To analyse plasma HIV-1 RNA levels as a marker of clinical stability and survival in a cohort of HIV-infected patients whose time of seroconversion is unknown. Retrospective cohort study. Retrovirology laboratory and AIDS Unit in a teaching hospital. A total of 916 samples from 302 patients, most on antiretroviral therapy, were analysed. Mean initial CD4 cell counts and HIV-1 RNA were 299 x 10(6)/l (range: 0-1600) and 134,261 copies/ml (range: < 200-4,300,000), respectively. Sixty-six cases had been diagnosed previously with AIDS. Analysis of progression to AIDS and survival, according to initial and longitudinal viral load (VL) and CD4 cell count measurements was performed by Kaplan-Meier test. Relative risks were calculated by Cox's proportional hazards model. During a mean follow-up of 444 +/- 309 days, 29 patients developed AIDS and 21 died. Relative risk (RR) of progression related to the group with VL < 35,000 was: 10.4 when CD4 > or = 250 x 10(6)/l and VL > or = 35,000 (P = 0.001); and 45.3 when CD4 < 250 x 10(6)/l and VL > or = 35,000 (P < 0.0001). Cumulative probability of progression was: 0%, 0% and 12.3%, at the first, second and third year respectively, for patients with all their sequential VL determinations < 60,000; and 13.3%, 34.7% and 79.3% for patients who did not maintain VL values always < 60,000 (RR = 23; P < 0.0001). The minimum value of VL that reached statistical significance for the survival analysis was 100,000 copies/ml (P < 0.0001). VL > or = or < 35,000 is a better discriminant for progression than a CD4 cell count > or = or < 250 x 10(6)/l. Sequential VL determinations < 60,000 are associated with a better prognosis.