Peer-led counselling with problem discussion therapy for adolescents living with HIV in Zimbabwe: A cluster-randomised trial

Background Adolescents living with HIV have poor virological suppression and high prevalence of common mental disorders (CMDs). In Zimbabwe, the Zvandiri adolescent peer support programme is effective at improving virological suppression. We assessed the effect of training Zvandiri peer counsellors known as Community Adolescent Treatment Supporters (CATS) in problem-solving therapy (PST) on virological suppression and mental health outcomes. Methods and findings Sixty clinics were randomised 1:1 to either normal Zvandiri peer counselling or a peer counsellor trained in PST. In January to March 2019, 842 adolescents aged 10 to 19 years and living with HIV who screened positive for CMDs were enrolled (375 (44.5%) male and 418 (49.6%) orphaned of at least one parent). The primary outcome was virological nonsuppression (viral load ≥1,000 copies/mL). Secondary outcomes were symptoms of CMDs measured with the Shona Symptom Questionnaire (SSQ ≥8) and depression measured with the Patient Health Questionnaire (PHQ-9 ≥10) and health utility score using the EQ-5D. The adjusted odds ratios (AORs) and 95% confidence intervals (CIs) were estimated using logistic regression adjusting for clinic-level clustering. Case reviews and focus group discussions were used to determine feasibility of intervention delivery. At baseline, 35.1% of participants had virological nonsuppression and 70.3% had SSQ≥8. After 48 weeks, follow-up was 89.5% for viral load data and 90.9% for other outcomes. Virological nonsuppression decreased in both arms, but there was no evidence of an intervention effect (prevalence of nonsuppression 14.7% in the Zvandiri-PST arm versus 11.9% in the Zvandiri arm; AOR = 1.29; 95% CI 0.68, 2.48; p = 0.44). There was strong evidence of an apparent effect on common mental health outcomes (SSQ ≥8: 2.4% versus 10.3% [AOR = 0.19; 95% CI 0.08, 0.46; p < 0.001]; PHQ-9 ≥10: 2.9% versus 8.8% [AOR = 0.32; 95% CI 0.14, 0.78; p = 0.01]). Prevalence of EQ-5D index score <1 was 27.6% versus 38.9% (AOR = 0.56; 95% CI 0.31, 1.03; p = 0.06). Qualitative analyses found that CATS-observed participants had limited autonomy or ability to solve problems. In response, the CATS adapted the intervention to focus on empathic problem discussion to fit adolescents’ age, capacity, and circumstances, which was beneficial. Limitations include that cost data were not available and that the mental health tools were validated in adult populations, not adolescents. Conclusions PST training for CATS did not add to the benefit of peer support in reducing virological nonsuppression but led to improved symptoms of CMD and depression compared to standard Zvandiri care among adolescents living with HIV in Zimbabwe. Active involvement of caregivers and strengthened referral structures could increase feasibility and effectiveness. Trial registration Pan African Clinical Trials Registry PACTR201810756862405.


Hypotheses
 The primary hypothesis is that the CATS-PST intervention will be more effective than standard CATS care in reducing the proportion of ALHIV who have died or who have virological failure at 48 weeks (defined as a viral load 1000 copies/ml).  The secondary hypotheses are that adolescents receiving the CATS-PST intervention will have (i) reduced prevalence and severity of depression and/or anxiety (using the SSQ-14 and PHQ-9 scales) and (ii) improved health-related quality of life, compared with adolescents receiving standard CATS care at 48 weeks.

Objectives
The trial will provide evidence on whether training and mentoring the existing CATS in problem-solving therapy will, compared with standard CATS care: i) reduce the proportion of adolescents with virological failure or death at 48 weeks (primary outcome); ii) reduce prevalence and severity of depression and/or anxiety at 48 weeks; iii) improve quality of life

Trial design
The trial is being conducted in ten districts in five provinces. The trial is a cluster randomized controlled trial. The clinic is the unit of randomization. Sixty clinics were randomized to either enhanced ART adherence support through the Zvandiri programme or CATS-PST in a 1:1 allocation stratified by district. Adolescents aged 11-19 years were recruited. Trial outcomes are assessed through a clinical, behavioural and psychological assessment conducted at baseline and after 48 weeks. The primary endpoint will be at 48 weeks post-enrolment, with an 8-week window period either side (ie 40-56 weeks).

Eligibility criteria
i) Inclusion criteria: HIV positive adolescents, aged 11-19 years, eligible for ART (i.e. either starting or already on ART) and able to provide informed assent and their caregiver is able to provide informed consent (those aged 18 and 19 years do not need caregiver consent).
ii) Exclusion criteria: Too physically or psychologically unwell to participate; or unable to give informed consent.

Outcomes
The primary outcome is the proportion who have died or have a viral load >1000 copies/mL at 48 weeks, assessed on a dried blood spot sample.
Secondary outcomes are assessed at 48 weeks, and are the proportion of participants i. with common mental disorders (depression and/or anxiety) ii.
with poor quality of life

Sample size and potential power of the trial
The total sample size of 840 participants recruited from 60 clusters will provide 85% power to detect a difference in virological non-suppression of 36% among participants in the standard care arm versus 24% in the intervention arm assuming 20% loss to follow-up and a coefficient of variation (k) between clusters of 0.25. For secondary outcomes, the sample size provides 87% power to detect a difference in the proportion with CMD at 12 months of 16% in the standard CATS arm and 8% in the CATS-PST arm.

