From Sex to Size: Rethinking Fairness in Liver Transplantation

In this issue of Hepatology, Tanaka and colleagues pose an important question regarding the longevity of equity surrounding the model for end-stage liver disease (MELD) 3.0 implementation.¹ Due to the increased observed waitlist mortality among female liver transplant (LT) candidates, MELD 3.0 was developed with the intent to improve mortality outcomes and equity for patients, specifically women, listed for LT. The final model included female sex, albumin, interaction terms between albumin and creatinine, a reduced maximum creatinine value and updated coefficients, and was adopted in the United States in July, 2023. The additional 1.3 points for female sex were intended to mitigate the disadvantage related to lower muscle mass and reduced creatinine production.² This study by Tanaka and colleagues evaluates the effects of the MELD 3.0 policy on the sex-based disadvantages in transplant access.

This study provides real-world data on 60 thousand LT candidates including nearly 10-million person-day observations to evaluate the important questions: “Does the MELD 3.0 policy change actually increase equity, are there still gaps, and how do we fix them?” Using a differences-in-differences (DiD) study design, the authors evaluate the impact of policy, sex, and time by comparing pre-MELD 3.0 (pre-7/2023), three months after MELD 3.0 implementation (7/2023-10/2023 or the “burn-in” period), and longer term outcomes after the burn-in period (10/2023-9/2024). Primary outcomes are the receipt of a top-5 organ offer on the match run and LT. By evaluating top-5 offers and transplantation as outcomes, the authors address liver allocation from a waitlist priority and practical perspectives, as disparities exist at multiple levels for women.³ Results show early improvements for women with a relative 25% increased top-5 organ offers rate at implementation. However, this advantage attenuates after the burn-in period to a 16% increase during the three-month implementation. Regarding transplantation, relative to the pre-MELD 3.0 baseline, odds are 36% higher at MELD 3.0 implementation and 18% higher during the initial burn-in period compared to men. However, in longer follow-up of greater than a year post-implementation, there is an observed return to the pre-MELD 3.0 disadvantage, with 0.53 fewer transplants per 1,000 person-days for women compared to men.

The authors also investigate height as the main surrogate for body size as a mediator for the male-female disadvantage. Prior studies show that lower height is associated with lower probability of transplantation and higher mortality.^4,5 Height, however, was not chosen for MELD 3.0, due collinearity between height and sex, posing risks to model reliability in predicting mortality. Given that the magnitude of the effect for sex exceeded that of height, only sex was included in the final MELD 3.0 model.² Nonetheless, the issue of whether height has a place in transplant allocation remains unresolved. Tanaka and colleagues’ current analysis shows that height, rather than sex, accounts for the observed differences in transplant access. After adjusting for sex and policy era, each 10-cm increase in height is associated with a 12% increase in odds of transplant. The implementation of MELD 3.0 initially benefits women of shorter stature (<166 cm), who had similar pre-MELD 3.0 disadvantages as men of shorter stature. After the 3-month burn-in period, women of shorter stature are observed to have an increase in longer-term transplant rates by 47%. Women of taller stature (≥166 cm), however, have a more modest increase in transplant rates of 16% during the same period. This observed benefit for shorter women post-MELD 3.0, when viewed in context, implies a relative disadvantage for men of short stature.

It’s important to emphasize that MELD 3.0 development was based on reducing sex disparity in short-term waitlist mortality. In this current manuscript, some of the most salient findings related to mortality are reported in the supplement. During a more contemporary study period than the one used to develop MELD 3.0, baseline waitlist mortality for women relative to men is 14% higher, but the difference does not quite reach statistical significance (95% CI 0.99-1.32). Introduction of MELD 3.0 is not observed to have a significant association with waitlist mortality. However, including height rather than sex in the model significantly decreases waitlist mortality risk by 10% for every 10-cm increase in height. The findings should stimulate active discussion of whether height rather than sex can better improve equity in access to transplantation.

The authors conduct multiple sensitivity analyses adjusting for additional variables, including MELD exception candidates, and stratifying by alcohol-associated liver disease to further understand the effect of MELD 3.0 in relevant subgroups. In general, findings are similar, showing a post-MELD 3.0 implementation spike in transplant rates for women relative to men with a subsequent decline and regression to pre-MELD 3.0 levels.

The regression and reversion to the pre-MELD 3.0 male-female disadvantage is a surprising finding that requires further scrutiny. The authors posit that the short-lived gains in transplant rates for women under MELD 3.0 relate to “front-loaded effects” whereby a subset of women close to the allocation threshold crossed that threshold right at policy implementation. Further analyses need to elucidate the reasons for the observed regression to pre-policy implementation by examining waitlist additions and other concurrent changes in transplant practices.

These findings highlight the potential limits of simulated allocation models in clinical implementation. The MELD 3.0 model was developed using the liver simulated allocation model (LSAM), which used historical data from 2013-2016. Refitting and recalibration may be in order. Additionally, this study reinforces the importance of evaluating policy changes using timely, real-world data. Difference-in-difference methodology to evaluate temporal impacts of policy or changes in care has long been used in health services research. These methods are welcome and more studies like this are needed in liver transplantation. Additional questions regarding the impact of MELD 3.0 and other allocation policies remain and will require high-quality data inputs and continued modeling sophistication. Secular trends are rapidly changing transplant practice. For example, the rapid rise and health equity impacts of machine perfusion, normothermic regional perfusion and increased used of donation after cardiac death require further study. Unfortunately, machine perfusion data are not collected by the Scientific Registry of Transplant Recipients making it difficult to study these changes in policy and practice in real time. Simultaneously, there has been a rise of transplantation for alcohol-associated hepatitis⁶ with notable gender disparity observed in access to the waitlist.⁷ Detailed studies of health equity in liver transplant for alcohol-associated liver disease related to top organ offers and waitlist outcomes should be performed. Alarming trends of rising out-of-sequence allocation practices must also be placed into context within organ allocation policy⁸ and any downstream healthcare disparities further be examined in that context. Continued attention to achieving equity in transplantation requires a commitment to constant vigilance, modernized data collection, sophisticated methods of evaluation, and a universally accepted model for continuous improvement.