the therapy. Unguided bibliotherapy or unguided internet interventions will thus be excluded. The IPT can be delivered in any setting (for example primary care, out-patient and in-patient psychiatric care) and we will place no restrictions on the number of IPT sessions and the duration of follow-up.

We will only include studies of acute-phase depression treatment. Thus, we will exclude, for example, studies that randomised participants to antidepressants and IPT as maintenance treatments after successful acute treatment. Only data from the relevant phases and comparisons of eligible studies will be included. For example, for studies involving augmentation of antidepressants and IPT following non-response to monotherapy, only study data up until the first triage point for augmentation will be included.

Participants must be at least 18 years old. Studies concerning older adult populations will be included as well. Participants will be considered to have depression if they meet specified criteria (for example DSM, ICD-10) for major depressive disorder or another unipolar mood disorder assessed by means of a semi-structured interview or clinicians’ assessment, or if they present an elevated score above the ‘no depression’ cut-off on an evaluator-assessed, clinician-assessed or self-reported measure of depression. Comorbid mental and somatic disorders will be allowed.

Participant criteria will be assessed at study level. Thus, eligible participants from a study including a wider population (for example adults from a study including both adolescents and adults) will not be included, because the integrity of randomisation within the subgroup of eligible participants could be compromised. No restrictions will be placed concerning the years when the study was conducted, or with regard to publication language, date or status.

Information sources, search strategy and selection process

To identify studies, we will search a database of randomised clinical trials examining the efficacy and effectiveness of psychological treatments for depression that has been used in a series of published meta-analyses (www.metapsy.org). This METAPSY database was developed through comprehensive literature searches in the bibliographic databases PubMed, PsycINFO, Embase and the Cochrane Library and is updated annually. The search strings use a combination of index terms and free-text words indicative of depression and psychotherapies, with filters for randomised clinical trials. The exact terms for the searches are available from https://osf.io/nv3ea/. Searches will be performed until 1 January 2020.

Two experienced depression treatment researchers will independently screen all records, assess full-text papers for METAPSY database eligibility, and rate the treatment comparison(s) examined. Next, two other experienced depression treatment researchers will independently assess all full-text papers of studies marked as comparing a psychotherapy monotreatment condition against another active monotreatment condition for meeting the inclusion criteria for this work. Disagreements will be resolved through consensus in each phase.

In addition, we will check the references of IPT^{Reference Feijo de Mello, de Jesus Mari, Bacaltchuk, Verdeli and Neugebauer5–Reference Cuijpers, Donker, Weissman, Ravitz and Cristea7,Reference van Hees, Rotter, Ellermann and Evers19} and of psychotherapy versus pharmacotherapy^{Reference Cuijpers, Noma, Karyotaki, Vinkers, Cipriani and Furukawa20} efficacy and effectiveness reviews, as well as the reference lists of the included studies. We will contact a listserv of members of the International Society of Interpersonal Psychotherapy to request ongoing or unpublished studies, and studies that were missed.

Data collection

We will contact authors of the included studies and invite them to contribute the participant-level data of their studies. Researchers who share their data will be offered co-authorship for all publications that are based on their study data, inasmuch as they meet internationally accepted criteria for authorship of scientific publications (www.icmje.org). In addition, the collected data will be made available to investigators who contribute data to examine other research questions in the combined data-set. This strategy, used in previous IPD meta-analyses concerning depression treatments,^{Reference Driessen, Abbass, Barber, Connolly Gibbons, Dekker and Fokkema12,Reference Weitz, Hollon, Twisk, van Straten, Huibers and David21} has successfully convinced researchers to share their data.

Contact details of all corresponding authors will be collected from the relevant publications, or if not reported there, through internet searches or personal contacts with other researchers. Corresponding authors will be contacted by email with an invitation outlining the project's goals and asking if they would be willing to collaborate by sharing their participant-level trial data. If the corresponding author does not respond after 3 weeks, a second and third email will be sent. If these attempts fail, the other authors will be contacted in the same way, in this order: first, last, second, third, fourth, etc. In case of non-response to email, a letter will be mailed to the corresponding author (again with three attempts). If still no response is received, we will try to contact the corresponding author by telephone. If the corresponding author does not respond, we will contact the other authors by letter and telephone. If none of the authors respond to these efforts, we shall seek other ways of contacting one of the authors (for example via colleagues who might know them).

Study data will be considered unavailable only if all these attempts fail, or if an author indicates that the participant-level data have not been retained or declines to share these data. This strategy has also been used in a previous IPD meta-analysis of depression treatment^{Reference Driessen, Abbass, Barber, Connolly Gibbons, Dekker and Fokkema12} and has proven to be successful in reaching authors.

