PURPOSE: To test the hypothesis that absence of statistically significant lung nodule enhancement (< or =15 HU) at computed tomography (CT) is strongly predictive of benignity. MATERIALS AND METHODS: Five hundred fifty lung nodules were studied. Of these, 356 met all entrance criteria and had a diagnosis. On nonenhanced, thin-section CT scans, the nodules were solid, 5-40 mm in diameter, relatively spherical, homogeneous, and without calcification or fat. All patients were examined with 3-mm-collimation CT before and after intravenous injection of contrast material. CT scans through the nodule were obtained at 1, 2, 3, and 4 minutes after the onset of injection. Peak net nodule enhancement and time-attenuation curves were analyzed. Seven centers participated. RESULTS: The prevalence of malignancy was 48% (171 of 356 nodules). Malignant neoplasms enhanced (median, 38.1 HU; range, 14.0-165.3 HU) significantly more than granulomas and benign neoplasms (median, 10.0 HU; range, -20.0 to 96.0 HU; P < .001). With 15 HU as the threshold, the sensitivity was 98% (167 of 171 malignant nodules), the specificity was 58% (107 of 185 benign nodules), and the accuracy was 77% (274 of 356 nodules). CONCLUSION: Absence of significant lung nodule enhancement (< or = 15 HU) at CT is strongly predictive of benignity.