Heart failure in <scp>SAVOR‐TIMI</scp> 53: The hindsight of diabetic retinopathySAVOR‐TIMI 53研究中的心脏衰竭：糖尿病视网膜病变的后尘

The SAVOR-TIMI 53 Trial (Saxagliptin Assessment of Vascular Outcomes Recorded in Patients with Diabetes Mellitus conducted by the Thrombolysis in Myocardial Infarction Group) was designed to evaluate cardiovascular safety and efficacy in patients with diabetes who, by virtue of a history of established cardiovascular disease or multiple risk factors for it, were at high risk of cardiovascular events. Although the primary and secondary composite endpoints were similar in placebo and saxagliptin groups, subjects assigned to receive saxagliptin experienced an unanticipated 27% relative increase (0.7% absolute risk over 2 years) in hospitalization for heart failure.1 Although conceivably a play of chance, several factors suggest that this may not necessarily be the case. Heart failure in SAVOR-TIMI 53 was centrally and blindly adjudicated with statistical adjustment for multiple comparisons. Furthermore, the EXAMINE (Examination of Cardiovascular Outcomes with Alogliptin versus Standard of Care in Patients with Type 2 Diabetes and Acute Coronary Syndrome) Trial, which compared another dipeptidyl peptidase (DPP)-4 inhibitor, namely alogliptin, with placebo also noted a similar numerical trend,2 so that when the findings of SAVOR and EXAMINE were combined, a significant odds ratio (OR) for heart failure hospitalization was apparent (OR 1.24; 95% confidence interval [CI] 1.07–1.44; P = 0.004).3 Probing the trial database, the TIMI Group has identified several factors most strongly associated with hospitalization for heart failure, regardless of treatment. These include prior heart failure, estimated glomerular filtration rate ≤60 mL/min per 1.73 m2, and elevated N-terminal pro-B-type natriuretic peptide (NT-proBNP).4 Importantly, the Group also conducted landmark analyses demonstrating that most of the increased risk of heart failure hospitalization with saxagliptin occurred early, with a 6 month rate of 1.1% compared with 0.6% in the placebo group (hazard ratio [HR] 1.80; 95% CI 1.29–2.55; P = 0.001) and 12 month rates of 1.9% vs 1.3% in the two groups, respectively (HR 1.46; 95% CI; 1.15–1.88; P = 0.002). The corresponding post-6 and 12 month rates were not different between the two treatment groups where HRs were 1.11 (P = 0.31) and 1.09 (P = 0.51), respectively.4 To further examine the duration-dependent attenuation in risk, a time varying coefficients model was also developed. This showed that the increased risk of heart failure hospitalization with saxagliptin subsided by 10–11 months after randomisation4 (Fig. 1). Time and risk of heart failure hospitalization where the number of events are equally distributed over 10 time intervals. The lower confidence interval of the log hazard ratio (HR) crosses 0 at around 314 days, suggesting that approximately 10–11 months is when the risk of hospitalization for heart failure with saxagliptin subsided in this trial. (Reproduced with permission from