Severe thrombocytopenia is a rare but life-threatening side effect of cytotoxic chemotherapy for which risk factors are not well known. Our objective was to delineate a risk model for chemotherapy-induced thrombocytopenia requiring platelet transfusions in cancer patients. Univariate and multivariate analysis of risk factors for chemotherapy-induced thrombocytopenia requiring platelet transfusions were performed on the cohort of the 1,051 patients (CLB 1996) treated with chemotherapy in the Department of Medicine of the Centre Léon Bérard (CLB) in 1996. In univariate analysis, performance status (PS) greater than 1, platelet count less than 150, 000/microL at day 1 (d1) before the initiation of chemotherapy, d1 lymphocyte count < or = 700/microL, d1 polymorphonuclear leukocyte count less than 1,500/microL, and the type of chemotherapy (high risk v others) were significantly associated (P < .01) with an increased risk of severe thrombocytopenia requiring platelet transfusions. Using logistic regression, d1 platelet count less than 150,000/microL (odds ratio [OR], 4.3; 95% confidence interval [CI], 1.9 to 9.6), d1 lymphocyte counts < or = 700/microL (OR, 3.37; 95% CI, 1.77 to 6.4), the type of chemotherapy (OR, 3.38; 95% CI, 1.77 to 6.4), and PS greater than 1 (OR, 2.23; 95% CI, 1.22 to 4.1) were identified as independent risk factors for platelet transfusions. The observed incidences of platelet transfusions were 45%, 13%, 7%, and 1.5% for patients with > or = 3, 2, 1, or 0 risk factors, respectively. This model was then tested in 3 groups of patients treated with chemotherapy used as validation samples: (1) the series of 340 patients treated in the CLB in the first 6 months of 1997, (2) the prospective multicentric cohort of 321 patients of the ELYPSE 1 study, and (3) the series of 149 patients with non-Hodgkin's lymphoma treated in the CLB within prospective phase III trials (1987 to 1995). In these 3 groups, the observed incidences of platelet transfusions in the above-defined risk groups did not differ significantly (P > .1) from those calculated in the model. This risk index could be useful to identify patients at high risk for chemotherapy-induced thrombocytopenia requiring platelet transfusions.