Estimating the effect of antiretroviral treatment during HIV seroconversion: impact of confounding in observational data.

摘要

OBJECTIVE: To assess whether treatment with antiretroviral drugs within the first 3 months of infection with HIV affects medium-term health outcomes. DESIGN AND METHODS: Data from 20 cohorts in Europe and Australia were used Concerted Action on SeroConversion to AIDS and Death in Europe (CASCADE). Analysis was restricted to persons seroconverting in 1988-1998 who started antiretroviral treatment in the first 3 months or 1-2 years from seroconversion. The relationship between times to low CD4 count, AIDS and death and time of initiation of treatment was estimated using proportional hazards models. RESULTS: Seroconversion illness was more common in those who began antiretroviral treatment in the first 3 months (73%) than in those who started treatment within 1-2 years post-seroconversion (33%). Subjects receiving early antiretroviral treatment had times to AIDS and to CD4 counts <200 cells/microL that were intermediate between those of subjects starting treatment within 1-2 years and those of the subset of these subjects starting treatment within 1-2 years who also had a prior CD4 count of >350 cells/microL and no prior AIDS diagnosis. CONCLUSIONS: On the basis of these analyses, the effect of antiretroviral treatment initiation during HIV seroconversion is uncertain. It may result in lower rates of progression compared with starting antiretroviral treatment at 1-2 years, but the early antiretroviral treatment group had a similar or even higher incidence of low CD4 counts and AIDS events than the group who started antiretroviral treatment within 1-2 years with CD4 counts over 350 cells/microL and no prior AIDS diagnosis. Estimates of the effect of early treatment are probably confounded with a number of factors, including, in particular, reasons for treatment initiation.