Programmed death-ligand 1 (PD-L1) expression is used for treatment prediction in non-small cell lung cancer (NSCLC). While cytology may be the only available material in the routine clinical setting, testing in clinical trials has mainly been based on biopsies.

We included 2 retrospective cohorts of paired, concurrently sampled, cytological specimens and biopsies. Also, the literature on PD-L1 in paired cytological/histological samples was reviewed. Focus was on the cutoff levels ≥1 and ≥50% positive tumor cells.

Using a 3-tier scale, PD-L1 was concordant in 40/47 (85%) and 66/97 (68%) of the paired NSCLC cases in the 2 cohorts, with kappa 0.77 and 0.49, respectively. In the former cohort, all discordant cases had lower score in cytology. In both cohorts, concordance was lower in samples from different sites (e.g., biopsy from primary tumor and cytology from pleural effusion). Based on 25 published studies including about 1,700 paired cytology/histology cases, the median (range) concordance was 81-85% (62-100%) at cutoff 1% for a positive PD-L1 staining and 89% (67-100%) at cutoff 50%.

The overall concordance of PD-L1 between cytology and biopsies is rather good but with significant variation between laboratories, which calls for local quality assurance.