OBJECTIVE: To compare the response to protease inhibitor (PI) and efavirenz-containing combination therapy among treatment-naive HIV-infected persons. DESIGN: Prospective observational cohort study. METHODS: Response to treatment was analysed according to the intent-to-treat principle among antiretroviral-naive patients who started either efavirenz (n = 89) or PI (n = 183) plus two nucleoside reverse transcriptase inhibitors between February 1999 and March 2000 using Kaplan-Meier and multivariable Cox proportional hazard regression methods. Primary endpoint was time to undetectable plasma viral load. Secondary endpoints included the number of CD4 cells gained, virological rebound, treatment change, and clinical progression. RESULTS: Patients on PI regimens had lower median CD4 counts (165 versus 216 x 5 106/l; P = 0.15) and were more likely to have AIDS at initiation of treatment (25% versus 15% P = 0.048) than patients starting efavirenz regimens. The probability of reaching plasma HIV-1 RNA < 400 copies/ml was higher with efavirenz- than with PI-containing regimens [adjusted hazard ratio, 1.75; 95% confidence interval (CI), 1.34-2.29]. Median times to undetectable viral load were 58 days (95% CI, 44-70 days) for efavirenz-treated and 88 days (95% CI, 79-98 days) for PI-treated patients. The median number of CD4 cells gained in the first 6 months (90 x 10(6) cells/l with efavirenz, 10(7) x 10(6) cells/l with PI; P = 0.63), time to and reasons for treatment change, time to viral rebound, drug intolerance and clinical progression rates were similar in the two treatment groups. CONCLUSIONS: Treatment with efavirenz-, compared with PI-based regimens, appeared to result in a superior virological response but no difference in immunological or clinical efficacy. The relevance of these observations remains to be determined in studies with longer follow-up.