Impact of recurrence on survival after gastric cancer curative surgery: a retrospective, single center, multistate model analysis.

BACKGROUND: Recurrence is the strongest determinant of mortality after curative gastrectomy for gastric cancer. Conventional Cox models do not explicitly account for recurrence as an intermediate event. Multistate models (MSMs) allow dynamic estimation of transition-specific risks across clinically relevant disease states. METHODS: Patients undergoing curative-intent gastrectomy between 2013 and 2023 were retrospectively analyzed. A three-state MSM (alive recurrence-free, alive after recurrence, death) was constructed. State occupation probabilities and transition probabilities were estimated using the Aalen-Johansen estimator. Subanalyses were conducted on high (≥0.25) and low (<0.25) lymph node ratio (LNR) subcohorts. RESULTS: Among 179 patients (median age 72 years), the probability of remaining alive and recurrence-free declined from 66.5% at 12 months to 33.6% at 58 months, while death increased from 22.2% to 57.2%. The probability of being alive after recurrence was 11.3% at 12 months and 9.2% at 60 months. All high-LNR patients developed recurrence within 3 years of follow-up, with marked differences in death and recurrence-free probabilities (88.8% vs 41.3%, and 0% vs 48.2%, respectively, at 60 months). High-LNR patients showed earlier and near-universal recurrence, reaching 100% by 36 months, unlike the relatively stable recurrence probabilities in low-LNR patients. Low-LNR patients who remained alive longer after recurrence had a lower probability of death compared with high-LNR patients (62.5% vs 80% at 24 months). CONCLUSIONS: MSMs provide a dynamic reconstruction of postoperative gastric cancer progression. Recurrence profoundly alters survival trajectories, and LNR emerges as a key determinant of prognosis. Multistate modeling may enhance risk-adapted surveillance after curative gastrectomy.