Abstract
Objective: To identify the predictors of distant metastasis in patients with cervical cancer treated with definitive radiotherapy and develop a model for predicting distant metastasis. Methods: We reviewed the clinical records of patients with cervical cancer treated with definitive radiotherapy (IMRT) at Peking Union Medical College Hospital between January 2011 and December 2015. Eligible patients were randomly assigned into model development cohort and validation cohort in a 2:1 ratio. Distant metastasis rate (DMR) was calculated with Kaplan-Meier method. Univariate and multivariate analyses using cox proportional hazard model was performed to identify the risk factors of distant relapse. Based on the identified risk factors for distant metastasis, a model for predicting distant metastasis was developed and validated. A two-side P<0.05 was defined as statistically significant. Results: A total of 1193 patients were eligible for this analysis including 797 patients in the model development cohort and 396 patients in the validation cohort. The median follow-up durations of the model development cohort and the validation cohort were 28.7 months (range: 2.5-83.9 months) and 30.9 months (1.9-83.5 months). The 2-year distant metastasis rates (DMR) for patients in the model development cohort and validation cohort were 13.3% and 12.8%. Non-squamous cell carcinoma (non-Scc), common iliac lymph nodes metastasis (LNM) and bilateral pelvic LNM (PLNM) were identified as risk factors for distant metastasis. In the model development cohort, significant difference between high-risk group (with 2-3 risk factors) and low-risk group (with 0-1 risk factor) regarding DMR was observed (39.3% vs 19.3%, P<0.001). Similar conclusions were observed in the validation cohort (high-risk group vs low-risk group, 47.6% vs 10.9%, P<0.001) Conclusion: We successfully developed a model for predicting distant metastasis in patients with cervical cancer receiving definitive radiotherapy based on the three identified risk factors for distant metastasis. This model would help us distinguish patients with high risk of distant relapse from others.