See related article, p 3399

Wechsler et al¹ present an opinion of the STAIR group, an international panel of stroke clinicians, industry, and regulatory representatives, who attended an invitation only workshop. Their principal argument is that further research is required in the setting of stroke to determine how reperfusion therapy—endovascular thrombectomy (EVT) and/or intravenous thrombolysis (IVT)—may be extended to larger groups of patients on the basis that initial trial inclusion criteria were too strict, and many more patients outside these criteria might benefit.

Randomized controlled trials, by their very nature, strictly limit study populations. This means that their results may only be applicable to narrowly defined groups of patients—an epidemiological challenge known as the generalizability gap. This has always posed the problem for promising new therapies of how to identify the outer limits of benefit. Traditionally, the zone of uncertainty about the applicability of a novel therapy has been gradually reduced through a variety of (often haphazardly applied) mechanisms, including incrementally expanding clinical trials and systematic reviews, off-label prescribing, formal and informal debates and arguments involving case studies, personal conversations among clinicians and patients, and media reports.

Recently, more sophisticated techniques have been developed in attempts to achieve a more systematic process of expanding evidence and reducing the field of uncertain knowledge. Clinical registries, platform studies, pragmatic and adaptable designs, together with enhanced access to data through compulsory registration of trials and wider knowledge of negative trial outcomes, have greatly enhanced the speed and efficiency of making new treatments available.

EVT and IVT have been important steps forward in the management of stroke, a major cause of morbidity, mortality, and social cost. However, the introduction of these therapies has not been completely straightforward, in part because of the difficulty in identifying patient populations who are most likely to benefit, and in part because of the well documented risks of, sometimes devastating, adverse effects. For this reason, the treatments have aroused significant concerns in both the clinical and broader communities. After an extended debate, the importance of the role of IVT, and now EVT, in the management of stroke has become widely accepted and there is general consensus internationally regarding the clinical application of EVT and IVT.

Indeed, in the early time window (sub-6 hours for EVT and sub-4.5 hours for IVT), many consider that the consensus guidelines are quite liberal. Despite this, in many countries, patients who are currently eligible for EVT and IVT do not receive it. In the United States, for example, only 10% to 15% of patients with ischemic stroke are receiving IV thrombolysis, and only 3% are undergoing thrombectomy. While Wechsler and colleagues suggest that this creates an argument for expanding the inclusion criteria for EVT and IVT, we suggest that before extending these treatments to clinical settings in which evidence is lacking, it would seem more prudent and appropriate to focus on ensuring availability to those who are already eligible and for whom there is already a favorable risk-benefit. This is, of course, not a technological problem but a socioeconomic and ethical one. The reasons why people are not accessing treatments at present are complex but relate to lack of access to timely expert care, underpinned by the intertwined issues of socioeconomic status, rural/regional dwelling, racial background, and insurance status, issues that are, of course, by no means restricted to the United States.^2–4

Such a focus on improving access and implementing reperfusion therapy for all currently eligible patients is more likely to lead to clinically meaningful benefits at a population level than expanding treatment indications for EVT and IVT in a speculative manner without supporting evidence. At the same time, Wechsler et al have identified significant uncertainties that remain regarding the clinical application of EVT and IVT, the identification of patients who may benefit or be harmed by these therapies, and the means by which these uncertainties may be addressed. For example, should we ask the interventionalist to retrieve a distal clot (ie, beyond the M1 segment of the middle cerebral artery), when the outcome with IVT alone is quite good, and knowing that there is higher risk of vessel perforation when chasing clots in smaller caliber vessels? At what volume of irreversibly damaged brain (ischemic core) does thrombectomy still offer a chance of a good outcome? Indeed, at what volume of ischemic core is reperfusion is harmful or futile (providing no meaningful benefit)? What level of prestroke functional or cognitive impairment prohibits a good outcome? In these patients at the fringes, who are unlikely to have a traditionally defined good outcome with reperfusion, what level of long-disability is acceptable? These are complex questions. Indeed, recent Australian registry data found that the Disability Adjusted Life Years saved in patients treated with EVT in routine practice were considerably lower than reported from key EVT trials that used perfusion imaging to select patients.⁵ Many of the patients treated in routine practice were at the outer limits of the treatment guidelines, in particular, with less favorable pretreatment imaging characteristics.

