Objective: Endometrial carcinoma is the most common gynecologic cancer. Although the prognosis for endometrial cancer is generally good, cancers identified at late stages are associated with high levels of morbidity and mortality. Therefore, prevention and early detection may further reduce the burden of this challenging disease.
Methods: A panel of 64 serum biomarkers was analyzed in sera of patients with stages I-III endometrial cancer and age-matched healthy women, utilizing a multiplex xMAP bead-based immunoassay. For multivariate analysis, four different statistical classification methods were used: logistic regression (LR), separating hyperplane (SHP), k nearest neighbors (KNN), and classification tree (CART). For each of these classifiers, a diagnostic model was created based on the cross-validation set consisting of sera from 115 patients with endometrial cancer and 135 healthy women.
Results: Our data have demonstrated that patients with endometrial cancer have significantly different expression patterns of several serum biomarkers as compared to healthy controls. Prolactin was the strongest discriminative biomarker for endometrial cancer providing 98.3% sensitivity and 98.0% specificity alone. Our results have revealed that serum concentration of cancer antigens, including CA 125, CA 15-3, and CEA are higher in patients with Stage III endometrial cancer as compared to those with Stage I. In addition, we have shown that the expression of CA 125, AFP, and ACTH is elevated in women with tumor grade 3 vs. grade 1. Furthermore, five-biomarker panel (prolactin, GH, Eotaxin, E-selectin, and TSH) identified in this study was able to discriminate endometrial cancer from ovarian and breast cancers with high sensitivity and specificity.
Conclusions: The ability of prolactin to accurately discriminate between cancer and control groups indicates that this biomarker could potentially be used for development of blood-based test for the early detection of endometrial cancer in high-risk populations. Combining the information on multiple serum markers using flexible statistical methods allows for achieving high cancer selectivity.