ACHIEVING MORE EQUITABLE ALLOCATION OF DONOR LIVER GRAFTS THROUGH BALANCING SURVIVAL AND MEDICAL URGENCY: A NOVEL MATHEMATICAL MODEL MONALISA 1.0

Background: Organ allocation based on the principles of urgency vs utility presents an ethical conundrum given the scarcity of liver grafts. The current Model for End-Stage Liver Disease (MELD) score prioritizes the sickest patients but fails to consider potential additional life-years gained from liver transplant (LT). We aimed to develop a new algorithm for equitable prioritization of liver grafts.

Methods: Using OPTN/UNOS data up to December 2021, we developed a robust model to estimate LT survival benefit defined as extra years of life that a candidate can achieve with LT versus remaining on the waitlist (life-years from transplant [LYFTs]). Weibull regression model was fitted using recipient-only characteristics at waitlist registration. The model was manually built based on a-priori clinical knowledge and flexible splines and interaction terms were generated for 86 estimated parameters. The model predicted the marginal median life expectancies of a patient under the counterfactual circumstances of receiving LT vs remaining waitlisted, and the LYFT score for the patient was calculated as the difference between life expectancy after LT vs waitlist. We proposed a novel score (MoNaLISA: Maximization Of Net Liver Survival benefit and medical Acuity) as the geometric average of the LYFT and MELD score which conceptually balances principles of utility (LYFT) and urgency (MELD). Monte Carlo simulations with bootstrap resampling were conducted to assess the impact of (i) MELD, MELD-Na, and MELD 3.0 score, (ii) MaxLYFT which prioritizes recipients by LYFT scores to maximize LT survival benefit at a population level, (iii) MoNaLISA score.

Results: 219,384 adult LT candidates were included in complete-case analyses with 1.64 million person-years follow-up. The model exhibited excellent goodness-of-fit and Harrell’s C-index of 0.784 (Figure 1). The model also affirmed numerous “clinically-expected” statistical interactions. For example, age was an important effect modifier, as the gain in life expectancy after LT was generally lesser for older recipients (predicted LYFT of 23.1 years for a recipient aged 30 years vs LYFT of 13.8 years for a recipient aged 55 years). Through resampling-based simulations, MELD, MELD-Na, MELD 3.0, MoNaLISA and MaxLYFT respectively yielded 9.5 vs 12.8 vs 12.9 vs 14.2 and 14.5 additional years-of-life per liver graft. Interestingly, MoNaLISA had a negligible adverse impact on 6-month waitlist mortality (22.5%) vs MELD schemes (21.9%-22.6%). MaxLYFT, unsurprisingly, had the highest 6-month waitlist mortality (27.8%) (Figure 2).

Conclusion: We present a novel liver allocation score that maximizes LT survival benefit while minimizing 6-month waitlist mortality. Assuming 13,000 waitlist registrations and 7,000 LTs occur annually in the USA, MoNaLISA adds 9,800 life-years with a negligible increase in waitlist deaths.