Clinical and CT body-composition model for major pathological response to neoadjuvant tislelizumab, albumin-bound paclitaxel, and nedaplatin in esophageal squamous cell carcinoma

Background Sarcopenia and osteoporosis are linked to poor treatment outcomes and are significantly underdiagnosed in esophageal cancer patients. Despite the promising efficacy of neoadjuvant immunochemotherapy (NACI) in esophageal squamous cell carcinoma (ESCC), predicting treatment response remains a major clinical hurdle. Objective This study aimed to evaluate the associations of sarcopenia and osteoporosis on pathological response and to develop a both clinical and computed tomography (CT)-based predictive model for NACI outcomes in ESCC patients. Methods A retrospective analysis was conducted in 179 patients with ESCC who underwent radical resection after neoadjuvant tislelizumab, albumin-bound paclitaxel, and nedaplatin. Within each of 500 stratified bootstrap resamples, five-fold cross-validation was used to determine the optimal LASSO penalty parameter, followed by Firth logistic regression. Aggregated out-of-bag predictions underwent five-fold cross-fitted recalibration. Model discrimination, calibration, and clinical utility were evaluated using the AUC, accuracy, sensitivity, specificity, Brier score, calibration intercept and slope, Hosmer–Lemeshow test, and decision curve analysis. Results Major pathological response (MPR) rates were 70.0%, 40.0%, 22.0%, and 8.3% in the control, osteoporosis-only, sarcopenia-only, and osteosarcopenia groups, respectively. In multivariable Firth logistic regression, lower tumor grading, absence of comorbidities, absence of lymphovascular invasion, absence of sarcopenia, and a higher intermuscular adipose tissue (IMAT) area percentage were associated with a greater likelihood of achieving MPR, whereas osteoporosis was not associated with MPR (P = 0.130). The cross-fitted recalibrated model achieved an AUC of 0.885 (95% CI, 0.835–0.934), with 0.818 accuracy, 0.840 sensitivity, and 0.800 specificity at a threshold of 0.405. Calibration was satisfactory (Brier score, 0.136; intercept, −0.009; slope, 0.960; Hosmer–Lemeshow P = 0.219), and decision curve analysis showed clinical net benefit across thresholds of 0.08–0.80. Conclusion Absence of sarcopenia and a higher IMAT area percentage were associated with a greater likelihood of achieving major pathological response. The combined clinical and CT-based model showed good internally validated performance.