Duration from seroconversion to eligibility for antiretroviral therapy and from ART eligibility to death in adult HIV-infected patients from low and middle-income countries: collaborative analysis of prospective studies

Background: Estimation of the number of people in need of antiretroviral therapy (ART) in resource-limited settings requires information on the time from seroconversion to ART eligibility and from ART eligibility to death. Objectives: To estimate duration from seroconversion to different ART eligibility criteria and from ART eligibility to death in HIV-infected adults in low-income and middle-income countries. Methods: Participants with documented seroconversion from five cohorts (two cohorts from Uganda, two from Thailand and one from Cote d'Ivoire) were analysed. We used Weibull survival models and Bayesian simulation methods to model true (unobserved) first time of treatment eligibility. We set a consistency constraint so that the mean duration from seroconversion to death was equal to the mean from seroconversion to ART eligibility plus the mean from eligibility to death. Results: We analysed data from 2072 participants, 16 157 person-years of follow-up and 794 deaths. For the criterion CD4 T-lymphocyte count < 200 cells x10(6)/l, the median duration from seroconversion to ART eligibility was 6.1 years (95% credibility interval 3.3-10.4) for all studies and 7.6 years (95% credibility interval 3.4-15.2) for all but the Thai cohorts. Corresponding estimates for the time from CD4 T-lymphocyte count < 200 cells x10(6)/l to death were 2.1 years (0.7-4.8) and 2.7 years (0.8-8.4). When including all cohorts, the mean time from serconversion to CD4 T-lymphocyte count, < 200 cells x10(6)/l and from CD4 T-lymphocyte count < 200 cells x10(6)/l to death represented 66% (38-87%) and 34% (13-62%), respectively of the total survival time. Conclusions: The duration of different ART eligibility criteria to death was longer than the estimates used in previous calculations of the number of people needing ART. However, uncertainty in estimates was considerable and heterogeneity across cohorts important.