Quantitative data analysis
The data will be analysed using Stata 15 or above. Statistical analyses will be conducted and reported in line with Consolidated Standards of Reporting Trials (CONSORT) guidelines (extended for cluster randomised trials [1] including a flowchart. Analysis programs ("do-files") will be prepared based on blinded data, and unblinding will not take place until after analysis and interpretation of blinded results has taken place.

Recruitment and representativeness of recruited patients and patients who completed follow up
Initial analyses will compare baseline characteristics of i) individuals who consented and did not consent, and ii) participants who did and did not complete outcome assessments at 48 weeks, compared using mixed-effects logistic regression (linear regression for continuous variables) with a random effect for clustering. CONSORT flow charts will be constructed (see appendix 1).

Baseline comparability of randomised groups
Baseline characteristics of enrolled participants will be compared between treatment arms and also reported overall, summarised using mean and standard deviation, median and interquartile range or numbers and proportions as appropriate. No significance testing will be done as any differences are due to chance if randomisation was correctly applied. For continuous outcomes, histograms will also be plotted within each arm to assess normality, and whether any transformation is required.
Among those enrolled, the following baseline characteristics will be compared by arm: -age -gender -in school -orphanhood status -HIV viral load at enrolment -aware of HIV status -ever disclosed HIV status -mean SSQ score -mean PHQ-9 score Any variables for which there is a substantial imbalance will be noted so that final analysis can take this into account.

Adherence to allocated intervention and intervention fidelity
To assess treatment fidelity in the CATS arm, the quantity/coverage of the active treatment will be described, as indicated by number of support group meetings attended as well as home visits and SMS reminders received.

Loss to follow-up and other missing data
The numbers and proportions actively withdrawing from the trial and passively lost to follow-up will be reported overall and by arm at 48 weeks (window period 40-56 weeks). The reasons for withdrawal from the trial will be summarised.

Description of counsellors
The CATS will be described in terms of age, gender, education and caseload.

Description of cohort at follow up
Selected characteristics of those seen and not seen at 48 weeks will be compared to assess generalisability of results. Attrition will be compared by arm to assess bias due to loss to follow-up, using mixed-effects logistic regression to adjust for within-clinic clustering. Variables associated with missing endline data will be adjusted for in subsequent analyses. Reasons for not attending follow up will be analysed by arm. The outcome measures will be summarised at baseline and 48 weeks by intervention arm, using proportions or means (SD) as appropriate.

Analysis of intervention effect
The primary analyses will be intention-to-treat, adjusted for baseline values of the outcome measure and for variables that are imbalanced between arms at baseline. The primary analysis will be complete case, with appropriate methods to impute missing outcome data as sensitivity analyses if necessary. Analyses will be based on individual-level measures, as there are sufficient clusters per arm to allow them to perform robustly, even with stratification by district [2].
For binary outcomes, a logistic regression random effects model will be used to estimate the odds ratio of the outcome, with a random effects term to allow for clustering by clinic. A quadrature check will be performed to evaluate the model fit. For continuous outcomes a mixed effects linear regression model will be used to estimate the mean difference between arms with a random effect to allow for clustering.
Pre-defined sensitivity analyses will include adjustment for baseline viral load and other variables imbalanced at baseline. We will conduct sensitivity analyses to investigate the effect of missing data by using multiple imputation, analysed using an individual level Poisson regression model, and a per-protocol analysis.

Analysis of primary outcome
The primary outcome is the proportion who have died or have a viral load >1000 copies/mL at 48 weeks (virological nonsuppression), assessed on a dried blood spot sample. This will be analysed using odds ratios as described above. Health Questionnaire (PHQ-9). Severity will be assessed using PHQ-9 as an ordinal score.

Analysis of secondary outcomes
Scale has been widely used in Africa, and recently validated for use in Zimbabwe among young people living with HIV (Rabia Khan, unpublished) [3,4,5] Common mental disorders Score of ≥8/14 on the Shona Symptom Questionnaire (SSQ-14); Severity of CMD will be assessed using SSQ-14 as a continuous score [6] Poor quality of life Measured using a previouslyvalidated Shona version of the EQ-5D scale, analysed as a continuous score [7,8]

Adjustment for multiple outcomes and reporting p-values
No p-value adjustment will be conducted. Interpretation of the intervention effect will be based on the strength of evidence of effect size and consistency of results for related outcomes.

Missing baseline and outcome data
The number (%) of participants with complete data will be reported. If scales have recommended methods for dealing with missing data, these will be applied. As outlined above, primary analyses will be complete case, with adjustments made for baseline variables which are associated with the outcome and/or missingness, to account for missing data. If necessary, in sensitivity analyses, we will apply appropriate methods to impute missing outcome data (see below).

Model assumption checks
For continuous outcomes, model residuals will also be plotted to check for normality and inspected for outliers. If substantial departures from normality occur, transformations will be considered. If a suitable transformation cannot be found, a non-parametric analysis will be considered.

Programme records
Programme data will be collected in intervention communities and regularly compiled to record uptake and attendance at community intervention, visits by CATS and use of clinical services.