Sikander 2019.

Study characteristics
Methods	Study design: single‐blind, cluster‐randomised controlled trial; unit of allocation: village cluster Duration of study: screening between 15 October 2014 and 25 February 2016. Last assessment at 6 months after childbirth; total duration for each participant is up to 9 months
Participants	Country: Pakistan  Income classification: lower‐middle income between 2014 and 2016   Geographical scope: Kallar Syedan, a rural subdistrict of Rawalpindi, Pakistan   Healthcare setting: individual THPP sessions were delivered by Razakaars at participants’ homes, and group sessions were delivered at the LHW’s health house or at a nearby place that was convenient for participants   Mental health condition: perinatal MD  Population  1. Age: 18+ years  2. Gender: women   3. Socioeconomic background: over 93% did not work; over 52% had no formal education   4. Inclusion criteria  a. Women aged 18 years or older  b. In third trimester of pregnancy c. Registered with local LHWs d. Intended to stay in the study area for at least 1 year  e. Had scored ≥ 10 on Patient Health Questionnaire‐9 (PHQ‐9). 5. Exclusion criteria  a. Did not speak Urdu, Punjabi, or Potohari b. Needed immediate medical or psychiatric inpatient care
Interventions	Stated purpose: to evaluate effectiveness and cost‐effectiveness of the adapted Thinking Healthy Programme peer‐delivered (THPP) compared with enhanced usual care (EUC) in Pakistan  INTERVENTION (n = 227)  Name: the Thinking Healthy Programme  (peer‐delivered) Delivered by: LHW  Title/name of PW and number: Razakaars, volunteer peers (lay women from the community) ‐ 66   1. Selection: local volunteers; married women, around ages 30 to 35 years, with good communication skills. Possible Razakaars were identified by LHWs and by community elders   2. Educational background: not specified, but stated similar educational background to participants   3. Training (contents, duration, by whom): peers received brief classroom training and field training, and were able to deliver the intervention to satisfactory fidelity. A high proportion of women (78%) completed ≥ 10 sessions of the possible 14 sessions, indicating the acceptability of peers in delivering these sessions   4. Supervision: brief classroom training of peers was supplemented with regular group training and field supervision by local THPP trainers, who were not mental health specialists; these THPP trainers were supervised by a specialist therapist, generating a cascade model of training and supervision   5. Incentives/remuneration: peers received no financial remuneration for this work   Intervention details   1. Duration/frequency: THPP consisted of 10 individual and 4 group sessions, each of which lasted 30 to 45 minutes, from third trimester of pregnancy to 6 months after childbirth   2. Content of intervention (by types of health workers and per patients/carers): THPP sessions were front‐loaded (i.e. greater frequency of sessions, intensity, and content were delivered during pregnancy until the third month after childbirth); 10 of the 14 sessions were delivered during pregnancy and in the first 3 months after childbirth. This approach was taken to ensure an early reduction in maternal depressive symptoms during this crucial phase of infant care   CONTROL: enhanced usual care (n = 226) Standard care from lady health workers (LHWs), which did not include treatment of perinatal depression. In addition, participants and LHWs who had registered them were informed of their screening results, and all doctors and midwives at primary healthcare centres were given the adapted mental health Gap Action Programme treatment guidelines for perinatal depression, which included information on how to refer patients with severe depression and those at high risk of suicide to specialist mental healthcare facilities; participants were provided with an information sheet that included details on where to seek appropriate health care during pregnancy and beyond CO‐INTERVENTIONS: enhanced usual care was given to participants from both groups
Outcomes	Patients 1. Symptom severity (PHQ‐9) at 6 months postpartum* 2. Remission (PHQ‐9 < 5) at 6 months postpartum* 3. PHQ‐9 at 3 months# 4. Remission at 3 months# 5. Recovery (PHQ‐9 < 5 at 3 and 6 months) 6. WHODAS 7. Number of days unable to work 8. Perceived social support# 9. Exclusive breastfeeding# 10. Infant weight‐for‐age and height‐for‐age scores# 11. Serious adverse events (death of participant due to any cause, death of the child, suicide attempts, hospital admissions, experiences of physical violence, infant abuse or neglect, social stigmatisation regarding the study, violence towards others)   Carers  None  Process/health workers Proportion of women attending at least 10 of 14 sessions  Economic outcomes  1. Average duration of health service use  2. Average time to access services 3. Any health service use, unit costs, health system costs, productivity costs, societal costs (asterisk for study's primary outcomes; star: outcomes that we have not reported in this review)  Time points post intervention: baseline, during intervention (3 months' postpartum), end of intervention (6 months' postpartum)
