Brief History of Endovascular Acute Ischemic Stroke Treatment

• endovascular recanalization
• endovascular surgery
• endovascular treatment
• stenting
• thrombectomy

The natural history of large-vessel extracranial and intracranial arterial occlusion has played a major role in the evolution of endovascular stroke therapy. Basilar artery occlusion was felt to have the worst prognosis, and this drove Zeumer et al¹ to first perform basilar thrombolysis via catheter and report excellent outcomes based on historical controls. However, early stroke thrombolysis clinical trials focused on intravenous (IV) and not intraarterial (IA) delivery. Gregory del Zoppo—a hematologist by training—led international angiography-based, dose escalation trials of double-stranded tissue-type plasminogen activator (tPA) given IV. This Burroughs Wellcome–sponsored research trial showed that distal M3 middle cerebral artery (MCA) occlusions often recanalized with IV tPA, but recanalization rates were low with more proximal large-vessel occlusions. Furthermore, the site of arterial occlusion could not be reliably predicted by clinical phenotype or National Institutes of Health Stroke Scale (NIHSS). Concurrent with this angiography-based effort, the National Institute of Neurological Disorders and Stroke (NINDS) investigators downplayed the need for angiography and instead focused on time to treatment with IV tPA. The NINDS Genentech IV tPA stroke trial established the critical importance of time to treatment and led to Food and Drug Administration (FDA) approval of IV tPA to treat acute ischemic stroke in 1996 but within a small window of 3 hours.

The major drivers for the development of endovascular stroke therapy were the desire to expand the 3-hour treatment window and the relatively low recanalization rates for large-vessel occlusions with IV tPA (which often caused the most severe strokes). Toward this end, Gregory Del Zoppo, Anthony Furlan, Randall Higashida, and Michael Pessin convinced Abbott Laboratories that the quickest way to FDA approval of their new thrombolytic agent recombinant prourokinase (r-proUK) was an IA trial focused only on MCA occlusion. Accordingly, Del Zoppo, Furlan, Higashida, and Pessin designed the Prolyse in Acute Cerebral Thromboembolism (PROACT, versions I and II) trials to demonstrate the safety, recanalization efficacy, and clinical benefit of IA r-proUK in patients with MCA occlusion treated within 6 hours of stroke onset.^2,3 PROACT-I, which used an IA saline control group, ended early because the FDA approved IV tPA during trial recruitment, leading to questions about the ethics of continuing with an IA placebo control.

When the 46 PROACT-I cases were reviewed, there was highly significant MCA recanalization efficacy and a strong signal of clinical benefit. Furthermore, the vast majority of patients were treated beyond 3 hours and therefore were not candidates for IV tPA. Abbott therefore agreed to fund PROACT-II, especially because the sample size estimate for proof of benefit was only 186 patients. PROACT-II introduced several novel endovascular trial design changes in addition to focusing only on MCA occlusion. Because of ethical concerns, the IA placebo control was replaced by IV heparin. Of more interest, PROACT-II was the first stroke trial to stratify patients by baseline stroke severity (later an issue with the NINDS trial) and was the first trial to use a dichotomized Rankin score of ≤2 as the primary end point rather than ≤1 for IV trials (allowing for a modest residual deficit in patients with more severe baseline strokes). Although the first controlled endovascular trial, PROACT-II ironically prohibited any mechanical manipulation of the clot because the goal was to show efficacy of r-proUK. Drug delivery also took place over 2 hours and could be started no later than 6 hours, establishing the 8 hour time window that was used in several subsequent device trials and one recent randomized trial.⁴

PROACT-II demonstrated that IA r-proUK safely opened M1 occlusions and improved clinical outcomes by 15% absolute (number needed to treat of 7). However, the FDA did not approve r-proUK or IA stroke therapy because of the small size and marginal significance (P=0.043) of PROACT-II. There was also physician backlash over FDA approval of IV tPA based on the NINDS trials (considered as 2 trials) and then a flurry of IV trials with longer time windows that were ineffective (ECASS I and ECASS II [European Cooperative Acute Stroke Study versions I and II]. The Association of Academic Emergency Medicine went as far as publishing that IV tPA was not standard of care by citing the lack of confirming evidence in subsequent trials. This opinion also served to limit liability of emergency physicians for not offering tPA.

