Baseline thyroid function and treatment-emergent thyroid dysfunction predict pathological response and survival after neoadjuvant PD-1 inhibitor plus platinum-based chemotherapy in locally advanced gastric and gastroesophageal junction adenocarcinoma: a multicenter cohort study.

BACKGROUND: Thyroid dysfunction is among the most frequent endocrine immune-related adverse events (irAEs) during PD-1/PD-L1 blockade, but whether baseline thyroid function and treatment-emergent thyroid dysfunction (TeTD) predict pathological response and survival after neoadjuvant immunochemotherapy in locally advanced gastric or gastroesophageal junction adenocarcinoma (LAGC/GEJC) remains unclear. METHODS: This multicenter retrospective cohort enrolled 420 patients with cT2-4bN_anyM0 LAGC/GEJC from five Hebei centers (2019-2024). All received four cycles of neoadjuvant PD-1 inhibitor plus SOX or XELOX, followed by D2 gastrectomy. Baseline TSH, FT3, FT4, TPOAb, and TgAb were measured. A favorable baseline thyroid profile was prespecified as TSH 0.55-2.50 mIU/L, FT3 ≥4.30 pmol/L, normal FT4, and negative autoantibodies. TeTD was defined as new-onset hypothyroidism, thyrotoxicosis, or biphasic thyroiditis during therapy. Logistic and Cox regression, subgroup and sensitivity analyses, and a pCR nomogram were applied. RESULTS: Of 420 patients (median age 60, 66.9% male), 114 (27.1%) had a favorable thyroid profile and 105 (25.0%) developed TeTD; tumor-intrinsic factors were balanced. Overall pCR was 23.3% and MPR 42.6%. Favorable profile predicted higher pCR (43.0% vs. 16.0%; adjusted OR 6.08, P<0.001) and MPR (61.4% vs. 35.6%; OR 3.42, P<0.001). TeTD independently predicted higher pCR (38.1% vs. 18.4%; OR 3.26, P<0.001) and MPR. Non-thyroid irAEs and discontinuation were similar (P = 0.082). At 39.2 months median follow-up, 3-year DFS was 77.7% vs. 49.8% (TeTD vs. non-TeTD; adjusted HR 0.35, P<0.001) and 3-year OS 99.0% vs. 78.4% (HR0.11, P<0.001). A thyroid-inclusive pCR nomogram achieved AUC 0.82 (95% CI 0.77-0.86) versus 0.73 for a staging-only model (P<0.001), with adequate calibration (Hosmer-Lemeshow P = 0.84). Effects were robust across all subgroups and sensitivity analyses. CONCLUSIONS: Pretreatment thyroid function and on-treatment thyroid dysfunction are independent and complementary host-factor predictors of pathological response and long-term survival after neoadjuvant PD-1 inhibitor plus platinum-based chemotherapy in LAGC/GEJC. Incorporating these inexpensive thyroid-axis variables into pretreatment risk stratification provides additional prognostic information beyond conventional staging and tissue biomarkers.