Appropriately designed and conducted randomized clinical trials (RCTs) are foundational for the practice of evidence-based medicine, including the regulatory approval of therapeutics and the inclusion of those therapies in clinical practice guidelines. Broadly defined, RCTs can be considered in 2 categories: explanatory or mechanistic trials that examine the effect of an intervention on biological measures and clinical outcome trials that are testing the effect of a treatment on what matters to patients and clinicians, including living longer, feeling better, avoiding unpleasant experiences, or spending less money. Many years ago, Topol and Califf noted the following:

Article, see p 372

…both types of trials are needed to advance the clinical treatment of acute myocardial infarction. Mega-trials can provide definitive evidence about the mortality reduction afforded by a class of therapy so that broad changes in clinical practice can be justified. Mini-trials can explain why a treatment is effective to allow development of more effective approaches attacking the identified mechanisms.¹

In the intervening years since Topol and Califf put forth this observation, cardiovascular clinical investigators have performed both types of trials, looking to reduce the risk of complications that arise from the development and progression of atherosclerotic coronary artery disease. RCTs, intended to guide or change clinical practice, have typically examined whether a proposed therapeutic strategy can reduce important clinical outcomes such as death, myocardial infarction, stroke, hospitalizations, revascularization procedures, and anginal symptoms. Trials, typically smaller and sometimes embedded in a larger outcomes RCT, have also been performed to establish a possible relationship between a biomarker and the more important clinical outcomes, although it is also well accepted that care must be taken in the interpretation of such relationships as being causal.²

Lipoproteins have been the center of much interest in the understanding of atherosclerosis and subsequently as therapeutic targets in the development of drugs intended to interfere with their biological actions. There are 3 classes of drugs that lower low-density lipoprotein (LDL) that are proven to also reduce the cardiovascular complications associated with coronary atherosclerosis. Multiple large-scale RCTs have definitively shown that statins, ezetimibe, and proprotein convertase subtilisin/kexin type 9 inhibitors all improve clinical outcomes among patients with or at risk for coronary artery disease and that this clinical benefit is attributable largely to the LDL-lowering effects of the drugs.³ Conversely, drugs intended to increase high-density lipoprotein have not been associated with clinical outcome improvements in large-scale RCTs despite epidemiological and biomarker data that would suggest that raising high-density lipoprotein could, as with LDL lowering, improve clinical outcomes among those with coronary artery disease⁴ and that RCTs of lipoprotein(a) lowering have not yet been completed.⁵ All of these experiences support the notion that it is insufficient to understand the biological mechanism of a therapeutic intervention to know its effect on human health; instead, appropriately designed RCTs are needed that measure and report meaningful clinical outcomes.⁶

Triglycerides also have been shown to be a marker of cardiovascular risk, even among patients with coronary artery disease already treated with LDL-lowering therapies.⁷ Less certain have been the clinical outcome effects of therapies that lower triglycerides, despite the epidemiological and biomarker observations that such a therapeutic intervention might have meaningful clinical benefits. In 2019, Bhatt and colleagues⁸ reported the primary results of REDUCE-IT (Reduction of Cardiovascular Events With Icosapent Ethyl–Intervention Trial) showing that treatment with icosapent ethyl in patients with increased cardiovascular risk and fasting triglycerides of 135 to 499 mg/dL lowered the risk of cardiovascular events by ≈25% compared with a mineral oil placebo. An accompanying editorial noted that “the cardiovascular benefits of icosapent ethyl were greater than would be predicted on the basis of the changes in triglyceride levels” and suggested that more work was needed both to understand the biological mechanisms associated with the therapy and to untangle any interfering effects of the mineral oil placebo.⁹ In late 2019, the US Food and Drug Administration approved icosapent ethyl for commercial use on the basis of the REDUCE-IT results.¹⁰

Subsequently, a second large trial testing a different omega-3 fatty acid supplement, omega-3 carboxylic acid, against a corn oil placebo was reported and showed no effect on lowering major cardiovascular events among a high-risk group of patients.¹¹ These 2 seemingly contradictory study reports from well-done large trials raised questions in the cardiovascular community about the clinical usefulness of these therapies in patients at high cardiovascular risk.¹² Included in these debates was the question of how much the use of mineral oil as the placebo comparator contributed to the difference observed between the active therapy and placebo in REDUCE-IT. In other words, did mineral oil have proinflammatory properties that might have amplified the benefits of icosapent ethyl, which has been shown to have anti-inflammatory effects?

