Lung function was measured at 3-month intervals for up to 1 yr in a group of Caucasian HIV-seropositive subjects. The objective was to document any deterioration in lung function and seek correlations between such deterioration and smoking history and Centers for Disease Control (CDC) status. Ninety-nine subjects were studied at enrollment; 43 were followed-up (mean duration 9 +/- 3 months). Ninety-five of the 99 enrolled subjects remained free of HIV-related respiratory disease and were included in the analysis. At enrollment, carbon monoxide diffusing capacity (TLCO) was significantly lower than predicted in non-smokers, smokers and ex-smokers (88, 77 and 88%, respectively, P < 0.001). The TLCO measurements in the smoking group were significantly lower than those of the life-long non-smoking subjects (P < 0.01). Residual volume (RV) was significantly higher than predicted in smokers (111%, P = 0.02). During follow-up, all three groups demonstrated significant declines in TLCO (7%, P = 0.01; 9%, P = 0.005; 13%, P < 0.001, respectively), and increases in RV (9%, P = 0.03; 13.5%, P = 0.02, 22%, P = 0.02, respectively). At enrollment, significantly lower than predicted values of TLCO were observed in groups stratified by CDC criteria: in asymptomatic HIV-seropositive subjects (CDC 11) 89%, P = 0.01; persistent generalized lymphadenopathy (PGL) 84%; AIDS-related complex (ARC) 81%; and in non-pulmonary AIDS (IV C1) 69%, P = 0.0001, respectively. Residual volume was significantly higher than predicted in CDC II (114%, P = 0.05). During follow-up, TLCO fell in groups PGL and ARC by 7 and 9%, respectively, while RV increased in groups CDC II, PGL and ARC by 17, 15 and 8%, respectively. Only the TLCO decline in PGL showed any linkage to clinical deterioration. This study demonstrates deficits at enrollment, and a continuing decline of TLCO and increase in RV in HIV-seropositive subjects without overt lung disease.