Fluoro-deoxy-glucose positron emission tomography for evaluation of indeterminate lung nodules: assigning a probability of malignancy may be preferable to binary readings

Abstract

To assess the diagnostic value of fluorine-18 fluoro-2-deoxy-D-glucose positron emission tomography (FDG-PET) using standard uptake values (SUV) in the differential diagnoses of indeterminate pulmonary nodules. Specifically, we assessed the probability of malignancy for various SUV ranges, and compared the diagnostic efficacy of SUV with and without correction for partial volume effects on the basis of lesion size. The FDG-PET scans performed on 158 patients with biopsy-proven pulmonary lesions seen on computed tomography (CT) scan were retrospectively reviewed. Histopathological confirmation was obtained to establish the diagnosis of the lesions. A region of interest (ROI) was drawn for each lesion, and FDG uptake was quantified (SUV(raw)). The SUV(raw) values were normalized for the "size" of the pulmonary lesions measured on CT (SUV(size)). Sensitivity and specificity of FDG-PET for pulmonary lesions <2 cm in diameter or > or =2 cm in diameter were determined at SUV cutoff values of 2.5. The areas under the receiver-operating characteristic (ROC) curve for SUV(raw) and SUV(size) regarding the presence of malignancy were compared for statistical differences. The frequency of malignant lesions for each range of SUVs was obtained to produce the probability of cancer (POC). The mean SUV(raw) was 3.17 +/- 2.76 and 9.18 +/- 6.72 for benign and malignant lesions, respectively. When a SUV(raw) value of 2.5 was used as a cutoff, sensitivity and specificity were 89% and 51%, respectively, for all lesion sizes. The sensitivity and specificity at a cutoff SUV(raw) of 2.5 for lesions less than 2 cm in diameter were 75% and 72%, respectively, and 92% and 41% for lesions 2 cm or greater, respectively. The sensitivity and specificity at a cutoff SUV(size) of 2.5 were 88% and 42%, respectively. The area under the ROC curves for SUV(raw) and SUV(size) was 0.816 and 0.743, respectively (P value 0.034). When the SUV(raw) was divided into three groups, the probability of malignancy was 26% when the SUV(raw) was <2, 57% for 2 < or = SUV(raw) < 6, and 89% for SUV(raw) > or = 6. The FDG-PET is a reasonably accurate and useful tool for characterizing the nature of indeterminate pulmonary lesions, although the specificity was not as high as that reported in the literature, probably owing in part to our patient population and selection bias. Our data suggest that reporting the results of PET studies as a probability rather than as positive or negative for malignancy would be more useful for further management decision making. Correction of SUVs for tumor size did not improve accuracy.