Artificial intelligence in liver transplant allocation: Promise, pitfalls, and ethical safeguards.

Abstract

Liver transplantation (LT) is an ethically complex procedure that involves allocating a limited, life-saving resource. For over 20 years, the Model for End-Stage Liver Disease (MELD) score and its variations have implemented a sickest-first ethical approach, enhancing waitlist survival rates and adding objectivity to the process. Nonetheless, ongoing disparities based on sex, race, and other structural factors reveal the shortcomings of relying on simplified, static models of disease severity. As LT progresses into an era increasingly shaped by artificial intelligence (AI) and machine learning (ML), these algorithms offer better predictions of waitlist mortality, post-transplant outcomes, and donor-recipient matching. However, higher predictive accuracy does not automatically lead to more ethical distribution. In this Perspective, we explore AI-assisted deceased donor liver allocation through the ethical principles of justice, respect for persons, and humanism. We review emerging AI/ML models proposed for LT, highlighting their potential advantages and risks, including amplification of historical inequities embedded in training data, lack of transparency, erosion of informed consent, and overreliance on algorithmic outputs at the expense of clinical judgment. We contend that AI should serve as decision support rather than a decision-maker, and that ethical implementation requires deliberate safeguards. We suggest a framework for responsibly integrating AI into liver allocation focused on fairness monitoring, transparency, explainability, human oversight, and multi-stakeholder governance. When applied properly, AI can potentially enhance both utility and equity in liver transplantation-but only if its use is guided by clear ethical principles and ongoing evaluation in real-world clinical settings.