Variability in the detection of enlarged mediastinal lymph nodes in staging lung cancer: a comparison of contrast-enhanced and unenhanced CT.

Abstract

Because CT protocols for staging lung cancer vary and little information exists regarding the diagnostic importance of using i.v. contrast material, our intent was to evaluate intra- and interobserver agreement in the detection of enlarged mediastinal lymph nodes, comparing i.v. contrast-enhanced and unenhanced CT. Fifty patients with known or suspected bronchogenic carcinoma underwent unenhanced thoracic CT followed by contrast-enhanced CT. Three observers noted enlarged lymph nodes (> 10 mm in the short axis) and assigned the enlarged nodes to American Thoracic Society nodal station designations. Enlarged lymph nodes were grouped two ways: by assigning the exact number of enlarged lymph nodes found (zero, one, two, three, four or more), and by assigning whether at least one, or no, enlarged mediastinal lymph nodes were found at a station ("one or none"). Agreement levels were determined for inter- and intraobserver interpretations using weighted kappa statistics and the McNemar test. The number of enlarged lymph nodes with enhanced CT was 11% higher than on unenhanced studies (418 versus 377; p = .044). Numbers of enlarged lymph nodes were different for five stations; however, the numbers were small except for the right upper paratracheal station (2R) (contrast-enhanced, 68 enlarged lymph nodes; unenhanced, 44 enlarged lymph nodes; p = .014). With regard to all stations together, intraobserver agreement between contrast-enhanced and unenhanced studies was almost perfect (kappa range, .85-.94), and no difference was found for any observer in the proportion of patients with at least one enlarged lymph node. Interobserver agreement was substantial or almost perfect for the total number of enlarged lymph nodes. For specific stations, the lowest kappa value was .48 at 2R. One observer reported more patients with at least one enlarged lymph node with contrast enhancement at station 2R (p = .031). Greater agreement existed between two observers at station 2R with contrast enhancement versus no enhancement (kappa = .85 versus .48; p = .02). Conclusions matched, and calculations of estimated kappa values gave similar results for determination of the specific number of enlarged lymph nodes at a station and the "one or none" category. We found high agreement for intra- and interobserver interpretations for contrast-enhanced and unenhanced CT, although contrast-enhanced CT revealed more enlarged lymph nodes, especially at station 2R.