See related article, p 1226

One of the most significant developments in modern neurology was the emergence of intravenous thrombolytic therapy as a powerfully effective treatment for acute ischemic stroke.¹ Multiple randomized clinical trials demonstrated and confirmed the benefit of thrombolysis, and it is now standard of care around the world.² Along the way to widespread acceptance, however, there were bumps in the road. One of those bumps involved a critical misunderstanding of the first trial by the National Institute of Neurological Disorders and Stroke (NINDS) r-tPA (recombinant tissue-type plasminogen activator) for Acute Stroke Investigators. In part 1 of the study, the primary end point assessed the fraction of treated patients who improved by 4 or more points within 24 hours following thrombolytic therapy. Today, this response is called early neurological improvement (ENI) but at the time we called it "on the table" or OTT because many patients with ENI responded within 2 hours of r-tPA infusion. The protocol specified that 24-hour response would be measured as a 4 or more-point NIHSS (National Institutes of Health Stroke Scale) improvement over baseline. The rationale for choosing 4 points, rather than some other definition, involved minimal data from small pilot studies and investigator consensus.^3,4 The lack of a prospective validation of the 4-point definition seemed inconsequential at the time, because the NINDS tPA for Acute Stroke Trial was intended to be a dose-confirming Phase 2b trial only, not the final, pivotal trial it became. The power analysis for the trial posited that 40% of the r-tPA treated group would show the 4-point NIHSS improvement but that only 20% of the placebo would. The planned sample of 140 subjects per group would give a 95% power to detect the hypothesized benefit at an α of 0.05 (Final Protocol, November 19, 1990). As it turned out, the percentage of patients with OTT response of 4 or more points in the r-tPA group was 47% compared with 39% in the placebo-treated group, for a Mantel-Haenszel test P value of 0.21, suggesting failure. In reality, the early response of the r-tPA treated group was impressive and highly statistically significant, as shown in the attached Figure. The issue was the 4-point definition: all other values for defining ENI—all other definitions of improvement by 24 hours—were statistically significant, with P values well below 0.05 (see the Figure)—only the 4-point definition failed to show statistical significance. Comparing the median NIHSS score at 24 hours (8 for tPA treated; 12 for placebo-treated) also showed statistical significance using ANCOVA (P<0.02).⁵ In addition, the first part of the study was highly statistically significant on the 3-month outcome, the modified Rankin Scale score. The prespecified primary outcome, however, was not statistically significant, and so part 2 of the trial was organized to extend and confirm the benefits seen in what then became called part 1. When the full trial was published, out of an overabundance of academic meekness, we stated that part 1 failed on its primary outcome without including the rest of the story: that for any other ENI definition, part 1 showed a statistically significant benefit. We published parts 1 and 2 together, and to our eyes, looking at the totality of the data, the NINDS r-tPA for Acute Stroke Trial indicated a powerful and persuasive treatment effect.

Figure. Early Neurological Improvement in part 1 of the NINDS tPA (tissue-type plasminogen activator) for Acute Stroke Trial. The proportion of patients treated with placebo or intravenous r-tPA (recombinant tissue-type plasminogen activator) whose National Institutes of Health Stroke Scale improved from baseline to 24 h is shown. To define early neurological improvement, definitions of improvement ranged from 4 points to 14 points. *P<0.05, Mantel-Haenszel test. Reprinted from Haley et al,⁵ with permission. Copyright ©1997, Elsevier.

