A pretreatment hormonal-nutritional-immune score for locally advanced gastric cancer receiving neoadjuvant immunochemotherapy: a multicenter retrospective cohort study

Background Pathological response and survival following neoadjuvant PD-1 inhibitor plus chemotherapy in locally advanced gastric cancer (LAGC) are heterogeneous, and tumor biomarkers do not capture host endocrine, nutritional, or inflammatory status. We developed a pretreatment Hormonal-Nutritional-Immune Score (HNIS) based on 12 routine blood indicators. Methods This multicenter retrospective cohort enrolled 342 patients with cT2-4bN_anyM0 gastric or gastroesophageal junction (GEJ) adenocarcinoma treated with at least 2 cycles of PD-1 inhibitor plus platinum-based chemotherapy at 4 Chinese centers between December 2019 and January 2023, with patients allocated to a training cohort (n=206), an internal validation cohort (n=68), and an external validation cohort (n=68). HNIS assigned 1 point for each of 12 hormonal, nutritional, and inflammatory abnormalities; patients were stratified as Low (0 to 2), Intermediate (3 to 5), or High risk (6 or greater). The primary endpoint was major pathological response (MPR). Logistic regression, Cox regression, Kaplan-Meier analysis, and discrimination analyses were used. Additional analyses assessed treatment interactions, individual components, LASSO-derived weights, and alternative group boundaries using repeated nested cross-validation. Results The median age was 62 years, 233 patients (68.1%) were male, and the median HNIS was 3 (IQR 2 to 5); 138 (40.4%) were Low risk, 121 (35.4%) Intermediate, and 83 (24.3%) High risk. MPR rates declined from 67.4% in Low risk to 46.3% in Intermediate and 20.5% in High risk patients (P<0.001), and grade 3 or higher immune-related adverse events (irAEs) increased from 7.2% to 9.9% to 26.5% (P<0.001). Each one-point HNIS increase was independently associated with lower odds of MPR (adjusted OR 0.68, 95% CI 0.61 to 0.76; P<0.001). At a median follow-up of 52.5 months, 3-year DFS was 69.6%, 58.7%, and 24.1%, and 3-year OS was 94.9%, 87.6%, and 69.9% across the 3 groups (both log-rank P<0.001). The HNIS score achieved AUCs of 0.725, 0.736, and 0.753 for MPR prediction in the training, internal, and external validation cohorts, outperforming a clinical model based on stage, PD-L1, and MMR (AUCs 0.554 to 0.611; DeLong all P<0.05). In supplementary repeated cross-validation, the equal-weight score had an AUC of 0.733, compared with 0.704 for a nine-component LASSO model. Conclusions A pretreatment HNIS based on routine blood tests independently predicts MPR, severe irAEs, DFS, and OS in patients with LAGC receiving neoadjuvant PD-1 inhibitor plus chemotherapy, and may refine pretreatment risk stratification beyond conventional staging and tumor biomarkers.