Background:  There are no long-term data concerning the safety and efficacy of protease inhibitor simplification with nevirapine (NVP), efavirenz (EFV), or abacavir (ABC).

Methods:  We randomized 460 patients taking 2 nucleosides plus at least 1 protease inhibitor with plasma HIV-1 RNA <200 copies/mL for ≥6 months to switch the protease inhibitor to NVP (n = 155), EFV (n = 156), or ABC (n = 149); and were followed them for at least 3 years regardless of discontinuation of assigned therapy. The primary end-point was death, progression to AIDS, or virological failure. Secondary end-points were side effects, plasma lipids and body fat abnormalities.

Results: At 3 years, 17 (11%) patients in the NVP, 23 (15%) in the EFV, and 34 (23%) in the ABC arm reached protocol-defined end-points according to an intent-to-treat analysis (generalized Log-rank test, p = 0.020). Patients who had received prior suboptimal ART had a higher risk of reaching end-points than those who had not (HR 2.794, 95%CI 1.673 to 4.668, p <0.001). There were no differences in the rates of virologic failure in patients without prior suboptimal therapy:  NVP, 4%; EFV, 2%; and ABC, 6% (p = 0.384). The incidence of adverse events leading to study discontinuation was significantly lower with ABC (n = 13, 9%) than with NVP (n = 29, 19%) or EFV (n = 39, 25%) (p = 0.005). Median fasting plasma cholesterol and the proportion of patients with plasma cholesterol above 240 mg/dL were significantly lower in the ABC than in the NVP and EFV arms. The distribution of patients with new clinically evident lipoatrophy during follow-up was not different among arms (NVP, n = 38, 34%; EFV, n = 39, 35%; ABC, n = 35, 31%, p = 0.684).

Conclusions:  There was a higher probability of maintaining the suppression of viral replication after 3 years of follow-up when NVP or EFV were substituted for protease inhibitors as compared with ABC. EFV had a higher incidence of adverse effects leading to discontinuation and ABC caused a greater decrease in plasma cholesterol.