The REDUCE-IT Biomarker Substudy, published by Ridker et al in this issue of Circulation, attempts to provide some mechanistic insights into this debate.¹³ In this substudy, the investigators compared samples in each of the randomized treatment groups (icosapent ethyl, n=4089; mineral oil, n=4090) at baseline, 12 months, 24 months, and the end of the study. The assays represented biomarkers in lipid metabolism and the inflammatory process, both shown to be associated with atherosclerotic disease and cardiovascular risk. All values were similar at baseline between the randomized treatment groups. In the patients randomized to icosapent ethyl, there were minimal changes in the inflammatory markers from baseline to 12 and 24 months and a very modest decrease in LDL, whereas in the mineral oil group, the inflammatory markers and LDL increased compared with baseline, so the differences between treatment groups at 12 and 24 months were explained by the mineral oil changes. Icosapent ethyl was associated with a significant decrease in triglycerides at 12 months compared with a modest increase in the mineral oil group. The authors use appropriate caution in their interpretation of the findings, noting that the trial was designed as a clinical outcome study and not as one to demonstrate which mechanistic finding contributed to that benefit. However, they also conclude that sorting out this disconnect between clinical outcomes and biological findings probably requires another RCT that does not involve mineral oil as the comparator.

So, what are the lessons to be learned from the REDUCE-IT experience? First, kudos to the investigators for the transparency in putting these biomarker data into the public space. Acknowledging a problem or limitation in a trial in which one dedicated up to a decade of one’s professional life is difficult but the right thing to do, especially because these data might prove informative for others in considering study designs, including a choice of placebo materials. Second, the investigators (and the sponsor) should be congratulated for embedding the collection of biosamples into this large, global RCT. This is not an easy task given the tensions between the desire for simplicity and pragmatism in large outcome studies to more efficiently and quickly answer the clinical question and the scientific desires of investigators (and sponsors) to better understand mechanisms that might add insights into how the therapy might ultimately be used or that might prove helpful through mechanistic substudies that further inform the broader field.¹⁴

The issues related to placebo control are perhaps the most important learnings here. In its “Guidance for Industry, Choice of a Control Group and Related Issues in Clinical Trials,” the US Food and Drug Administration notes that “the placebo control design, by allowing blinding and randomization and including a group that receives an inert treatment, controls for all potential influences on the actual or apparent course of the disease other than those arising from the pharmacologic action of the test drug.”¹⁵ The data presented by Ridker and colleagues raise the difficult question not of whether icosapent ethyl improves clinical outcomes in high-risk cardiovascular patients but of the magnitude of that benefit and whether that benefit is sufficiently robust for recommendations to guide clinical practice. Icosapent ethyl was superior to mineral oil in a large, well-designed RCT in reducing important clinical events. However, the mechanistic work presented here by the investigators also shows that mineral oil was not inert, with a number of proinflammatory markers increased over time and no measured anti-inflammatory effects associated with icosapent ethyl, only a decline, as expected, in triglycerides. In addition, although modeling of the contributions of various biomarkers to the outcomes to create a hypothetical framework for parsing the clinical benefit of icosapent ethyl is possible, it is just that: a modeling exercise that forms a hypothesis. The hard reality is that we are left with uncertainties and the questions raised by the use of the mineral oil as placebo that can be answered only by another RCT. Although this recommendation comes with the acknowledgment of the time, effort, and financial resources required to carry out another trial, it is important from a public health perspective given the large number of high-risk patients who might be considered for treatment with icosapent ethyl.

The author thanks David J. Maron, MD, and Kenneth W. Mahaffey, MD, for their careful review, edits, and expert clarifications.

None.

Disclosures Dr Harrington reports grants to Stanford University from CSL Behring.

Circulation is available at www.ahajournals.org/journal/circ

The opinions expressed in this article are not necessarily those of the editors or of the American Heart Association.

The podcast and transcript are available as Supplemental Material at https://www.ahajournals.org/doi/suppl/10.1161/CIRCULATIONAHA.122.060649.

For Sources of Funding and Disclosures, see page 382.

