Patterns, predictors, and consequences of initial regimen type among HIV-infected women receiving highly active...

It is important to elucidate differences among initial highly active antiretroviral therapy (HAART) regimen types in comparative studies of therapy effectiveness. We aimed to identify predictors of initiation with different HAART regimen types and the effect of initial regimen type on switching and immunologic response to therapy--controlling for those predictors--among human immunodeficiency virus (HIV)-infected women in the United States.

Participants in the Women's Interagency HIV Study underwent semiannual interview, venipuncture, and clinical examination. Those beginning with protease inhibitor-based, nonnucleoside reverse-transcriptase inhibitor (NNRTI)-based, or triple-nucleoside reverse-transcriptase inhibitor (NRTI)-based HAART during April 1996-March 2005 were eligible for analysis. Predictors of initial regimen type were assessed with polytomous logistic regression. Correlates of switching were assessed with discrete-time proportional hazards models, and immunologic response to therapy was assessed with linear regression.

Among 1555 HAART initiators, CD4(+) lymphocyte count and HIV load were significant predictors of initial regimen type during 1996-2002; only sociodemographic predictors were significant during 2002-2005. Initial regimen type was not a significant predictor of subsequent regimen switching. Compared with those whose initial treatment was protease inhibitor-based HAART, those who began with triple-NRTI-based regimens had significantly lower CD4(+) cell counts at 1 year (P=.006) and 2 years (P=.004) after initiation; NNRTI initiators had lower CD4(+) cell counts after 2 years (P=.05).

We demonstrate that predictors of initial regimen type among women in the United States have been changing over time. Protease inhibitor initiators had significantly higher CD4(+) cell counts than did NNRTI or triple-NRTI initiators up to 2 years after HAART initiation. Adjustment for biological predictors of initial regimen is important to avoid confounding in the study of treatment effectiveness.