Despite the lack of FDA approval, PROACT-II ended equipoise for many (but not all) neurointerventionalists and stroke neurologists, solidifying reluctance to ever randomize such patients again. PROACT-II launched the era of endovascular stroke therapy, providing fuel to device companies to design mechanical techniques to remove clot. Indeed, the first acute stroke endovascular device clinical trial (TIME [Thrombectomy in Middle Cerebral Artery Embolism] trial; stopped in 2003 because of negative results and never published) was launched by Possis Medical using an Angiojet and a modified version of the PROACT-II protocol. Many failed devices (eg, lasers) followed.

In addition to improving recanalization and reducing hemorrhage risk by limiting thrombolytics usage, mechanical endovascular devices were developed to theoretically reduce the risk of brain hemorrhage and improve outcomes. Dr Y. Pierre Gobin at University of California Los Angeles had patented a corkscrew-like, endovascular retrieval device for lost endovascular coils; this device could be extruded from a catheter and ensnare a foreign body (typically coils) accidentally released into the cerebral vasculature. The device was cleared for foreign body removal by the FDA device branch based on a handful of cases but did not get much use. The device was subsequently licensed by Concentric Medical, Inc. and called the Mechanical Embolus Removal in Cerebral Ischemia (MERCI) retriever. Concentric used the FDA 510-K regulatory pathway to clear the device (devices are cleared not approved). This pathway requires a new device be considered substantially equivalent to a cleared, predicate device and also that it is safe during use for the disease process.⁵ Efficacy for other medical devices had rarely been required at that time this research was initiated because efficacy is as much a function of the device as it is the user of the device, and it is not possible for the FDA to approve an endovascular surgeon. So, the MERCI device approval pathway was substantial equivalence to itself (for removing foreign bodies and clot was considered a foreign body), that it could open vessels better than natural history, and that it was safe when used in stroke patients.

Using the PROACT-II control recanalization rate of 18% as a benchmark, the single arm MERCI trial enrolled 153 patients and showed the device alone opened 48% of intracranial blood vessels, including the basilar artery, and the rate of intracerebral hemorrhage was <3% when considering parenchymal hematoma type 2 intracerebral hemorrhages.⁶ This recanalization rate was significant and met the primary outcome of the study. Subgroup analysis revealed a remarkable difference in the rate of good clinical outcome and mortality in favor of those patients who were recanalized, and multivariate models show that revascularization was highly predictive of good clinical outcome. These data were presented to an FDA advisory panel in February 2004 by the study PI Wade Smith. The panel was critical that the trial was not randomized comparing to medical therapy, and they were somewhat shocked by the high mortality rate overall (42%).⁷ Because few angiographically confirmed large-vessel occlusion patients had ever been studied at that time, there was little comparison data for mortality, especially considering carotid terminus and basilar occlusions. In response to the concerns, additional data (comparison to PROACT-II and other studies) was submitted that supported efficacy, and review of the little literature at the time that reported mortality for large-vessel occlusion, the FDA committee cleared the MERCI retriever as the first device to be used for clot removal in acute ischemic stroke in August 2004. The approved label did not indicate that the device treated stroke, however, because that would require contemporaneous randomized controls.

FDA approval of the MERCI device caused remarkable controversy at the time for several reasons: (1) neurologists had little experience with the FDA 510-K pathway for clearance (even though mandated by Congress),⁵ in part, because medical devices had not played a major role in neurology, (2) PROACT-II provided higher quality, randomized data on IA proUK to support drug approval, but the FDA refused to approve it,⁸ and (3) the panel had recommended nonapproval of the MERCI retriever; yet, the FDA cleared it anyway (because the panel never saw the supplemental data submitted by the sponsor, although this was published later⁹). Werner Hacke stated publically during a press conference, “Have MERCI on me.” Finally, Concentric Medical successfully negotiated a diagnosis-related group reimbursement through the Centers for Medicare & Medicaid Services for stroke treated with embolectomy equal to that reimbursed for craniotomy. This caught the attention of hospital administrators because IV tPA reimbursed significantly less. Many have criticized this compensation as dissuading the study of embolectomy in a randomized fashion.

