To reassess the natural history of polycythemia vera and to obtain reliable estimates of both incidence of thrombosis and survival for use in defining the sample size for therapeutic clinical trials. Retrospective cohort study of patients with polycythemia who had been followed for 20 years. 11 Italian hematology institutions. 1213 patients with polycythemia vera, which was diagnosed according to criteria established by the Polycythemia Vera Study Group and commonly used in clinical practice. All-cause mortality, venous and arterial thrombosis, and hematologic and nonhematologic neoplastic disease. Myocardial infarction and stroke were classified as major thrombotic events, and venous and peripheral arterial thrombosis were considered minor thrombotic events. The number of patients who died and the number of those who had major thrombotic events (combined end point) were used as a comprehensive measure of the benefit-risk ratio associated with the use of myelosuppressive agents. 634 fatal and nonfatal arterial and venous thromboses were recorded in 485 patients (41%); 36% of these episodes occurred during follow-up in 230 patients (19%), and 64% occurred either at presentation or before diagnosis. Thrombotic events occurred more frequently in the 2 years preceding diagnosis, suggesting a causal relation between the latent myeloproliferative disorder and the vascular event. The incidence of thrombosis during follow-up was 3.4%/y; older patients or those with a history of thrombosis had a higher risk for thrombosis. Overall mortality was 2.9/100 patients per year; thrombotic events and hematologic or nonhematologic cancers had similar effects on mortality. Patients receiving chemotherapy died three to four times more frequently than those not receiving chemotherapy. The increased risk for cancer in patients receiving myelosuppressive agents was seen approximately 6 years after diagnosis. In addition, the combined end point, computed as the sum of the hardest available events (death, nonfatal myocardial infarction, or stroke), suggests that myelosuppressive agents have an overall unfavorable effect. Cytoreduction favorably affects the incidence of thrombotic events, but aggressive treatment seems to be associated with increased risk for neoplasm. These results provide a basis for reevaluating the therapeutic strategy in patients with polycythemia vera and for estimating the size of clinical trials aimed at testing new therapeutic approaches.