29.04 Clinicopathologic Score Predicting Lymph Node Metastasis in T1 Gastric Cancer

T. B. Tran1, D. J. Worhunsky1, M. H. Squires2, L. X. Jin3, G. Spolverato4, K. I. Votanopoulos7, C. S. Cho5, S. M. Weber5, C. Schmidt6, E. A. Levine7, R. C. Fields3, T. Pawlik4,6, S. Maithel2, J. A. Norton1, G. A. Poultsides1  2Emory University,Atlanta, GA, USA 3Washington University In St. Louis,St. Louis, MO, USA 4John Hopkins Hospital,Baltimore, MD, USA 5University Of Wisconsin,Madison, WI, USA 6The Ohio State University,Columbus, OH, USA 7Wake Forest University,Winston-Salem, NC, USA 1Stanford University,Palo Alto, CA, USA

Introduction:  While gastrectomy with D2 lymphadenectomy is considered the standard treatment for invasive gastric adenocarcinoma, endoscopic resection (ER) has been described by Asian authors in select patients with T1 gastric cancer. Accurate preoperative prediction of lymph node (LN) metastasis in this setting is critical, since ER omits LN harvest. The objective of this study is to identify preoperative predictors of LN metastasis in US patients with T1 gastric cancer.

Methods:  Patients who underwent surgical resection for T1 gastric cancer (T1a: into lamina propria or muscularis mucosa, and T1b:  into submucosa) between 2000 and 2012 in 7 US academic institutions were identified. Clinicopathologic predictors of LN metastasis were determined using univariate and multivariate logistic regression. A preoperative score was created assigning points based on each variable’s beta-coefficient.

Results: Among 965 patients with gastric cancer undergoing surgical resection, 198 patients (20.5%) had T1 disease confirmed on final pathology.  Of those, 40 patients (20%) had LN metastasis. Independent predictors of LN involvement on multivariate analysis were poor differentiation (OR 4.5, P=0.002, beta 1.5), T1b stage (OR=4.5, P=0.02, beta 1.5), lymphovascular invasion (OR 2.8, P=0.049, beta 1.4), and tumor size > 2 cm (OR 2.8, P=0.026, beta 1.0). A clinicopathologic risk score predicting LN metastasis was created, assigning 3 points for the first 3 variables and 2 points for the last variable. The performance of the score was evaluated with an ROC curve (Figure) showing excellent discrimination (AUC = 0.79) and 100% sensitivity in detecting LN metastasis in patients with a score of 3 or less.

Conclusion: In this cohort of US patients with T1 gastric adenocarcinoma, lack of LN involvement could be predicted if none or one of the following unfavorable factors is present (T1b, poor differentiation, lymphovascular invasion, size > 2 cm). For these patients, endoscopic resection may be a potential treatment option provided it could be achieved with negative margins.