Formalin-fixed, paraffin-embedded (FFPE) tissue sections of 4-mm thickness were stained for PD-L1 with an anti-human PD-L1 rabbit monoclonal antibody (clone SP142; Ventana, Tucson, AZ) on an automated staining platform (Benchmark; Ventana) using a concentration of 4.3 mg ml21 (link), with signal visualization by diaminobenzidine; sections were counter-stained with haematoxylin. PD-L1 expression was evaluated on tumour cells and tumour-infiltrating immune cells. For tumour cells the proportion of PD-L1-positive cells was estimated as the percentage of total tumour cells; tumour cells typically showed membranous staining with avariably strong component of cytoplasmic staining. The distribution of PD-L1-positive tumour cells within a given tumour sample was typically very focal; in tumours growing as solid aggregates positive tumour cells were more commonly observed at the interface between malignant cells and stroma containing tumour-infiltrating immune cells. For tumour-infiltrating immune cells, the percentage of PD-L1-positive tumour-infiltrating immune cells occupying the tumour was recorded; tumour-infiltrating immune cells with clearly discernible cytoplasm, such as macrophages and dendritic cells, showed a membranous staining pattern for PD-L1—this was more difficult to determine for cells of small lymphoid morphology with scant amounts of cytoplasm. PD-L1-positive tumour-infiltrating immune cells were typically seen as variably-sized aggregates towards the periphery of the tumour mass or in stromal bands dissecting the tumour mass or as single cells scattered in stroma or within tumour-infiltrating immune cell aggregates. Specimens were scored as IHC 0, 1, 2, or 3 if, 1%, $1% but, 5%, $5% but, 10%, or $10% of cells per area were PD-L1 positive, respectively. PD-L1 scores in patients with multiple specimens were based on the highest score. Based on the complexity of our scoring algorithm, we determined concordance between individual reads by different pathologists; in a cohort of >200 NSCLC samples, concordance between two pathologists was >90%. CD8 (clone SP16 (Epitomics)) IHC was performed on a Discovery XT autostainer (Ventana) using CC1 antigen retrieval and OmniMap (Ventana) detection technology.