Prognostic nomograms for gastric carcinoma after D2 + total gastrectomy to assist decision-making for postoperative treatment: based on Lasso regression.

Objective: This study aimed to establish novel nomograms that could be used to predict the prognosis of gastric carcinoma patients who underwent D2 + total gastrectomy on overall survival (OS) and progression-free survival (PFS).

Methods: Lasso regression was employed to construct the nomograms. The internal validation process included bootstrapping, which was used to test the accuracy of the predictions. The calibration curve was then used to demonstrate the accuracy and consistency of the predictions. In addition, the Harrell's Concordance index (C-index) and time-dependent receiver operating characteristic (t-ROC) curves were used to evaluate the discriminative abilities of the new nomograms and to compare its performance with the 8th edition of AJCC-TNM staging. Furthermore, decision curve analysis (DCA) was performed to assess the clinical application of our model. Finally, the prognostic risk stratification of gastric cancer was conducted with X-tile software, and the nomograms were converted into a risk-stratifying prognosis model.

Results: LASSO regression analysis identified pT stage, the number of positive lymph nodes, vascular invasion, neural invasion, the maximum diameter of tumor, the Clavien-Dindo classification for complication, and Ki67 as independent risk factors for OS and pT stage, the number of positive lymph nodes, neural invasion, and the maximum diameter of tumor for PFS. The C-index of OS nomogram was 0.719 (95% CI: 0.690-0.748), which was superior to the 8th edition of AJCC-TNM staging (0.704, 95%CI: 0.623-0.783). The C-index of PFS nomogram was 0.694 (95% CI: 0.654-0.713), which was also better than that of the 8th edition of AJCC-TNM staging (0.685, 95% CI: 0.635-0.751). The calibration curves, t-ROC curves, and DCA of the two nomogram models showed that the prediction ability of the two nomogram models was outstanding. The statistical difference in the prognosis between the low- and high-risk groups further suggested that our model had an excellent risk stratification performance.

Conclusion: We reported the first risk stratification and nomogram for gastric carcinoma patients with total gastrectomy in Chinese population. Our model could potentially be used to guide treatment selections for the low- and high-risk patients to avoid delayed treatment or unnecessary overtreatment.