deceased供肾移植后移植物失功的动态风险预测
该研究开发并验证了一个利用eGFR变化动态预测 deceased供肾移植受者3年移植物失功的风险模型。模型在开发队列(N=1114)中3个月AUC为0.70,30个月达0.90,验证队列AUC为0.60-0.64至0.72-0.78。模型可动态更新预测风险,有助于识别高危受者进行个性化干预。
要点
- 该模型利用随访eGFR动态预测移植后3年移植物失功风险。
- 开发队列3个月AUC为0.70,30个月达0.90,验证队列性能有所衰减。
- 模型可在移植后3年内任何时间点使用,随每次eGFR测量更新风险。
核心结论
- 该动态风险预测模型有潜力通过识别高危受者来加强移植后护理,使其可能受益于更密切监测和个性化干预。
研究设计与样本
- 模型在2010-2013年前瞻性多中心 deceased供肾移植受者队列中开发,并在三个独立队列(一个登记队列和两个电子病历队列)中验证。
- 采用两阶段方法:先使用线性混合效应模型估计eGFR趋势,再将这些趋势用于逻辑回归模型预测3年移植物失功。
- eGFR采用无种族方程计算,移植物失功定义为恢复透析或再次移植,死亡为删失。
主要发现
- 开发队列(N=1114)中94例(8%)在移植后3年内发生移植物失功。
- 模型预测准确性随时间增加而提高,乐观校正的ROC曲线下面积(AUC)在移植后3个月为0.70(95%CI 0.63-0.77),在30个月达到0.90(95%CI 0.85-0.95)。
- 验证队列中性能有所衰减,AUC范围从3个月的0.60-0.64到30个月的0.72-0.78,可能源于确定eGFR的数据收集方法差异。
临床怎么用
- 该模型可在移植后3年内任何时间点使用,利用每次随访eGFR测量动态更新3年移植物失功预测风险。
- 有助于识别高危受者,从而实施更密切监测和个性化干预以改善结局。
局限性
- 验证队列中预测性能有所衰减,可能由于确定eGFR的数据收集方法差异。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
KEY POINTS: This risk prediction model can enhance post-transplant care by identifying high-risk recipients for personalized interventions. The predicted risk of 3-year graft failure is dynamically updated with each follow-up eGFR measurement. BACKGROUND: The identification of kidney transplant recipients at high risk of graft failure enables timely interventions to improve outcomes. We developed and validated a risk prediction model that uses changes in eGFR to dynamically predict 3-year graft failure. METHODS: The risk prediction model was developed in a prospective multicenter cohort of deceased donor kidney transplant recipients (2010–2013) and validated in three independent cohorts (a registry cohort and two electronic medical record cohorts). The two-stage approach first estimated eGFR trends using a linear mixed-effects model and then used these trends in a logistic regression model to predict 3-year graft failure. eGFR was calculated using race-free equation, and graft failure was defined as the return to dialysis or retransplantation, censored for death. The model can be used at any time within the first 3 years after transplant, and the predicted risk is dynamically updated with each additional eGFR measurement. RESULTS: In the development cohort (N=1114), 94 (8%) experienced graft failure within 3 years of transplant. The model's predictive accuracy improved over time with the increase in available eGFR measurements. The optimism-corrected area under the ROC curve (AUC) was 0.70 (95% confidence interval, 0.63 to 0.77) at 3 months post-transplant and reached an AUC of 0.90 (95% confidence interval, 0.85 to 0.95) at 30 months post-transplant. Performance was attenuated in the validation cohorts (AUC range 0.60–0.64 at 3 months to 0.72–0.78 at 30 months), likely due to differences in data collection approaches in ascertaining eGFR. CONCLUSIONS: This risk prediction model has the potential to enhance post-transplant care by identifying high-risk recipients who may benefit from closer monitoring and personalized interventions.
引用
- Thiessen Philbrook H, Hu DG, Mohan S, et al. Dynamic Risk Prediction of Graft Failure after Deceased Donor Kidney Transplant. Clin J Am Soc Nephrol. 2025. PMID: 41052470. doi:10.2215/CJN.0000000883
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。