If the author is willing and able to share their trial IPD, the author will transfer the participant-level data-set, including all outcome variables assessed during and after treatment, both in the antidepressant medication and the IPT condition of the study. Data-sets will be anonymised by the authors prior to transfer, so that no personal data are transferred.

Data integrity check

After the data-set has been transferred, the file will be checked to examine whether the data received match the data reported in the publication, following Driessen and colleagues.^{Reference Driessen, Abbass, Barber, Connolly Gibbons, Dekker and Fokkema12} For both the antidepressant medication and the IPT condition, sample size, number of female participants, mean age, observed mean pre-treatment depression scores, observed mean post-treatment score for the primary depression outcome and the number of missing cases for the latter will be calculated from the data-set received and checked against the published article for this purpose. Discrepancies will be resolved with the authors. In addition, the data will be checked for invalid, out-of-range or inconsistent items.

For each study, we will list all outcome measures and assessment points. From the published articles, we will extract multiple study-level characteristics (country, recruitment method, target group, depression inclusion criteria) and treatment characteristics (antidepressant type, number of IPT sessions, IPT format).

To ascertain antidepressant treatment quality, studies will be assessed for the use of a therapeutic dosage and titration schedule (i.e. therapeutic dose achieved within 3 weeks). Study pharmacotherapy will be deemed adequate if both criteria are met. To ascertain IPT quality, studies will be assessed with respect to using a treatment manual, provision of therapy by specially trained therapists and verification of treatment integrity.^{Reference Chambless and Hollon22,Reference Cuijpers, van Straten, Bohlmeijer, Hollon and Andersson23}

Two raters will independently assess the Cochrane risk-of-bias tool for randomised trials at outcome level.^{Reference Higgins, Savović, Page, Elbers, Sterne, Higgins, Thomas, Chandler, Cumpston, Li, Page and Welch24} Disagreements will be resolved by consensus. When publications omit reporting relevant information, we will ask the authors. For studies for which IPD were not obtained, we will extract effect-size data at baseline, post-treatment and follow-up for each outcome measure reported.

After checking the data, the data-sets will be standardised.^{Reference Driessen, Abbass, Barber, Connolly Gibbons, Dekker and Fokkema12} For this purpose, a copy of each trial's raw data file will be recoded into a data file that matches the IPD meta-analysis database in terms of variables. Next, the individual study data files will be merged into a single database in which each participant is identified by a study ID and a unique individual participant ID. After all data files have been recoded and entered, the data for each study will be checked with the original data file for accuracy.

Measures

The primary outcome for this study is treatment efficacy as assessed by a continuous measure of depression symptom severity at post-treatment. We have chosen depressive symptoms at post-treatment as the main outcome for this study as we consider this the primary target of antidepressant and IPT treatment. For each trial, we will identify the primary continuous measure of depressive symptom severity defined by the study authors. All instruments explicitly measuring depression qualify in this regard. To account for the likelihood that different primary depression measures will have been used, we will standardise depression outcomes by converting depression scores into z-scores within each study^{Reference Driessen, Abbass, Barber, Connolly Gibbons, Dekker and Fokkema12} as well as by exploring novel approaches based on equipercentile linking^{Reference Furukawa, Reijnders, Kishimoto, Sakata, DeRubeis and Dimidjian25} and item-response theory.^{Reference Wahl, Löwe, Bjorner, Fisher, Langs and Voderholzer26} Sensitivity analyses will be conducted using unstandardised scores for each specific measure that is assessed in two or more studies included in the meta-analysis (for example the Hamilton Rating Scale for Depression,^{Reference Hamilton27} the Beck Depression Inventory^{Reference Beck, Ward, Mendelson, Mock and Erbaugh28}).^{Reference Driessen, Abbass, Barber, Connolly Gibbons, Dekker and Fokkema12} Sensitivity analyses will also be conducted for studies that include outcomes within specific time windows (for example 12–16 weeks).

Secondary outcomes for this study are depression outcomes at follow-up as well as all outcome measures other than depression symptom severity that are assessed at post-treatment or follow-up in at least two of the included studies. These include quality of life, social functioning and health status. As individual studies will likely assess these domains differently, they will be standardised by converting scores into z-scores within each study. The sensitivity analyses described previously for the primary outcome will also be conducted for the secondary outcomes.

Data analysis

Following Driessen et al,^{Reference Driessen, Dekker, Peen, Van, Maina and Rosso29} all data-sets received that contain IPD and at least one relevant outcome measure will be considered for quantitative synthesis. We will conduct IPD meta-analyses according to the one-stage approach, because that provides a more exact likelihood estimate in the case of small studies.^{Reference Burke, Ensor and Riley30} We will conduct IPD meta-analyses using mixed models with restricted maximum likelihood to estimate between-study heterogeneity, which is recommended when there are few studies in the meta-analysis or studies have small sample sizes.^{Reference Higgins, Whitehead, Turner, Omar and Thompson31} Analyses will be based on intention-to-treat samples including all randomised participants. Follow-up data will be excluded from post-treatment analyses, because additional treatment cannot be controlled. We will assess heterogeneity with the I ² statistic, which describes the variance between studies as a proportion of the total variance. Mixed-model analyses will be conducted with MLwiN (version 2.35 for PC, Centre for Multilevel Modelling, University of Bristol).

