Risk-scoring models for individualized prediction of overall survival in low-grade and high-grade endometrial cancer

•Endometrial cancer consists of two disparate patient populations.•Risk scoring models were developed to predict overall survival in endometrial cancer.•Enhanced discrimination of risk scoring models enables personalized counseling and treatment.

ObjectiveOverall survival (OS) in endometrial cancer (EC) is dependent on patient-, disease-, and treatment-specific risk factors. Comprehensive risk-scoring models were developed to estimate OS in low-grade and high-grade EC.MethodsPatients undergoing primary surgery for EC from 1999 through 2008 were stratified histologically according to the International Federation of Gynecology and Obstetrics (FIGO) as either (i) low grade: grades 1 and 2 endometrioid EC or (ii) high grade: grade 3, including non-endometrioid EC. Associations between patient-, pathological-, and treatment-specific risk factors and OS starting on postoperative day 30 were assessed using multivariable Cox regression models. Factors independently associated with OS were used to construct nomograms and risk-scoring models.ResultsEligible patients (N = 1281) included 925 low-grade and 356 high-grade patients; estimated 5-year OSs were 87.0% and 51.5%, respectively. Among patients alive at last follow-up, median follow-up was 5.0 (low grade) and 4.6 years (high grade), respectively. In low-grade patients, independent factors predictive of compromised OS included age, cardiovascular disease, pulmonary dysfunction, stage, tumor diameter, pelvic lymph node status, and grade 2 or higher 30-day postoperative complications. Among high-grade patients, age, American Society of Anesthesiologists score, stage, lymphovascular space invasion, adjuvant therapy, para-aortic nodal status, and cervical stromal invasion were independent predictors of compromised OS. The two risk-scoring models/nomograms had excellent calibration and discrimination (unbiased c-indices = 0.803 and 0.759).ConclusionPatients with low-grade and high-grade EC can be counseled regarding their predicted OS using the proposed risk-scoring models. This may facilitate institution of personalized treatment algorithms, surveillance strategies, and lifestyle interventions.