A panel of antibodies against keratins, epithelial membrane antigen (EMA), epithelial antigen (Ber-EP4), carcinoembryonic antigen (CEA), tumour-associated glycoprotein (B72.3), vimentin and LeuM1 was applied to sections of adenocarcinoma of the lung and malignant mesothelioma in a randomized design. The proportion of stained tumour cells within each section was estimated independently in five categories by three pathologists (no positive tumour cells, 1-10%, 11-33%, 34-66% and more than 67% positive tumour cells). The kappa values representing the chance corrected interobserver agreement for the different antibodies in such a five group assessment were between 0.38 and 0.72. In two group assessment the kappa values were between 0.53 and 0.94. Nosological sensitivity and nosological specificity were calculated for all antibodies, and diagnostic sensitivity and diagnostic specificity (predictive values) were calculated for the Ber-EP4, CEA, B72.3, LeuM1 and vimentin. The difference between nosological sensitivity and nosologic specificity and the clinically relevant predictive values of positive and negative tests were demonstrated. In respect of the reproducibility and the diagnostic power defined by the predictive values, we demonstrated that a panel of antibodies, including CEA, Ber-EP4 and B72.3 and, to a lesser degree, LeuM1 and vimentin is applicable for the histopathological distinction between adenocarcinoma of the lung and malignant mesotheliomas. Before introduction of new diagnostic tests, including new antibodies, the prevalence of the tested tumours should be estimated. Nosological sensitivity and nosological specificity should be converted to predictive values.