Based on clinical and histologic features, differentiating metastatic carcinomas from benign or malignant meningiomas usually is not difficult. Occasionally, however, in some patients without a clinical history of carcinoma, malignant meningiomas can morphologically simulate metastatic carcinoma, necessitating an immunohistochemical study for cytokeratin to make a correct diagnosis. However, the utility of immunohistochemical markers to separate malignant meningioma from metastatic carcinoma has not been investigated. The immunoperoxidase method with antigen retrieval was used to characterize the expression of three cytokeratins (AE1/AE3, CAM 5.2, and Pan cytokeratin), EMA, CEA, Ber-EP4, CD 15, and B72.3 in 12 previously diagnosed malignant meningiomas, 20 benign meningiomas, and 20 metastatic carcinomas. Cytokeratin expression was detected in 75% of malignant meningiomas, 0% of benign meningiomas, and 100% of metastatic carcinomas. While epithelial markers of Ber-EP4, CEA, B72.3 and CD-15 were positive in 90, 80, 70 and 65% of the metastatic carcinoma, respectively, they were negative in all 12 malignant meningioma examined. Vimentin immunoreactivity was seen in all benign and malignant meningiomas, and in 20% of metastatic carcinomas. Our results indicated that cytokeratin is not a reliable immunohistochemical marker to separate a malignant meningioma from metastatic carcinoma. A panel of epithelial markers including Ber-EP4, CEA, B72.3 and CD-15, and vimentin may be needed to separate malignant meningioma from metastatic carcinoma. Cytokeratin expression can be a potential pitfall for confusing a malignant meningioma with a metastatic carcinoma.