Multivariate analysis of prognostic factors in chronic lymphocytic leukemia

Abstract

Prognostic factors in a group of 90 patients with chronic lymphocytic leukemia were studied by methods of survival analysis. The relationship between survival and a set of demographic, clinical and laboratory variables, and identification of subsets of variables that are associated with survival, was tested by multivariate analysis, which is based upon Cox proportional hazards regression models in a stepwise procedure. Six variables showed significant correlation with survival: lymph node enlargement, splenomegaly, hepatomegaly, increased percentage (greater than 80%) of lymphocytes, hyperuricemia, and anemia. Stepwise analysis showed that the number of coexistent risk factors was a better predictor of survival than any single risk marker (P less than 0.001). Median survival of patients with 0 or 1 risk marker was 120 months; with 2 or 3, 96 months; with 4, 36 months; and with 5 or 6, only 24 months. Comparison of staging by number of risk markers with staging of the same patients by the Rai system showed a significant trend of decreasing survival with increasing number of risk markers within the same Rai stage. Staging by the number of coexistent risk markers is a simple and readily available method, which may complement existing methods to provide a more accurate assessment of prognosis in patients with chronic lymphocytic leukemia.