Development and validation of an online dynamic prognostic nomogram for incidental gallbladder adenocarcinoma patients without distant metastasis after surgery: a population-based study

Abstract

Background: Gallbladder cancer is the most common malignant tumor of the biliary system, most of which is adenocarcinoma. Our study explored developing and validating a nomogram to predict overall and cancer-specific survival probabilities internally and externally for incidental gallbladder adenocarcinoma patients without distant metastasis after surgery.

Methods: Patients screened and filtered in the Surveillance, Epidemiology, and End Results (SEER) database, whose years of diagnosis between 2010 and 2015 were collected as a derivation cohort, while those between 2016 and 2019 were a temporal validation cohort. Overall survival (OS) and cancer-specific survival (CSS) were chosen as the primary and secondary endpoints of the retrospective study cohort. Potential clinical variables were selected for a Cox regression model analysis by performing both-direction stepwise selection to confirm the final variables. The performance of final nomograms was evaluated by Harrell's C statistic and Brier score, with a graphical receptor operating characteristic (ROC) curve and calibration curve.

Results: Seven variables of age, race, tumor size, histologic grade, T stage, regional lymph nodes removed, and positive regional lymph nodes were finally determined for the OS nomogram; sex had also been added to the CSS nomogram. Novel dynamic nomograms were established to predict the prognosis of incidental gallbladder adenocarcinoma patients without distant metastasis after surgery. The ROC curve demonstrated good accuracy in predicting 1-, 3-, and 5-year OS and CSS in both derivation and validation cohorts. Correspondingly, the calibration curve presented perfect reliability between the death or cancer-specific death probability and observed death or cancer-specific death proportion in both derivation and validation cohorts.

Conclusion: Our study established novel dynamic nomograms based on seven and eight clinical variables separately to predict OS and CSS of incidental gallbladder adenocarcinoma patients without distant metastasis after surgery, which might assist doctors in advising and guiding therapeutic strategies for postoperative gallbladder adenocarcinoma patients in the future.

Keywords: SEER; cancer-specific survival; dynamic nomogram; incidental gallbladder adenocarcinoma; overall survival.

Figure 1

Flowchart of patient selection in the SEER database.

Figure 2

Optimal cutoff values of age, tumor size, and regional lymph nodes removed by X-tile software analysis. (A) X-tile plot of age in the derivation cohort; (B) optimal cutoff value of age highlighted by a histogram; (C) Kaplan–Meier plot of prognosis determined by the optimal cutoff value of age; (D) X-tile plot of tumor size in the derivation cohort; (E) optimal cutoff value of tumor size highlighted by a histogram; (F) Kaplan–Meier plot of prognosis determined by the optimal cutoff value of tumor size; (G) X-tile plot of regional lymph nodes removed in the derivation cohort; (H) optimal cutoff value of regional lymph nodes removed highlighted by a histogram; (I) Kaplan–Meier plot of prognosis determined by the optimal cutoff value of regional lymph nodes removed.

Figure 3

Forest plot for the hazard ratio of selected nomogram variables to predict OS and CSS. (A) Selected nomogram variables to predict OS; (B) selected nomogram variables to predict CSS.

Figure 4

Nomogram to predict OS and CSS probability. (A) Prognostic nomogram of OS; (B) prognostic nomogram of CSS. 1-, 3-, and 5-year baseline OS probability: 0.9906716, 0.986361, 0.9809064; 1-, 3-, and 5-year baseline CSS probability: 0.997877, 0.9966245, 0.9946404.

Figure 5

Web dynamic nomogram calculator with a clinical example. (A) Web dynamic nomogram calculator of OS with a clinical example; (B) web dynamic nomogram calculator of CSS with a clinical example.

Figure 6

ROC curves of the nomogram model. (A) ROC curve of OS in the derivation cohort; (B) ROC curve of OS in the validation cohort; (C) ROC curve of CSS in the derivation cohort; (D) ROC curve of CSS in the validation cohort.

Figure 7

Calibration curves of the nomogram model. (A) Calibration curve of OS in the derivation cohort; (B) calibration curve of OS in the validation cohort; (C) calibration curve of CSS in the derivation cohort; (D) calibration curve of CSS in the validation cohort.