Assessing risk of morbidity and mortality is an important aspect of caring for patients with heart disease that guides patient counselling and implementation of therapies. As the pathophysiological mechanisms involved in cardiovascular disease are multiple and complex, combining factors that reflect different components of the pathophysiology into a risk score can help provide an overall indication of risk. Ideally a clinical risk score is comprised of data that are readily available to clinicians. Most of the risk scores for predicting stroke or bleeding in atrial fibrillation (AF) follow this rule (i.e. CHADS-VASC, HAS-BLED, ORBIT, and ATRIA). In this issue of the journal, Dr Hijazi and colleagues have derived a clinical risk score, the ABC death risk score, which includes measurement of GDF-15 (growth differentiation factor 15)—a biomarker recently made available for clinical laboratories. The results of their study leave us to consider the questions of whether our hospitals and clinics should make the investment to begin measuring GDF-15, and would a risk score to predict mortality in anti-coagulated patients with AF be useful to our practice. In this study, data from 14 611 participants enrolled in the ARISTOTLE trial (apixaban vs. warfarin) were used to derive a risk score to predict all-cause mortality. Using 13 standard clinical variables, 4 standard biomarkers, and GDF-15, a backward elimination algorithm selected 5 variables to include in the final model [age, congestive heart failure, troponin, brain natriuretic peptide (BNP), and GDF-15]. The model predicted all-cause mortality with an impressive c-statistic of 0.74 [95% confidence interval (CI) 0.73–0.76] in the derivation cohort and 0.74 in an independent validation cohort (the RELY trial; dabigatran vs. warfarin). This model is called the ABC death risk score, and along with the ABC bleeding and ABC stroke risk scores, is the third clinical risk score proposed by Dr Hijazi and colleagues using data from the ARISTOTLE Biomarker Sub-study (Figure 1). Interestingly, they also show that the present score performed better than a comprehensive model of only clinical variables [cstatistic of 0.68 (95% CI 0.66–0.70)], demonstrating the potential power for quantitative biomarkers to improve risk prediction significantly. For some biomarkers there is a clear pathophysiological link by which they add valuable information about disease severity compared with the binary definitions used for clinical variables. For example, in the ABC death risk score, the addition of BNP levels presumably helped to distinguish between the low-risk well-compensated heart failure patients and the higher risk volume-overloaded heart failure patients. Similarly, troponin levels helped differentiate the patients with prior myocardial infarction (MI) who were completely revascularized vs. those with severe residual obstructive coronary artery disease. Less obvious, however, is the method by which GDF-15 levels improved the ability of the ABC death risk score to predict mortality in AF. For readers without a special interest in biomarkers, GDF-15 may be unfamiliar. Originally called macrophage inhibitory cytokine (MIC-1), it was described over two decades ago. It has structural similarities to proteins of the transforming growth factor-beta (TGF-b) superfamily, but actually represents a distinct class of its own. It is up-regulated by pro-inflammatory cytokines (e.g. interferon-c, interleukin-1, TGF-b) and is believed to inhibit the later phases of macrophage activation, thereby exerting an overall anti-inflammatory effect. While GDF-15/MIC-1 remains incompletely understood, levels of circulating GDF-15 can be viewed as a measure of systemic inflammation and oxidative stress. Disorders such as hypertension, obesity, diabetes, and vascular disease are well established risk factors for AF, and there is substantial evidence that inflammation and oxidative stress are major mechanisms by which these disorders promote AF pathogenesis. From this perspective it can be speculated that GDF15 may be a quantitative measure of the severity and cumulative effect of a variety of co-morbidities associated with mortality risk in AF (hypertension, obesity, diabetes, and vascular disease). GDF-15 has been associated with bleeding risk and has also been

中文翻译：

---

ABC 死亡风险评分：是时候开始测量 GDF-15 了吗？

评估发病率和死亡率风险是照顾心脏病患者的一个重要方面，可指导患者咨询和实施治疗。由于涉及心血管疾病的病理生理机制是多重和复杂的，将反映病理生理学不同组成部分的因素结合到风险评分中可以帮助提供风险的整体指示。理想情况下，临床风险评分由临床医生容易获得的数据组成。大多数用于预测房颤 (AF) 中风或出血的风险评分都遵循此规则（即 CHADS-VASC、HAS-BLED、ORBIT 和 ATRIA）。在本期期刊中，Hijazi 博士及其同事得出了临床风险评分，即 ABC 死亡风险评分，其中包括测量 GDF-15（生长分化因子 15）——一种最近可供临床实验室使用的生物标志物。他们的研究结果让我们思考以下问题：我们的医院和诊所是否应该投资开始测量 GDF-15，以及预测抗凝治疗 AF 患者死亡率的风险评分是否对我们的实践有用。在这项研究中，来自参加 ARISTOTLE 试验（阿哌沙班与华法林）的 14 611 名参与者的数据用于推导风险评分以预测全因死亡率。使用 13 个标准临床变量、4 个标准生物标志物和 GDF-15，后向消除算法选择了 5 个变量以包含在最终模型中 [年龄、充血性心力衰竭、肌钙蛋白、脑钠肽 (BNP) 和 GDF-15]。该模型在推导队列中以 0.74 [95% 置信区间 (CI) 0.73–0.76] 和独立验证队列（RELY 试验；达比加群与华法林）中的 0.74 的 c 统计量预测全因死亡率，令人印象深刻。该模型称为 ABC 死亡风险评分，与 ABC 出血和 ABC 中风风险评分一起，是 Hijazi 博士及其同事使用来自 ARISTOTLE 生物标志物子研究的数据提出的第三个临床风险评分（图 1）。有趣的是，他们还表明，目前的评分比仅包含临床变量的综合模型表现更好 [cstatistic 为 0.68 (95% CI 0.66–0.70)]，证明了定量生物标志物显着改善风险预测的潜在能力。对于某些生物标志物，存在明确的病理生理学联系，与用于临床变量的二元定义相比，它们添加了有关疾病严重程度的宝贵信息。例如，在 ABC 死亡风险评分中，增加 BNP 水平可能有助于区分低风险的代偿性心力衰竭患者和高风险的容量超负荷心力衰竭患者。同样，肌钙蛋白水平有助于区分完全血运重建的既往心肌梗死 (MI) 患者与严重残留阻塞性冠状动脉疾病的患者。然而，不太明显的是 GDF-15 水平提高 ABC 死亡风险评分预测 AF 死亡率能力的方法。对于对生物标志物没有特殊兴趣的读者，GDF-15 可能比较陌生。最初称为巨噬细胞抑制性细胞因子 (MIC-1)，它在二十多年前被描述。它与转化生长因子-β (TGF-b) 超家族的蛋白质具有结构相似性，但实际上代表了它自己的一个独特类别。它被促炎细胞因子（例如干扰素-c、白细胞介素-1、TGF-b）上调，据信可抑制巨噬细胞活化的后期阶段，从而发挥整体抗炎作用。虽然 GDF-15/MIC-1 仍未完全了解，但循环 GDF-15 的水平可被视为全身炎症和氧化应激的衡量标准。高血压、肥胖、糖尿病和血管疾病等疾病是公认的 AF 危险因素，并且有大量证据表明炎症和氧化应激是这些疾病促进 AF 发病机制的主要机制。从这个角度来看，可以推测 GDF15 可能是衡量与 AF 死亡风险（高血压、肥胖、糖尿病和血管疾病）相关的各种合并症的严重程度和累积效应的定量指标。GDF-15 与出血风险相关，并且