Design of a bio-mathematical prediction model using serum tumor markers and immunohistochemistry in peritoneal carcinomatosis with ovarian involvement: a pilot study

Abstract

The diagnosis of disseminated intra-abdominal malignancy in women with ovarian involvement can be problematic. Whilst both blood tumor markers and use of immunohistochemical staining on tissue can help decide the origin of the tumor, this is done separately. This study looked at the blood and tissue marker profiles of 198 cases of disseminated malignancy to construct a model, which may help to determine tumor origin. The original histology material from 198 cases of disseminated intra-abdominal epithelial malignancy were reviewed, blind, and reassessed as to the likely site of origin. These cases had immunohistochemical (IHC) staining for cytokeratins (CK) 7 and 20, carcinoembryonic antigen (CEA) and CA125. Blood values for CEA and CA125 were also known at diagnosis. The histologic types of the tumors in this pilot study were of ovarian type morphologically in 130 cases (65.7%), nonovarian in 32 (16.1%), and not assigned in 36 cases (18.2%). The majority of the nonovarian cases were of mucinous type or too poorly differentiated to classify. Analysis showed an overall sensitivity and specificity of 93% and 69%, and positive predictive and negative predictive value of 92% and 71%, respectively, for a diagnosis of ovarian vs nonovarian origin using histology alone vs histology and IHC. Use of an ordinal regression developed a model which uses tissue staining for CK 7 and CEA along with blood levels of CEA to help determine the site of tumor origin