Defining safe thresholds for risk-adapted surveillance after liver transplantation for hepatocellular carcinoma

BACKGROUND: Recurrence of hepatocellular carcinoma (HCC) after liver transplantation (LT) is a major concern, but it is unclear whether existing models can safely "rule out" patients from intensive imaging strategies. METHODS: We retrospectively studied 493 LT recipients transplanted for HCC (2002-2023). Competing-risk Fine-Gray regression with bootstrap stability selection was used to derive an illustrative score and compared with established models. Rule-out performance was evaluated by 5-year cumulative incidence. Decision curve analysis (DCA) and surveillance simulations quantified trade-offs between CT scans saved and recurrences missed under reduced protocols. RESULTS: Our model identified total tumor diameter, microvascular invasion, satellite nodules, and log(AFP) as stable predictors. DCA showed all models outperformed uniform surveillance only when 5-year recurrence exceeded 4%, with 7.5% as the upper acceptable threshold. Within this range, low-risk groups across models had 5-year recurrence risks of 2.5-5.7%. Reduced-intensity strategies could save 160-856 CTs per 100 patients while missing ≤1 recurrence with semiannual or annual two-year protocols. Cost-benefit analysis supported 4-7.5% as the optimal threshold. False negatives were uncommon (n = 13), with only two patients misclassified by all models. CONCLUSIONS: Widely used models can identify patients suitable for reduced surveillance, though none delineated a true "no-screening" group.