Immunocytochemistry in the differential diagnosis of effusions: use of logistic regression to select a panel of antibodies to distinguish adenocarcinomas from mesothelial proliferations.

Many studies have shown immunohistochemistry to be beneficial in discriminating between adenocarcinoma and mesothelial reactions in effusions. Although many of these studies suggest using a panel of antibodies, none of them used statistical methods to optimize their choice of assays. In the current study, stepwise logistic regression was applied to our data to select an appropriate panel of antibodies to differentiate between adenocarcinoma and all types of mesothelial proliferations. One hundred effusions (64 cases of adenocarcinoma, 27 cases of benign mesothelial proliferations, and 9 cases of malignant mesothelial proliferations) were analyzed for their reactivity with anti-EMA, anti-MFG, anti-CEA, Leu-M1, B72.3, and the newly described epithelial membrane marker BER-EP4. An abbreviated panel consisting of anti-CEA, EMA, and B72.3 was shown to be sufficient in over 95% of our cases to accurately characterize a given effusion. When all three assays are negative, a diagnosis of adenocarcinoma is extremely unlikely, while when two or three of the assays are positive a diagnosis of adenocarcinoma is almost certain. The use of stepwise logistic regression has proven useful in the design of antibody panels as an adjunct to the differential diagnosis of effusions and may be applicable to the selection of panels in other diagnostic problems.