Nomogram to predict peritoneal dissemination in gastric cancer.

Abstract

465 Background: Peritoneal metastasis is a poor prognostic factor in patients with gastric cancer and cannot be predicted, that can tailor oncological treatment. We retrospectively analyzed the clinicopathological data among 110 gastric cancer patients. Methods: Univariate logistic regression was performed with peritoneal disease as the outcome to assess potential risk factors associated with it. We found that age, gastric outlet obstruction, VEGF, Ki67, p53 and NLR ratio were significant factors associated with peritoneal disease. Using multiple logistic regression, the adjusted odds ratio and 95% confidence interval were estimated. The predictive accuracy of this regression model was estimated using ROC curve analysis and Area under the curve, sensitivity and specificity were estimated. A nomogram was plotted based on the regression coefficient for the prediction of peritoneal disease. Results: It was estimated that age less than 40 years had 0.19 (0.05, 0.79), GOO positive had 2.04 (0.69, 6.03), VEGF positive had 4.69 (1.79, 12.33), elevated Ki67 had 3.94 (1.55, 9.97), positive p53 had 0.43 (0.17, 1.11) and NLR ratio had 1.15 (0.88, 1.51) odds of having peritoneal disease from multiple logistic regression. The regression model had significant goodness of fit for the data (Hosmer-Lemeshow test P value < 0.001). On estimating the AUC from the ROC curve for the model, we found that it had a good discriminatory ability with AUC = 0.81. Sensitivity was estimated to be 71.15%, whereas specificity was estimated to be 72.41% for the model. Subsequently, a nomogram was plotted based on the regression estimated from this model. Conclusions: The nomogram could risk-stratify those at risk of developing peritoneal disease and modify therapeutic strategies like selecting patients for prophylactic intraperitoneal chemotherapy or other novel therapeutic strategies. However, the nomogram needs to be validated prospectively.