Three MRI features that flag a small HCC likely to recur

Design

Retrospective diagnostic model development with internal validation cohort and survival analysis

Population

315 patients with resected hepatocellular carcinoma 3 cm or less and preoperative Gd-EOB-DTPA-enhanced MRI

Primary outcome

Prediction of macrotrabecular-massive subtype and 3-year recurrence

Effect

Validation AUC 0.75 (95% CI 0.65 to 0.84); two or three features present, recurrence HR 2.55 (1.24 to 5.22)

Three hundred and fifteen patients with surgically confirmed hepatocellular carcinoma of 3 cm or less had preoperative Gd-EOB-DTPA-enhanced MRI reviewed against pathology. Three features independently predicted the macrotrabecular-massive subtype: a hypovascular component of 20 per cent or more, an intratumoral artery, and an incomplete capsule. Combined into a model, they gave an AUC of 0.75 in validation, with sensitivity 77.4 per cent and specificity 72.4 per cent.

The outcome link is what makes this reportable. Patients with two or three of the features had a higher three-year recurrence risk after hepatectomy (HR 2.55), and recurrence-free survival separated in both cohorts.

So the practical change is to the report, not the protocol. These three features are assessable on studies already being done for small HCC, and naming them tells the surgical team which patient warrants closer surveillance, or a wider margin discussion, before the operation rather than after the histology.

• Report hypovascular component, intratumoral artery and capsule integrity explicitly in every small HCC staging study.
• State how many of the three are present; the risk was tied to two or more.
• Raise the finding at the multidisciplinary meeting before resection, when it can still affect the plan.
• Do not use it to select against resection — this predicts recurrence, it does not compare treatments.
• Note that hepatobiliary-phase contrast is required; the features were defined on Gd-EOB-DTPA studies.

Why it matters

It moves a prognostic subtype from the pathology report, where it arrives too late, to the preoperative scan.

Don't overread it

Retrospective, single-series development and validation — the model has not been tested prospectively or externally.

The statistics, in plain English

An AUC of 0.75 with sensitivity 77 per cent and specificity 72 per cent is a useful signal and a poor rule — roughly a quarter of subtypes are missed and a quarter of calls are wrong. The hazard ratio of 2.55 has a wide interval (1.24 to 5.22), which keeps the direction but not the magnitude. The model was built and validated within one retrospective series, so its performance elsewhere is unknown.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.