BACKGROUND: Prior studies show severe acute-on-chronic liver failure (ACLF) at liver transplantation (LT) negatively impacts short-term, but not long-term, post-LT outcomes. However, not accounting for ACLF's time-varying effect on the waitlist may underappreciate its dynamic nature. Moreover, excluding those who died or dropped off the waitlist raises concerns about selection bias. METHODS: This US nationwide retrospective cohort study estimated the effect of severe ACLF (grade 3) (ACLF-3) on post-LT outcomes, including adult, first-time deceased donor LT candidates listed from June 2013 to May 2023. A marginal structural model (MSM) to address selection bias and time-varying exposure (ACLF-3) was applied, with extended Cox proportional hazard models using a Heaviside step function to assess the hazard of death after LT. RESULTS: Among 31,267 eligible candidates for LT (baseline cohort), 11.3% (n = 3518) had ACLF-3 at listing
13.6% (n = 4243) died or dropped out while on the LT waitlist. Of the 27,024 patients who received LT (transplanted cohort), 12.3% (n = 3333) had ACLF-3 at LT. ACLF-3 at LT (but not at waitlisting) was associated with a higher hazard of death, with the hazard ratio of 1.80 (95% CI: 1.09-2.97) within 1 year after LT but not thereafter. This marginal structural model effect size was 9% higher than conventional multivariable Cox proportional hazard models. Sensitivity analyses corroborated these findings. CONCLUSIONS: Compared to previous studies, ACLF-3 at LT in our marginal structural model was associated with a discernible increase in short-term mortality after transplant, presumably due to our addressing of selection bias, while long-term survival was similar to those without severe ACLF at LT. However, potential vulnerability to posttransplant complications warrants further investigation.