Key points are not available for this paper at this time.
Adenocarcinoma that metastasizes from an unknown primary site is a significant oncologic problem. With the exception of prostate-specific antigen and thyroglobulin, no single immunohistochemical marker is entirely site-specific. A retrospective study was undertaken to determine whether a panel of markers could accurately predict the site of origin of common metastatic adenocarcinomas. On the basis of reports of their relatively restricted specificity for carcinomas of colon, breast, lung, ovary, and upper gastrointestinal tract (stomach, pancreas, and bile duct), eight markers were selected for simultaneous evaluation: gross cystic disease fluid protein-15, breast cancer antigen 225 (BCA225), B72.3, DF3 (CA15-3), carcinoembryonic antigen (CEA), CA19-9, CA125, and estrogen receptor. The study population consisted of 128 metastatic nonmucinous adenocarcinomas for which the primary site was known. Staining was performed on formalin-fixed, paraffin-embedded tissue using an enhanced-sensitivity avidin-biotin peroxidase complex detection system. The most informative markers were CEA, CA19-9, CA125, and BCA225. With this four-marker panel, the most predictive multiple-marker phenotypes, as determined by a combination of area under the receiver operating characteristic curve, specificity, and percent correct predictions, were CEA+, BCA225-, and CA125- for colon tumors; BCA225+, CEA-, and CA125- for breast tumors; BCA225+, CEA+, and CA19-9- for lung tumors; CA125+ and CEA- for ovarian tumors; and CEA+, CA19-9+, and CA125+ for upper gastrointestinal tract tumors. Overall, these phenotypes correctly predicted the known primary site in 66% of cases. Until single highly sensitive and specific markers are developed for adenocarcinomas other than prostate and thyroid tumors, the origin of a metastatic adenocarcinoma can best be suggested or excluded with clinicopathologic data combined with a panel of selected immunohistochemical markers.
Brown et al. (Wed,) studied this question.