BACKGROUND: Adolescents living with HIV face challenges to their wellbeing and antiretroviral therapy adherence and have poor treatment outcomes. We aimed to evaluate a peer-led differentiated service delivery intervention on HIV clinical and psychosocial outcomes among adolescents with HIV in Zimbabwe. METHODS: 16 public primary care facilities (clusters) in two rural districts in Zimbabwe (Bindura and Shamva) were randomly assigned (1:1) to provide enhanced HIV care support (the Zvandiri intervention group) or standard HIV care (the control group) to adolescents (aged 13-19 years) with HIV. Eligible clinics had at least 20 adolescents in pre-ART or ART registers and were geographically separated by at least 10 km to minimise contamination. Adolescents were eligible for inclusion if they were living with HIV, registered for HIV care at one of the trial clinics, and either starting or already on ART. Exclusion criteria were being too physically unwell to attend clinic (bedridden), psychotic, or unable to give informed assent or consent. Adolescents with HIV at all clinics received adherence support through adult counsellors. At intervention clinics, adolescents with HIV were assigned a community adolescent treatment supporter, attended a monthly support group, and received text messages, calls, home visits, and clinic-based counselling. Implementation intensity was differentiated according to each adolescent's HIV vulnerability, which was reassessed every 3 months. Caregivers were invited to a support group. The primary outcome was the proportion of adolescents who had died or had a viral load of at least 1000 copies per μL after 96 weeks. In-depth qualitative data were collected and analysed thematically. The trial is registered with Pan African Clinical Trial Registry, number PACTR201609001767322. FINDINGS: Between Aug 15, 2016, and March 31, 2017, 500 adolescents with HIV were enrolled, of whom four were excluded after group assignment owing to testing HIV negative. Of the remaining 496 adolescents, 212 were recruited at Zvandiri intervention sites and 284 at control sites. At enrolment, the median age was 15 years (IQR 14-17), 52% of adolescents were female, 81% were orphans, and 47% had a viral load of at least 1000 copies per μL. 479 (97%) had primary outcome data at endline, including 28 who died. At 96 weeks, 52 (25%) of 209 adolescents in the Zvandiri intervention group and 97 (36%) of 270 adolescents in the control group had an HIV viral load of at least 1000 copies per μL or had died (adjusted prevalence ratio 0·58, 95% CI 0·36-0·94; p=0·03). Qualitative data suggested that the multiple intervention components acted synergistically to improve the relational context in which adolescents with HIV live, supporting their improved adherence. No adverse events were judged to be related to study procedures. Severe adverse events were 28 deaths (17 in the Zvandiri intervention group, 11 in the control group) and 57 admissions to hospital (20 in the Zvandiri intervention group, 37 in the control group). INTERPRETATION: Peer-supported community-based differentiated service delivery can substantially improve HIV virological suppression in adolescents with HIV and should be scaled up to reduce their high rates of morbidity and mortality. FUNDING: Positive Action for Adolescents Program, ViiV Healthcare.