Risk prediction model of peritoneal seeding in advanced gastric cancer: A decision tool for diagnostic laparoscopy

Abstract

BackgroundSelective diagnostic laparoscopy in gastric cancer patients at high risk of peritoneal metastasis is essential for optimal treatment planning.In this study available clinicopathologic factors predictive of peritoneal seeding in advanced gastric cancer (AGC) were identified, and this information was translated into a clinically useful tool. MethodsTotally 2833 patients underwent surgery for AGC between 2003 and 2013. The study identified clinicopathologic factors associated with the risk of peritoneal seeding for constructing nomograms using a multivariate logistic regression model with backward elimination. A nomogram was constructed to generate a numerical value indicating risk. Accuracy was validated using bootstrapping and cross-validation. ResultsThe proportion of seeding positive was 12.7% in females and 9.6% in males. Of 2833 patients who underwent surgery for AGC, 300 (10.6%) were intraoperatively identified with peritoneal seeding. Multivariate analysis revealed the following factors associated with peritoneal seeding: high American Society of Anesthesiologists score, fibrinogen, Borrmann type 3 or 4 tumors, the involvement of the middle, anterior, and greater curvature, cT3 or cT4cN1 or cN2 or cN3, cM1, and the presence of ascites or peritoneal thickening or plaque or a nodule on the peritoneal wall on computed tomography. The bootstrap analysis revealed a robust concordance between mean and final parameter estimates. The area under the ROC curve for the final model was 0.856 (95% CI, 0.835–0.877), which implies good performance. ConclusionsThis nomogram provides effective risk estimates of peritoneal seeding from gastric cancer and can facilitate individualized decision-making regarding the selective use of diagnostic laparoscopy.