OBJECTIVE The diagnostic criteria of the Polycythemia Vera Study Group (PVSG), although generally acknowledged as the gold standard for establishing a diagnosis of polycythemia vera (PV), do not consider bone marrow features. It may, therefore, be speculated that initial-early stages of PV are overlooked. In this retrospective study we tried to investigate whether bone marrow morphology of patients with an only borderline to slight increase in hemoglobin/hematocrit not conforming with the postulates of the PVSG enabled a clear-cut differentiation between PV and secondary (reactive) polycythemias (SP). METHODS From a series of 348 patients with a borderline to pronounced erythrocytosis and representative pre-treatment bone marrow trephine biopsies a cohort of 86 cases was selected showing only a borderline increase in hemoglobin (males < 18.5 g/dL, females < 16.5 g/dL). Biopsies and clinical records were evaluated independently and following histologic and clinical work-up a straightforward consensus was reached. The diagnostic impact of histologic findings was tested by means of discriminate analysis of 20 standardized morphologic features based on histochemical and immunohistochemical staining techniques. RESULTS Bone marrow histopathology in 47 patients diagnosed as having SP was characterized by a minimal to slight increase in cellularity with predominance of the erythroid lineage. Neutrophil granulocytopoiesis was prominent and left-shifted and small to medium-sized megakaryocytes without maturation defects were scattered throughout the bone marrow. There was an increased number of eosinophils, marked perivascular plasmacytosis, histiocytic reticular cells with accumulated cell debris and many iron-laden macrophages. Contrasting this appearance in SP our 39 patients with initial-early stage PV revealed a hypercellular bone marrow with trilineage proliferation (pan-myelosis) showing confluent sheets of erythropoiesis and loose clusters of megakaryocytes. Megakaryocytopoiesis was characterized by a pleomorphous appearance, i.e. giant cells were lying adjacent to small ones, but lacked an obvious cytologic abnormality. There was usually no prominent inflammatory reaction of the interstitial compartment. In ten patients lymphoid nodules were found, but no conspicuous iron deposits and in six patients a borderline to minimal increase in reticulin fibers was present. Following stepwise discriminate analysis of histologic features a set of parameters emerged including increase in megakaryocyte size, perivascular plasma cells, overall bone marrow cellularity and cellular debris. This pattern exerted a significant impact on separation (Wilks' lambda statistics = 0.110, p < 0.0001) of early stage PV from SP. Most patients with SP had an underlying bronchopulmonary condition, frequently associated with heavy smoking or rarely renal pathology. In addition to the histopathologic features, splenomegaly, thrombocyte count, LDH, LAP and erythropoietin levels proved to be different in the two groups of patients. CONCLUSIONS Initial-early PV is characterized by a specific pattern of bone marrow histopathology. Clinical features of distinctive impact include splenomegaly, thrombocyte count, LDH, LAP and in particular erythropoietin level. Taking these clinical and histologic findings into consideration, reactive-secondary causes of polycythemia (SP) are clearly distinguishable from autonomous ones (PV).