In advanced non-small cell lung cancer (NSCLC), small biopsy specimens from endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA) are often the only available material from cancer tissue for the analysis of programmed death ligand-1 (PD-L1) expression. We aim to assess the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of PD-L1 expression at ≥ 1% and ≥ 50% on EBUS-TBNA samples compared with their corresponding surgically resected tumor. We retrospectively reviewed all patients who underwent EBUS-TBNA followed by surgical resection of NSCLC between July 2006 and September 2016. Demographic information and periprocedural/surgical data were collected. The archived specimens were retrieved and assessed for PD-L1. A positive PD-L1 stain was defined using two separate cutoff points: ≥ 1% and ≥ 50% of tumor cell positivity. EBUS-TBNA aspirates were compared with the surgically resected specimen to calculate the sensitivity, specificity, PPV, and NPV. Sixty-one patients were included. For PD-L1 ≥ 1%, the sensitivity, specificity, PPV, and NPV were 72%, 100%, 100%, and 80%, respectively. For PD-L1 ≥ 50%, the sensitivity, specificity, PPV, and NPV were 47%, 93%, 70%, and 84%, respectively. The concordance rates for PD-L1 ≥ 1% and ≥ 50% were 87% and 82%, respectively. A PD-L1 cutoff of ≥ 1% on EBUS-TBNA has a strong correlation with resected tumor specimen. For PD-L1 ≥ 50%, there is a significant decrease in the sensitivity and PPV of EBUS-TBNA specimen when compared with resected tumor. When analyzing for PD-L1 expression using a cutoff of ≥ 50%, EBUS-TBNA specimens may misclassify the status of PD-L1.