Nomogram for Predicting Postoperative Cancer-specific Early Death in Patients with Epithelial Ovarian Cancer based on the SEER Database: A Large Cohort Study

Abstract

Background: Ovarian cancer is a common gynecological malignant tumor. Poor prognosis is strongly associated with early death, but there is no effective tool to predict this. This study aimed to construct a nomogram for predicting cancer-specific early death in ovarian cancer patients.Methods: Our study used data from the Surveillance, Epidemiology, and End Results (SEER) database of ovarian cancer patients registered from 1988 to 2016. Important independent prognostic factors were determined by univariate and multivariate logistic regression and LASSO Cox regression. Several risk factors were considered in constructing the nomogram. Nomogram discrimination and calibration were evaluated using C-index, internal validation, and receiver operating characteristic (ROC) curves.Results: A total of 4769 patients were included. Patients were assigned to the training set (n = 3340; 70%) and validation set (n = 1429; 30%). Based on the training set, eight variables were shown to be significant factors for early death and were incorporated in the nomogram: AJCC (American Joint Committee on Cancer) stage, residual lesion size, chemotherapy, serum CA125 level, tumor size, number of lymph nodes examined, surgery of primary site, and age. The C-indices and ROC curves showed that the nomogram had better predictive ability than the AJCC staging system and good clinical practicability. Internal validation based on validation set showed good consistency between predicted and observed values for early death. Conclusions: Compared with predictions made using AJCC stage or residual lesion size, the nomogram was able to provide more robust predictions for early death in ovarian cancer patients.