BACKGROUND: Plasma HIV-1 RNA measurements are used for initiation of antiretroviral treatments. Whether the viral-load association with prognosis is similar in women and men is unknown. METHODS: We studied 812 specimens from 650 injection-drug users (IDUs) participating in a continuous observational study of patients based in a community clinic. HIV-1 load was measured by branched-chain DNA on samples from 527 IDUs from the baseline visit, and by reverse-transcriptase PCR and quantitative microculture on samples from 285 IDUs at a follow-up visit 3 years later. FNDINGS: Women had lower median viral-load measurements than men by branched-chain DNA (3365 vs 8907 copies/mL; p=0.001), reverse-transcriptase PCR (45416 vs 93130 copies/mL; p=0.02), and quantitative microculture (5 vs 8 infectious units per million peripheral blood mononuclear cells; p=0.015). This association remained even after adjustment for CD4 cell count, race, and drug use within the previous 6 months. Time to AIDS was statistically similar for men and women in a univariate proportional-hazards model and in a model adjusting for CD4 cell count. Proportional-hazards models showed that women with the same viral load as men had a 1.6-fold higher risk of AIDS (95% CI 1.10-2.32); or, equivalently, that women with half the viral load of men had a similar time to AIDS as men. INTERPRETATION: Although a biological mechanism remains unclear, these data suggest that current recommendations for HIV-1 viral-load thresholds to initiate antiretroviral therapy should be revised downwards for women.