Multivariable analysis of factors that predict OS. For the variable treatment, the 4 categories were mutually exclusive. The Cox proportional survival analysis only displays the best model based on the following factors that were checked: treatment (this variable should be forced into this model), age group, sex, type of cancer, risk classification at diagnosis, BMF grade, high molecular risk mutation at diagnosis, transfusion dependent at index date, stem cell transplant (yes/no), and splenectomy (yes/no). The risk categories were pooled from multiple risk classification systems. Patients with missing data on age and undifferentiated sex were not included in the analysis because of low frequency. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001.