Transplant Surgery

Offline reinforcement learning optimizes liver waitlist donor-offer decisions

An offline reinforcement learning model used SRTR waitlist trajectories to optimize when to wait, delist, or transplant liver candidates. Applied retrospectively, the learned policy would have avoided 73% of donor–recipient matches ending in graft failure or death. It preserved 93% of successful transplants, indicating limited sacrifice of good outcomes for safety gains. The model also identified potentially suitable donors for 47% of candidates who died on the waitlist. Analysis suggested it inferred features associated with successful liver donor–recipient matches, supporting prospective evaluation as clinical decision support. 1

cPRA ≥96% sharply reduces deceased-donor kidney access in France

This retrospective multicenter cohort included 14,485 French kidney transplant candidates listed between 2011 and 2021. Access to deceased-donor transplantation declined progressively with increasing sensitization and dropped sharply from cPRA ≥96%. Compared with cPRA 0%, adjusted subdistribution hazard ratios for transplant fell to 0.353 at cPRA 99% and 0.082 at 100%. Waitlist attrition increased from cPRA ≥97%, reaching an adjusted subdistribution hazard ratio of 2.713 at cPRA 100%. Candidates with cPRA 85–95% retained preserved access, defining cPRA 96–100% as an ultrasensitized group needing refined prioritization strategies. 2

Candidemia after solid organ transplant is infrequent but highly lethal

Among 10,249 adult solid organ recipients in Ontario, 135 developed candidemia during follow-up through 2023. Cumulative candidemia probability was 0.87% at 1 year, 1.33% at 5 years, and 1.67% at 10 years. Lung recipients had the highest 10-year risk at 4.17%, with earlier onset; median 34 days versus 174 days for abdominal allografts. Thirty-day and 90-day mortality after candidemia were 39.3% and 47.4%, respectively. Candidemia was associated with markedly increased mortality, with adjusted hazard ratios around 7–12 regardless of fluconazole susceptibility, underscoring the need for aggressive prevention and early treatment. 3

Neighborhood disadvantage dampens liver transplant referral for alcohol-associated hepatitis

This multicenter cohort included 325 patients hospitalized with severe alcohol-associated hepatitis (MELD >20) at five US transplant centers. Only 36.9% were referred for liver transplant evaluation despite high overall severity. Higher MELD scores increased referral odds, but generalized additive models showed significant MELD × Area Deprivation Index interactions for referral and outcomes. At MELD 20–30, referral probability was 40–60% for ADI <30 versus about 20% for ADI ≥30. Among MELD 20–30 patients, 180-day mortality exceeded 20% at ADI ≥60 compared with 10–20% at ADI <20, indicating structurally mediated under-referral. 4