Abstract:

INTRODUCTION: Programmed death ligand 1 (PD-L1) expression is a predictive biomarker for patient response to nivolumab in nonsquamous NSCLC. However, the number of biopsy samples and tumor cells (TCs) required to assess PD-L1 expression remains unclear.
METHODS: A total of 222 biopsy samples from 80 patients with nonsquamous NSCLC treated with nivolumab were collected. Number of TCs and PD-L1 score were compared among the sample containing the largest number of TCs (Max-TC), the sample containing the smallest number of TCs (Min-TC), and the total samples from each patient. The impact of the number of samples and TCs on the prediction of patient response to nivolumab with use of PD-L1 scores was evaluated.
RESULTS: There was a mismatch in PD-L1 scores less than 1% and those of at least 1% between Max-TC and the total samples in one patient (1%) and between Max-TC and Min-TC in six patients (8%). The optimal number of TCs to match PD-L1 expression less than 1% versus at least 1% between Max-TC and Min-TC was 100 (sensitivity = 0.676 and 1 - specificity = 0.333). PD-L1 expression of at least 1% in Min-TCs containing at least 100 TCs was associated with longer progression-free survival (median 7.6 versus 1.8 months [p < 0.01]) and overall survival (median not reached versus 9.9 months [p = 0.04]) compared with PD-L1 expression less than 1%. However, there were no differences in progression-free survival (median 3.9 versus 2.3 months [p = 0.37]) or overall survival (median 9.7 versus 7.6 months [p = 0.60]) between PD-L1 expression of at least 1% and PD-L1expression less than 1% in Min-TCs containing fewer than 100 TCs.
CONCLUSION: Single biopsy samples containing at least 100 TCs are required to evaluate PD-L1 expression for predicting patient response to nivolumab.

日本語要旨: