BackgroundReducing attrition in paediatric HIV-positive patients using combined antiretroviral therapy (cART) programmes in sub-Saharan Africa is a challenge. This study explored the rates and predictors of attrition in children started on cART in Asmara, Eritrea.MethodsThis was a retrospective cohort study using data from all paediatric patients on cART between 2005 and 2020, conducted at the Orotta National Referral and Teaching Hospital. Kaplan-Meier estimates of the likelihood of attrition and multivariate Cox proportional hazards models were used to assess the factors associated with attrition. All p values were two sided and p<0.05 was considered statistically significant.ResultsThe study enrolled 710 participants with 374 boys (52.7%) and 336 girls (47.3%). After 5364 person-years’ (PY) follow-up, attrition occurred in 172 (24.2%) patients: 65 (9.2%) died and 107 (15.1%) were lost to follow-up (LTFU). The crude incidence rate of attrition was 3.2 events/100 PY, mortality rate was 2.7/100 PY and LTFU was 1.2/100 PY. The independent predictors of attrition included male sex (adjusted HR (AHR)=1.6, 95% CI: 1 to 2.4), residence outside Zoba Maekel (AHR=1.5, 95% CI: 1 to 2.3), later enrolment years (2010–2015: AHR=3.2, 95% CI: 1.9 to 5.3; >2015: AHR=6.1, 95% CI: 3 to 12.2), WHO body mass index-for-age z-score <−2 (AHR=1.4, 95% CI: 0.9 to 2.1), advanced HIV disease (WHO III or IV) at enrolment (AHR=2.2, 95% CI: 1.2 to 3.9), and initiation of zidovudine+lamivudine or other cART backbones (unadjusted HR (UHR)=2, 95% CI: 1.2 to 3.2). In contrast, a reduced likelihood of attrition was observed in children with a record of cART changes (UHR=0.2, 95% CI: 0.15 to 0.4).ConclusionA low incidence of attrition was observed in this study. However, the high mortality rate in the first 24 months of treatment and late presentation are concerning. Therefore, data-driven interventions for improving programme quality and outcomes should be prioritised.