The authors suggest the use of platform trials to increase efficiency and make the best use of individual patient outcomes. Also referred to as MAMS (multiarm, multistage) design trials, these studies (mostly used in cancer) commonly evaluate several interventions against a common control group and can be perpetual.⁶ The design incorporates prespecified adaptation rules to allow dropping of ineffective intervention(s) along with the flexibility of adding new intervention(s) during the trial. In regards to EVT and IVT it is not, however, obvious that platform trial methodologies would effectively resolve the uncertainties associated with their application. This is because their main contribution is to allow addition or subtraction of different treatments by adapting data across the trial arms, whereas the main questions about EVT and IVT relate largely to fringe populations with discrepant characteristics (eg, large core versus distal occlusion), for which borrowing information would be less likely to be applicable.

We suggest that to establish the limits of reperfusion therapy in stroke that it may be better, and more cost-effective, to evaluate the safety and efficacy issues for multiple well-defined subgroups of patients through a set of nimble and lean focused single-group trials with clear yes/no decisions, before taking these into an effectiveness story. This may be effectively achieved through the so-called basket/umbrella trial design of many subgroups that need to be circumscribed, rather than through a large perpetual platform.⁶ Considering we already know quite well what would happen to these pockets of patients under usual care scenarios, an objective performance criterion comparator structure like that used in device trials could be used to increase efficiency through only having a single group per study.

The authors also suggest that large pragmatic trials have a role in addressing generalizability and expected results in routine practice. While this may be true in general terms it does not help resolve the current problem. Patients for whom a treatment is not demonstrated to be safe and efficacious should not be included in large pragmatic effectiveness trials. This is simply because the risks are too great: without the safety mechanisms built into rigorously conducted RCTs the potential to cause harm would be unconscionably great. If patients are harmed the entire treatment approach will likely be compromised, with ongoing, far-reaching effects.

Relevant to trial design, of course, is the choice of outcome measure. The authors suggest we determine a Minimum Clinically Important Difference to help determine the outer limits of therapy. Typically, this would be established by consensus of a panel of experts (perhaps by invitation only). In the aforementioned large core patients, such an approach would seek agreement on what should be accepted as the lowest possible chance of achieving a good outcome compared with standard of care (with which the chance of a good outcome is close to zero). For example, a Minimum Clinically Important Difference of 5% means that 20 such patients would need to be treated (at considerable cost) to achieve one good outcome. Interestingly, the authors regard the inclusion of cost in determining treatment limits with diffidence, suggesting that this variable might be factored in separately.

In this regard, the authors seem to underappreciate the potential utility of Health Technology Assessment and related mechanisms to systematically establish the properties, effects and impacts of health technologies including their efficacy, safety, cost-utility, and cost-effectiveness.⁷ This process of Health Technology Assessment is, therefore, not simply a clinical one, but also an economic and ethical one. It includes the identification and appraisal of evidence about the intervention(s) in question, the determination of what counts as evidence, the possibility and conduct of risk-benefit analysis, the processes by which a value is placed on clinical outcomes, the points at which therapy may be regarded as burdensome or (even) futile, the identification and involvement of relevant stakeholders, the possibility of review and appeal, and the process by which a decision is made by public and private funders to subsidise the costs of an intervention or drug. Sometimes referred to as the bridge between evidence and policy-making, many countries have established structures and processes for performing health technology assessment. While Health Technology Assessment takes various forms and remains a highly contested process, it provides a more systematic way of establishing the clinical and social values and consequences of EVT and IVT than that proposed by Wechsler et al. Here, they suggest that a consensus on Minimum Clinically Important Difference formed on the basis of the opinions of eminent experts would be sufficient, as their opinion would consciously or subconsciously consider the benefits, harms and costs of EVT and IVT. However, conscious and subconscious biases can sometimes affect impartiality.