The meek portrayal of the part 1 outcome led to considerable mischief among deniers, who either failed to grasp the totality of the data presented or harbored an ulterior reason for blocking the approval of thrombolytic therapy. Suffice to say that in 1996—when there were no Code Stroke teams, no systems of care and no Vascular Neurologists ready to take call and respond immediately to patients with acute stroke —the reaction among seasoned General Neurologists and Emergency Medicine specialists ranged from lukewarm to openly hostile. To more accurately portray the truth about the study, Clarke Haley published (on behalf of all the authors) the full spectrum of early responses to r-tPA, as redrawn by me in the attached Figure and addressed a few other myths that had grown up around the trial.⁵ Broderick⁶ led an analysis looking for the best early predictor of 3-month outcome and found that although many definitions of early response at 24 hours were all good, a change from baseline to 1 week more powerfully predicted outcome at 3 months. Sadly, these results were not enough to overcome the clinical inertia among doctors adhering to the status quo or the doubts stoked by a few deniers. Intravenous thrombolytic therapy emerged very slowly as standard of care over many years following the first publication showing its powerful benefits.

In this issue of Stroke, Agarwal et al⁷ return to the problem of predicting 3-month outcomes using the OTT or ENI response. In this new analysis of the published NINDS Trial data set, they redefined ENI using percent decrement NIHSS score from baseline to 24 hours, rather than the absolute value of the NIHSS difference from baseline. The authors found the best definition to be 30% decrement from baseline NIHSS, compared with other cut points and verified the value of their definition across all levels of stroke severity. A 30% decrement in NIHSS at 24 hours, compared with baseline, showed improved power—as measured with receiver operating characteristics curves—to predict 3-month modified Rankin Scale score. Percent decrement in NIHSS also correlated well with 3-month disability (measured with the Barthel index) and with infarct volume. The authors are to be congratulated for this advance, because improved predictive power at early time points may have considerable utility in streamlining clinical trials that seek early markers to drive an adaptive dose selection process. The authors also focus on using 30% NIHSS decrement at 24 hours to study other prognostically relevant clinical and radiological variables such as 24-hour brain imaging. Evidence of ENI—such as 30% NIHSS score decrement—may potentially prove valuable as a marker of reperfusion, however further correlation of the 30% decrement with post-recanalization angiography will be needed to fully validate this potential new biomarker.

How will practice change, knowing this new definition of ENI? Most importantly, we should keep in mind the results apply only to the derivation data set, that is, the original NINDS Trial data set. Future prospective confirmations of any predictive method are required. Ideally, a putative biomarker that predicts long-term outcome should be prospectively validated in >1 trial. We certainly know already that patients with early, dramatic response tend to do well over the long-term. We do not know which, if any, definition of ENI is sufficient to substitute for 3-month outcome in a clinical trial; whether ENI is useful when triaging patients to rehabilitation strategies; or whether, ENI reliably identifies patients eligible for early discharge. Such early biomarkers—assuming they were validated in well-executed prospective studies—could radically alter current clinical trial design toward faster, more efficient protocols.

One worry, however, is that the best end point for stroke clinical trials remains undefined. The 3-month modified Rankin Scale score continues as the de facto standard outcome, but only because it was used in the first positive trial of thrombolysis and no alternative outcome appears superior, yet. Investing effort to design a biomarker that predicts an imperfect outcome—one we all hope to replace someday—may be less impactful than working to improve long-term assessment of stroke outcome.

The 30% NIIHSS score decrement, as demonstrated by Agarwal et al,⁷ better predicted 3-month outcomes in the original NINDS Trial data set. While its important to prospectively confirm this result, perhaps effort would be better spent improving our measures of long-term stroke outcome. Most importantly, as shown in the Figure, in part 1 of the original trial thrombolysis showed a clear and persuasive benefit on early neurological improvement.

This study was supported by U24 NS113452 and R01 NS075930.

Dr Lyden received royalties from the book Thrombolytic Therapy for Acute Ischemic Stroke, third edition and income from occasional expert testimony regarding the use of thrombolytic therapy for acute ischemic stroke.

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

Guest Editor for this article was Kazunori Toyoda, MD, PhD.

