美国供体心脏分配成人医学紧迫度评分的开发与验证
Development and Validation of an Adult Medical Urgency Score for US Donor Heart Allocation
基于美国器官移植系统 2019 年 1 月至 2025 年 8 月全部成年心脏移植候选人的注册表观察研究,开发并验证了基于 6 周无移植死亡率的 US-CRS 评分。
- 背景
- 在美国,每5名心脏移植候选人中就有1人在未接受移植的情况下死亡或被移出等待名单。供体心脏目前依据一套基于治疗的6级状态系统分配,该系统被批评为风险分层能力有限、且容易被策略性操作。本研究要建立一套基于客观医疗紧急度评分的分配政策,按照生理状态与机械循环支持之间的交互对候选人排序。
- 做法
- 这是一项基于登记数据的观察性研究,使用美国器官移植系统的数据。纳入2019年1月至2025年8月全部成年(≥18岁)心脏移植候选人,并按时间划分为训练集(2019—2023年,n=21859)、验证集(2023—2024年,n=7897)和测试集(2024—2025年,n=8256)。以未接受移植的6周死亡为结局,估计边际结构模型,即美国候选人风险评分(US-CRS)。在留出数据中比较8个模型,评价指标为加权时间依赖AUC和按状态的校准度。最后把US-CRS转换为分配政策,并进行匹配运行模拟,对等待名单重新排序,与现行6状态系统作对比。
- 结论
- 候选人的平均年龄为53(SD 13)岁,74%为男性。最终选定的US-CRS模型包含5项实验室指标和4类治疗装置:耐久性左心室辅助装置、主动脉内球囊反搏、微轴流泵和高级机械循环支持。在同一日历日期仍处于等待状态的候选人测试集横断面中,US-CRS的时间依赖AUC与现行6状态系统相当或显著更高(平均AUC分别为0.807与0.693),按状态的校准度可接受,在以例外方式获得状态的候选人中尤其如此。匹配运行分析(n=45530个供体心脏)显示,基于US-CRS的系统使67%的移植受者排序位次下移,并使匹配运行中第一位候选人的观察6周移植删失死亡率从现行系统的11.2%(95% CI,10.1%—12.5%)升至基于US-CRS系统的15.7%(95% CI,13.9%—17.6%)。
- 意义
- 与现行医疗紧急度状态相比,基于US-CRS的分配系统把未接受移植时死亡率更高的更重候选人排到更前。对制定美国供体心脏分配政策的机构而言,这给出了一套以客观评分取代状态分类的方案及其名次重排模拟结果,可用于评估改换排序规则后哪些候选人位次上升、哪些下移,以及匹配运行首位候选人的结局分布如何变化。
原文摘要
出版方登记 · CrossrefImportance One in 5 heart transplant candidates in the US die or are delisted without transplant. Donor hearts are currently allocated by a 6-status therapy-based system criticized for limited risk stratification and susceptibility to gaming. Objective To develop a new heart allocation policy based on an objective medical urgency score that ranks candidates based on the interaction between physiology and mechanical circulatory support. Design, Setting, and Participants Registry-based observational study using data from the US organ transplant system. All adult (aged ≥18 years) heart transplant candidates from January 2019 to August 2025 were included, split into training (2019-2023), validation (2023-2024), and test (2024-2025) datasets. Main Outcomes and Measures A marginal structural model of 6-week mortality without transplant was estimated (the US Candidate Risk Score [US-CRS]). Eight models were compared in held-out data by weighted time-dependent area under the curve (AUC) and calibration by status. Finally, US-CRS was converted into an allocation policy and a match-run simulation of waitlist reordering was performed compared with the current 6-status system. Results Adult heart transplant candidates were partitioned into consecutive training (n = 21 859), validation (n = 7897), and test (n = 8256) datasets (mean [SD] age, 53 [13] years; 74% male). The chosen US-CRS model included 5 laboratory values and 4 therapeutic devices (durable left ventricular assist device, intra-aortic balloon pump, microaxial flow pump, and advanced mechanical circulatory support). In test-set cross-sections of candidates actively waiting on the same calendar date, US-CRS had similar or significantly higher time-dependent AUC than the current 6-status system (mean AUC, 0.807 vs 0.693, respectively) and acceptable calibration by status, especially for candidates who received their status by exception. In match-run analysis (n = 45 530 donor hearts), a US-CRS–based system led to 67% of transplant recipients moving down in rank order and increased observed 6-week transplant-censored mortality for the first candidate on the match run from 11.2% (95% CI, 10.1%-12.5%) under the current system to 15.7% (95% CI, 13.9%-17.6%) under the US-CRS–based system. Conclusions and Relevance In this registry-based study of US heart transplant candidates, a US-CRS–based allocation system prioritized sicker candidates with higher mortality without transplant compared with the current medical urgency statuses.
页面未能读取全文,以上为出版方登记的摘要;方法与图表见原文
本页是 AI 精读,不转载论文原文;完整论文(含方法、图表与参考文献)见原文
基于注册表的观察研究把心脏分配从 6 级状态改为客观紧迫度评分,配型模拟显示首批候选人的 6 周死亡率更接近真实风险。
来源:JAMA · doi.org