Objectives: To identify clinical or radiologic predictors of mediastinal ly
mph node involvement in patients with non-small cell lung cancer, and to de
fine the indications of preoperative mediastinoscopy.
Methods: From August 1992 through April 1997, 387 patients with lung cancer
(290 adenocarcinoma and 97 squamous cell carcinoma) underwent surgical res
ection, We retrospectively measured all mediastinal lymph node sizes both i
n the shortest and longest axes on contrast-enhanced CT scan to determine t
he optimal size criteria. Using multivariate logistic regression analysis,
we identified clinical or radiologic predictors of N2 disease.
Results: We could not identify reliable size criteria for nodal involvement
. We found two significant predictive factors of N2 disease on the basis of
multivariable analysis: maximum tumor dimension and serum carcinoembryonic
antigen (CEA) concentrations. The lymph node size did not prove to be a si
gnificant factor. Among 50 patients with serum CEA concentrations < 5.0 mg/
mL and maximum turner dimension < 20 mm, pathologic N2 disease was proven o
nly in three patients (6%), regardless of the lymph node size on CT scan. A
mong 140 patients with serum CEA concentrations greater than or equal to 5.
0 ng/mL and maximum tumor dimension greater than or equal to 20 mm, approxi
mately one third (n = 46) showed N2 disease.
Conclusion: Serum CEA concentrations and maximum tumor dimension were more
valuable in predicting N2 disease than the lymph node size on CT scan. Medi
astinoscopy is indicated in patients with serum CEA concentrations < 5.0 ng
/mL and maximum tumor dimension greater than or equal to 20 mm, and not ind
icated in patients with serum CEA concentrations < 5.0 ng/mL and maximum tu
mor dimension < 20 mm.