适当设计和实施的随机临床试验 (RCT) 是循证医学实践的基础，包括对疗法的监管批准以及将这些疗法纳入临床实践指南。从广义上讲，随机对照试验可分为两类：检查干预措施对生物学指标的影响的解释性或机制性试验，以及测试治疗对患者和临床医生重要事项的影响的临床结果试验，包括寿命更长、感觉更好，避免不愉快的经历，或花更少的钱。许多年前，Topol 和 Califf 注意到以下几点：

文章，见第 372 页

…两种类型的试验都需要推进急性心肌梗死的临床治疗。大型试验可以提供关于一类疗法降低死亡率的明确证据，从而证明临床实践的广泛变化是合理的。小型试验可以解释为什么治疗有效，可以开发更有效的方法来攻击已识别的机制。^1个

在 Topol 和 Califf 提出这一观察结果后的几年里，心血管临床研究人员进行了两种类型的试验，以期降低因动脉粥样硬化性冠状动脉疾病的发展和进展而引起的并发症的风险。旨在指导或改变临床实践的随机对照试验通常检查所提出的治疗策略是否可以减少重要的临床结果，例如死亡、心肌梗塞、中风、住院、血运重建手术和心绞痛症状。还进行了通常规模较小且有时嵌入较大结果 RCT 中的试验，以确定生物标志物与更重要的临床结果之间可能存在的关系，^2个

脂蛋白一直是理解动脉粥样硬化的中心，随后在开发旨在干扰其生物学作用的药物中作为治疗靶点。有 3 类药物可降低低密度脂蛋白 (LDL)，这些药物已被证明还可减少与冠状动脉粥样硬化相关的心血管并发症。多项大规模随机对照试验明确表明，他汀类药物、依折麦布和前蛋白转化酶枯草杆菌蛋白酶/kexin 9 型抑制剂均可改善冠心病患者或有冠心病风险的患者的临床结果，并且这种临床益处主要归因于 LDL 降低作用的药物。^3个相反，旨在增加高密度脂蛋白的药物与大规模随机对照试验中的临床结果改善无关，尽管流行病学和生物标志物数据表明，提高高密度脂蛋白与降低低密度脂蛋白一样，可以改善患有这些疾病的患者的临床结果。冠状动脉疾病⁴和降低脂蛋白 (a) 的随机对照试验尚未完成。⁵所有这些经验都支持这样一种观点，即了解治疗干预的生物学机制不足以了解其对人类健康的影响；相反，需要适当设计的随机对照试验来衡量和报告有意义的临床结果。^6个

甘油三酯也被证明是心血管风险的标志，即使在已经接受低密度脂蛋白降低疗法治疗的冠状动脉疾病患者中也是如此。⁷降低甘油三酯疗法的临床结果效果不太确定，尽管流行病学和生物标志物观察表明这种治疗干预可能具有有意义的临床益处。2019 年，Bhatt 及其同事⁸报告了 REDUCE-IT（减少心血管事件与 Icosapent 乙基干预试验）的主要结果表明，使用 icosapent 乙基治疗心血管风险增加且空腹甘油三酯为 135 至 499 mg/dL 的患者可将心血管事件的风险降低 ≈ 25% 与矿物油安慰剂相比。随附的一篇社论指出，“二十碳五烯乙酯对心血管的益处大于根据甘油三酯水平的变化所预测的”，并建议需要做更多的工作来了解与该疗法相关的生物学机制并理清任何干扰因素矿物油安慰剂的作用。⁹2019 年底，美国食品药品监督管理局根据 REDUCE-IT 结果批准了 icosapent ethyl 的商业用途。¹⁰

随后，第二项大型试验报道了一种不同的 omega-3 脂肪酸补充剂 omega-3 羧酸与玉米油安慰剂的对比，结果表明它对降低高危患者的主要心血管事件没有影响。¹¹这 2 份来自完成良好的大型试验的看似矛盾的研究报告在心血管界提出了关于这些疗法在高心血管风险患者中的临床实用性的问题。¹²这些争论中包括的问题是，使用矿物油作为安慰剂对照物对 REDUCE-IT 中观察到的活性疗法和安慰剂之间的差异有多大影响。换句话说，矿物油是否具有促炎特性，可能会放大已被证明具有抗炎作用的二十碳五烯乙酯的益处？

