Statistical analysis

All patients in the unmatched dataset met the inclusion criteria. Propensity score matching (PSM) can help to achieve balanced covariates across treatment groups. Patients in the two groups were matched 1:1 using the nearest propensity score (PS) on the logit scale. A matched dataset was created using PS of age, gender, race, tumor differentiation, histological type, and year of diagnosis. After PSM, differences in categorical clinical characteristics were tested for significance by chi‐square tests.

Five‐year OS and CSS were calculated and expressed as months. The OS was right censored if the patient was alive at study termination or was lost to follow‐up, and patient death was considered an event. In CSS analysis, surviving patients or those that died from other reasons were censored, while death from esophageal cancer was considered an event. The Kaplan–Meier method was used to generate the survival curve. A log‐rank test was performed to compare OS and CSS among ES, ET, and RT groups. A multivariate Cox proportional hazards model was constructed to assess the hazard ratios (HRs) and 95% confidence intervals (CI) of eight covariates: age, gender, race, marital status, T staging, tumor differentiation, histological type, and year of diagnosis. SPSS version 22.0 was used for statistical analyses. All tests were two‐sided with a significance level of P < 0.05.