A Chinese prospective cohort research developed and validated a risk prediction model for patients with cervical cancer

Abstract

Abstract Objective Cervical cancer constitutes a formidable health challenge imperiling the well-being and lives of women globally, particularly in underdeveloped nations. The survival rates among patients diagnosed with cervical cancer manifest considerable heterogeneity, shaped by a myriad of variables. Within the scope of this inquiry, a predictive model for projecting overall survival (OS) in cervical cancer patients was formulated and subsequently validated. Methods Clinicopathological and follow-up information of patients diagnosed with cervical cancer were prospectively collected from May 1, 2015, to December 12, 2019, as part of an ongoing longitudinal cohort study conducted at Chongqing University Cancer Hospital. Subsequent to the acquisition of follow-up data, the sample was randomly divided into two cohorts: a training cohort (n = 2788) and a validation cohort (n = 1194). The predictors for the model were selected through least absolute shrinkage and selection operator (LASSO) regression analysis. Cox stepwise regression analysis was then employed to identify independent predictive indicators. The study results were subsequently presented in the form of static and web-based dynamic nomograms. To elucidate the objective validation of the prognosis and anticipated survival, the concordance index (C-index) was computed. The model's discriminatory ability across various variables and its predictive performance were assessed through calibration plots. Additionally, the predictive model's capacity for outcome prediction and its net benefit were evaluated using the Net Reclassification Index (NRI) and Decision Curve Analysis (DCA) curves. Results The final model regarded the following variables from the training cohort as independent risk factors for cervical cancer patients: age, medical insurance, pathology, HPV infection status, chemotherapy, β2-microglobulin, neutrophil-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR). The C-indices of OS for the training group were 0.769 (95% CI, 0.748–0.789) and for the validation cohort were 0.779 (95% CI, 0.751–0.808). In both the training and validation cohorts, the calibration curve for estimating the chance of survival exhibited a significant agreement between prediction by nomogram and actual observation. In the training cohort, the areas under the curve (AUC) of the receiver operating characteristic (ROC) curves for 1-year, 3-year, and 5-year OS were 0.811, 0.760, and 0.782, respectively, while in the validation cohort, they were 0.818, 0.780, and 0.778, respectively. The Net Reclassification Index (NRI) and Decision Curve Analysis (DCA) provided evidence of the model's superior predictive ability and net benefit when compared to the FIGO Staging system. Conclusion The prediction methods effectively forecasted the outcomes of cervical cancer patients. Due to the model's excellent calibration and discrimination, it provided a clear and reliable approach for predicting patient survival, potentially facilitating the implementation of individualized treatment strategies.