3.1 Demographic data

A total of 469 patients were included in the study. Among these, 27 patients (5.76%) were aged 80 years or older. Preoperative assessments showed that 373 patients (79.5%) had locally advanced disease, with cT3 or cT4 stage based on preoperative esophageal ultrasound and CT scans. Only 81 patients (17.3%) were considered without lymph node (LN) metastasis before surgery. Of the 469 patients, 358 (76.3%) underwent minimally invasive esophagectomy (MIE) (Table 1).

Table 1 Demographic characteristics of the groups

In terms of pathological features, 164 patients (35.0%) exhibited lymphovascular invasion, and 204 patients (43.5%) had nerve invasion. A total of 452 patients (96.4%) successfully underwent R0 resection, indicating complete surgical resection with no residual cancer at the resection margin (Table 2).

Table 2 Demographic characteristics after surgery

The patients were distributed across the four seasonal groups as follows: 147 patients (31.3%) were in the Spring Group, 116 patients (24.7%) in the Summer Group, 106 patients (22.6%) in the Autumn Group, and 100 patients (21.3%) in the Winter Group (Fig. 1).

Fig. 1

CONSORT diagram of patient selection

3.2 Overall survival and disease free survival

Among the 469 patients, the median follow-up duration was 47.5 months, with a median OS time of 51.6 months (95% CI 38.10–65.10). Across the four seasonal groups, no statistically significant differences were observed in OS (p-values ranged from 0.206 to 0.863). The Spring Group demonstrated the longest median OS at 61.3 months (95% CI 40.5–82.0), with 1-, 3-, and 5-year OS rates of 86%, 57%, and 52%, respectively. The Summer Group had a median OS of 43.1 months (95% CI 28.5–57.7), and the 1-, 3-, and 5-year OS rates were 89%, 53%, and 42%. The Autumn Group showed a median OS of 40.2 months (95% CI 25.6–54.8), with OS rates at 1, 3, and 5 years of 80%, 54%, and 40%. Similarly, the Winter Group’s median OS was 40.2 months (95% CI 22.7–52.8), with 1-, 3-, and 5-year OS rates of 88%, 51%, and 43%. Despite the observed numerical differences, no significant survival advantage was linked to seasonal timing of surgery (Fig. 2A).

Fig. 2

A Overall survival curves of participants in 4 groups; B Disease-free survival curves of participants in 4 groups

DFS also showed no significant differences among the seasonal groups, with p-values ranging from 0.493 to 0.963. The Spring Group had a median DFS time of 33.0 months (95% CI 13.1–52.9), while the Summer Group exhibited a slightly longer median DFS of 34.6 months (95% CI 17.8–51.4). The Autumn Group’s median DFS was 33.5 months (95% CI 26.6–40.4), and the Winter Group had the shortest DFS at 27.1 months (95% CI 15.8–38.4). Despite these variations, no significant seasonal differences in DFS were identified, suggesting that the timing of surgery within a given season did not notably influence recurrence-free outcomes for elderly patients with ESCC (Fig. 2B).

3.3 Restricted mean survival time and restricted mean disease-free survival time

Analysis of RMST revealed that all four subgroups exhibited RMST values exceeding 40 months, demonstrating a consistent trend across seasons. The Spring Group demonstrated the longest RMST at 44.7 months (Crude 95% CI: 40.42–48.95; Adjusted 95% CI 33.93–55.43). The Summer Group had an RMST of 42.1 months (Crude 95% CI 37.60–46.56; Adjusted 95% CI 32.26–51.87), followed by the Winter Group at 41.0 months (Crude 95% CI 36.25–45.65; Adjusted 95% CI 31.95–49.96), and the Autumn Group at 40.6 months (Crude 95% CI 35.48–45.67; Adjusted 95% CI 28.91–52.25) (Fig. 3A, B).

