There is no established methodology to help select cost effective antibody panels. We used Bayesian statistics and an evidence-based pathology (EBP) approach to retrospectively review the use of immunohistochemistry (IHC) in 153 consecutive pleural effusions evaluated in our laboratory from 2005-2007 for the differential diagnosis of malignant mesothelial cells versus carcinoma cells and to estimate the likely site of origin of a carcinoma. The results in this "training" set were used to design antibody panels and test their clinical applicability on a "test set" of 44 pleural effusions collected in early 2008. Cytopathologists had used 6 +/- 4.5 IHC tests per case for the diagnosis of malignant mesothelioma (n = 9) and carcinomas of lung (n = 60), breast (n = 47), Müllerian (n = 25), and other origins in the "training set". The sensitivity and specificity of pleural cytology using all these IHC tests were 32% and 95%, respectively. Sensitivity, specificity and post-test odds (PTO) of a positive IHC result were calculated for each antibody and by the following classes: malignant mesothelial cells and carcinoma cells by primary site of origin. The antibodies that provided the best PTO to diagnose the most prevalent tumors in our population were included in diagnostic panels for male (calretinin, TTF-1, PSA and CDX-2) and female (calretinin, TTF-1, ER and CA125) patients. These panels provided 100% specificity and 77% and 50% sensitivity, respectively, for the pleural effusions from female and male patients in the "test set." The use of an EBP approach for test selection in cytopathology is discussed.
(c) 2009 Wiley-Liss, Inc.