OBJECTIVE To prospectively validate the AIDS Clinical Trials Group (ACTG) staging classification for AIDS-associated Kaposi's sarcoma (KS). METHODS Two hundred ninety-four consecutive patients enrolled in eight ACTG therapeutic trials for AIDS-associated KS were staged prospectively according to tumor extent (T), severity of immunosuppression (I), and other systemic human immunodeficiency virus type 1 (HIV-1)-associated illness (S) and were observed for survival. Patients were classified as good risk (subscript 0) or poor risk (subscript 1) for each variable according to published ACTG criteria. Univariate and multivariate analyses were used to evaluate the associations between TIS variables and survival; additional analyses were conducted to improve the predictive value of the staging system. RESULTS Survival was significantly shorter for patients in the poor-risk category for each of the TIS variables. Respective median survivals for patients in the good- and poor-risk categories were 27 and 15 months for T (P < .001); 40 and 13 months for I (P < .001) when I0 included CD4 counts > or = 200/microL and 22 and 16 months for S (P = .04). Multivariate analysis indicated that severity of immunosuppression gave the most predictive information but also showed that T provided significant additional predictive information in patients whose immune function was least impaired. Refined Cox models using a CD4 count of 150/microL rather than 200/microL to distinguish I0 and I1 yielded a simplified model with better fit to the observed data. CONCLUSIONS The ACTG TIS classification predicts survival in patients with AIDS-associated KS; CD4 count and tumor stage provide the most predictive information. However, a lower CD4 count than the one originally proposed provides better discrimination between prognostic groups.