Background: In the United States, patients with early-stage hepatocellular carcinoma (HCC) listed for liver transplant (LT) receive MELD exception points on the waiting list, as they often do not generate a laboratory MELD score to be competitive for LT. Despite the heterogeneity in dropout risk and LT urgency, the 20% of candidates on the waiting list with HCC receive uniform MELD exception points after a mandatory 6-month waiting period (MMAT -3). In contrast, laboratory MELD prioritizes patients by urgency or “sickest first” as outlined by the Final Rule. Previous HCC risk scores predate the 6-month waiting period, MMAT-3, HCV therapeutic advances, and both MELD-Na and MELD 3.0.
Aims: We propose to risk stratify LT candidates with HCC using a simple and parsimonious model using objective variables available at the time of listing.
Methods: We analyzed data from 7143 adult patients with stage T2 HCC added to the LT waitlist between 10/2015 and 06/2022 in the OPTN/UNOS database. The dataset was divided into 70% training and 30% validation datasets. Cumulative probability of dropout after 1 year was estimated using Kaplan-Meier methods, with censoring at LT. Multivariable Cox proportional hazards analysis was used to identify predictors of waitlist dropout, which were combined into a points-based score. Model performance was evaluated by discrimination, calibration, and net reclassification in the validation set, in comparison to previously derived HCC risk scores.
Results: The median MELD 3.0 was 11 (IQR 8-13); median AFP was 7 ng/mL (IQR 4-20); 51.4% of patients had a solitary HCC between 2-3 cm; and 23.4% had multifocal disease at the time of listing. By the end of follow-up, 64.8% had received deceased donor LT and 20.3% experienced waitlist dropout. The points-based score was derived awarding 1 point for each MELD 3.0 point, 2 points for either a solitary lesion 3-5 cm or 2-3 lesions regardless of size, and 0, 4, and 8 points for AFP categories of <20, 21-40, and ≥41 ng/dL, respectively. The median score was 14 (IQR 10-18). Cumulative probability of dropout ranged from 35.1% (95% CI 29.9-39.9%) at 1 year for those in the highest risk tercile and 11.1% (95% CI 8.1-14.1%) in the lowest risk tercile (log-rank p <0.01, Figure). Our model outperformed existing scores including deMELD, HALT-HCC, and HCC-MELD, with an overall C-statistic of 0.67 (95% CI 0.64-0.70) and improved Brier Score and net-reclassification index (Table).
Conclusion: Our model uses MELD 3.0, tumor burden, and AFP at listing to risk stratify patients with T2 HCC and may allow for more nuanced allocation of MELD exception points and distribution of scarce organs in the modern era. These findings can improve waiting list outcomes; maximize liver allocation utility; and help to inform the continuous distribution framework that will replace the MELD-based system for liver allocation.