Fig. 3

Restricted mean survival time (RMST) and Restricted Mean Disease-Free Survival Time (RMDFST) estimates patients. A Crude RMST estimates different patients; B Adjust RMST estimates different patients; C Crude RMDFST estimates different patients; D Adjust RMDFST estimates different patients

Similarly, analysis of RMDFST showed that all four seasonal groups had RMDFST values above 40 months, maintaining a comparable trend. The Spring Group showed the longest RMDFST at 37.8 months (Crude 95% CI 33.32–42.32; Adjusted 95% CI 25.73–49.90), followed by the Summer Group at 37.4 months (Crude 95% CI 32.53–42.18; Adjusted 95% CI 27.50–47.20). The Autumn Group had an RMDFST of 36.0 months (Crude 95% CI: 30.90–41.16; Adjusted 95% CI 23.57–48.49), and the Winter Group had the shortest RMDFST at 35.3 months (Crude 95% CI 30.69–39.96; Adjusted 95% CI 26.55–44.11) (Fig. 3C, D).

3.4 Short-term outcomes and adverse events (Clavien-Dindo, 2009)

Of the 469 patients included in the study, 233 (49.7%) experienced Clavien-Dindo grade III-IV postoperative complications. Three patients died within 30 days of surgery, and 14 patients died within 90 days. A breakdown by season reveals the following rates of Clavien-Dindo grade III-IV complications: Spring Group, 76 patients (51.70%); Summer Group, 59 patients (50.86%); Autumn Group, 52 patients (49.06%); and Winter Group, 46 patients (46.00%). Two patients died during the perioperative period (Tables 2 and  3).

Table 3 Adverse events (Clavien–Dindo ≥ III, 2009)

Analysis of specific complication types revealed significant seasonal variations. The Summer Group showed a significantly higher incidence of hydrothorax compared to other groups (P < 0.05). The Summer Group also exhibited a higher rate of pulmonary infection than other groups, with the difference being statistically significant when compared to the Winter Group (P = 0.041). The Autumn Group demonstrated a significantly higher incidence of anastomotic leakage compared to the Spring Group (P = 0.011) (Fig. 4). Following PSM, the Summer Group continued to show a significantly higher incidence of hydrothorax compared to the non-Summer groups (P = 0.025) (Table 4 and Fig. 5).

Fig. 4

Adverse events of participants with Clavien-Dindo grade III-V. A Adverse events of Spring Group and Summer Group before PSM; B Adverse events of Spring Group and Autumn Group before PSM; C Adverse events of Spring Group and Winner Group before PSM; D Adverse events of Summer Group and Autumn Group before PSM; E Adverse events of Summer Group and Winner Group before PSM; F Adverse events of Autumn Group and Winner Group before PSM

Table 4 Adverse events (Clavien–Dindo ≥ III, 2009)Fig. 5

A Adverse events of Summer Group and Non-Summer Group before PSM; B Adverse events of Summer Group and Non-Summer Group after PSM

3.5 Risk factors

Univariate analysis revealed that smoking status, anastomosis location (intrathoracic vs. cervical anastomosis), lymphovascular invasion, nerve invasion, R0 resection status, occurrence of adverse events, and tumor stage/grade were all associated with OS. Multivariate Cox proportional hazards regression analysis identified lymphovascular invasion (P < 0.001), pT4 stage (P = 0.023), pN3 stage (P < 0.001), pTNM stage IV (P = 0.023), and the occurrence of severe adverse events (Clavien-Dindo ≥ III) (P = 0.029) as independent prognostic factors for OS (Fig. 6).

Fig. 6

Univariate and multivariate Cox regression analyses regarding factors affecting OS of patients

Similarly, univariate analysis for DFS showed associations between smoking status, R0 resection status, adverse events, and tumor stage/grade. Multivariate analysis confirmed R0 resection status (P = 0.013), pN3 stage (P = 0.003), pN3 stage (P < 0.001), pTNM stage IV (P = 0.020), and severe adverse events (Clavien-Dindo ≥ III) (P < 0.001) as independent predictors of DFS (Fig. 7).

Fig. 7

Univariate and multivariate Cox regression analyses regarding factors affecting DFS of patients