Liangpunsakul Suthat, Krebs William B, Kwong Allison J, Kwo Paul Y, Brown Robert S, Lin WeiQi, Sussman Norman L
Department of Medicine, Division of Gastroenterology and Hepatology, Indiana University School of Medicine, Indianapolis, Indiana, USA.
Krebs Consulting Statistician, Santa Rosa, CA, USA.
Hepatol Commun. 2025 Apr 30;9(5). doi: 10.1097/HC9.0000000000000709. eCollection 2025 May 1.
Alcohol-associated hepatitis (AH) is a serious liver disease caused by heavy alcohol consumption with severe cases exhibiting a 90-day mortality rate of ~30%. No drugs have been approved for AH, and regulatory approval currently requires evidence of improved survival. The lack of effective drug therapies and high mortality rates have fueled interest in early liver transplantation (LT), which has a survival rate that exceeds 90%. However, LT is resource-intensive and is available only in expert centers, where most AH trials are conducted. As a result, LT is overrepresented in recent AH studies, leading to confounding and unresolved questions regarding valid endpoints in therapeutic AH trials.
We propose methodological approaches to address the inclusion of LT in AH trials, supported by power calculations and data from the AHFIRM trial, a 300-patient multicenter study completed in late 2023. We demonstrate the impact of effect size, trial size, and statistical methods on trial design and interpretation.
Effect size plays a crucial role in power calculations. While 90-day survival is the most efficient endpoint, competing risk analysis, primary stratum analysis, and win ratio are valuable tests for assessing the role of LT. The combined endpoint of death or LT is the least efficient method and requires the largest trial population to achieve statistical significance. We recommend using multiple statistical methods with adjustments for multiplicity.
The adoption of early LT complicates the assessment of new therapies for AH. Statistical methods and endpoints are critical in power calculations and when assessing the efficacy of new therapeutic agents. We recommend mortality as the primary analysis complemented by hierarchical secondary analyses that avoid problems of multiplicity.
酒精性肝炎（AH）是一种由大量饮酒引起的严重肝脏疾病，严重病例的90天死亡率约为30%。目前尚无药物被批准用于治疗AH，而监管部门批准药物需要有生存率提高的证据。缺乏有效的药物治疗方法和高死亡率引发了人们对早期肝移植（LT）的兴趣，肝移植的生存率超过90%。然而，肝移植资源密集，且仅在大多数AH试验开展的专家中心才可进行。因此，在近期的AH研究中，肝移植的比例过高，导致在治疗性AH试验中关于有效终点的混淆和未解决的问题。
我们提出了在AH试验中处理肝移植纳入问题的方法，并通过效能计算以及AHFIRM试验（一项于2023年末完成的纳入300例患者的多中心研究）的数据加以支持。我们展示了效应大小、试验规模和统计方法对试验设计和解读的影响。
效应大小在效能计算中起着关键作用。虽然90天生存率是最有效的终点指标，但竞争风险分析、主要分层分析和胜率是评估肝移植作用的有价值的检验方法。死亡或肝移植的联合终点是效率最低的方法，需要最大规模的试验人群才能达到统计学显著性。我们建议使用多种统计方法并对多重性进行调整。
早期肝移植的采用使AH新疗法的评估变得复杂。统计方法和终点指标在效能计算以及评估新治疗药物的疗效时至关重要。我们建议将死亡率作为主要分析指标，并辅以分层二级分析以避免多重性问题。