Ridker 等人在本期Circulation上发表的 REDUCE-IT Biomarker Substudy试图对这场辩论提供一些机制上的见解。¹³在这项子研究中，研究人员在基线、12 个月、24 个月和研究结束时比较了每个随机治疗组（二十碳五烯酸乙酯，n=4089；矿物油，n=4090）的样本。这些测定代表了脂质代谢和炎症过程中的生物标志物，两者都被证明与动脉粥样硬化疾病和心血管风险有关。随机化治疗组之间的所有基线值都相似。在随机分配给二十碳五烯乙酯的患者中，从基线到 12 个月和 24 个月炎症标志物的变化很小，LDL 有非常适度的降低，而在矿物油组中，炎症标志物和 LDL 与基线相比有所增加，因此12 个月和 24 个月时治疗组之间的差异可以用矿物油的变化来解释。与矿物油组适度增加相比，二十碳五烯乙酯与 12 个月时甘油三酯的显着降低有关。作者在解释研究结果时采取了适当的谨慎态度，并指出该试验是作为一项临床结果研究而设计的，而不是用来证明哪种机制发现有助于该益处。然而，他们还得出结论，解决临床结果和生物学发现之间的这种脱节可能需要另一项不涉及矿物油作为比较对象的随机对照试验。注意到该试验被设计为一项临床结果研究，而不是证明哪种机制发现有助于该益处的研究。然而，他们还得出结论，解决临床结果和生物学发现之间的这种脱节可能需要另一项不涉及矿物油作为比较对象的随机对照试验。注意到该试验被设计为一项临床结果研究，而不是证明哪种机制发现有助于该益处的研究。然而，他们还得出结论，解决临床结果和生物学发现之间的这种脱节可能需要另一项不涉及矿物油作为比较对象的随机对照试验。

那么，从 REDUCE-IT 经验中可以吸取哪些教训呢？首先，感谢调查人员将这些生物标志物数据公开的透明度。在一项试验中承认问题或局限性是一件困难的事情，在一项试验中，一个人投入了长达十年的职业生涯，但却是正确的做法，尤其是因为这些数据可能对其他人在考虑研究设计（包括安慰剂材料的选择）时提供信息。其次，应该祝贺研究人员（和赞助商）将生物样本的收集嵌入到这个大型的全球 RCT 中。¹⁴

与安慰剂对照相关的问题可能是这里最重要的知识。美国食品和药物管理局在其“行业指南、对照组的选择和临床试验中的相关问题”中指出，“安慰剂对照设计，通过允许盲法和随机化，并包括接受惰性治疗的组，控制除试验药物的药理作用引起的影响外，对疾病的实际或表观过程的所有潜在影响。” ¹⁵Ridker 及其同事提出的数据提出了一个棘手的问题，不是二十碳五烯乙酯是否能改善高危心血管患者的临床结果，而是这种益处的大小，以及这种益处是否足够强大，可以推荐用于指导临床实践。在一项精心设计的大型随机对照试验中，二十碳五烯乙酯在减少重要临床事件方面优于矿物油。然而，研究人员在此展示的机理研究还表明，矿物油并不是惰性的，随着时间的推移，许多促炎标志物会增加，并且没有测量到与二十碳五烯酸乙酯相关的抗炎作用，正如预期的那样，甘油三酯只会下降。此外，尽管可以对各种生物标志物对结果的贡献进行建模，以创建一个假设框架来解析 icosapent ethyl 的临床益处，但这只是：形成假设的建模练习。残酷的现实是，我们面临着不确定性和使用矿物油作为安慰剂所引发的问题，这些问题只能通过另一项 RCT 来回答。尽管这一建议伴随着对开展另一项试验所需的时间、努力和财政资源的认可，但从公共卫生的角度来看，考虑到可能考虑使用 icosapent ethyl 治疗的大量高危患者，这一建议很重要。残酷的现实是，我们面临着不确定性和使用矿物油作为安慰剂所引发的问题，这些问题只能通过另一项 RCT 来回答。尽管这一建议伴随着对开展另一项试验所需的时间、努力和财政资源的认可，但从公共卫生的角度来看，考虑到可能考虑使用 icosapent ethyl 治疗的大量高危患者，这一建议很重要。残酷的现实是，我们面临着不确定性和使用矿物油作为安慰剂所引发的问题，这些问题只能通过另一项 RCT 来回答。尽管这一建议伴随着对开展另一项试验所需的时间、努力和财政资源的认可，但从公共卫生的角度来看，考虑到可能考虑使用 icosapent ethyl 治疗的大量高危患者，这一建议很重要。

作者感谢医学博士 David J. Maron 和医学博士 Kenneth W. Mahaffey 的仔细审查、编辑和专家澄清。

没有任何。

披露Harrington 博士报告了 CSL Behring 对斯坦福大学的资助。

流通可在 www.ahajournals.org/journal/circ

本文表达的观点不一定代表编辑或美国心脏协会的观点。

播客和成绩单可作为补充材料在 https://www.ahajournals.org/doi/suppl/10.1161/CIRCULATIONAHA.122.060649 获得。

有关资金来源和披露，请参见第 382 页。