Noninvasive Staging of Non-small Cell Lung Cancer*

To determine the test performance characteristics of CT scanning, positron emission tomography (PET) scanning, MRI, and endoscopic ultrasound (EUS) for staging the mediastinum, and to evaluate the accuracy of the clinical evaluation (ie, symptoms, physical findings, or routine blood test results) for predicting metastatic disease in patients in whom non-small cell lung cancer or small cell lung cancer is diagnosed.Systematic searches of MEDLINE, HealthStar, and Cochrane Library databases to July 2001, and of print bibliographies. Studies evaluating the staging results of CT scanning, PET scanning, MRI, or EUS, with either tissue histologic confirmation or long-term clinical follow-up, were included. The performance of the clinical evaluation was compared against the results of brain and abdominal CT scans and radionuclide bone scans.Pooled sensitivities and specificities for staging the mediastinum were as follows: for CT scanning: sensitivity, 0.57 (95% confidence interval [CI], 0.49 to 0.66); specificity, 0.82 (95% CI, 0.77 to 0.86); for PET scanning: sensitivity, 0.84 (95% CI, 0.78 to 0.89); specificity, 0.89 (95% CI, 0.83 to 0.93); and for EUS: sensitivity, 0.78 (95% CI, 0.61 to 0.89); specificity, 0.71 (95% CI, 0.56 to 0.82). For the evaluation of brain metastases, the summary estimate of the negative predictive value (NPV) of the clinical neurologic evaluation was 0.94 (95% CI, 0.91 to 0.96). For detecting adrenal and/or liver metastases, the summary NPV of the clinical evaluation was 0.95 (95% CI, 0.93 to 0.96), and for detecting bone metastases, it was 0.90 (95% CI, 0.86 to 0.93).PET scanning is more accurate than CT scanning or EUS for detecting mediastinal metastases. The NPVs of the clinical evaluations for brain, abdominal, and bone metastases are > or = 90%, suggesting that routinely imaging asymptomatic lung cancer patients may not be necessary. However, more definitive prospective studies that better define the patient population and improved reference standards are necessary to more accurately assess the true NPV of the clinical evaluation.