Heterogeneity in Postoperative Survival Benefit Associated With Neoadjuvant Therapy in Esophageal Squamous Cell Carcinoma: A Causal Survival Forest Analysis

ABSTRACT The treatment effect of neoadjuvant therapy (NAT) in esophageal squamous cell carcinoma (ESCC) may vary across individuals. We aimed to estimate individualized treatment effects (ITEs) to identify patients most likely to benefit from NAT to characterize postoperative benefit stratification. This was a cohort‐based study for ITE estimation and treatment benefit prediction. The primary outcomes were 5‐year overall survival (OS) and disease‐free survival (DFS). We applied causal survival forests (CSF) to estimate ITEs, defined as the potential survival outcome differences under NAT before surgery and surgery alone. Patients were stratified into tertiles based on predicted ITEs. Survival outcomes were descriptively compared between patients whose actual treatment was concordant or discordant with the predicted treatment benefit pattern. A total of 838 patients were included, the predicted ITE ranged from −480 to 312 days for OS and from −435 to 442 days for DFS. Patients were stratified into lower, middle, and upper benefit groups. The C‐for‐benefit values were 0.532 for OS and 0.544 for DFS, and the AUTOC values were 34.86 and 89.49 days, respectively. IPTW‐weighted survival curves showed more favorable OS and DFS among patients whose actual treatment was concordant with the predicted treatment benefit pattern than among those with discordant treatment patterns. Age, tumor size, and clinical N stage were the most influential variables. CSF‐based ITE estimation suggested heterogeneity in NAT benefit and may support future individualized treatment research in ESCC, but external validation is needed before clinical implementation.