Role of computed tomography in predicting adrenal adenomas with cortisol hypersecretion.

Abstract

To investigate performance of adrenal CT-derived multivariate prediction models in differentiating adenomas with cortisol hypersecretion from the other subtypes. This retrospective study included 127 patients who underwent adrenal CT and had a surgically proven adrenal adenoma. Adenoma subtypes were defined according to biochemical test results: Group A, overt cortisol hypersecretion; Group B, mild cortisol hypersecretion; Group C, aldosterone hypersecretion; and Group D, non-function. Two independent readers analyzed size, attenuation, and washout properties of adenomas, and performed quantitative and qualitative analyses for assessing contralateral adrenal atrophy. Actual and internally validated areas under the curves (AUCs) of adrenal CT-derived multivariate prediction models for differentiating adenomas with cortisol hypersecretion from the other subtypes were assessed. In differentiating Group A from the other groups, the actual and internally validated AUCs of the prediction model were 0.856 (95% confidence interval [CI]: 0.786, 0.926) and 0.847 (95% CI: 0.695, 0.999) for Reader 1, respectively, and 0.901 (95% CI: 0.845, 0.956) and 0.897 (95% CI: 0.783, 1.000) for Reader 2, respectively. In differentiating Group B from groups C and D, the actual and internally validated AUCs of the prediction model were 0.777 (95% CI: 0.687, 0.866) and 0.760 (95% CI: 0.552, 0.969) for Reader 1, respectively, and 0.783 (95% CI: 0.690, 0.875) and 0.765 (95% CI: 0.553, 0.977) for Reader 2, respectively. Adrenal CT may be useful in differentiating adenomas with cortisol hypersecretion from the other subtypes. Adrenal CT may benefit in adrenal adenoma subtyping.