An unexpectedly high incidence of blast transformation after splenectomy has been reported in patients with myelofibrosis with myeloid metaplasia. However, whether this was associated with spleen removal after adjustment for risk factors was not determined. We conducted a multicenter historical cohort study of patients with myelofibrosis with myeloid metaplasia diagnosed from January 1970 through January 1994. A total of 549 patients (325 men and 224 women from 22 to 92 years of age; median age, 63 years) were included in the final data set. The Cox's proportional-hazards model was used to identify factors associated with blast transformation and death. To further adjust for factors related to spleen removal assignment, a propensity score for splenectomy was estimated using recursive-partitioning analysis. Blast transformation developed in 78 patients (14.2%). Patients who underwent splenectomy developed more blast transformations than those who were not splenectomized (23 of 87 [26.4%] v 55 of 462 [11.9%]; P < .001). The cumulative incidence of blast transformation 12 years after diagnosis was 27.0% in nonsplenectomized patients and 55.0% in splenectomized ones (P = . 01). The risk factors independently predictive of blast transformation included prior splenectomy (relative risk = 2.61), platelet count less than 100 x 10(9)/L at diagnosis (relative risk = 2.45), and the presence of blasts in peripheral blood at diagnosis (relative risk = 2.31). The relative risk of blast transformation in splenectomized patients increased from 2.2 at 48 months from diagnosis to 14.3 at 12 years. Patients with the same propensity score for splenectomy showed a higher risk for blast transformation on the basis of having undergone splenectomy (P = .02). In conclusion, the risk of blast transformation is significantly increased in subjects who underwent splenectomy and appears to be independent of factors related to spleen removal assignment.