The authors consider the idea of inviting consumer input into this debate, although mostly as a way of attempting to justify that an acceptable outcome after reperfusion therapy might include severe disability. Even if some patients/families have views that death is the worst possible outcome, it may still be unethical to offer treatment that is likely to lead to a patient being severely disabled (not to mention extremely costly to the community in the long-term). Again, conscious and subconscious clinician biases can influence opinions on these issues. Such less satisfactory outcomes might be more likely if there is either significant prestroke functional/cognitive impairment, and/or there is a large pretreatment ischemic core. These are 2 groups in whom the authors suggest the treatment limits be tested in trials, and we do not disagree, but we cannot at the same time shift the goalposts to accept outcomes that are simply not acceptable on a societal basis.

If successful, the expanded application of EVT and IVT could produce significant benefits, with reduced morbidity, mortality, and societal costs associated with caring for people with lasting disabilities caused by strokes. However, significant risks and caveats must be carefully considered before major steps are taken. Indeed, the construction of the infrastructure of checks and balances, which includes the entire research ethics and government regulatory systems, was developed to impose such caution and circumspection. Among these checks and balances are processes to limit ways in which the personal, financial, or professional interests of clinicians and researchers may compromise the impartiality of clinical decision-making.

There will always be a zone of uncertainty that has to be negotiated when clinical decisions are contemplated. This is the ethical territory that involves the prudent and thoughtful consideration of risks and benefits, personal values, and social and cultural attitudes and beliefs. Such reflection is an essential, enduring, and irreducible part of clinical practice. However, at the outer reaches of practice, there is a different terrain of uncertainty, where knowledge of benefits is significantly more limited and the possibility of harm is greatly magnified, and where social and communal values about what we want from our health systems must be taken into account. Here, it important to resist the urge to be swept up by the hubris of new technologies, to exercise prudence and circumspection, to seek expert perspectives beyond the clinic, and to create and welcome public discussion on the goals of medicine and the design and delivery of health care.

Disclosures Dr Parsons reports research collaborations (discounted software) with Siemens, Canon, and Apollo Medical Imaging (MIStar) and Advisory Board member for Tenecteplase (Boehringer Ingelheim). The other authors report no disclosures.

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

For Disclosures, see page 3406.

参见相关文章，第 3399 页

Wechsler 等人¹发表了 STAIR 小组的意见，该小组是一个由卒中临床医生、行业和监管代表组成的国际小组，他们参加了仅限受邀参加的研讨会。他们的主要论点是，需要在卒中的背景下进行进一步研究，以确定如何在初始试验纳入标准的基础上将再灌注治疗——血管内血栓切除术 (EVT) 和/或静脉溶栓 (IVT)——扩展到更大的患者群体太严格了，更多超出这些标准的患者可能会受益。

随机对照试验，就其本质而言，严格限制了研究人群。这意味着他们的结果可能仅适用于狭义的患者群体——一种被称为普遍性差距的流行病学挑战。这一直给有希望的新疗法带来问题，即如何确定益处的外部限制。传统上，通过各种（通常是随意应用的）机制，一种新疗法的适用性的不确定性区域已经逐渐缩小，包括逐步扩大临床试验和系统评价、标签外处方、正式和非正式辩论以及涉及病例的争论。研究、临床医生和患者之间的个人对话以及媒体报道。

最近，已经开发了更复杂的技术，以尝试实现一个更系统的过程来扩展证据和减少不确定知识领域。临床注册、平台研究、务实和适应性强的设计，以及通过强制注册试验和更广泛地了解阴性试验结果来增强对数据的访问，大大提高了提供新疗法的速度和效率。

EVT 和 IVT 是卒中管理的重要步骤，卒中是发病率、死亡率和社会成本的主要原因。然而，这些疗法的引入并不是完全直接的，部分原因是难以确定最有可能受益的患者群体，部分原因是有充分记录的风险，有时是破坏性的不良反应。出于这个原因，这些治疗引起了临床和更广泛社区的重大关注。经过长时间的争论，IVT 和现在的 EVT 在卒中管理中的作用的重要性已被广泛接受，国际上对 EVT 和 IVT 的临床应用已达成普遍共识。

