Very late initiation of HAART impairs treatment response at 48 and 96 weeks: results from a meta-analysis of randomized clinical trials

Initiation of highly active antiretroviral therapy (HAART) with low CD4 lymphocyte counts is associated with AIDS-related and non-AIDS-related events and increased mortality. However, no clear association has been found with an increased rate of treatment failure. We conducted a meta-analysis including randomized clinical trials of currently recommended HAART in naive patients to evaluate treatment response in very late starters (VLSs). Studies with information on response in at least one of the two strata (≤ 50 versus >50 CD4 cells/mm(3) and/or ≤ 200 versus >200 CD4 cells/mm(3)) and follow-up of at least 48 weeks were analysed. A pooled odds ratio of the effect of starting HAART with ≤ 50 versus >50 or ≤ 200 versus >200 CD4 cells/mm(3) for each arm by fitting a random-effect logistic regression model was computed. Sources of heterogeneity [sex, age, year of study initiation, nucleos(-t)ide pair and third drug] were investigated. We included 25 treatment arms from 13 randomized clinical trials. Being a VLS consistently impairs treatment outcomes at 48 and 96 weeks. Only hepatitis C virus (HCV)/hepatitis B virus (HBV) coinfection was associated with a reduced impact of late initiation of HAART; at 48 weeks for 50 and 200 cells/mm(3) thresholds (P = 0.013 and P = 0.032, respectively). None of the remaining sources of heterogeneity explored was significantly associated with the impact of being a VLS. We found that initiation of antiretroviral therapy with very low CD4 lymphocyte counts is consistently associated with poorer outcomes of HAART. This effect could be modulated by HBV/HCV coinfection, but not by the individual components of the HAART regimen.