Biomarkers of Kidney Injury: Statistical Methods for Risk Model Development and Evaluation

 描述（由申请人提供）：急性肾损伤（AKI）是心脏手术的常见并发症，对长期健康有潜在的严重影响，包括需要透析以及心血管事件和死亡风险增加。目前 AKI 的诊断、预防和治疗方面存在差距，人们对开发 AKI 风险预测模型来解决这些问题抱有极大的兴趣。传统的临床和人口统计学变量对 AKI 的预测能力有限。因此，人们越来越关注使用生物标志物和生物标志物组合来预测 AKI。虽然生物标志物有潜力识别 AKI 高风险患者，但问题仍然存在。拟议的研究将解决两个这样的问题：多水平 AKI 结果的使用和多中心研究中生物标志物组合的开发。首先，虽然定义了多级别 AKI 结局（例如无、轻度和重度），但严重 AKI 往往是令人感兴趣的结局，因为该结局与发病率和死亡率密切相关。将使用多项建模方法和专门的模型选择技术来利用无 AKI 的个体和轻度 AKI 的个体之间生物标志物水平的变化，以改善对严重 AKI 风险的预测。其次，虽然多中心生物标志物研究通常提供更大的功效和更高的结果概括性，但由于 AKI 患病率的不同和/或生物标志物测量的差异，可能会存在中心差异。如果忽略这些差异，可能会导致生物标志物组合性能评估出现偏差。因此，拟议的研究将创建在多中心研究中开发生物标志物组合的方法，包括表征中心差异的工具和识别代表中心的预测生物标志物组合的技术。解决这两个问题将提高生物标志物用于 AKI 风险预测的潜力。能够预测心脏手术中 AKI 风险的生物标志物可能是 用于通过 (1) 提供更准确的诊断来减轻 AKI 的负担； (2)及早诊断AKI，打开治疗窗口； （三）确定应采取预防措施的高风险人群； (4) 丰富临床试验入组情况，协助开发新疗法和预防工具； (5) 为临床医生和患者提供更好的决策依据信息。这些结果中的任何一个都将改变肾脏病学的临床护理并改善接受心脏手术的患者的健康。

 DESCRIPTION (provided by applicant): Acute kidney injury (AKI) is a common complication of cardiac surgery with potentially serious effects on long- term health, including need for dialysis and increased risk of cardiovascular events and mortality. Current gaps exist in the diagnosis, prevention and treatment of AKI, and there is great interest in developing AKI risk prediction models to address these issues. Traditional clinical and demographic variables have limited predictive capacity for AKI. As a result, there is increasing interest in using biomarkers and biomarker combinations to predict AKI. While biomarkers have the potential to identify patients at high risk for AKI, issues remain. The proposed research will address two such issues: the use of multi-level AKI outcomes and the development of biomarker combinations in multi-center studies. First, while multi-level AKI outcomes (such as no, mild and severe) are defined, severe AKI is often the outcome of interest, as this outcome is most strongly associated with morbidity and mortality. Multinomial modeling methods and specialized model selection techniques will be used to leverage variation in biomarker levels between individuals with no AKI and those with mild AKI to improve prediction of severe AKI risk. Second, though multi-center biomarker studies typically offer greater power and increased generalizability of results, i is possible to have center differences due to varying AKI prevalence and/or differences in biomarker measurements. Such differences, if ignored, can lead to bias in the assessment of the performance of biomarker combinations. Thus, the proposed research will create methods for developing biomarker combinations in multi-center studies, including tools to characterize differences by center and techniques to identify predictive biomarker combinations that account for center. Addressing these two issues will advance the potential of biomarkers for AKI risk prediction. Biomarkers capable of predicting risk of AKI in the setting of cardiac surgery could be used to reduce the burden of AKI by (1) providing a more accurate diagnosis; (2) diagnosing AKI earlier, opening a therapeutic window; (3) identifying high risk individuals for whom preventative measures should be implemented; (4) enriching clinical trial enrollment, aiding in the development of novel therapies and tools for prevention; and (5) providing the clinician and patient with better information on which to base decisions. Any of these outcomes would transform clinical care in nephrology and improve the health of patients undergoing cardiac surgery.