Notes	Source of funding: National Institute of Mental Health (USA)  Notes on validation of instruments (screening and outcomes): all instruments were validated   Additional information: declarations of interest ‐ study authors declared no competing interests Handling the data: as per footnotes in data and analysis  Prospective trial registration number: trial was registered with ClinicalTrials.gov (NCT02111915)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Judgement comment: randomisation was conducted using a 'computerised randomisation sequence'
Allocation concealment (selection bias)	Low risk	"Randomisation list for village clusters, which was stratified by 11 union councils (the smallest administrative unit of the subdistrict), was prepared by an independent statistician (HAW, who had no subsequent involvement in the trial) by use of a computerised randomisation sequence"; "all members of the Trial Steering Committee, except for the data manager (AZ), remained masked to allocation status until the data were unmasked after interpretation of the results" Assignment to groups was concealed and conducted by an independent researcher
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Single‐blind, cluster‐randomised controlled trial. Participants and personnel not blinded to intervention; this is unlikely to affect outcomes
Blinding of outcome assessment (detection bias) all outcomes	Low risk	"Outcome assessors were masked to treatment allocation of participants during baseline and follow‐up assessments, had no interaction with the intervention team, and resided outside of the study area" Outcomes were assessed blindly by assessors, and precautions were taken to ensure masking
Baseline outcome measurements similar	Low risk	"There was no evidence of a difference in baseline characteristics in women for whom we had 6‐month outcome data and those for whom we did not, except for a slightly longer time between screening and childbirth for those from whom we were missing primary outcome data (which was therefore adjusted for in outcome analyses), and there was no difference in the baseline characteristics of those who had a study visit during the protocol‐defined window compared with those who did not" Baseline outcome measurements were similar between groups, and any differences were adjusted for in analyses
Baseline characteristics similar?	Low risk	"There were no major imbalances between the groups, except a slightly higher proportion of women with duration of depression of more than 12 weeks in the intervention group versus the control group (which was therefore adjusted for in outcome analyses)" Any differences in baseline characteristics were adjusted for in analysis
Incomplete outcome data (attrition bias) Efficacy data	Low risk	Missing outcome data were adjusted for during analysis "Sensitivity analyses for primary outcomes included random‐effects models to account for missing outcome data with multiple imputation (which assumed data were missing at random) and alternative models for PHQ‐9 score"
Incomplete outcome data (attrition bias) Safety data (e.g. adverse events)	Low risk	"Overall, 43 (15%) women in the intervention group and 47 (16%) women in the control group had at least one serious adverse event, which were evenly distributed between the groups (P = 0·72; appendix). The most common serious adverse events were death of the child, hospital admissions (mainly of the child), and experience of physical violence; however, there was no evidence of any difference between the groups" Adverse events were similar between groups; there was no difference
Protection against contamination	Low risk	"To minimise contamination, we used eligible village clusters that were geographically separate" Precautions were taken to prevent contamination "We assumed a more conservative effect size to allow for the possibility of contamination between groups" However, this was also adjusted for during analyses
Selective reporting (reporting bias)	Low risk	Outcomes described in trial registry: remission (i.e. recovery from depression) measured with Patient Health Questionnaire‐9 (PHQ‐9). Remission (i.e. recovery from depression) measured with Patient Health Questionnaire‐9 (PHQ‐9). Maternal disability measured with World Health Organization Disability Assessment Schedule (WHO‐DAS). Maternal support measured with Multidimensional Scale of Perceived Social Support (MSPSS). Breastfeeding rates, infant height, all outcomes described in trial registry were reported
Other bias	Low risk	No other evidence of bias