A Risk-scoring Model for Predicting Post-recurrence Survival in Patients With Endometrial Carcinoma

The survival time of patients with recurrent endometrial carcinoma is generally short. However, considerable interindividual variation exists. We developed a risk-scoring model for predicting post-recurrence survival in patients with endometrial carcinoma.

Patients with endometrial carcinoma treated at a single institution between 2007 and 2013 were identified. Pearson chi-squared analyses were used to compute odds ratios for the associations between risk factors and short survival after cancer recurrence. The results for biochemical analyses represented values at diagnosis of disease recurrence or values at initial diagnosis for those patients who had a primary refractory disease. Logistic regression models were constructed for the identification of variables that independently predict short post-recurrence survival. The models were used to assign points based on odds ratios for risk factors and risk scores were derived.

In total, 236 patients with recurrent endometrial carcinoma were included in the study. Based on overall survival analysis, 12 months was selected as the cut-off for short post-recurrence survival. Factors associated with short post-recurrence survival were platelet count, serum CA125 concentration and progression-free survival. A risk-scoring model with an area under the receiver operating characteristic curve (AUC) of 0.782 (95% confidence interval 0.713-0.851) was developed in patients without missing data (n = 182). When patients with a primary refractory disease were excluded, age and blood haemoglobin concentration were identified as additional predictors of short post-recurrence survival. For this subpopulation (n = 152), a risk-scoring model with an AUC of 0.821 (95% confidence interval 0.750-0.892) was developed.

We report a risk-scoring model that shows acceptable to excellent accuracy in predicting post-recurrence survival in patients with endometrial carcinoma, with primary refractory diseases included or excluded. This model has potential applications in precision medicine in patients with endometrial carcinoma.