Can virtual noncontrast computed tomography improve the diagnostic uncertainty of adrenal incidentalomas?

Abstract

Although most adrenal incidentalomas are benign, many are identified by single-phase contrast-enhanced computed tomography, which is unreliable for excluding malignancy. Virtual noncontrast computed tomography is a newer modality with the potential to better characterize adrenal nodules. Virtual noncontrast computed tomography of adrenal nodules with available reference standard of true noncontrast computed tomography were identified (2016-2024). Images were evaluated for nodule characteristics including Hounsfield unit attenuation and variability. Nodules were classified as benign (≤10 Hounsfield units) or indeterminate/suspicious (>10 Hounsfield units) by true noncontrast computed tomography. Hounsfield units were compared between virtual noncontrast computed tomography and true noncontrast computed tomography. Variability in attenuation measurements was compared by evaluating Hounsfield unit differences 1 slice up and down from the chosen mid-depth image. Receiver operating characteristic analysis was used to define optimal virtual noncontrast computed tomography accuracy thresholds. After excluding 5 adrenal nodules due to suboptimal imaging, 67 nodules were identified. Based on true noncontrast computed tomography Hounsfield units, 23 nodules (34.3%) were benign, and 44 (65.7%) were indeterminate/suspicious. Hounsfield unit measurements for each nodule exhibited wide variability by both virtual noncontrast computed tomography and true noncontrast computed tomography. Virtual noncontrast computed tomography and true noncontrast computed tomography were significantly correlated with moderate effect size (Pearson coefficient 0.69, P < .001). Conflicting impressions occurred for 6 nodules (9.0%). Overall, virtual noncontrast computed tomography exhibited outstanding discrimination between benign and indeterminate/suspicious nodules (area under the curve 0.94). Maintaining a threshold of ≤10 Hounsfield units achieved 93% sensitivity, 76% specificity, and 84% negative predictive value, whereas ≤7 Hounsfield units achieved 100% negative predictive value. The functional utility of virtual noncontrast computed tomography as a rule-out test applied to 16% of nodules. Despite wide variability in Hounsfield unit measurements, adrenal nodules are well defined by both virtual noncontrast computed tomography and true noncontrast computed tomography. Well-reconstructed virtual noncontrast computed tomography images can accurately rule out malignancy in selected patients, potentially obviating the need for additional imaging.