Validation of the immunohistochemical expression of programmed death ligand 1 (PD-L1) on cytological smears in advanced non small cell lung cancer

Introduction The assessment of PD-L1 expression by immunohistochemistry is mandatory for the administration as first-line therapy of the anti PD-1 check-point inhibitor Pembrolizumab in patients with advanced non-small-cell lung cancer (NSCLC). Currently, only formalin-fixed paraffin-embedded samples are acceptable for PD-L1 immunostaining with the anti-PD-L1 antibodies 22-C3 and SP263. We investigated retrospectively the accuracy of the anti PD-L1 antibodies 22-C3, 28-28, SP263 in 50 paired histological samples and cytological smears of NSCLC patients. Results The accuracy of the three antibodies for the detection of PD-L1 in histological samples was higher for the antibody SP263 (AUC/ROC = 1) compared to the clones 28-8 (AUC/ROC =,991) and 22-C3 (AUC/ROC =,942). The overall concordance between histological samples and cytological smears using the SP263 clone was moderate (kappa = 0,364). However when the cyto-histological concordance was calculated using just the <50% vs ≥50% cut-off the agreement (kappa = 0.626) was good. The accuracy of the antibody SP263 in cytological smears was good (AUC/ROC =,921). A fluorescent in situ hybridization analysis on 10 histological cases positive for PD-L1 at immunohistochemistry showed amplification of the CD274 gene only in one case. Conclusions Immunocytochemical staining for PD-L1 in diagnostic cytological smears of NSCLC is feasible and applicable at least using the >50% cancer cell cut-off. The three antibodies SP263, 22-C3 and 28-8 are all suitable for the diagnostic detection of PD-L1 on tissue sections with a superiority of the SP263 clone. The implementation of PD-L1 immunocytochemistry on cytological smears will likely expand the pool of NSCLC patients candidate to first-line immunotherapy.