Background: How to best evaluate the disease-specific survival (DSS) of gastric cancer (GC) survivors over time is unclear.
Methods: Clinicopathological data from 22 265 patients who underwent curative intend resection for GC were retrospectively analyzed. Changes in the patients' 3-year conditional disease-specific survival (CS3) were analyzed. We used time-dependent Cox regression to analyze which variables had long-term effects on DSS and devised a dynamic predictive model based on the length of survival.
Results: Based on 1-, 3-, and 5-year survivorships, the CS3 of the population increased gradually from 62% to 68.1%, 83.7%, and 90.6%, respectively. Subgroup analysis showed that the CS3 of patients who had poor prognostic factors initially demonstrated the greatest increase in postoperative survival time (eg, N3b: 26.6%-84.1%, Δ57.5% vs N0: 84.1%-93.3%, Δ9.2%). Time-dependent Cox regression analysis showed the following predictor variables constantly affecting DSS: age, the number of examined lymph nodes (LNs), T stage, N stage, and site (P < .05). These variables served as the basis for a dynamic prediction model.
Conclusions: The influence of prognostic factors on DSS and CS3 changed dramatically over time. We developed an effective model for predicting the DSS of patients with GC based on the length of survival time.
Keywords: conditional survival; disease-specific survival; gastric cancer; length of survivorship; predictive model.
© 2019 Wiley Periodicals, Inc.