Metastatic adenocarcinoma of an unknown primary site. A comparison of the relative contributions of morphology, minimal essential clinical data and CEA immunostaining status.

Measurement of the relative contributions of morphology alone; minimal essential clinical data; immunohistologic reactivity of a prototypic tumor marker, carcinoembryonic antigen (CEA); and the process by which a pathologist can identify the origin of a metastatic adenocarcinoma of unknown primary site is the subject of this report. To standardize the case material, we used an image digitizing and archival system to present 100 metastatic adenocarcinomas of known primary site as unknowns to two pathologists. The images were selected to show only gland-forming areas of the carcinomas and excluded all normal tissue elements. They were viewed, initially without, and then with, identification of gender and metastatic site. Subsequently, the results of immunoperoxidase staining for CEA, assessed independently by a third pathologist, were provided. Our analysis showed that, overall, the correct primary site was chosen as choice 1, 2, or 3 in 72% and 76%, and as choice 1 in 49% and 47% of cases, respectively. Accuracy was highest for prostatic, ovarian, and breast carcinomas, and lowest for upper-gastrointestinal tract, biliary tract, and pancreatic adenocarcinoma. Statistical analysis showed the largest increments in accuracy in the choice 1 prediction in each tumor category were achieved by provision of minimal essential clinical data. Knowledge of CEA status did not affect overall accuracy; however, it increased the odds of making the correct diagnosis for ovarian, colorectal, and endometrial (both pathologists) carcinomas, and for prostatic, pulmonary and esophago-gastric adenocarcinomas (one pathologist). The study exemplifies a model for the objective measurement of the contribution of ancillary studies, such as immunoperoxidase markers, to the accuracy of pathologic diagnosis.