TOPLINE

Two MRI-based models — the preoperative and postoperative models of MRI for Very Early Recurrence Prediction (MERP-pre and MERP-post) — predicted very early recurrence within 1 year after resection in patients with single Barcelona Clinic Liver Cancer (BCLC) stage 0/A hepatocellular carcinoma (HCC); these models outperformed other systems. The models also stratified patients into high- and low-risk groups with distinct 1-year recurrence-free survival (RFS) rates.

METHODOLOGY

• Multiparametric MRI is effective for predicting very early recurrence as it thoroughly evaluates tumor morphology, vasculature, function, and metabolism; however, evidence on its effectiveness for predicting early HCC recurrence is limited.
• Researchers conducted a retrospective cohort study across 14 tertiary-care hospitals in China, South Korea, Singapore, France, and the US, including 1851 patients (median age, 57 years) who underwent curative-intent resection for single BCLC stage 0/A HCC between 2011 and 2024 with preoperative contrast-enhanced MRI or CT.
• Investigators developed MERP-pre (preoperative model) and MERP-post (postoperative model) in a derivation cohort of 628 patients from West China Hospital and externally validated the models in 775 patients from 11 hospitals across China, South Korea, and Singapore (Eastern cohort), 178 patients from hospitals in France and the US (Western cohort), and 230 patients who underwent CT imaging (CT cohort).
• The primary endpoint was 1-year RFS, calculated from the date of resection to the first documented tumor recurrence or all-cause death within a year of resection.
• Clinical, laboratory, imaging, surgical, and histopathologic parameters were recorded. The biological underpinnings of the models were explored using whole-exome sequencing data from 146 patients in the derivation cohort and RNA sequencing data from 40 patients in The Cancer Genome Atlas Liver Hepatocellular Carcinoma dataset, with patients stratified into low- and high-risk groups using thresholds from the models.

TAKEAWAY

• In the MERP-pre model, the natural logarithm of alpha-fetoprotein (P = .001), tumor size (P = .005), portal venous phase peritumoral hypoenhancement (P = .011), and blood products in the mass (P = .022) predicted 1-year RFS, and in the MERP-post model, tumor size was replaced by microvascular invasion among the predictors.
• MERP-pre model demonstrated C-indexes ranging from 0.685 to 0.742 across test cohorts, consistently outperforming the BCLC, American Joint Committee on Cancer Tumor-Node-Metastasis, and modified Union for International Cancer Control systems (C-indexes, 0.524-0.551; P < .001 for all).
• Both the MERP models consistently stratified patients into high- and low-risk groups with distinct 1-year RFS rates across all cohorts (all P < .001). RFS in general and overall survival also differed significantly between the high- and low-risk groups according to MERP-pre and MERP-post models.
• MERP-pre model improved risk reclassification when integrated with major staging systems and microvascular invasion.

IN PRACTICE

“[T]he MERP-post model may be applied to direct individualized follow-up strategies for high-risk patients since these patients may benefit from more intensified monitoring with advanced imaging (eg, contrast-enhanced MRI > CT > US). Beyond recurrence prevention, early identification of high-risk patients can also facilitate realistic expectation setting, reinforce adherence to postoperative care, and support proactive measures such as antiviral therapy, alcohol cessation, metabolic control to curb de novo tumorigenesis, and further clinical trials for adjuvant systemic treatment,” the authors wrote.

SOURCE

The study was led by Hong Wei, Department of Radiology, Functional and Molecular Imaging Key Laboratory of Sichuan Province, West China Hospital, Sichuan University in Chengdu, China. It was published online in the Journal of Hepatology.

LIMITATIONS

The study took place for over more than 10 years in different locations, which may have caused differences in surgical methods and imaging techniques that could affect the results. The retrospective nature of the study may have led to population bias. The usefulness of MERP models as tools for decision-making remains hypothetical owing to the observational nature of the study, and therefore, the models cannot yet be recommended for treatment allocation.

DISCLOSURES

This study received support from the National Natural Science Foundation of China and the Sichuan Science and Technology Program. Some authors disclosed research collaboration or serving as an advisory member and receiving support for attending meetings, research grants, honoraria, educational fees, or consulting fees paid to self or institution from various pharmaceutical and biotechnology companies.

This article was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.