Long-term Comparison of Robotic and Laparoscopic Gastrectomy for Gastric Cancer: A Propensity Score-weighted Analysis of 2084 Consecutive Patients

Abstract

OBJECTIVE: To compare long-term outcomes between robotic and LG approaches using propensity score weighting based on a generalized boosted method to control for selection bias. SUMMARY OF BACKGROUND DATA: Minimally invasive surgical approaches for GC are increasing, yet limited evidence exists for long-term outcomes of robotic gastrectomy (RG). METHODS: Patients (n = 2084) with GC stages I-III who underwent LG or RG between 2009 and 2017 were analyzed. Generalized boosted method was used to estimate a propensity score derived from all available preoperative characteristics. Long-term outcomes were compared using the adjusted Kaplan-Meier method and the weighted Cox proportional hazards regression model. RESULTS: After propensity score weighting, the population was balanced. Patients who underwent RG showed reduced blood loss (16 mL less, P = 0.025), sufficient lymph node harvest from the initial period, and no changes in surgical outcomes over time. With 52-month median follow-up, no difference was noted in 5-year overall survival in unweighted [91.5% in LG vs 94% in RG; hazard ratio (HR), 0.71; 95% confidence interval (CI), 0.46-1.1; P = 0.126] and weighted populations (94.2% in LG vs 93.2% in RG; HR, 0.88; 95% CI, 0.52-1.48; P = 0.636). There were no differences in 5-year recurrence-free survival (RFS), with unweighted 5-year RFS of 95.4% for LG and 95.2% for RG (HR, 0.95; 95% CI, 0.55-1.64; P = 0.845) and weighted 5-year RFS of 96.3% for LG and 95.3% for RG (HR, 1.24; 95% CI, 0.66-2.33; P = 0.498). CONCLUSIONS: After balancing covariates, RG demonstrated reliable surgical outcomes from the beginning. Long-term survival after RG and LG for GC was similar.