Programmed death-ligand 1 expression influenced by tissue sample size. Scoring based on tissue microarrays' and cross-validation with resections, in patients with, stage I-III, non-small cell lung carcinoma of the European Thoracic Oncology Platform Lungscape cohort.

PD-L1, as assessed by immunohistochemistry, is a predictive biomarker for immuno-oncology treatment in lung cancer. Different scoring methods have been used to assess its status, resulting in a wide range of positivity rates. We use the European Thoracic Oncology Platform Lungscape non-small cell lung carcinoma cohort to explore this issue. PD-L1 expression was assessed via immunohistochemistry on tissue microarrays (up to four cores per case), using the DAKO 28-8 immunohistochemistry assay, following a two-round external quality assessment procedure. All samples were analyzed under the same protocol. Cross-validation of scoring between tissue microarray and whole sections was performed in 10% randomly selected samples. Cutoff points considered: ≥1, 50 (primarily), and 25%. At the two external quality assessment rounds, tissue microarray scoring agreement rates between pathologists were: 73% and 81%. There were 2008 cases with valid immunohistochemistry tissue microarray results (50% all cores evaluable). Concordant cases at 1, 25, and 50% were: 85, 91, and 93%. Tissue microarray core results were identical for 70% of cases. Sensitivity of the tissue microarray method for 1, 25, and 50% was: 80, 78, and 79% (specificity: 90, 95, 98%). Complete agreement between tissue microarrays and whole sections was achieved for 60% of the cases. Highest sensitivity rates for 1% and 50% cutoffs were detected for higher number of cores. Underestimation of PD-L1 expression on small samples is more common than overestimation. We demonstrated that classification of PD-L1 on small biopsy samples does not represent the overall expression of PD-L1 in all non-small cell cancer carcinoma cases, although the majority of cases are 'correctly' classified. In future studies, sampling more and larger biopsies, recording the biopsy size and tumor load may permit further refinement, increasing predictive accuracy.