Influence of the Learning Effect on the Diagnostic Yield of Endobronchial Ultrasound-guided Transbronchial Needle Aspiration of Mediastinal and Hilar Lymph Nodes.

Background: The diagnostic yield of conventional transbronchial needle aspiration (TBNA) is characterized by a learning effect. The aim of this retrospective study was to verify whether a learning curve similarly affected the yield of endobronchial ultrasound-guided (EBUS)-TBNA. To this end, we evaluated the sensitivity and diagnostic accuracy of EBUS-TBNA during the first 3 years of activity.

Methods: EBUS-TBNA was performed by 2 operators with no previous experience in this technique. Cytologic samples were obtained from mediastinal and hilar lymph nodes enlarged at a chest computed tomography scan and/or with increased fluorodeoxyglucose uptake at computed tomography/positron emission tomography scan in patients with suspected lung cancer. The cytologic diagnosis of EBUS-TBNA samples has been compared with the final diagnosis obtained from further diagnostic procedures, surgery, or clinical-radiologic follow-up.

Results: From October 2012 to October 2015, we collected 408 EBUS-TBNA cytologic samples from 313 patients: 223 samples were positive for metastatic involvement and 185 were nonmetastatic. The latter included 137 true-negative and 48 false-negative results. The final diagnosis comprised 271 metastatic and 137 nonmetastatic lymph nodes. The overall sensitivity for cancer was 82% and diagnostic accuracy was 88%. Sensitivity and accuracy per year were as follows: first year, 78% and 82% in 90 nodal samples; second year, 83% and 89% in 144 nodal samples; third year, 85% and 91% in 174 nodal samples.

Conclusions: EBUS-TBNA can be considered as a reliable tool even if performed by operators without previous experience in this procedure, and the diagnostic yield continues to increase progressively over a long time.