事实上，在早期的时间窗口（EVT 低于 6 小时，IVT 低于 4.5 小时），许多人认为共识指南非常宽松。尽管如此，在许多国家/地区，目前符合 EVT 和 IVT 条件的患者并未接受。例如，在美国，只有 10% 到 15% 的缺血性卒中患者正在接受静脉溶栓，只有 3% 正在接受血栓切除术。虽然 Wechsler 及其同事认为这为扩大 EVT 和 IVT 的纳入标准创造了一个论据，但我们建议在将这些治疗扩展到缺乏证据的临床环境之前，专注于确保这些治疗的可用性似乎更为谨慎和适当。谁已经符合资格并且已经有有利的风险收益。这当然是 不是技术问题，而是社会经济和伦理问题。人们目前无法获得治疗的原因很复杂，但与无法获得及时的专家护理有关，其基础是社会经济地位、农村/地区居住、种族背景和保险状况等相互交织的问题，这些问题当然是，绝不仅限于美国。^2–4

与在没有支持证据的情况下以推测的方式扩大 EVT 和 IVT 的治疗适应症相比，这种专注于改善所有当前符合条件的患者的再灌注治疗的可及性和实施再灌注治疗更有可能在人群水平上带来具有临床意义的益处。与此同时，Wechsler 等人已经确定了关于 EVT 和 IVT 的临床应用、确定可能受益或受到这些疗法伤害的患者以及解决这些不确定性的方法方面仍然存在的重大不确定性。例如，当单独使用 IVT 的结果非常好时，我们是否应该要求介入医生取出远端凝块（即，超出大脑中动脉 M1 段），并且知道在小口径血管中追逐凝块时血管穿孔的风险更高吗？在不可逆损伤的大脑（缺血核心）的体积是多少时，血栓切除术仍然有机会获得良好的结果？事实上，在缺血核心的多少体积下，再灌注是有害的或徒劳的（没有提供有意义的益处）？何种程度的卒中前功能或认知障碍妨碍了良好的结果？在这些边缘患者中，他们不太可能通过再灌注获得传统定义的良好结果，什么样的长期残疾水平是可以接受的？这些都是复杂的问题。事实上，最近的澳大利亚注册数据发现，在常规实践中接受 EVT 治疗的患者的残疾调整生命年比使用灌注成像来选择患者的关键 EVT 试验报告的要低得多。⁵在常规实践中接受治疗的许多患者处于治疗指南的外部界限，尤其是治疗前影像学特征较差。

作者建议使用平台试验来提高效率并充分利用个体患者的结果。也称为 MAMS（多臂、多阶段）设计试验，这些研究（主要用于癌症）通常针对一个共同的对照组评估几种干预措施，并且可以是永久性的。⁶该设计结合了预先指定的适应规则，以允许在试验期间删除无效干预以及添加新干预的灵活性。然而，对于 EVT 和 IVT，平台试验方法论能否有效解决与其应用相关的不确定性并不明显。这是因为他们的主要贡献是允许通过调整跨试验组的数据来增加或减少不同的治疗，而关于 EVT 和 IVT 的主要问题主要涉及具有差异特征的边缘人群（例如，大核心与远端闭塞），对于哪些借款信息不太可能适用。

我们建议建立脑卒中再灌注治疗的限制，通过一组灵活、精益的聚焦单组来评估多个明确定义的患者亚组的安全性和有效性问题。在将这些纳入有效性故事之前，进行明确的是/否决定的试验。这可以通过对许多需要限制的亚组进行所谓的篮子/伞式试验设计来有效实现，而不是通过一个大型的永久平台。⁶考虑到我们已经非常清楚在常规护理方案下这些患者群体会发生什么情况，可以使用像设备试验中使用的客观性能标准比较器结构来提高效率，因为每个研究只有一个组。

作者还建议大型实用试验在解决常规实践中的普遍性和预期结果方面发挥作用。虽然这在一般情况下可能是正确的，但它无助于解决当前的问题。对于未证明治疗安全有效的患者，不应包括在大型实用有效性试验中。这仅仅是因为风险太大：如果没有严格实施的 RCT 中内置的安全机制，造成伤害的可能性将非常大。如果患者受到伤害，整个治疗方法可能会受到影响，并产生持续的、深远的影响。

