Mediastinal and Hilar Lymph Node Measurements. Comparison of Multidetector-Row Computed Tomography and Endobronchial Ultrasound

Abstract

Rationale: Multidetector-row chest computed tomography scan is a common initial imaging modality and endobronchial ultrasound is a minimally invasive diagnostic tool used to evaluate enlarged lymph nodes, but comparisons of imaging results are lacking.

Objectives: To determine the size of thoracic lymph nodes and the strength of agreement between each measurement from coronal plane computed tomography and static endobronchial ultrasound images.

Methods: A retrospective review of consecutive patients who underwent endobronchial ultrasound-transbronchial needle aspiration of their lymph nodes because of clinical suspicion of benign or malignant thoracic disease.

Measurements and main results: One hundred and twenty-four lymph nodes from the mediastinal (74.2%) and hilar (25.8%) stations were measured in 59 patients (mean age, 64.5 yr; 33 males). The mean (standard deviation) short-axis diameter on computed tomography was 14.1 (6.7) mm compared with 12.6 (6.6) mm on endobronchial ultrasound. Benign lymph nodes (n = 42) were larger on computed tomography than on endobronchial ultrasound (14.1 [6.2] vs. 11.5 [6.2] mm). Malignant lymph nodes (n = 35) were larger on endobronchial ultrasound than on computed tomography (17.3 [6.4] vs. 16.2 [6.7] mm). Sixty-five percent of the lymph nodes that were initially interpreted as not enlarged on axial computed tomography images measured greater than 10 mm on each imaging modality (12.5 [5.9] mm on computed tomography and 10.5 [5.6] mm on endobronchial ultrasound) and 24% of the sampled lymph nodes from this group contained malignant cells. Random-effects maximal likelihood linear regression showed a statistically significant difference between endobronchial ultrasound and the computed tomography method for measuring short-axis diameter in all 124 lymph nodes. There was a weak agreement (intraclass correlation, rho: 0.44 [95% confidence interval, 0.31-0.59]) between short-axis diameter measurements from each imaging modality.

Conclusions: Our single-center study shows that there was poor correlation between computed tomography and endobronchial ultrasound for the measurement of mediastinal and hilar lymph nodes. Malignant cells were recovered by ultrasound-guided needle aspiration from a substantial fraction of lymph nodes that were initially interpreted as normal in size. If these findings are confirmed, new criteria may be needed for lymph node measurement on computed tomography that will guide selection of lymph nodes for endobronchial ultrasound-transbronchial needle aspiration.

Keywords: bronchoscopy; interventional ultrasonography; lymph nodes; multidetector-row computed tomography.