Scirica et al.4) This early but unsustained worsening of a diabetes complication with an antihyperglycemic intervention, as seen with heart failure hospitalization in SAVOR-TIMI 53, is not entirely new. Some 30 years ago, the Kroc Collaborative Study Group reported in their trial of 70 individuals with type 1 diabetes and non-proliferative diabetic retinopathy, that institution of tight glycaemic control, with an approximate 2% reduction in HbA1c, seemed to worsen disease at 8 months.5 However, by 2 years the extent of retinopathy in the conventional and intensive groups was indistinguishable.6 A similar early, 3–6 month exacerbation of retinopathy with stabilization at 1 year with intensive glycemic control was also reported in the smaller Oslo study.7 Indeed, this early acceleration of disease was further noted in the Diabetes Control and Complications Trial (DCCT), in which at 6 and/or 12 months, worsening in retinopathy was evident in 13.1% of those assigned to intensive treatment compared with only 7.6% in the conventionally treated group (odds ratio [OR] 2.06; P < 0.001).8 By 18 months, these between-group differences had dissipated, and after 2–3 years the intensively treated group was beginning to clearly fare better than those assigned to conventional management.9 Although neither SAVOR TIMI 35 nor EXAMINE were studies of glycemic control, HbA1c was approximately 0.3% lower in DPP-4- than placebo-treated patients. Although a 0.3% variance in glycemia may seem small in comparison with the Kroc Study and DCCT, the size of the study population in SAVOR TIMI needs to be considered. In the Outcome Reduction with an Initial Glargine Intervention (ORIGIN) study, a similarly large study, subjects whose baseline HbA1c was above the 6.4% median and who were randomized to intervention with insulin glargine achieved a 0.32% reduction in HbA1c with a 10% reduction in microvascular complications.10 These findings suggest that the consequences of a small reduction in HbA1c may become apparent in a sufficiently large study. Although we tend to generalize the findings of antihyperglycemic interventions within the micro- and macrovascular categories, within these not all complications necessarily respond with the same time frame. In contrast with the early exacerbation of retinopathy, albumin excretion rates fell within the first 8 months of improved glycemic control in the Kroc Study.5 Accordingly, glucose reduction alone may well explain the positive effects of albuminuria in SAVOR-TIMI 53 without the need to invoke any potential pleiotropic actions of DPP-4 inhibition. Although the failure to include heart failure as an outcome along with its under-representation in many trials of antihyperglycemic agents and glycemic control is lamentable, some have done so. In reviewing the findings of these studies, it is important to note that unlike the recent time-dependent analysis of SAVOR-TIMI 53, outcomes are traditionally reported for the entire study. This is particularly relevant when we consider the findings