The aim was to evaluate whether the three-month CD4 cell counts are a reliable predictor of the long-term clinical outcome of HAART-treated patients, by an observational study of 585 patients initiating HAART in a clinical setting. Clinical failure was defined as the occurrence of new or recurrent AIDS-defining events or death, and was analysed by means of intention-to-treat, univariate and multivariate analyses. An adjusted Cox regression model was used to evaluate the effect of three-month CD4+ counts on clinical outcome. Clinical failure occurred in 65 patients (11.1%) during a median follow-up of 31 months (1-65) as a result of new AIDS-defining events (ADEs) in 48 patients, ADE recurrence in six, and death in 11. The mean (median; range) CD4+ counts were 156/microL (155; 4--529) in patients with and 362/microL (326; 18--1162) in patients without clinical failure (P < .0001). Moreover, the proportion of patients with mean CD4+ counts < 200 microL was higher in those experiencing subsequent clinical failure (chi2: 41.11; P< .00001). Multivariate analysis showed that baseline CD4+ counts < 50 microL, HIV-RNA > 100,000 copies/mL and AIDS at baseline predicted failure; after adjusting for three-month CD4+ counts, this marker was the only one independently associated with clinical failure (HR 2.93; 95% Cl: 1.16--7.38). The three-month immunologic response is a reliable predictor of long-term clinical outcome.