当然，与试验设计相关的是结局指标的选择。作者建议我们确定最小临床重要差异以帮助确定治疗的外部限制。通常，这将通过专家小组的协商一致确定（可能仅通过邀请）。在上述大型核心患者中，这种方法将寻求就与护理标准（良好结果的机会接近于零）相比应该接受的最低可能获得良好结果的机会达成一致。例如，5% 的最小临床重要差异意味着需要治疗 20 名这样的患者（以相当大的成本）才能获得一个好的结果。有趣的是，作者对在确定治疗限制时包含的成本持怀疑态度，

在这方面，作者似乎低估了卫生技术评估和相关机制的潜在效用，以系统地确定卫生技术的特性、效果和影响，包括其功效、安全性、成本效用和成本效益。⁷因此，这一卫生技术评估过程不仅是临床过程，也是经济和伦理过程。它包括对有关干预措施的证据的识别和评估、确定什么是证据、风险收益分析的可能性和进行、对临床结果进行评估的过程、在哪种疗法可能被视为繁重或（甚至）徒劳，相关利益相关者的识别和参与，审查和上诉的可能性，以及公共和私人资助者做出决定以补贴干预或药品。有时被称为证据和决策之间的桥梁，许多国家已经建立了执行卫生技术评估的结构和流程。虽然卫生技术评估采取多种形式并且仍然是一个备受争议的过程，但它提供了一种比 Wechsler 等人提出的更系统的方法来确定 EVT 和 IVT 的临床和社会价值和后果。在这里，他们建议在著名专家意见的基础上形成的关于最小临床重要差异的共识就足够了，因为他们的意见会有意或无意地考虑 EVT 和 IVT 的好处、危害和成本。然而，有意识和潜意识的偏见有时会影响公正性。他们认为，基于知名专家的意见形成的关于最小临床重要差异的共识就足够了，因为他们的意见会有意或无意地考虑 EVT 和 IVT 的益处、危害和成本。然而，有意识和潜意识的偏见有时会影响公正性。他们认为，基于知名专家的意见形成的关于最小临床重要差异的共识就足够了，因为他们的意见会有意或无意地考虑 EVT 和 IVT 的益处、危害和成本。然而，有意识和潜意识的偏见有时会影响公正性。

作者考虑了邀请消费者参与这场辩论的想法，尽管主要是为了证明再灌注治疗后可接受的结果可能包括严重残疾。即使一些患者/家庭认为死亡是最糟糕的结果，但提供可能导致患者严重残疾的治疗仍然是不道德的（更不用说从长远来看对社区来说极其昂贵） . 同样，有意识和无意识的临床医生偏见会影响对这些问题的看法。如果存在显着的卒中前功能/认知障碍和/或存在较大的治疗前缺血核心，则这种不太令人满意的结果可能更有可能发生。作者建议在这 2 个组中测试治疗限度，我们并不反对，

如果成功，EVT 和 IVT 的扩大应用可以产生显着的好处，降低发病率、死亡率和与照顾因中风导致的永久性残疾相关的社会成本。但是，在采取重大步骤之前，必须仔细考虑重大风险和注意事项。事实上，包括整个研究伦理和政府监管体系在内的制衡基础设施的建设就是为了强加这种谨慎和谨慎。在这些制衡过程中，包括限制临床医生和研究人员的个人、财务或专业利益可能损害临床决策公正性的方式。

在考虑临床决策时，总会有一个必须协商的不确定区域。这是涉及对风险和收益、个人价值观以及社会和文化态度和信仰进行审慎和深思熟虑的考虑的道德领域。这种反思是临床实践中必不可少的、持久的和不可减少的部分。然而，在实践的外围，存在着不同的不确定性，在那里对好处的了解明显更加有限，伤害的可能性被大大放大，关于我们想要从我们的卫生系统中得到什么的社会和公共价值观必须是考虑在内。在这里，重要的是要抵制被新技术的傲慢席卷的冲动，谨慎行事，寻求临床以外的专家观点，

披露Parsons 博士报告了与 Siemens、Canon 和 Apollo Medical Imaging (MIStar) 以及 Tenecteplase (Boehringer Ingelheim) 顾问委员会成员的研究合作（折扣软件）。其他作者报告没有披露。

本文中表达的观点不一定是编辑或美国心脏协会的观点。

有关披露，请参阅第 3406 页。