BACKGROUND: The prognostic significance of a response to highly active antiretroviral therapy (HAART) that is immunologically and virologically discordant is not well understood. METHODS: Four hundred four antiretroviral-naive patients initiating HAART at an urban HIV outpatient clinic in 1995 to 2004 were analyzed. The association of treatment responses at 3 to 9 months after HAART initiation with time to development of an opportunistic infection (OI) or death was determined using Cox proportional hazards modeling. Logistic regression modeling was used to examine the association between discordant responses and patient characteristics. RESULTS: Of 404 patients, 70.5% experienced favorable concordant responses (CD4 cell count [CD4]+/viral load [VL]+: increase in CD4 count of >or=50 cells/microL and achievement of undetectable plasma HIV RNA level), 15.8% an immunologic response only (CD4+/VL(-)), 8.7% a virologic response only (CD4(-)/VL+), and 5.0% a concordant unfavorable response (CD4(-)/VL(-)). Both types of discordant responses (CD4+/VL(-) and CD4(-)/VL+), nonresponse (CD4(-)/VL(-)), and baseline CD4 cell count were significantly associated with earlier development of an OI or death (relative hazard [RH] = 2.81, 95% confidence interval [CI]: 1.31 to 3.97; RH = 4.83, 95% CI: 2.10 to 11.12; and RH = 0.93, 95% CI: 0.88 to 0.99, respectively). CD4+/VL(-) and CD4(-)/VL(-) were associated with nonwhite race in multivariate logistic regression models (adjusted OR = 2.83, 95% CI: 1.46 to 5.47 and adjusted OR = 6.50, 95% CI: 1.65 to 25.69, respectively). CONCLUSION: Discordant immunologic and virologic responses at 3 to 9 months after HAART initiation play important roles in predicting long-term clinical outcomes in treatment-naive patients.