Development of a prediction model of adverse events after stent placement for esophageal cancer

Background and Aims: Self-expandable metal stent (SEMS) positioning is the recommended method for palliation of dysphagia from esophageal cancer, although it is not adverse event-free. The present study was aimed at identifying predictors for adverse events and at proposing a statistical model to predict them. Methods: We performed a retrospective analysis of a prospectively collected database. All patients who underwent SEMS placement for stricture due to esophageal cancer between 2002 and 2011 in a tertiary-care center were identified. Multivariable regression analysis in the presence of competing risk events was used to identify factors associated with SEMS-related adverse events and to build a prediction model. Results: A total of 267 patients were included. According to the competing risk regression analysis, only 2 variables were significantly associated with the risk of SEMS-related adverse events: prior chemoradiotherapy (CRT), yielding a hazard ratio (HR) of 1.687 (95% confidence interval [CI], 1.076-2.644), and the SEMS length (HR 0.884; 95% CI, 0.798-0.980) for every 10-mm length increase. Based on the estimated probability curves, after 4 months from SEMS placement, the probability of an adverse event in patients who did receive prior CRT was 50.9% compared with 34.4% in those who did not receive prior therapy, which was reduced to 9.2% and 15.1%, respectively, if a 180 mm-length stent was used. The ability of the predictive model to differentiate between patients who did and did not experience the adverse event was moderate (c-index: 0.617). Conclusion: The rate of SEMS-related adverse events was higher in patients with previous CRT and lower in patients receiving longer stents. Both factors were used to build an accurate predictive model.