We shall start with a basic model including a main effect for time and a time × treatment interaction.^{Reference Driessen, Dekker, Peen, Van, Maina and Rosso29} Twisk and colleagues recommend this approach,^{Reference Twisk, Bosman, Hoekstra, Rijnhart, Welten and Heymans32} because it adequately accounts for baseline values and has favourable properties concerning missing data, allowing participants with only a baseline value but missing post-treatment and/or follow-up assessments to remain included in the analyses. Time will be treated as a categorical variable to facilitate treatment comparison at the different time points.

To account for clustering of participants within studies, we will estimate a random intercept with respect to the study. We will also estimate a random intercept with respect to participants, and fixed slopes. If adding a random slope for the time × treatment interaction on study level results in a model improvement, models including this random slope will be used for effect estimation.

Using this approach, treatment effects can be directly obtained from the regression coefficient of the time × treatment interaction.^{Reference Driessen, Dekker, Peen, Van, Maina and Rosso29,Reference Twisk, Bosman, Hoekstra, Rijnhart, Welten and Heymans32} The regression coefficients of the time × treatment interactions at post-treatment and follow-up represent the treatment comparisons at these assessment moments. These can be interpreted as Cohen's d effect sizes for analyses with z-scores as outcome measure, and as mean differences for analyses with unstandardised scores as outcome measure.

A number of sensitivity analyses will be conducted to examine the robustness of the findings. Because outcomes of IPT for depression have been found to be associated with the number of sessions,^{Reference Cuijpers, Donker, Weissman, Ravitz and Cristea7} we will repeat the analyses in subgroups of studies examining IPT and the briefer interpersonal counselling. Because there are indications that psychological treatments of dysthymia are less effective than psychological treatments of major depressive disorder,^{Reference Cuijpers, van Straten, Schuurmans, van Oppen, Hollon and Andersson33} we will repeat the analyses in subgroups of studies that include participants meeting diagnostic criteria for dysthymia versus studies that include participants meeting diagnostic criteria for other unipolar mood disorders.

To examine potential effects of study setting, depression inclusion criteria, age and antidepressant type, we will also repeat the analyses in the following subgroups of studies: primary versus secondary or tertiary care, studies that enrolled participants meeting diagnostic criteria for depression versus studies that enrolled participants presenting an elevated score on a standardised measure of depression, studies that enrolled adult versus studies that enrolled older adult populations, and for subgroups of studies examining different classes of antidepressants (such as SSRI, SNRI, TCA, MAOI).

We will also conduct sensitivity analyses to control for the effects of additional non-study treatment during the trial and in the follow-up period by adding collected data in this regard as covariates to the mixed-effects models. To examine the impact of risk of bias, we will conduct analyses in which we add risk-of-bias items as covariates to the mixed-effects models.^{Reference Higgins, Thomspon and Spiegelhalter34} In addition, we will conduct sensitivity analyses only including those studies that score negative on all risk-of-bias criteria assessed. We will examine the impact of treatment quality in the same way.

Furthermore, we will examine possible data-availability bias following prior work.^{Reference Driessen, Abbass, Barber, Connolly Gibbons, Dekker and Fokkema12} We will conduct t-tests and χ² analyses, comparing the included studies with studies for which no participant-level data were obtained with regard to the extracted study characteristics using SPSS Statistics (version 25 for PC). For studies for which IPD were not obtained, we will calculate Cohen's d effect sizes for each relevant outcome measure based on data extracted from the publications. Conventional meta-analysis techniques will be utilised to examine differences in effect sizes between studies that contributed data and studies that did not. More specifically, we will conduct subgroup analyses comparing studies for which IPD were versus studies for which IPD were not obtained using Comprehensive Meta-analysis (version 3.3.070 for PC, Biostat). We will apply a fully random-effects analysis and pool study-to-study variance across subgroups, as is recommended when subgroups involve small numbers of studies.^{Reference Borenstein, Hedges, Higgins and Rothstein35} We will assess potential publication bias by assessing asymmetry in funnel plots for meta-analyses including ten or more trials (regardless of whether or not IPD were obtained), as is recommended by Sterne and colleagues.^{Reference Sterne, Sutton and Ioannidis36} We will assess the strength of the body of evidence for each outcome measure based on the number of included studies and participants as well as the quality of the included studies.