LAUSR

Purpose To develop and validate a quantitative model based on gross tumor volume (GTV) of gastric adenocarcinoma (GA) corresponding to N-stage measured at multidetector computed tomography (CT) for preoperative determination of resectability Materials and methods 493 consecutive patients with confirmed GA undergoing contrast-enhanced CT two weeks before treatments were randomly enrolled into the training cohort (TC, n = 271), internal validation cohort (IVC, n = 107) and external validation cohort (EVC, n = 115). GTV was measured on CT by multiplying sums of all tumor areas by section thickness. In TC, univariate and multivariate analyses were performed to select factors associated with resectability. Receiver operating characteristic (ROC) analysis was to determine if N-stage based GTV could identify resectability. In IVC and EVC, unweighted Cohen’s Kappa tests were to evaluate performances of the ROC models. Results According to univariate analysis, age, cT stage, cN stage and GTV were related to resectability in TC (all P-values < 0.05), and multivariate analysis suggested that cN stage and GTV were independent risk factors with odds ratios of 1.594 (95% confidence interval [CI]: 1.105–2.301) and 1.055 (95%CI: 1.035–1.076), respectively. ROC analysis in TC revealed the cutoffs of 21.81, 21.70 and 36.93 cm3 to differentiate between resectable and unresectable cancers in stages cN0-3, cN2 and cN3 with areas under the curves of more than 0.8, respectively, which was validated in IVC and EVC with average Cohen k-values of more than 0.72. Conclusions GTV and cN stage can be independent risk factors of unresectable GA, and N-stage based GTV can help determine resectability.