The role of computed tomographic scanning in diagnosing mediastinal node involvement in non-small cell lung cancer

The reliability of computed tomographic scanning in evaluating mediastinal node involvement is controversial because of the high false result rate. We attempted to identify significant factors responsible for false-positive and false-negative scans.

From August 1992 through April 1997, 401 patients with lung cancer who underwent major lung resection and systematic lymph node dissection were enrolled in this study. We retrospectively examined mediastinal node size, tumor location, maximum tumor dimension, the presence or absence of obstructive pneumonia, atelectasis, pulmonary fibrosis, and lymph node calcification on contrast-enhanced computed tomographic scans. We identified clinical and radiologic factors responsible for the false results by using univariate and multivariable analysis.

Central tumor location proved to be a significant factor of false-positive scans. Elevated carcinoembryonic antigen level and larger tumor dimension were significant factors of false-negative scans. In patients with a peripheral tumor smaller than 40 mm and normal levels of serum carcinoembryonic antigen, sensitivity, specificity, positive predictive value, and negative predictive value were 6%, 93%, 8%, and 90%, respectively. The reliability of computed tomographic scanning in this low-risk subgroup was high in detecting N0-1 disease but low in diagnosing N2 disease.

It is not possible to accurately diagnose N2 disease by using lymph node size on computed tomographic scanning alone, especially in patients with a central tumor, an elevated serum carcinoembryonic antigen level, or a tumor of 40 mm or larger. A preoperative invasive staging procedure is indicated in these populations and may not be indicated in the population with normal computed tomographic scan results without any of these risk factors.