A subsequent single arm trial allowed approval of a newer more effective device (Multi-MERCI trial¹⁰), and by 2010, over 10 000 cases had been treated worldwide using this family of devices. Many of us continued to advocate for a randomized trial of device versus medical therapy but had difficulty convincing interventionalists to randomize, mostly because they were swayed by PROACT-II and they had their own anecdotes of cure on the table. In the meantime, Joseph Broderick and colleagues initiated the Interventional Management of Stroke-III to study intervention versus medical therapy in an IV tPA eligible group of moderate stroke patients.¹¹ This trial failed to show benefit and when presented in 2013 to the International Stroke Congress dealt a hard blow to the field. Two other trials of IV versus IA therapy were published in the same New England Journal of Medicine issue showing no benefit (SYNTHESIS [Intra-Arterial Versus Systemic Thrombolysis for Acute Ischemic Stroke Expansion] and MR-RESCUE [Mechanical Retrieval and Recanalization of Stroke Clots Using Embolectomy]).

These 3 trials had design limitations, however. Interventional Management of Stroke-III, which was the largest trial of the 3, used an NIHSS threshold rather than computed tomographic angiography (CTA) to select patients for therapy (because CTA was not used routinely when this 8-year trial initiated). Not surprisingly, 7.8% of patients randomized to IA therapy had fully open vessels once they arrived at the angiography suite and had their initial angiogram. This led to dilution of the true target population studied and compromised power. As CTA became mainstream, however, patient were selected by CTA in some, then eventually all, centers. Subgroup analysis of Interventional Management of Stroke-III showed that IA therapy in CTA-positive patients was marginally significant,¹² but unfortunately the trial was stopped early by the data safety monitoring board for futility without having this subgroup analysis available.

Likely frustrated by the lower recanalization rates of various devices at the time, neurointerventionalists began using coronary or self-expanding stents off-label to open large intracranial vessels as a bail out procedure in 2005–2006. Soon, experience with using the stent as a clot snare was reported (ie, stenting the clot but pulling the clot out with the stent and not leaving the stent behind), and the era of stentrievers was born.^13–15 Two stentrievers were designed (Solitaire and Trevo) and tested in 2 randomized trials showing that these newest generation devices could have significantly better, ≤90%, recanalization compared with the MERCI predicate.^16,17

Moving quickly, 6 trials designed to test the clinical efficacy of these devices versus medical control (some pretreated with IV tPA and some tPA was optional) began, all for anterior circulation stroke and most except 1 within 6 hours of stroke onset. MR CLEAN was the first to report and publish that the number needed to treat was 3 to 4 patients to make a good outcome, and the intracranial hemorrhage rate was negligible.¹⁸ This prompted the data safety monitoring boards of the 5 other trials to take an interim look at their data, and all separately stopped their trials for efficacy^4,19–21 (and THRACE-NCT01062698). Most recently, the American Heart Association/American Stroke Association has published revised guidelines for acute ischemic stroke, indicating that mechanical embolectomy has Class 1, Level A evidence for efficacy and should be provided to patients at centers with endovascular capability.²² Further research in progress will explore wake-up strokes (NCT02143383), and considerable efforts will be directed at treating patients more quickly much like our cardiology colleagues concluded with myocardial infarction.

From the time of the first endovascular device clearance by the FDA in 2004, it took 11 years to improve device design and ultimately prove efficacy of this treatment method. Although it is exciting to be where we are now, this decade-long debate left many patients untreated because some physicians were skeptical of device efficacy and wanted Class 1, Level A evidence before instituting this in their centers. Ultimately, having high-quality data from randomized trials is what we all seek, but these authors feel that this took too long in part because of the regulatory and reimbursement process currently in use in the United States and other countries. We join others who recommend that Centers for Medicare & Medicaid Services not cover newly cleared devices unless the device is being used in a randomized trial sufficient to prove or disprove clinical efficacy. The 510-K process is fine as is because it allows companies to rapidly innovate promising new devices. However, these devices should not be sold for use outside of clinical trials until they are proven effective. Such a regulatory/reimbursement strategy will lead to more rapid determination of efficacy and the rejection of promising but ineffective devices more quickly. This adheres to the least-burdensome rule currently used by the device arm of the FDA and conforms to the desire to improve the health state of patients by Centers for Medicare & Medicaid Services. In our estimation, we may have shaved off 4 to 5 years of time to obtain Class 1, Level A evidence of embolectomy had this Centers for Medicare & Medicaid Services strategy been in place, and likely this would have improved the health state of thousands of patients. Our hats are off to those interventionalists, neurologists, clinical research coordinators, nurses, institutional review boards, and industry sponsors who had the guts to randomize patients in these trials and prove what many of us hoped was right.

• Received July 16, 2015.
• Revision received August 10, 2015.
• Accepted August 11, 2015.

• © 2015 American Heart Association, Inc.