To determine the prognostic value of plasma viremia in long-term zidovudine (AZT)-treated HIV-infected patients, HIV-1 plasma viremia (PV) was quantified in 28 HIV-infected patients before and during AZT long-term treatment; the follow-up also included p24 antigenemia and CD4 cell counts. The variations of these markers during the follow-up period, the correlation with the clinical outcome (progressors versus nonprogressors), and the discrepancies between PV and surrogate markers were then analyzed. A significant and stable decrease in PV titer was observed in only nonprogressors (Friedman test, p < 0.005). At the end of follow-up, 11 (73%) of the 15 non-progressors were PV responders (patients who remained or became PV- long-term), whereas all the 13 progressors were PV nonresponders (patients who remained or became PV+). These results indicated a strong correlation between PV and clinical outcome (Fischer's exact test, p < 0.0001). The persistence, increase, or reappearance of viral replication appeared to be an important predictor of poor clinical outcome in HIV-infected patients under AZT treatment. This finding could provide a rational basis to help the clinician's decision in the clinical treatment of HIV-infected patients.