Background and study purpose:
Gallbladder cancer (GBC) is a highly aggressive malignancy, and surgery is the primary curative option. However, postoperative survival of patients with GBC remains limited. This study aimed to develop a practical model for predicting the postoperative overall survival (OS) of patients with GBC. The model aims to guide surgical decisions and benefit-risk assessments, addressing an unmet need in current practice.
Methods
A total of 287 patients from three medical institutions were analyzed. Univariate Cox regression analysis was used to screen for prognostic factors. Bidirectional stepwise multivariate Cox regression analysis was used for the feature selection. A nomogram was constructed to predict 1-, 3-, and 5-year postoperative survival rates. The predictive performance of the nomogram was assessed using Harrell's concordance index (C-index), receiver operating characteristic (ROC) curves, and calibration curves. Kaplan-Meier curves and log-rank tests were used to compare survival between the high-risk and low-risk groups determined by the nomogram. Decision curves were used to evaluate the clinical benefits of the nomograms. After training in one institution, internal and external validation were performed in the other two institutions to assess the reliability of the nomogram.
Results
Eight factors were selected via stepwise regression: TNM stage, serum carbohydrate antigen 125 (CA125), carbohydrate antigen 199 (CA199), R0 resection, body mass index (BMI), serum albumin, age-adjusted Charlson Comorbidity Index (aCCI), and serum platelet count. The C-index values were 0.770 and 0.757 before and after bootstrap resampling, respectively. The time-dependent C-index consistently exceeded 0.70 from 6 months to 5 years postoperatively, significantly outperforming TNM staging and CA199 levels. Time-dependent ROC analysis showed an area under the curve (AUC) of over 75% when predicting 1-, 3-, and 5-year postoperative survival. The calibration curves demonstrated good concordance between the predicted and observed 1-, 3-, and 5-year postoperative survival rates. The high-risk group identified by the nomogram exhibited significantly better survival than the low-risk group in both the overall population and in late-stage patients. Decision curves indicated the superior clinical benefits of the novel model compared to TNM staging and CA199 levels. The model performance in the validation process was comparable to that in the training process, demonstrating good reliability.
Conclusions
Our innovative multi-factor nomogram exhibits excellent discriminative and predictive efficacy, along with robust generalizability. The nomogram model has the potential to be a high-quality tool for forecasting postoperative survival in GBC, aiding clinical decision-making.
