Hepatocellular carcinoma (HCC) represents around 80% of primary liver cancer cases and is the third leading cause of cancer-related mortality globally. Despite its diverse etiology, early detection and curative treatments (CT) such as surgical resection (SR) and percutaneous ablation (PA) offer promising survival benefits for early-stage HCC patients. However, disease recurrence (DR) remains a critical issue, with rates reaching up to 70% within five years. Real-world data from early HCC patients in France are sparse, with currently no consensus on the definition of high-risk (HR) recurrence following CT. We aim to: (i) describe the characteristics of early HCC patients who received their CT in two regional centers in France and (ii) explore and compare three different approaches to define high-risk DR in this cohort.

A retrospective observational study (OPRAH) was conducted based on the review of records from patients with early-stage HCC who received their first CT (either SR or PA) between 2017 and 2021 in two French regional centers. Patients were included at the date of their first CT and followed for up to two years, until December 2023 at the latest. Data collected at baseline included socio-demographic, clinical and biological characteristics, etiology and comorbidities.Two definitions of HR of recurrence were used, one from the IMbrave-050 clinical trial (IM) and another provided by the OPRAH's study Scientific Committee recommendations (SC). Both definitions were based on variables such as tumor type (infiltrative vs nodular) and size, the number and location of nodules, type of vascular invasion (microvascular vs macrovascular), tumor differentiation grade and Alfa-Fetoprotein levels. We assessed the performance of the two HR definitions (sensitivity, specificity and overall accuracy) in classifying the patients in the study with respect to their DR status. A data-driven definition using machine learning methods will be developed as a perspective to contrast with the existing HR definitions. We estimated median and two-year recurrence-free survival (RFS), using Kaplan-Meier curves, overall and stratified by HR status.

Of 371 patients included with a mean (SD) age of 67.3 years (10.4), 317 (85.4%) were male. For their first CT, 237 (63.9%) underwent PA, 117 (31.5%) SR, and 17 (4.6%) received both. 359 patients (96.8%) had pre-existing comorbidities at the time of their first CT. Metabolic etiologies of HCC were identified in 68 (18.3%) patients while 85 (22.9%) were linked to viral etiologies (hepatitis B or C). The median (IQR) follow-up was 24.0 (13.6) months and 173 (43.6%) patients experienced DR within two years. In the time-to-event analysis, the RFS survival at two years was 36.5% (95% CI: 31.4%- 42.5%). Among the 371 patients, 206 (55.5%), 195 (52.6%), 250 (67.4%) and 151 patients (40.7%) were classified as HR of DR according to the IM, SC, at least one (IM or SC) or both definitions (IM and SC), respectively. Performance metrics (sensitivity, specificity and overall accuracy) were calculated on 335 patients with known recurrence status. For identifying DR within 2 years, these were 58%,48%,53% (IM), 55%,54%,55% (SC), 69%,37%,54% (at least one) and 44%,65%,54% (both).

Neither of the two compared approaches for HR of DR in early-stage HCC clearly outperformed the other. Results from the data-driven definition, to be presented, will explore its alignment with the IM and SC definitions or its potential to complement them.