Noncontrast abbreviated MRI demonstrates superior diagnostic accuracy compared to ultrasound for hepatocellular carcinoma detection: Interim results from a prospective surveillance trial

BACKGROUND AND AIMS: To prospectively evaluate the diagnostic performance of a rapid abbreviated noncontrast MRI (AMRI) protocol compared to ultrasound (US) for HCC surveillance in a high-risk population with cirrhosis. APPROACH AND RESULTS: This prospective, single-center, diagnostic accuracy study (ClinicalTrials.gov: NCT05716620) enrolled patients with cirrhosis and annual HCC risk >5%. Participants underwent paired screening with US and noncontrast AMRI across 2 rounds, 6 months apart. Patients with positive findings on either imaging modality or clinical suspicion of HCC underwent multiphasic contrast-enhanced MRI (CE-MRI) as the reference standard. The primary outcome was the HCC detection rate (per-patient sensitivity) comparing AMRI and US. In 614 paired screening examinations across 404 patients, 97 underwent CE-MRI (based on positive screening), identifying 39 HCCs in 37 patients. AMRI demonstrated significantly superior sensitivity (94.6% [95% CI: 83.3-98.9] vs. 51.4% [95% CI: 34.7-67.8]; p <0.001) and specificity (96.6% [95% CI: 88.9-99.5] vs. 69.5% [95% CI: 55.8-80.8]; p <0.001). AUROC was 0.956 [95% CI: 0.913-0.997] vs. 0.604 [95% CI: 0.469-0.738] ( p <0.001). In the per-lesion analysis, AMRI detected 37 of 39 lesions (94.9%) versus US 20 of 39 (51.3%). Of HCCs detected by AMRI, 97.3% were early-stage Barcelona Clinic Liver Cancer Staging System 0 or A]. Interobserver agreement was "almost perfect" for AMRI (κ=0.929) versus "moderate" for US (κ=0.631). CONCLUSIONS: A rapid, noncontrast AMRI protocol shows superior per-patient sensitivity compared to US for HCC detection in patients with cirrhosis under surveillance. While these diagnostic findings are encouraging, prospective trials evaluating patient-level outcomes are essential before definitive guideline recommendations can be made.