Initial highly-active antiretroviral therapy with a protease inhibitor versus a non-nucleoside reverse transcriptase inhibitor: discrepancies between direct and indirect meta-analyses.

Abstract

The optimum treatment choice between initial highly-active antiretroviral therapy (HAART) with a protease inhibitor (PI) versus a non-nucleoside reverse transcriptase inhibitor (NNRTI) is uncertain. An indirect analysis reported that PI-based HAART was better than NNRTI-based HAART. However, direct evidence for competing interventions is deemed more reliable than indirect evidence for making treatment decisions. We did a meta-analysis of head-to-head trials and compared the results with those of indirect analyses. 12 trials of at least 24 weeks' duration directly compared NNRTI-based versus PI-based HAART in HIV-infected patients with limited or no previous exposure to antiretrovirals. We also identified six trials of NNRTI-based HAART and eight trials of PI-based HAART, each versus two NRTI regimens. We analysed the outcomes of virological suppression, death or disease progression, and withdrawals due to adverse events. In the direct meta-analysis, NNRTI-based regimens were better than PI-based regimens for virological suppression (OR 1.60, 95% CI 1.31-1.96). The difference was reduced in higher-quality trials, but still favoured NNRTI-based HAART. There were no differences in death or disease progression (0.87, 0.56-1.35) or withdrawal because of adverse events (0.68, 0.43-1.08). By contrast, in indirect analyses NNRTI-based HAART was worse than PI-based HAART for virological suppression (0.26, 0.07-0.91). There were no significant differences for death or disease progression (1.28, 0.56-2.94) and withdrawals because of adverse events (1.46, 0.66-3.24). When trials of delavirdine were excluded, similar results were produced. Results from direct analyses suggested that NNRTI-based HAART was more effective than PI-based HAART for virological suppression and was similar to PI-based HAART for clinical outcomes. Indirect comparisons could be unreliable for complex and rapidly evolving interventions such as HAART.