Prediction of lymph node and distant metastasis in patients with endometrial carcinoma: a new model based on demographics, biochemical factors, and tumor histology.

OBJECTIVE: To develop a model that might predict the probability of lymph node and distant metastasis (stages IIIC-IV) in endometrial carcinoma.
METHODS: We studied 774 patients with endometrial carcinoma treated in a single institution. Demographic factors, biochemical factors and preoperative tumor characteristics, identified as potential risk factors for advanced carcinoma in unadjusted analyses, were used to create a logistic regression model with lymph node and distant metastasis as the dependent variable. Statistically significant odds ratios in the regression model were rounded to the nearest whole number. These rounded values were the estimated weights for each factor that were summed to generate a score that might predict the probability of stage IIIC-IV carcinoma.
RESULTS: Biochemical factors and preoperative tumor characteristics predicted lymph node and distant metastasis in the regression model, whereas demographic factors were without effect. The score combining weighted risk factors was: (2 × leukocytosis)+(3 × thrombocytosis)+(7 × elevated CA125)+(4 × high-risk histology). The area under curve (AUC) for this total score was 0.823, with 71.6% sensitivity, 75.2% specificity, 25.9% positive predictive value, and 95.7% negative predictive value, using 6 as cut-point. After excluding stage IV carcinomas from the dataset, the AUC was 0.813 for the total score in predicting nodal involvement (P=0.82 vs. total score in predicting stage IIIC-IV carcinomas in the complete dataset).
CONCLUSIONS: Based on the high negative predictive value, this prediction model could be applied for identifying patients who may not benefit from lymphadenectomy for endometrial carcinoma staging.