Assessment of PD-L1 on matched surgical specimen and biopsy in resectable esophagogastric cancer: The APEROL multicentric study

Background Programmed death-ligand 1 (PD-L1) expression defines eligibility for immunotherapy in patients with esophagogastric (EG) adenocarcinoma. Limited data exist on the agreement of PD-L1 combined positive score (CPS) between diagnostic biopsies and matched surgical specimens. We do not know the impact of therapy on PD-L1 expression. We aimed to evaluate the reliability of PD-L1 assessment on biopsies compared with matched surgical specimens. Materials and Methods The multicentric observational study included patients with resectable esophagogastric adenocarcinoma (cTNM II–IVa) who were treated at Istituto Oncologico Veneto–IRCCS and Azienda Ospedaliero-Universitaria Pisana. The study collected and analyzed with immunohistochemistry tissue microarray sections from the biopsies and, from the matched resected specimens. PD‑L1 CPS was defined as negative (<1), low (1–4), intermediate (5–9), or high (≥10). Results From May 2014 to November 2024 254 patients who had both a biopsy and a surgical specimen were enrolled. The median age of the patients was 62 years [Interquartile Range (IQR) 58 -76 years]. The patients were 73% male. ECOG performance status (PS) of patients was 0 in 52%. The cases of gastric or gastroesophageal junction tumors were 93%. Patients underwent surgery upfront (n=109, 43%) or were treated with neoadjuvant oncological treatment (n=126, 49.6%) or concomitant chemoradiotherapy (n=19, 7.4%). Distribution of PD‑L1 CPS categories was similar for biopsies and surgical specimens: negative (n=20 vs. n=26), low (n=72 vs. n=75), intermediate (n=46 vs. n=37), and high (n=116 vs. n=116). Agreement between biopsy and surgical tumor samples was low in patients with fewer than six biopsies (κ = 0.125; p = 0.054; N = 114) and significantly higher in those with six or more biopsies (κ = 0.316; p < 0.001). The probability of PD-L1 CPS falling into different categories did not differ between the Surgery and Neoadjuvant groups (OR 0.86; 95% CI 0.46–1.61; p = 0.6). PD-L1 CPS values tended to be higher when assessed with the 22C3 antibody compared with SP263 (estimated difference +4.40 CPS; 95% CI −0.62 to 9.40; p = 0.086). Conclusions PD-L1 CPS demonstrates limited concordance between diagnostic biopsy and matched surgical specimens in resectable EG adenocarcinoma, reflecting substantial intratumoral and sampling-related heterogeneity. This variability may reduce the reliability of baseline PD-L1 assessment. Therefore, caution is warranted when using PD-L1 CPS as a sole biomarker to guide treatment decisions, and adequate biopsy sampling should be ensured to improve its clinical applicability.