请参阅相关文章，第1226页

现代神经病学最重要的发展之一是静脉溶栓治疗的出现，它是治疗急性缺血性中风的有效方法。¹多项随机临床试验证明并证实了溶栓治疗的益处，目前它已成为世界范围内的标准治疗方法。²然而，在被广泛接受的过程中，道路上遇到了坎bump。这些障碍之一涉及对急性中风研究者的国家神经疾病和中风研究所（NINDS）r-tPA（重组组织型纤溶酶原激活物）的首次试验的严重误解。在研究的第1部分中，主要终点评估了溶栓治疗后24小时内改善4点或更多点的治疗患者比例。今天，这种反应被称为早期神经功能改善（ENI），但在我们称之为“就餐”或OTT时，是因为许多ENI患者在r-tPA输注后2小时内反应良好。该协议规定，将24小时反应作为相对于基线的4分或更多点NIHSS（美国国立卫生研究院卒中量表）的改善来衡量。^3,4当时缺乏对4点定义的前瞻性验证似乎是无关紧要的，因为NINDS tPA急性中风试验仅打算作为剂量确认的2b期试验，而不是最终的关键性试验。该试验的功效分析认为，r-tPA治疗组的40％将显示4点NIHSS改善，但安慰剂仅20％。计划的每组140名受试者的样本将提供95％的功效，以α为0.05来检测假设的收益（最终方案，1990年11月19日）。事实证明，患者的r的tPA组中的4个或更多点OTT响应的百分比为47％，而安慰剂治疗组中39％，对于曼特尔-Haenszel检验比较P值为0.21，表明失败。实际上，如附图所示，r-tPA治疗组的早期反应令人印象深刻，并且具有高度统计学意义。问题是4点定义：用于定义ENI的所有其他值（所有其他24小时改善的定义）在统计上均具有显着意义，P值远低于0.05（见图），仅4点定义未显示统计学意义。使用ANCOVA比较24小时的NIHSS中位评分（tPA治疗为8；安慰剂治疗为12），也显示出统计学意义（P <0.02）。⁵此外，研究的第一部分在3个月结局，改良的Rankin量表评分上具有高度统计学意义。但是，预先确定的主要结局在统计上并不显着，因此组织了该试验的第二部分，以扩展并确认后来被称为第一部分的益处。当发表完整的试验时，出于学术上的过分热情，我们指出，第1部分的主要结果失败了，而没有包括故事的其余部分：对于任何其他ENI定义，第1部分都显示出统计学上的显着收益。我们一起发布了第1部分和第2部分，在我们的眼中，查看数据的整体信息，急性卒中试验的NINDS r-tPA表明了强大而有说服力的治疗效果。

数字。NINDS tPA（组织型纤溶酶原激活物）急性卒中试验第1部分的早期神经功能改善。美国国立卫生研究院卒中量表从基线提高到24小时的安慰剂或静脉内r-tPA（重组组织型纤溶酶原激活剂）治疗的患者比例已显示。为了定义早期的神经功能改善，改善的定义范围为4点至14点。* P＜0.05，Mantel-Haenszel检验。经许可，转载自Haley等⁵。版权所有©1997，Elsevier。

对第1部分结果的温和刻画导致拒绝者中的许多恶作剧，他们要么无法掌握所提供的数据总量，要么是阻止溶栓治疗获得批准的别有用心的原因。可以说，在1996年（当时没有Code Stroke小组，没有护理系统，没有血管神经病学家愿意接听电话并立即对急性中风患者做出反应），经验丰富的General Neurologists和急诊医学专家的反应不一，公开敌对。为了更准确地描述这项研究的真相，克拉克·海利（代表所有作者）发表了对r-tPA的早期反应的全貌，如我在附图中所重述的，并谈到了其他一些长大的神话各地审判。⁵布罗德里克^6进行了一项分析，以寻找3个月结果的最佳早期预测指标，并发现尽管许多24小时早期反应的定义都不错，但从基线到1周的变化更能有效预测3个月的结果。可悲的是，这些结果不足以克服坚持现状或少数否认者所引起的疑虑的医生的临床惯性。首次发表后，静脉溶栓治疗作为治疗标准出现了许多年，进展缓慢，显示出其强大的益处。

