Provider bias in the selection of non-nucleoside reverse transcriptase inhibitor and protease inhibitor-based highly active antiretroviral therapy and HIV treatment outcomes in observational studies.

Objective: To compare the characteristics of patients prescribed non-nucleoside reverse transcriptase inhibitors (NNRTI) and protease inhibitors (PI), and evaluate treatment outcomes in a setting in which nevirapine has been preferentially recommended since 1998.

Methods: A population-based analysis of antiretroviral-naive adults who started highly active antiretroviral therapy (HAART) between 1 August 1996 and 31 July 2000, and who were followed until 31 March 2002. We compared baseline characteristics, and evaluated virological responses and mortality.

Results: Overall, 439 patients (28.8%) started HAART with NNRTI (94.1% used nevirapine), 100 (6.6%) used a double PI, and 983 (64.6%) used a single PI-based regimen. Substantial differences were observed between the baseline clinical characteristics of these populations. In adjusted analyses, in comparison with single PI therapy, only the use of NNRTI was associated with more rapid HIV-RNA suppression [relative hazard (RH) 1.42; 95% confidence interval (CI) 1.22-1.65; P < 0.001]. A total of 204 deaths were identified in the study population [42 (9.6%) NNRTI; 11 (11%) double PI; 151 (15.4%) single PI, respectively]. In adjusted analysis, NNRTI (RH 1.01; 95% CI 0.71-1.45) and double PI-based HAART (RH 0.74; 95% CI 0.40-1.39) had similar mortality rates to the single PI reference category.

Conclusion: NNRTI use was associated with more rapid virological suppression, whereas similar rates of rebound and mortality were found. Nevertheless, major baseline differences existed between patients prescribed the various initial regimens. As such, it is likely that similar selection factors may explain why our findings contrast with several non-randomized studies showing worse clinical outcomes of patients prescribed nevirapine.