Claudin 18.2 Positivity Thresholds and Candidate Populations for Targeted Therapy in Pancreatic Ductal Adenocarcinoma: A Real-World Retrospective Cohort.

BACKGROUND/OBJECTIVES: Claudin 18.2 (CLDN18.2) is expressed in a substantial proportion of pancreatic ductal adenocarcinoma (PDAC) and has emerged as a promising therapeutic target. However, the positivity threshold for enrollment in CLDN18.2-directed trials remains undefined, and ongoing studies use different criteria. We quantified how different CLDN18.2 thresholds influence the size of the candidate population within the same PDAC cohort. METHODS: In this retrospective single-center study, CLDN18.2 was assessed by immunohistochemistry (43-14A clone, Ventana BenchMark) in 99 PDAC patients (metastatic, n = 71; non-metastatic, n = 28). The candidate population was evaluated at three thresholds: two used in current CLDN18.2 trials (any-intensity [1+/2+/3+] staining in ≥40% and moderate-to-strong [2+/3+] staining in ≥75% of tumor cells) and an exploratory lower bound of any detectable expression (≥5%). Between-threshold reclassification was analyzed as paired data; clinicopathological and survival analyses were exploratory. RESULTS: The candidate proportion was 40.4% (40/99) at ≥75%, 43.4% (43/99) at ≥40%, and 64.6% (64/99) at ≥5%; among metastatic patients, it was 36.6%, 38.0%, and 59.2%, respectively. Lowering the threshold from ≥75% to ≥40% caused virtually no reclassification (1/71 metastatic, 1.4%; McNemar p = 1.0), whereas lowering it to ≥5% reclassified 16 metastatic patients (22.5%; 95% CI, 14.4-33.5; p < 0.001) as candidates. Among the 64 tumors with detectable expression, median staining was 80% (IQR, 20-90%). Survival analyses were exploratory and underpowered (52 deaths; 80% power for HR ≥ 2.24); no association between CLDN18.2 and overall or progression-free survival was detected at any threshold. CONCLUSIONS: The stringent thresholds used in current drug development identified nearly identical candidate populations, whereas relaxation to any detectable expression substantially enlarged the pool; CLDN18.2 behaved as a targetable rather than a prognostic biomarker. Our findings highlight the need for a PDAC-specific, response-based CLDN18.2 threshold.