We retrospectively analyzed the clinical records of patients with esophageal cancer who received neoadjuvant therapy followed by surgical resection at the Affiliated Hospital of North Sichuan Medical College between 2020 and 2024. The institutional ethics committee approved the study (File Number: 2025ER240-1), which was conducted in accordance with the 2013 Declaration of Helsinki; written informed consent was waived owing to its retrospective design. Inclusion criteria were: (i) age 18–80 years and a preoperative diagnosis of locally advanced, resectable ESCC staged at least cT3 or N+; (ii) receipt of at least two cycles of NICT at our institution, followed by minimally invasive McKeown esophagectomy. Exclusion criteria included: (i) prior antitumor treatments or distant metastases; (ii) palliative resection or exploratory surgery only; (iii) lack of abdominal CT images before or after treatment, or incomplete clinical records. Pathological staging was determined following the 8th edition of the Union for International Cancer Control/American Joint Committee on Cancer staging system. Figure 1 shows the patient selection process.

We quantified changes in patients’ skeletal muscle mass, using the Skeletal Muscle Index (SMI) and semi-automatically delineated regions of interest (ROIs) in muscle tissue, defined by Hounsfield unit thresholds of −29 to +150 HU, using the SliceOmatic software. We calculated SMI as the total cross-sectional area of all skeletal muscles at the L3 level on pre- and post-treatment abdominal CT scans (cm²) divided by height squared (m²). Using large-scale population data (21), we defined sarcopenia as an SMI < 52.4 cm²/m² in men and <38.5 cm²/m² in women. Changes in SMI before and after NICT were expressed as a percentage: ΔSMI% = (SMI_post-NICT – SMI_pre-NICT)/SMI_pre-NICT × 100%. Supplementary Figure 1 presents a single patient’s CT images showing normal muscle mass pre-NICT and sarcopenia post-NICT.

The NICT regimen comprised a platinum agent [80 mg/m² intravenously (IV) on day 1] combined with albumin-bound paclitaxel (200 mg/m² IV on day 2), followed by an ICI (200 mg IV on day 3). All patients underwent at least two treatment cycles, with an inter-cycle interval of more than 3 weeks. Upon completing NICT, a multidisciplinary team assessed tumor resectability. For patients deemed suitable for curative resection, a minimally invasive three-incision approach (right thoracic, upper abdominal, and left cervical) was employed, together with a 2.5-field lymphadenectomy (22). Postoperatively, we scheduled follow-up visits every 3 months for the first 2 years and every 6 months thereafter. Nutritional interventions were not standardized in this study; in routine practice, patients with dysphagia underwent dietitian assessment and, if indicated, received oral nutritional supplements or nasogastric tube feeding.

The primary endpoints were overall survival (OS) and disease-free survival (DFS). We used a Cox proportional hazards regression model to adjust for clinical covariates and evaluate their prognostic significance. OS was defined as the interval between surgery and death from any cause, while DFS was defined as the interval between surgery and first recurrence or death from any cause. Secondary endpoints comprised neoadjuvant treatment-related adverse events, treatment response rate, pathological complete response (pCR), and postoperative complication rate. We employed a multivariate logistic regression model to identify independent risk factors for sarcopenia before and after neoadjuvant therapy. Post-treatment responses were classified following the Response Evaluation Criteria in Solid Tumors (RECIST) as complete response (CR), partial response (PR), stable disease (SD), or progressive disease (PD); CR, PR, and SD were collectively defined as disease control (DC). Treatment-related adverse events (TRAEs) were graded following the guidelines in version 5.0 of the National Cancer Institute’s Common Terminology Criteria for Adverse Events.

Continuous variables that are normally distributed are expressed as mean ± standard deviation (SD). Non-normally distributed variables are expressed as median (interquartile range [IQR]). We compared continuous variables between groups using the Student’s t-test or the Wilcoxon rank-sum test, depending on their distribution. The chi-square test or Fisher’s exact test was used to compare categorical variables. OS and DFS were estimated using the Kaplan–Meier method; between-group differences were assessed with the log-rank test, and survival curves were plotted in R version 4.3.2. The Kolmogorov–Smirnov (K–S) test indicated that ΔSMI% followed a normal distribution (p < 0.001). We used the Maxstat package in R to determine the optimal ΔSMI% cutoff for OS discrimination. A threshold of −2.8% was identified (p < 0.05), and ΔSMI% < −2.8% was defined as excessive skeletal muscle loss. Statistical analyses were performed using SPSS version 25.0 and R version 4.3.2. Two-sided p-values < 0.05 were considered statistically significant.

We included 272 patients with ESCC treated with NICT followed by surgical resection, with imaging data available for 253 patients before neoadjuvant therapy and for 252 patients after treatment. The baseline clinical characteristics of all patients are shown in Supplementary Table 1. The median age of the patients was 65 years, and the median BMI was 22.9 (IQR 20.8–24.9); most patients were male (73.2%) and reported a history of smoking (52.9%) or alcohol consumption (46.7%). Tumors were located predominantly in the middle third of the esophagus (71.3%) and in the lower third (16.5%). Prior to NICT, clinical staging was predominantly stage III (44.5%), followed by stage II (36.8%).

