MRI-BASED models demonstrated improved prediction of very early hepatocellular carcinoma (HCC) recurrence following surgical resection, outperforming established prognostic systems across international patient cohorts.

The findings suggest that combining imaging and clinical information could help identify patients at increased risk of recurrence within one year of surgery.

Developing MRI-Based Prediction Models

The international, retrospective cohort study included 1,851 patients who underwent curative-intent resection for a single Barcelona Clinic Liver Cancer (BCLC) stage 0/A HCC across 14 hospitals. Researchers developed two MRI for Very Early Recurrence Prediction (MERP) models using data from a derivation cohort of 628 patients.

The preoperative model, MERP-pre, incorporated serum alpha-fetoprotein (AFP), tumour size, portal venous phase peritumoral hypoenhancement, and intratumoral blood products. The postoperative model, MERP-post, included AFP, peritumoral hypoenhancement, intratumoral blood products, and microvascular invasion.

The models were externally validated in Eastern and Western cohorts comprising 775 and 178 patients, respectively, and further evaluated in a CT cohort of 230 patients.

Models Outperform Established Prognostic Systems

Both MERP models outperformed major staging systems and the IMbrave050 criteria across the test cohorts. Their concordance indices ranged from 0.685–0.790, compared with 0.524–0.682 for the established prognostic systems. The models also improved risk reclassification, with net reclassification improvement values ranging from 0.004–0.372.

Patients were consistently separated into high- and low-risk groups with significantly different one-year recurrence-free survival (RFS) outcomes across all cohorts. In the CT cohort, MERP-post achieved a concordance index of 0.790, compared with 0.742 for MERP-pre.

Imaging Features Reflect Tumour Biology

Researchers also investigated the biological characteristics associated with MERP-pre risk groups using whole-exome and RNA sequencing. Tumours classified as low risk demonstrated features associated with the non-proliferative class, increased lipid catabolism, and an immunoactive tumour microenvironment.

Researchers concluded that the MERP models could provide robust and generalisable tools for predicting very early HCC recurrence and help refine risk-stratified management strategies. However, the findings were derived from patients undergoing surgical resection and cannot be extrapolated to the broader BCLC 0/A population receiving non-surgical treatments.

Reference:

Wei H et al. MRI-based prediction of very early recurrence within 1 year after resection of HCC: An international cohort study. J Hepatol. 2026;85(5): 912-29.