BACKGROUND. Pituitary carcinomas are rare adenohypophysial neoplasms, the definition, diagnosis, therapy, and prognosis of which are controversial. METHODS. Pituitary carcinomas were defined as primary adenohypophysial neoplasms with documented craniospinal and/or systemic metastases. The authors report a clinicopathologic study of 15 examples examined by light microscopy, immunohistochemistry, and image analysis. Both proliferative activity and p53 tumor suppressor gene expression were studied. RESULTS. The study group consisted of 15 patients, including 8 males and 7 females ranging in age from 34-71 years (mean, 56 years). Of these patients, seven had adrenocorticotropic hormone (ACTH)-producing tumors (four in the context of Nelson's syndrome), seven had prolactin-producing tumors, and one had a non- functioning tumor. No evidence of diabetes insipidus was seen in any case. Fourteen tumors were initially considered macroadenomas. Of the ten cases for whom tumor extent was known, all had invasive tumors. The interval from the initial diagnosis adenoma to that of carcinoma ranged from 0.3 to 18.0 years (mean, 6.6 years; median, 5.0 years); the longest mean interval (15.3 years) occurred for patients with Nelson's syndrome. The latency was twice as long for ACTH-producting tumors as for prolactin (PRL) cell tumors (9.5 vs. 4.7 years). All carcinomas showed a greater tendency toward systemic metastasis than craniospinal metastasis; the rate of systemic metastasis was 71% for PRL cell tumors and 57% for ACTH-producing tumors. Thirteen percent of tumors showed both patterns of metastasis. Fully 50% of primary tumors and the majority of metastases showed nuclear pleomorphism and/or hyperchromasia. The mean mitotic, MIB-1, and proliferating cell nuclear antigen indices for primary tumors and metastases were as follows: 2/10 high-power field (hpf), 2.6% and 11%, respectively; 6/10 hpf, 7.8% and 16%, respectively. Staining for p53 protein was noted in 57% of primary tumors and 88% of metastatic tumors; a relative increase in p53 expression in metastases was noted in 83%. All but one of the primary and metastatic tumors were aneuploid. The most common treatments were radiation therapy and, for PRL cell carcinomas, dopamine agonist administration. Both treatments provided only palliation. Eighty percent of the patients died of metastatic disease 7 days to 8 years after the diagnosis of carcinoma; of these, 66% died within 1 year. At least follow-up, 20% of patients were alive with metastases 9-18 months after diagnosis. CONCLUSIONS. Nearly all pituitary carcinomas present as functioning, microscopically atypical or mitotically active, invasive macroadenomas. By definition, after an interval related to their immunotype, all metastasize. The tumors show a greater tendency toward systemic metastasis than craniospinal metastasis and are associated with poor prognosis. Radiation and dopamine agonist therapy generally provide only palliation. Proliferation indices and p53 expression tend to be higher in metastases than in primary tumors. The current definition of pituitary carcinoma requires the demonstration of metastasis; however, high mitotic and MIB-1 labeling indices as well as p53 immunoreactivity suggest the diagnosis and appear to be of prognostic significance. A redefinition of aggressive pituitary tumors is proposed-one that facilitates the recognition of tumors prone to metastasis.