Three hundred twenty-five previously untreated patients with chronic lymphocytic leukemia were analyzed to identify significant prognostic factors for survival. Univariate analysis identified the following characteristics associated with survival: (1) clinical characteristics: age, race, sex, performance status, lymphadenopathy, and hepatosplenomegaly; (2) hematologic parameters: WBC count, absolute lymphocyte and granulocyte counts, hemoglobin level, and platelet count; and (3) biochemical parameters: serum albumin, calcium, uric acid, lactate dehydrogenase, alkaline phosphatase, BUN, and creatinine. Multivariate regression analysis in a randomly selected training subset of 217 patients demonstrated that the combination of uric acid, alkaline phosphatase, lactate dehydrogenase, external lymphadenopathy, and age had the strongest predictive relation to survival time. The resulting model was validated in the remaining independent subset of 108 patients and led to classification of patients into low, intermediate, and high-risk groups with five-year survival rates of 75%, 59%, and 14%, respectively, and with distinctively different annual mortality rates (P less than .01). Both the regression model and Rai staging were highly effective in identifying risk groups among the entire patient population (P less than 0.001). Overall the regression model was superior to Rai staging in defining prognostic risk groups. In addition, it was able to separate patients into significantly different risk categories within each Rai stage, thus improving on the prognostic prediction of individual patients with chronic lymphocytic leukemia.