of diabetic retinopathy, where an early exacerbation was followed by a later stabilization and subsequent improvement, so that length of the study becomes paramount. In line with this, a retrospective cohort study the UK General Practice Research Database assessed the incidence of heart failure in 25 690 patients with newly diagnosed diabetes.11 Patients with any drug use during the first year after diabetes diagnosis had a 4.75-fold increased risk of heart failure compared with those who remained drug free. However, no increased risk was evident in subsequent years. Notably, the drugs administered did not include thiazolidinediones, DPP-4 inhibitors or glucagon-like peptide-1 (GLP-1) analogues. Hypothesizing possible mechanisms that may have contributed to the increased likelihood of hospitalization for heart failure, some investigators have focused on a range of non-incretin proteins that, like GLP-1, are also degraded by DPP-4.12, 13 However, although increased activity of certain DPP-4 substrates, such as substance P and neuropeptide Y, may be detrimental, others may be beneficial so that the totality of effects remains speculative. Alternatively, rather than an effect of GLP-1 or the non-incretin substrates of DPP-4, the key issue may be glucose itself. As a consequence of reduced glucose transporter GLUT4 activity in the setting of insulin resistance,14 glucose, the preferred substrate of the stressed cardiac myocyte, is less avidly transported in the failing diabetic heart.15 Accordingly, a diminution in the transmembrane glucose gradient, as a consequence of blood glucose lowering, may lead to a further reduction in the myocyte's supply of energy for contraction and the active phase of diastolic relaxation. However, with time, adaptive processes, such as increased expression of sodium–glucose cotransporter 1 (SGLT-1), would, in theory, mitigate these changes.16 Clearly, we will have to wait quite some time before fully understanding the mechanisms that contribute to the development of heart failure in studies such as SAVOR-TIMI 53. In the meantime, we can watchfully await the findings of the numerous cardiovascular safety studies that are soon to report, keenly focusing on heart failure and its time course. Hopefully, these studies will provide insight as to whether an early increase in heart failure risk is truly a feature of the DPP-4 class or whether similar risks also apply to GLP-1 mimetics, SGLT-2 inhibitors, insulins, and sulfonylureas,17 the potentially confounding issues of study duration and glycemic differences notwithstanding. RG is the Canadian Research Chair in Diabetes Complications. RG has received research grants to his institution, participated in advisory boards and received honoraria for continuing medical education engagements from Merck, Astra Zeneca, Bristol-Myers Squibb, Boehringer-Ingelheim, and Eli Lilly and Co. SAVOR-TIMI 53试验（评估使用沙格列汀后在糖尿病患者心肌梗死溶栓组中记录到的血管结果）的设计目的是在既往有明确心血管疾病史或者有多个心血管疾病的危险因素因而处于心血管事件高危风险的糖尿病患者中评估心血管的安全性与有效性。虽然安慰剂组与沙格列汀组的主要以及次要复合终点的发生率相似，但是未曾预料到的是接受沙格列汀治疗的患者组因为心脏衰竭而住院的相对风险增加了27%（2年绝对风险为0.7%）1。