Two regression models and a scoring system for predicting survival and planning treatment in myelodysplastic syndromes: a multivariate analysis of prognostic factors in 370 patients

Abstract

Therapy planning in patients with myelodysplastic syndromes (MDSs) is complicated by its high prognostic heterogeneity. Forty-one patient and disease characteristics at onset of 370 patients with MDS were analyzed to identify significant prognostic factors for survival and transformation to acute myeloblastic leukemia (AML), and to develop and validate a regression model for predicting survival. Multivariate regression analysis showed that the total bone marrow percentage of blast cells, age, platelet count, WBC count, and hemoglobin level were the characteristics more significantly associated with survival in the overall series. The bone marrow percentage of type I blast cells was the most important factor predicting transformation into AML. Proportional hazards regression analysis in a randomly selected training sample of 240 patients demonstrated that the combination of total bone marrow percentage of blast cells, platelet count, and age had the strongest predictive relation to survival length. The resulting regression models, continuous and categorized, were validated in the remaining test sample of 130 patients by demonstrating its capability of segregating patients into low-, intermediate-, and high-risk groups, with distinctively different survival curves (P less than .0001). A scoring system derived from the categorized model also had a great prognostic value (P less than .0001). These regression models and the simpler scoring system may be accurately used for decision-making regarding therapy in MDS patients.