Nomogram models for the prognosis of cervical cancer: A SEER-based study

Abstract

BackgroundCervical cancer (CC) is one of the most common cancers in women. This study aimed to investigate the clinical and non-clinical features that may affect the prognosis of patients with CC and to develop accurate prognostic models with respect to overall survival (OS) and cancer-specific survival (CSS).MethodsWe identified 11,148 patients with CC from the SEER (Surveillance, Epidemiology, and End Results) database from 2010 to 2016. Univariate and multivariate Cox regression models were used to identify potential predictors of patients’ survival outcomes (OS and CSS). We selected meaningful independent parameters and developed nomogram models for 1-, 3-, and 5-year OS and CSS via R tools. Model performance was evaluated by C-index and receiver operating characteristic curve. Furthermore, calibration curves were plotted to compare the predictions of nomograms with observed outcomes, and decision curve analysis (DCA) and clinical impact curves (CICs) were used to evaluate the clinical effectiveness of the nomograms.ResultsAll eligible patients (n=11148) were randomized at a 7:3 ratio into training (n=7803) and validation (n=3345) groups. Ten variables were identified as common independent predictors of OS and CSS: insurance status, grade, histology, chemotherapy, metastasis number, tumor size, regional nodes examined, International Federation of Obstetrics and Gynecology stage, lymph vascular space invasion (LVSI), and radiation. The C-index values for OS (0.831 and 0.824) and CSS (0.844 and 0.841) in the training cohorts and validation cohorts, respectively, indicated excellent discrimination performance of the nomograms. The internal and external calibration plots indicated excellent agreement between nomogram prediction and actual survival, and the DCA and CICs reflected favorable potential clinical effects.ConclusionsWe constructed nomograms that could predict 1-, 3-, and 5-year OS and CSS in patients with CC. These tools showed near-perfect accuracy and clinical utility; thus, they could lead to better patient counseling and personalized and tailored treatment to improve clinical prognosis.