Background: Endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA) is an important tool in the diagnosis of mediastinal and hilar pathology. We describe our experience with EBUS-TBNA performed in a teaching institution primarily under conscious sedation.
Methods: Patients who underwent EBUS-TBNA were included in this retrospective review. We focused on the diagnostic yield of EBUS-TBNA in relationship to the nature of the mediastinal or hilar lesions (suspected malignancy vs. benign disease), incremental 25 procedures aliquots, lymph node (LN) station, LN size, and the number of needle aspirations per LN station.
Results: Of the 212 patients who underwent EBUS-TBNA, 200 patients had adequate follow-up information and were included in this analysis. The procedure was performed under conscious sedation in 97 % of patients and 133 patients (67 %) were suspected to have malignancy before the procedure. A total of 690 TBNAs were performed from 294 LN stations. The mean number of LN stations sampled per procedure was 1.47 ± 0.6. The mean number of TBNAs per LN station was 2.35 ± 0.91. The mean number of TBNAs per procedure was 3.45 ± 1.2. The overall sensitivity, specificity, negative predictive value (NPV), and diagnostic accuracy for all procedures were 87.41 % (CI 80.76-91.99), 100 % (CI 93.12-100), 75.36 % (CI 64.04-84.01), and 90.91 % (CI 85.92-94.25), respectively. The NPV increased significantly after the initial 25 procedures and remained high thereafter. EBUS-TBNA was more accurate (96.12 % (CI 91.25-98.33)) with higher NPV (90.74 % (CI 80.09-95.98)) in patients with suspected malignancy compared with patients with suspected benign disease (79.31 % (CI 67.23-87.75), 20 % (7.05-45.19)). Samples from relatively smaller LN (>5 to ≤20 mm) and from all analyzed LN stations were similarly accurate with high sensitivity and NPV.
Conclusions: EBUS-TBNA allows safe real-time sampling of mediastinal and hilar lesions under conscious sedation with high diagnostic accuracy. The NPV is high and increased significantly after the initial 25-50 procedures. This is comparable to available surgical techniques, including mediastinoscopy, when malignancy is suspected. The NPV for specific benign disease remains low in our experience. The diagnostic yield is not affected by the LN station, size, or the number of passes per LN station.