Developing High-Quality Tools to Characterize Allograft Quality, Predict Transplant Outcomes and Expand Access to Kidney and Liver Transplantation

项目摘要/摘要 在美国，近9.5万名患者正在等待肾脏移植，但在2018年，只有14700名患者接受了肾移植 有近8,500人死亡或病情加重。器官短缺同样也是 对肝脏移植的热情；2018年，在14,000多名等待名单的患者中，只有7,700人接受了肝移植 死亡的供体肝移植患者中有2500人死亡或病情加重。不幸的是，超过5000个肾脏 2018年，2000名已故捐赠者的肝脏被提供进行移植，但从未进行过移植。尽管一个 这些器官的子集不适合移植，数据清楚地表明，无法准确地 评估移植物的质量直接导致许多器官被丢弃和/或破坏我们引导器官 病人。在器官被提供进行移植之前，已故的捐赠者通常会住院数日。 有许多可用于评估器官功能的纵向数据点(例如，实验室数值)。然而， 现有的移植物质量模型存在以下主要缺陷：1)依赖于横断面临床和实验室 直接在采购前的数据；2)忽视来自捐赠者终端的与生物相关的纵向数据 住院治疗，如连续血流动力学(肾脏和肝脏)和尿量(肾脏)；以及3)未能 整合捐赠者和接受者特征之间的相互作用。因此，现有的肾脏和肝脏捐赠者 风险模型的预测精度不足(C-统计量仅为0.6-0.65)。我们小组建议 通过开发利用广泛的纵向捐赠者数据的最先进的模型来推动该领域的发展 在供者住院期间--器官损伤的实验室生物标志物和器官的测量 功能和灌注量。其次，我们将使用联合建模开发高度稳健的同种异体移植风险模型 方法，该方法可以考虑供体纵向暴露数据和移植等事件发生时间结果 失败，而不是标准技术(例如，考克斯回归)。第三，我们将突出现实世界的影响 在人口健康方面的结果。我们有以下具体目标：1)建立肾移植衰竭模型 使用关节模型预测移植物失效，相对于电流具有更高的识别率和校准 肾供者风险指数；2)利用联合建模建立肝移植衰竭风险模型以预测移植肝衰竭 高分辨率和高校准；3a)从更好的配对模拟同种异体移植物寿命的变化 基于计划器官和患者生存的对准将器官发送给接受者；以及3b)模拟变化 通过实施改进的方案减少移植数量和移植人群的同种异体移植寿命 器官分配中的器官质量指标，以减少丢弃。这些模型将使用 全面的美国移植数据，并通过加拿大两个省的数据进行了外部验证。助学金 还将通过链接到Medicare的数据来包括对移植并发症的重要探索性分析。 我们最终将开发一个基于网络的工具，以实现对器官结果的实时预测，并将结果输入 临床医生和其他研究人员的手。

Project Summary/Abstract In the US, nearly 95,000 patients are waitlisted for a kidney transplant, yet in 2018, only 14,700 received a deceased donor kidney transplant, while nearly 8,500 died or became too sick. The organ shortage is equally intense for liver transplant; in 2018, among more than 14,000 waitlisted patients, only 7,700 received a deceased donor liver transplant while 2,500 died or became too sick. Unfortunately, more than 5,000 kidneys and 2,000 livers from deceased donors were offered for transplant in 2018, but never transplanted. Although a subset of these organs was unsuitable for transplant, data clearly demonstrate that the inability to accurately assess graft quality directly led to many discards and/or undermined our ability to guide organs to appropriate patients. Prior to their organs being offered for transplant, deceased donors are hospitalized for days, often with numerous longitudinal data points (e.g., laboratory values) available to assess organ function. Yet, existing models of graft quality have these major flaws: 1) a reliance on cross-sectional clinical and laboratory data directly prior to procurement; 2) neglect of biologically-relevant, longitudinal data from the donor terminal hospitalization such serial hemodynamics (kidney and liver) and urine output (kidney); and 3) failure to integrate interactions between donor and recipient characteristics. As a result, existing kidney and liver donor risk models have inadequate prediction accuracy (C-statistics of only 0.6-0.65). Our group proposes to advance the field by developing state-of-the art models that make use of extensive, longitudinal donor data during the donor's terminal hospitalization—laboratory biomarkers of organ injury, and measures of organ function and perfusion. Second, we will develop highly robust allograft risk models using the joint modeling approach, which can account for longitudinal donor exposure data and time-to-event outcomes such as graft failure, instead of standard techniques (e.g., Cox regression). Third, we will highlight the real-world impact of the results in terms of population health. We have these specific aims: 1) Develop kidney graft failure models using joint modeling to predict graft failure with higher discrimination and calibration relative to the current kidney donor risk index; 2) Develop liver graft failure risk models using joint modeling to predict graft failure with high discrimination and calibration; 3a) Simulate the change in allograft life years from better pairing organs to recipients based on alignment of projected organ and patient survival; and 3b) Simulate the change in the number of transplants and allograft life years for the transplant population by implementing improved organ quality metrics in organ allocation to decrease discards. The models will be constructed using comprehensive US transplant data and externally validated with data from two Canadian provinces. The grant will also include important exploratory analyses of transplant complications by linking to data from Medicare. We will finally develop a web-based tool to enable real-time predictions of organ outcomes to put the results in the hands of clinicians and other investigators.