Predictive model of pheochromocytoma based on the imaging features of the adrenal tumours

Abstract

The purpose of our study was to develop a predictive model to rule out pheochromocytoma among adrenal tumours, based on unenhanced computed tomography (CT) and/or magnetic resonance imaging (MRI) features. We performed a retrospective multicentre study of 1131 patients presenting with adrenal lesions including 163 subjects with histological confirmation of pheochromocytoma (PHEO), and 968 patients showing no clinical suspicion of pheochromocytoma in whom plasma and/or urinary metanephrines and/or catecholamines were within reference ranges (non-PHEO). We found that tumour size was significantly larger in PHEO than non-PHEO lesions (44.3 +/- 33.2 versus 20.6 +/- 9.2 mm respectively; P < 0.001). Mean unenhanced CT attenuation was higher in PHEO (52.4 +/- 43.1 versus 4.7 +/- 17.9HU; P < 0.001). High lipid content in CT was more frequent among non-PHEO (83.6% versus 3.8% respectively; P < 0.001); and this feature alone had 83.6% sensitivity and 96.2% specificity to rule out pheochromocytoma with an area under the receiver operating characteristics curve (AUC-ROC) of 0.899. The combination of high lipid content and tumour size improved the diagnostic accuracy (AUC-ROC 0.961, sensitivity 88.1% and specificity 92.3%). The probability of having a pheochromocytoma was 0.1% for adrenal lesions smaller than 20 mm showing high lipid content in CT. Ninety percent of non-PHEO presented loss of signal in the out of phase MRI sequence compared to 39.0% of PHEO (P < 0.001), but the specificity of this feature for the diagnosis of non-PHEO lesions low. In conclusion, our study suggests that sparing biochemical screening for pheochromocytoma might be reasonable in patients with adrenal lesions smaller than 20 mm showing high lipid content in the CT scan, if there are no typical signs and symptoms of pheochromocytoma.