Enabling Clinical Trials for AMR in the Era of Precision Medicine.
Enabling Clinical Trials for AMR in the Era of Precision Medicine.
复制标题
DOI：
10.1097/tp.0000000000003275
复制
发表时间：
2021-03-01
期刊：
影响因子：
6.2
通讯作者：
Menon MC
中科院分区：
文献类型：
--
作者：
Cumpelik A;Zhang Z;Menon MC
晚期抗体介导的排斥反应（AMR）仍然是长期同种异体移植失败的重要决定因素。尽管在不断完善 AMR 诊断标准 1 和了解潜在分子机制方面取得了重大进展，但对晚期 AMR 进行预测或风险分层以及随后改变管理的策略尚未在临床上使用。尽管多种治疗方案已用于晚期 AMR，但这些方法并未始终如一地转化为有意义的长期结果。因此，需要新的临床适用方法来检查 AMR 的自然过程，定义早期替代结果（以评估缓解率），并确定潜在靶向治疗的亚组。检查晚期 AMR 中这些指标的新颖研究可以为后续稳健的干预试验提供宝贵的基础。在本期《移植》中，Irish 等人回顾性地研究了晚期 AMR 发作后前 12 个月内估计肾小球滤过率 (eGFR) 斜率的效用，作为早期替代终点，使用 91 名患者的历史队列来转述死亡审查移植失败 (DCGF) 的风险。包括来自 4 个北美和欧洲中心的抗 HLA 供体特异性抗体 (DSA) 阳性、经活检证实 AMR（移植后 > 1 年且随访 3 年）的患者。这个主要是白人的队列在每个中心都按照通常的护理标准进行治疗，而 37% 的人没有接受特殊治疗。 DCGF（定义为肾脏替代治疗或持续 eGFR<15 mL/min/1.73 m2）在随访期间在 59% 的患者中观察到。使用平滑回归，作者证实了直观的观察结果，即基线 eGFR 以及代表性 eGFR 斜率，尤其是在 AMR 后的第一年，与 DCGF 事件显着相关。尽管一些测试变量与基线 eGFR 相关，但没有变量与 eGFR 斜率显着相互作用。应用一种新颖的联合建模方法，作者随后优雅地估计，AMR 后第一年 eGFR 斜率改善 30%，意味着 5 年内 DCGF 改善 10%。作者得出结论，AMR 诊断后 1 年 eGFR 斜率有助于设计晚期 AMR 的未来临床试验。使用 eGFR 等时间依赖性协变量进行 DCGF 建模需要仔细考虑。 Cox 模型中时间相关协变量的典型处理假设 (1) 协变量没有测量误差，并且 (2) 纵向观察中可能缺失的值与移植失败无关（即随机缺失的值）。这些假设对于基于肌酐的 eGFR 数据并不成立。首先，肌酐和 eGFR 的测量误差经常存在。接下来，后续 eGFR 观察中的任何缺失值本身都可能是由于移植物损失（即，非随机缺失的值）可能导致有偏差的结果。本研究中应用的联合模型的优点如下：（1）使用线性混合模型，通过考虑每个时间点的测量误差以及受试者特定的随机效应，更准确地估计eGFR随时间的“真实”轨迹；（2）在DCGF的Cox模型中使用这种估计的“真实”轨迹； (3) 将最终模型中的这 2 个子模型与特定于受试者的随机效应联系起来。因此，在此数据集中，联合模型允许利用纵向 eGFR 测量值（此处用 eGFR 斜率表示）动态预测每位患者的 DCGF 风险。这些预测可以针对每个主题进行，这对于……来说是一个特别有用的功能。
Late antibody-mediated rejection (AMR) remains an important determinant of long-term allograft loss. Although significant progress has been made in continuously refining AMR diagnostic criteria1 and understanding underlying molecular mechanisms, strategies to prognosticate or risk stratify late AMR, and thereafter alter management, are not being clinically utilized. Although several treatment protocols have been used for late AMR, these approaches have not consistently translated into meaningful long-term outcomes. New clinically applicable approaches are thus required to examine the natural course of AMR, define early surrogate outcomes (to assess response rates), and identify subgroups for potential targeted therapies. Novel studies examining these metrics in late AMR could provide a valuable foundation for subsequent robust interventional trials. In the current issue of Transplantation, Irish et al2 retrospectively examine the utility of estimated glomerular filtration rate (eGFR) slope within the first 12 months after an episode of late AMR as an early surrogate end point to relay the risk for death-censored graft failure (DCGF) using a historical cohort of 91 patients. Patients with anti-HLA donor-specific antibody (DSA) positive, biopsy-proven AMR (> 1 y posttransplant and with 3 y of follow-up) across 4 North American and European centers were included. This predominantly Caucasian cohort was treated per usual standard of care at each center, while 37% did not receive specific treatment. DCGF, defined as renal replacement therapy or persistent eGFR< 15 mL/min/1.73 m2, was observed in 59% of patients during follow-up. Using smoothed regression, the authors confirmed the intuitive observation that baseline eGFR, as well as a representative eGFR slope, especially in the first year post-AMR, was significantly associated with DCGF events. Although some of the tested variables were associated with baseline eGFR, no variable significantly interacted with eGFR slope. Applying a novel joint modeling approach, the authors then elegantly estimated that a 30% improvement of eGFR slope in the first year post-AMR translates to a 10% improvement in DCGF for 5 years. The authors conclude that 1-year eGFR slope post-AMR diagnosis could help design future clinical trials for late AMR.The modeling of DCGF using a time-dependent covariate such as eGFR needs careful considerations. A typical treatment of a time-dependent covariate in a Cox model assumes (1) the covariate has no measurement errors and (2) the possible missing values in the longitudinal observations are not related to graft failure (ie, values missing at random). These assumptions do not hold true for creatinine-based eGFR data. First, measurement errors in creatinine and therefore eGFR often exist. Next, any missing values in follow-up eGFR observations could themselves be due to graft loss (ie, values missing not at random) potentially leading to biased results. Advantage of the joint model applied in this study are the following:(1) the use of a linear mixed model to more accurately estimate the “true” trajectory of eGFR overtime by taking into account measurement errors at each time point, as well as subjectspecific random effects;(2) the use of this estimated “true” trajectory in the Cox model for DCGF; and (3) linking of these 2 submodels in the final model with subject-specific random effects. Therefore, in this dataset, the joint model allows the risk for DCGF to be dynamically predicted in each patient by utilizing the longitudinal eGFR measurements (represented here by eGFR slope). These predictions can be made specific to each subject, a particularly useful feature for a …