Background. To determine the relative utility of positron emission tomograp
hy (PET), computed tomography (CT), and magnetic resonance imaging with Com
bidex (MRI-C) in the non-invasive staging of non-small cell lung cancer (NS
CLC) mediastinal lymph nodes (MLN), we compared the three tests' individual
performance with surgical mediastinal sampling. In contrast to prior studi
es, cytology was not used.
Methods. The MLN were evaluated using PET and CT in 64 NSCLC patients. MRI-
C was performed in 9 of these patients. MLN with a PET standard uptake valu
e greater than or equal to 2.5, or greater than 1 cm in the short axis by C
T or lack of MRI-C signal change were considered positive for metastatic di
sease. All MLN were sampled and subjected to standard pathologic analysis.
PET, CT, and MRI-C scans were interpreted blinded to the histopathological
results. Sensitivity, specificity, and accuracy for each scan type to appro
priately stage MLN was determined using pathologic results as the standard.
Results. Thirty patients had stage I disease, 8 stage II, 9 stage IIIA, 7 s
tage IIIB, and 10 stage IV. Two-hundred-and-thirty MLN were sampled. Sixtee
n patients had metastatic mediastinal disease. Compared to the pathological
results, PET, CT, and MRI-C had a sensitivity, specificity, and accuracy o
f 70%, 86%, 84%; 65%, 79%, 76%; 86%, 82%, and 83%, respectively. PET and MR
T-C were statistically more accurate than CT (p < 0.001). In cases where PE
T and CT did not identify MLN involvement with NSCLC, 8% (2/25) were pathol
ogically positive.
Conclusions. FET and MRI-C are statistically more accurate than CT. However
, the differences are small and may not be clinically relevant. No techniqu
e was sensitive or specific enough to change the current recommendation to
perform mediastinoscopy for MLN staging in NSCLC. (Ann Thorac Surg 1999;68:
1022-8) (C) 1999 by The Society of Thoracic Surgeons.