Excellent outcomes among HIV plus children on ART, but unacceptably high pre-ART mortality and losses to follow-up: a cohort study from Cambodia

Background: Although HIV program evaluations focusing on mortality on ART provide important evidence on treatment effectiveness, they do not asses overall HIV program performance because they exclude patients who are eligible but not started on ART for whatever reason. The objective of this study was to measure mortality that occurs both pre-ART and during ART among HIV-positive children enrolled in two HIV-programs in Cambodia. Methods: Retrospective cohort study on 1168 HIV-positive children < 15 years old registered in two HIV-programs over a four-year period. Mortality rates were calculated for both children on treatment and children not started on ART. Results: Over half (53%) of children were 5 years or above and only 69(6%) were < 18 months. Overall, 9% (105/1168) of children died since the set-up of the programs. By the end of the observation period, 66(14.5%) patients not on ART had died compared to 39(5.5%) of those under treatment, and 100(22%) who did not start ART were lost-to-follow-up compared to 13(2%) on ART. 66/105 (62.8%) of all in-program deaths occurred before starting ART, of which 56% (37/66) and 79% (52/66) occurred within 3 and 6 months of enrolment respectively. Mortality rate ratio between children not on ART and children on ART was 4.1 (95% CI: 2.7-6.2) (P < 0.001). The most common contributing cause of death in first 3 months of treatment and in first 3 months of program enrollment was tuberculosis. 41/52 (79%) children who died within 6 months of enrollment had met the ART eligibility criteria before death. Conclusion: HIV-positive children experienced a high mortality and loss-to-follow-up rates before starting ART. These program outcomes may be improved by a more timely ART initiation. Measuring overall in-program mortality as opposed to only mortality on ART is recommended in order to more accurately evaluate pediatric HIV-programs performance.