Table 1 shows the clinicopathological characteristics of patients with sarcopenia versus those without sarcopenia before and after neoadjuvant therapy. Prior to NICT, 129 patients (50.9%) had sarcopenia, which increased to 139 patients (55.1%) at treatment completion. Supplementary Figure 2 shows that, according to RECIST criteria, those without sarcopenia did not experience PD before or after treatment, and those achieving DC had significantly higher SMI values than did the patients with PD (pre-treatment: 48.71 ± 8.39 vs. 41.60 ± 7.24, p = 0.04; post-treatment: 48.43 ± 8.71 vs. 39.57 ± 9.36, p = 0.02). Additionally, the pCR rate was higher in patients without sarcopenia before treatment (25.0% vs. 14.7%; p = 0.04), although the difference was not significant after treatment (p = 0.081 and p = 0.197).

In the multivariable binary logistic regression (forest plot in Figure 2), age (OR 1.05; 95% CI, 1.01–1.10; p = 0.02), BMI (OR 0.64; 95% CI, 0.56–0.73; p = 0.01), and male sex (OR 2.61; 95% CI, 1.02–6.71; p = 0.04), were identified as independent predictors of pre-NICT sarcopenia. Furthermore, age (OR 1.06; 95% CI, 1.02–1.11; p = 0.01), BMI (OR 0.65; 95% CI, 0.57–0.74; p = 0.01), male sex (OR 4.32; 95% CI, 1.63–11.46; p = 0.01), and clinical N0 stage (OR 0.52; 95% CI, 0.28–0.96; p = 0.03) emerged as independent predictors of post-NICT sarcopenia.

Patients with sarcopenia and those without demonstrated favorable safety profiles and manageable adverse events before and after NICT. As shown in Table 2, among grade 1–2 treatment-related adverse events before (145, 57.3%) and after (143, 56.7%) NICT, anemia was the most frequent, with a significantly higher incidence in the sarcopenia group than in the non-sarcopenia group (65.1% vs. 49.2%, p = 0.012; 63.3% vs. 48.7%, p = 0.024). The incidence of other TRAEs did not significantly differ between groups. In the pre- (30.8%) and post- (31.7%) NICT cohorts, postoperative pulmonary infection was the most common complication, with no significant difference between patients with sarcopenia and those without (p = 0.304; p = 0.972). We found no significant differences between groups in operative time, intraoperative blood loss, or anastomotic leakage (p > 0.05) (see Table 3).

After a median follow-up of 20.4 months (95% CI: 19.2–21.6), 61 patients had died. The 1- and 2-year OS rates were 87 and 74%, respectively, whereas the DFS rates were 77 and 62%. Kaplan–Meier curve (Figure 3) shows that OS or DFS did not significantly differ between patients with sarcopenia and those without, either before or after NICT (p > 0.05). To assess the prognostic impact of dynamic skeletal muscle changes during NICT, we analyzed 233 patients with complete pre- and post-treatment imaging to calculate ΔSMI%. We used the Maxstat package in R to determine the optimal ΔSMI% cutoff for OS; a threshold of −2.8% (Figure 4) was used to stratify patients into low (<−2.8%) and high (≥−2.8%) ΔSMI% groups (see Supplementary Table 2 for baseline characteristics). Kaplan–Meier curve (Figure 5) shows that the 2-year OS rate was significantly higher in the ΔSMI% ≥ −2.8% group than in the ΔSMI% < −2.8% group (78% vs. 67%; p = 0.015). However, the 2-year DFS did not significantly differ between the two groups (63% vs. 57%; p = 0.22). Finally, we constructed a Cox proportional hazards regression model to further evaluate the prognostic value of relevant variables (Table 4). In the univariate analysis, age, chronic obstructive pulmonary disease, and ΔSMI% were associated with poorer OS; after multivariate adjustment, ΔSMI% remained an independent predictor of OS (high vs. low: HR 0.47; 95% CI, 0.25–0.91; p = 0.03).

Supplementary Table 2 shows that, apart from a significantly higher BMI in the high-level group versus the low-level group (23.2 [IQR 21.3–24.9] vs. 21.6 [19.8–24.5]; p = 0.01), baseline clinical characteristics were otherwise balanced between cohorts. Tumors in the high-level group were predominantly in the mid-esophagus (75.9%) and lower esophagus (15.8%), whereas those in the low-level group were mainly in the mid-esophagus (62.7%) and upper esophagus (21.3%). A significant difference in pCR rates was observed between groups (high vs. low: 22.8% vs. 12.0%; p = 0.05), with the high-level group exhibiting significantly better pathological downstaging after two NICT cycles (p = 0.03). In Supplementary Table 3, besides anemia (p = 0.01), leukopenia incidence also significantly differed between groups (p = 0.04). The high-level group had a lower rate of postoperative pulmonary infection than did the low-level group, although the difference was not statistically significant (28.5% vs. 37.5%; p = 0.17) (see Supplementary Table 4).