Abstract
<title>Abstract</title> <p>Background Endoscopic biopsy PD-L1 Combined Positive Score (CPS) guides gastric cancer (GC) immunotherapy at clinical cutoffs of 1 or 5. However, high intratumoral heterogeneity causes biopsy CPS underestimation compared to definitive surgical specimens, leading to false negatives and missed therapeutic opportunities. This study evaluated this diagnostic discordance and identified independent clinicopathological predictors of underestimation in biopsy-negative patients. Methods From 1,454 retrospective GC cases dignosed at Tianjin Cancer Hospital between December 2021 and December 2025, a cohort of 257 patients with paired pre-treatment biopsies and surgical specimens was analyzed. "Biopsy underestimation" (false-negative) was defined as biopsy CPS below clinical cutoffs (CPS ≥ 1 and CPS ≥ 5) despite specimen CPS meeting or exceeding them. Logistic regression analyses identified independent predictors of missed PD-L1 positivity in biopsy-negative sub-cohorts. Results Biopsies systematically underestimated CPS compared to surgical specimens (p < 0.01; median 2.0 [IQR: 0.0, 5.0] vs. 3.0 [IQR: 1.0, 8.0]). False-negative rates were 19.1% at cutoff 5 and 17.5% at cutoff 1, both of which were unaffected by neoadjuvant chemotherapy. Lauren classification significantly influenced underestimation at both cutoffs in univariate analysis. On multivariable regression for the cutoff 1 biopsy-negative population, diffuse-type Lauren classification was a strong independent protective factor against underestimation (OR = 0.18, 95% CI: 0.05–0.54, p < 0.001). Furthermore, diffuse-type GC consistently exhibited lower overall CPS scores and a minimal median difference between biopsy and specimen compared to other subtypes. Conclusion Endoscopic biopsies systematically underestimate gastric cancer PD-L1 CPS. Lauren classification is a critical determinant of biopsy diagnostic reliability.</p>