Background Metastases to the liver are often the first finding in patients with uveal melanoma with extraocular disease, but little has been published on the utility of staging MRI at initial diagnosis. We aimed to evaluate the proportion of abnormal hepatic findings on baseline MRI and accuracy of MRI in patients with newly diagnosed uveal melanoma.

Methods This is a single-centre, retrospective, institutional review board-approved study of 145 consecutive patients diagnosed with uveal melanoma, at Memorial Sloan Kettering Cancer Center between 2004 and 2016, who had staging MRI within 1  month of diagnosis. Scans were classified as normal or abnormal, and further distinguished as abnormal non-metastatic, uncharacterisable lesions and suspicious for metastasis. Where available, follow-up MRI (at ~1  year) or biopsies were reviewed.

Results MRI in 145 patients revealed 62% (90) with abnormal hepatic findings; out of these 87% (78) had non-metastatic benign findings, 6.7% (6) had unclassifiable lesions and 6.7% (6) were suspicious for metastasis (6). Abnormal non-metastatic findings included 72 focal (36 solitary and 36 multiple) and 12 diffuse lesions. Lesions suspicious for metastases were found in 6 of 145 patients (4%), despite normal liver function tests. Of these, five had confirmed liver metastases and one patient had a stable, presumed non-metastatic lesion on follow-up. In this study, the sensitivity and specificity of staging MRI for all findings were 83.3% (95% CI 35.9 to 99.6) and 99.0% (95% CI 94.3 to 99.9), respectively.

Conclusion Staging MRI of patients with newly diagnosed uveal melanoma accurately identified early metastases. Furthermore, imaging revealed hepatic abnormalities in the majority of patients, although as expected few of these represented metastatic disease.