PD-L1 CPS in gastroesophageal cancer: differences in routine care versus Checkmate 649 and implications for biopsy-site choice and assay standardisation.

[PURPOSE] To explore why PD-L1 scores in metastatic gastro-esophageal adenocarcinomas (GEAs) were significantly lower in a real-world cohort compared with the CheckMate 649 (CM649) trial.

[METHODS] PD-L1 combined positive scores (CPS) were evaluated using validated assays in 100 consecutive patients with advanced/metastatic GEA at St Bartholomew's Hospital (SBH) and compared with CM649 (n = 1567). Clinicopathological factors and biopsy site were analysed to assess their impact on PD-L1 results.

[RESULTS] CPS ≥ 5 was substantially less frequent in SBH patients (30%) compared with CM649 (61%). Older age ( ≥ 65 years), non-diffuse histology, and MMR deficiency were associated with CPS ≥ 5 across both cohorts, yet these factors were more common at SBH and therefore did not explain the lower positivity rate. Metastatic biopsies were more frequent in CM649 (21% vs. 9%), but CPS ≥ 5 was lower in metastases (50%) than in primary tumors (60%). Importantly, PD-L1 positivity varied by metastatic site: lymph node metastases showed the highest rate (80%), while liver (50%) and other sites (44%) were significantly lower than primaries (60%).

[CONCLUSION] PD-L1 CPS is shaped by clinicopathological context and biopsy site. The persistently lower CPS ≥ 5 prevalence at SBH despite validated testing highlights assay variability and reinforces the urgent need for assay standardisation. Preferential use of primary tumor tissue may help reduce metastasis-specific bias.