在本期《中风》中，Agarwal等人⁷回到使用OTT或ENI响应预测3个月结果的问题。在对已发布的NINDS试验数据集的这一新分析中，他们使用从基线到24小时的NIHSS减量百分比分数，而不是与基线相比NIHSS差异的绝对值，重新定义了ENI。作者发现，与其他切点相比，最佳定义是相对于基线NIHSS减少30％，并验证了其定义在中风严重程度所有水平上的价值。与基线相比，NIHSS在24小时内降低了30％，表明功率得到了改善（通过接收器工作特性曲线进行了测量），可以预测3个月修正的Rankin量表评分。NIHSS的减少百分比还与3个月的残疾（通过Barthel指数测量）和梗死面积密切相关。作者应为此取得的进展表示祝贺，因为在早期时间点提高预测能力在简化寻求早期标记物以驱动适应性剂量选择过程的临床试验中可能具有相当大的实用性。作者还专注于在24小时使用30％NIHSS减量来研究其他与预后相关的临床和放射学变量，例如24小时脑成像。ENI的证据（例如30％的NIHSS评分降低）可能被潜在地证明是再灌注的标志物，但是30％的降低值与再通血管造影后血管造影的进一步关联将需要充分验证这种潜在的新生物标记物。作者还专注于在24小时使用30％NIHSS减量来研究其他与预后相关的临床和放射学变量，例如24小时脑成像。ENI的证据（例如30％的NIHSS评分降低）可能被潜在地证明是再灌注的标志物，但是30％的降低值与再通血管造影后血管造影的进一步关联将需要充分验证这种潜在的新生物标记物。作者还专注于在24小时使用30％NIHSS减量来研究其他与预后相关的临床和放射学变量，例如24小时脑成像。ENI的证据（例如30％的NIHSS评分降低）可能被潜在地证明是再灌注的标志物，但是30％的降低值与再通血管造影后血管造影的进一步关联将需要充分验证这种潜在的新生物标记物。

知道ENI的新定义后，实践将如何变化？最重要的是，我们应该记住，结果仅适用于派生数据集，即原始的NINDS试用数据集。需要任何预测方法的未来前瞻性确认。理想情况下，预测长期结果的推定生物标志物应在> 1个试验中进行前瞻性验证。我们当然已经知道，具有早期，戏剧性反应的患者从长期来看往往表现良好。我们不知道哪种ENI定义足以代替3个月的临床试验结果；在对患者进行康复策略分类时，ENI是否有用；或ENI是否能可靠地识别出有资格提前出院的患者。

然而，人们担心的是中风临床试验的最佳终点仍然不确定。为期3个月的改良兰金量表评分继续作为事实上的标准结局，但这仅是因为它已在溶栓治疗的第一个阳性试验中使用，并且尚无其他结局出现。投资设计可预测不理想结果的生物标记物（我们都希望有一天能替代）的努力可能不如努力改善对中风结果的长期评估有效。

如Agarwal等人所示，在原始NINDS试验数据集中^7个预测的3个月预后更好，NIIHSS分数降低了30％。虽然前瞻性地确认该结果很重要，但也许最好将精力花费在改进我们对中风的长期预后的评估上。如图所示，最重要的是，在原始试验的第1部分中，溶栓治疗对早期神经功能改善显示出明显而有说服力的益处。

这项研究得到了U24 NS113452和R01 NS075930的支持。

Lyden博士从《急性缺血性卒中溶栓治疗》（第三版）一书中获得了特许权使用费，偶尔从专家证词中获得关于溶栓治疗急性缺血性中风的收益。

本文表达的观点不一定是编辑者或美国心脏协会的观点。

这篇文章的客座编辑是丰田章男（Kazunori Toyoda），医学博士。