虽然有人认为这可能是一个巧合，但是一些因素却提示并非如此。在SAVOR-TIMI 53试验中有关心脏衰竭的统计结果是对多重比较的统计结果进行中心化和盲后调整后得出的。此外，在EXAMINE试验（在合并急性冠状动脉综合征的2型糖尿病患者中比较阿格列汀治疗与标准治疗对心血管结果的影响）中，对另外一种二肽基肽酶（DPP）-4抑制剂，即阿格列汀，与安慰剂进行了对比，结果也发现了相似的数值趋势2，因此将SAVOR与EXAMINE试验的结果相结合之后，因为心脏衰竭而住院的优势比（OR）就更加明显了（OR为1.24；95% CI为1.07-1.44；P = 0.004）3。 调查试验数据库后发现，无论是哪种治疗方案，在TIMI组中都有几个明确与心脏衰竭住院相关的危险因素。这些因素包括既往心脏衰竭、估算的肾小球滤过率≤ 60 mL/min每1.73m2、以及N末端B型尿钠肽（NT-proBNP）升高4。重要的是，对这一组患者还进行了具有里程碑意义的分析，结果表明在沙格列汀组中因心脏衰竭而住院的风险增加最明显的是在早期，在第6个月时比例为1.1%，而对照的安慰剂组为0.6%（危险比[HR]为1.80；95% CI为1.29-2.55；P = 0.001），在第12个月时两组的比例分别为1.9%与1.3%（HR为1.46；95% CI为1.15–1.88；P = 0.002）。而在两个治疗组之间相应的6个月以及12个月之后的比例就没有差异了，HR分别为1.11（P = 0.31）与1.09（P = 0.51）4。为了进一步研究风险是否呈疗程依赖性衰减，目前已经建立了一个时变系数模型。这个模型显示沙格列汀治疗组患者在随机化10–11个月之后增加的心脏衰竭住院风险就消失了（图1）4。 这种由降糖药物治疗后导致的早期但是持续时间不长的糖尿病并发症恶化，就像在SAVOR-TIMI 53试验中看到的心脏衰竭住院风险增加，并不是第一次才发现。大约在30年前，Kroc协作研究组就已经报告了在他们的试验中有70名合并非增殖性糖尿病视网膜病变的1型糖尿病患者严格控制血糖8个月之后（HbA1c大约下降了2%）病情似乎恶化了5。然而，2年之后常规组与强化组之间的视网膜病变程度就没有差别了6。在一个较为小型的Oslo研究中也报告了相似的结果，他们发现强化血糖控制后，视网膜病变在早期亦即3–6个月时出现了恶化，但是1年之后就稳定下来了7。确实，除此之外在糖尿病控制与并发症试验（Diabetes Control and Complications Trial，DCCT）中也观察到了这种早期的病情恶化，在这个试验中，在第6个月和/或第12个月时强化治疗组的患者中有13.1%的视网膜病变明显恶化，而在对照的常规治疗组这个比例只有7.6%（优势比[OR]为2.06；P < 0.001）8。到了第18个月，两组之间的这些差异消失，而2-3年后强化治疗组患者与常规治疗组患者相比病情开始变得明显好转9。 虽然SAVOR TIMI 53和EXAMINE都不是控制血糖的研究，但是与使用安慰剂治疗的患者相比DPP-4抑制剂治疗组的HbA1c大约都降低了0.3%。虽然与Kroc研究以及DCCT研究相比0.3%的血糖控制差异看起来可能很小，但是应该考虑到SAVOR TIMI的研究人群规模。在使用甘精胰岛素初始干预改善结果（Outcome Reduction with an Initial Glargine Intervention，ORIGIN）研究（同样是一项大型研究）中，受试者的基线HbA1c中位数为6.4%以上，被随机分配到甘精胰岛素干预组的患者HbA1c下降了0.32%，微血管并发症也减少了10%10。这些研究结果表明，在足够大型的研究中HbA1c即使下降很少也可能导致明显的结果改善。 虽然我们倾向于在微血管与大血管范畴内概括降糖药物干预的结果，但是并非在同一时间框架内要求所有的并发症都必须有所改善。与Kroc研究中血糖控制改善后早期发生的视网膜病变恶化相反，白蛋白排泄率在前8个月内就会下降5。因此，在SAVOR-TIMI 53试验中单纯的降糖作用就可以很好地解释蛋白尿减少的阳性结果并且不需要使用任何DPP-4抑制剂潜在的多效作用来解释。 虽然不能将许多降糖药物与血糖控制试验的心脏衰竭归纳为代表性不足的结果，这一点很可悲，但是有一些试验就是这样做的。在回顾这些研究的结果时，值得注意的是它们与近期SAVOR-TIMI 53试验所做的随时间变化的分析不一样，它们的结果都是传统地按照整个研究来报告。当我们考虑到糖尿病视网膜病变的结果时这一点特别重要，因为在血糖改善之后的视网膜病变都是早期先恶化稍后就稳定了下来，而最终都得到了改善，因此研究时长就变得特别重要。与此相一致的是，在一项利用英国综合医疗研究数据库的回顾性队列研究中，评估了25690名新诊断糖尿病患者的心脏衰竭的发生率11。与不使用药物的患者相比，使用任何药物进行治疗的患者在诊断糖尿病之后的第一年期间出现心脏衰竭的风险增加了4.75倍。然而，在随后的几年中风险的增加却不明显。值得注意的是，使用的药物并不包括噻唑烷二酮类、DPP-4抑制剂或者胰高血糖素样肽-1（GLP-1）类似物。 有一些研究者对DPP-4抑制剂增加心脏衰竭住院风险的可能机制进行了假设，他们已经开始聚焦于对一系列非肠促胰素蛋白的研究，就像GLP-1一样，它们也能够被DPP-4所降解12, 13。然而，虽然DPP-4的某一个底物活性增加了，例如P物质与神经肽Y，它们可能是有害的，但是其他底物却可能是有益的，因此总体效应仍然有待推测。或者，这既不是一种GLP-1效应也不是一种DPP-4的非肠促胰岛素底物的作用结果，关键还是血糖本身的问题。在胰岛素抵抗的环境下葡萄糖转运体GLUT4活性下降的结果是14，葡萄糖，心肌细胞应激的首选底物，在衰竭的糖尿病心脏中急需的转运更少了15。因此，作为血糖降低的结果，跨膜葡萄糖浓度梯度下降以后可能导致心肌细胞收缩以及舒张活跃期所需要的能量供应进一步地减少。然而，随着时间的过去，适应这些过程后，例如随着钠-葡萄糖共转运体1（SGLT-1）表达的增加，理论上这些变化将会减轻16。 目前很明确，在我们能够完全了解诸如SAVOR-TIMI 53研究中导致心脏衰竭发生的机制之前，我们将不得不继续等待相当长的一段时间。在此期间，我们可以小心翼翼地等待众多即将报告的心血管安全性研究结果，特别是要聚焦于心脏衰竭及其随时间变化的情况。我们希望这些研究能够进一步明确早期的心脏衰竭风险增加是否真的就是DPP-4抑制剂类药物的特点，或者是使用GLP-1类似物、SGLT-2抑制剂、胰岛素以及磺脲类药物17治疗也有相似的风险，尽管还混杂了潜在的研究期限与血糖控制情况各有差异等问题。 图1 心脏衰竭住院风险与时间的关系图，事件的数目平均分布在10个时间间隔中。在第314天左右对数危险比（HR）可信区间的下限与0相交叉，这意味着在这个试验中使用沙格列汀治疗的患者因为心脏衰竭而住院的风险大约经过10–11个月就消退了（转载得到了Scirica等4的许可）。

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.