Objective: To investigate whether the diagnostic accuracy of unenhanced computed tomography (CT) regarding the differentiation of adrenal adenomas from adrenal metastases is increased by applying a combination of morphologic criteria instead of only measuring the density values of the tumor.
Patients and Methods: Unenhanced CT scans of 56 patients with an adrenal mass and a history of an extra-adrenal malignancy were analyzed for size, attenuation, contour, and structure characteristics of the adrenal tumor. Coefficients yielded by multiple logistic regression analysis were used for the construction of an additive total score (score S) that included several diagnostic criteria. The reliability of the total score and all parameter combinations was tested by receiver operating characteristic (ROC) analysis. The nature of the adrenal lesion was determined by follow-up CT (40 patients), percutaneous biopsy (15 patients), or surgery (1 patient). Twenty-four of the neoplasms were adenomas, and 32 were found to be metastases.
Results: The score of the combined CT parameters showed the largest area under the ROC curve. The highest predictive power indicated by the model was calculated at a cutoff value of 7.05, with a sensitivity of 100% and a specificity of 96.8% for the detection of metastases. At 6.85 points as the cutoff value, the scoring system still maintained a sensitivity of 95.8% and a specificity of 96.9%.
Conclusion: The differentiation between adrenal adenomas and metastases is improved by applying our scoring system compared with any single parameter alone. The total score is obtained by adding 10% of the density values to the size in centimeters, plus 2 if the contour of the lesion is blurred and plus 1 if the structure is inhomogeneous. By setting the threshold at 7 points, all but 1 lesion were classified correctly.
From the *Department of Radiology, University of Giessen, Giessen, Germany, †Institute of Mathematics, University of Giessen, Giessen, Germany, ‡Department of Diagnostic and Interventional Radiology, University of Rostock, Rostock, Germany, and §Department of Biostatistics, University of Rostock, Rostock, Germany.
Reprints: Dr. Hubert Gufler, Department of Diagnostic Radiology, University of Giessen, Klinik Strasse 36, D-35385 Giessen, Germany (e-mail: Hubert.Gufler@radiol.med.uni-giessen.de).