Development and validation of pretreatment nomogram for disease‐specific mortality in gastric cancer‐A competing risk analysis

Background In several reports, gastric cancer nomograms for predicting overall or disease‐specific survival have been described. The American Joint Committee on Cancer (AJCC) introduced the attractiveness of disease‐specific mortality (DSM) as an endpoint of risk model. This study aimed to develop the first pretreatment gastric cancer nomogram for predicting DSM that considers competing risks (CRs). Methods The prediction model was developed using data for 5231 gastric cancer patients. Fifteen prognosticators, which were registered at diagnosis, were evaluated. The nomogram for DSM was created as visualizations of the multivariable Fine and Gray regression model. An independent cohort for external validation consisted of 389 gastric cancer patients from a different institution. The performance of the model was assessed by discrimination (Harrell's concordance (C)‐index), calibration, and decision curve analysis. DSM and CRs were evaluated, paying special attention to host‐related factors such as age and Eastern Cooperative Oncology Group performance status (ECOG PS), by using Gray's univariable method. Results Fourteen prognostic factors were selected to develop the nomogram. The new nomogram for DSM exhibited good discrimination. Its C‐index of 0.887 surpassed that of the American Joint Committee on Cancer (AJCC) clinical staging (0.794). The C‐index was 0.713 (AJCC, 0.582) for the external validation cohort. The nomogram showed good performance internally and externally, in the calibration and decision curve analysis. Host‐related factors including age and ECOG PS, were strongly correlated with competing risks. Conclusions The newly developed nomogram accurately predicts DSM, which can be used for patient counseling in clinical practice. The newly developed gastric cancer nomogram for disease‐specific mortality accurately predicts treatment outcomes and thereby improve risk stratification and patient selection for routine clinical decision‐making and future clinical trials. This model will improve on the traditional categorical AJCC system in the future.