In February the New England Journal of Medicine published the long-awaited results of three prospective stroke trials. The Interventional Management of Stroke III (IMS III) trial, the Mechanical Retrieval and Recanalization of Stroke Clots Using Embolectomy (MR RESCUE) trial and the SYNTHESIS Expansion trial all concluded that endovascular therapy (EVT) was not superior to medical management in the treatment of acute stroke.1–3

A critical appraisal of these trials, however, raises serious questions as to their applicability to the contemporary treatment of acute ischemic stroke. In evaluating the results, it is of paramount importance to keep two seemingly obvious fundamental concepts in mind. First, endovascular techniques are not designed to treat all acute ischemic stroke; they are specifically designed to address stroke secondary to large vessel occlusion (LVO). Any clinical trial testing the efficacy of a procedure must be meticulously designed to exclude patients who do not have the index disease process—in this case LVO. Any trial of EVT which includes a significant percentage of patients without the index disease will have little or no capacity to demonstrate a treatment effect. The IMS III and SYNTHESIS Expansion trials both fail in this regard.

Second, in order to evaluate a treatment paradigm within in a clinical trial—in this case EVT—it is a prerequisite that the allocated treatments are actually performed in those subjects randomized to the treatment arm and that the technical success rates achieved by the devices and operators are acceptable within the context of current practice. The IMS III and MR RESCUE trials both fail in this regard. Unfortunately, these basic criteria for clinical trials are sometimes lost in the overriding momentum to report a final definitive conclusion.

IMS III sought to compare the efficacy of combined intravenous tissue plasminogen activator (tPA) administration plus ‘protocol-approved’ EVT with that of intravenous tPA administration alone.1 The study, which began enrolling patients in 2006, was halted in 2012 because of futility. It failed to show the superiority of combined therapy over that of intravenous tPA alone. Nevertheless, major issues with patient inclusion and endovascular techniques call into question this conclusion.4

Of the total group of 656 patients enrolled, only 216 (33%) had CT angiograms (CTA). The majority of patients were enrolled after screening with plain head CT scans and without imaging confirmation of a LVO. Thus, the majority of patients enrolled in IMS III had no imaging confirmation that they actually had the index disease process. Not surprisingly, nearly 20% of patients assigned to the endovascular arm demonstrated no angiographic evidence of LVO or had ‘inaccessible’ thrombus and therefore did not receive intra-arterial therapy.

Patients without an accessible LVO in the EVT arm were not only exposed to potentially unnecessary procedural risks, but the majority also received less than the standard intravenous tPA dose prior to angiography (a protocol change in 2011 allowed standard dosing of all patients). At the same time, similar patients in the intravenous tPA group simply received full-dose intravenous tPA. This represents an important consideration since these patients with angiographically occult or distal small vessel occlusions typically do very well with full-dose intravenous tPA. So not only did the inclusion of a large percentage of patients without LVO potentially obscure the treatment effect of EVT, but this design also exposed these EVT-allocated patients to excess procedural risk (ie, angiography, general anesthesia) and resulted in them getting lower doses of intravenous tPA.

The importance of CTA-based patient selection is evidenced by the subgroup analyses which revealed a significant benefit of intra-arterial therapy, as measured by modified Rankin shift, among patients with an LVO documented on CTA (p=0.01).5 This subgroup constitutes the patients with early acute stroke who are actually being taken for EVT at the vast majority of centers. Thus, to some extent one could conclude that IMS III actually affirms the benefit of EVT in the patient population who are currently being treated—those with early acute stroke and a confirmed LVO.

Patients in IMS III were excluded if they demonstrated an unambiguous large area of hypodensity, suggestive of a completed infarct, on CT. Other standard signs of early acute completed infarction such as loss of gray-white distinction and sulcal effacement were not exclusionary. Furthermore, roughly 40% of enrolled patients had Alberta Stroke Program Early CT scores (ASPECTS) of ≤7, with lower ASPECTS scores being correlated with poor functional outcomes. Modern radiographic screening modalities such as CTA-based ASPECTS scoring, CT perfusion and magnetic resonance (MR) diffusion would probably have excluded a significant percentage of patients enrolled in the endovascular arm of IMS III.6–9 The inclusion of patients with completed infarcts almost certainly further diluted any advantage of intra-arterial therapy while increasing the risks of adverse events such as hemorrhagic conversion and complications attendant to the intervention.

Finally, beyond the concerns of patient selection criteria stated above, the endovascular devices used in the IMS III trial are largely no longer used in current practice. Only 59 patients were treated with modern devices such as the Penumbra aspiration system (Penumbra, Alameda, California, USA) (n=54) and the Solitaire Flow Restoration device (Covidien/ev3, Dublin, Ireland) (n=5). The primary device used in IMS III was the Merci Retriever (Stryker Neurovascular, Mountain View, California, USA) which was used in 28.4% of patients. In IMS III, Thrombolysis In Cerebral Infarction (TICI) 2b–3 reperfusion occurred in 23–44% of patients, depending on the occlusion site. This result differs dramatically from recent studies assessing the effectiveness of Trevo (Stryker Neurovascular) and Solitaire.10–12 In the Trevo 2 trial, the Trevo device achieved TICI 2b–3 reperfusion in 68% of cases.10 Similarly, the SWIFT trial, the recent comparative study of Solitaire and the Merci Retriever, showed the Solitaire to have far superior revascularization rates (61% vs 24%; p=0.0001) and fewer procedural complications than the Merci Retriever.11 ,12 Moreover, modern devices and access products (eg, flexible distal intracranial guide catheters) have also markedly improved revascularization times to the point where less than 60 min is considered standard.13 ,14 Time to revascularization was an important determinant of clinical outcome in IMS III, leading one to conclude that, by employing modern devices, outcomes with intra-arterial therapy would be considerably better.

In the MR RESCUE trial the investigators sought to determine whether penumbral-based imaging was useful in selecting patients with acute ischemic stroke who might benefit from thrombectomy. They concluded that penumbral-based imaging was not useful and that EVT was not beneficial compared with standard medical therapy. However, the data are not sufficient to support such conclusions.

The MR RESCUE trial enrolled only 127 patients (approximately 1 per month) from 2004 to 2011.2 As a result, as in the IMS III trial, the geologic enrollment rate in the MR RESCUE trial created a situation in which the technologies used during trial had become completely obsolete by trial completion. Embolectomies were performed employing the Merci Retriever, the early generation Penumbra aspiration catheters or both, achieving TICI 2b–3 reperfusion in approximately 25% of patients (16/64) randomized to EVT. Obviously, this rate of revascularization is unacceptably low when compared with those achievable in core laboratory adjudicated analyses with the current generation of stent retrievers and newer aspiration thrombectomy catheters.10–12 ,15 ,16 Unfortunately, these data were not featured in the main paper results but are available only in the online supplement. In fact, interventional therapy in the MR RESCUE trial was so ineffectual that the 7-day revascularization and reperfusion rates were no different in the EVT and medical management cohorts.

So it is challenging to draw any conclusions regarding the applications of penumbral-based patient selection for endovascular intervention from data derived within the context of a trial in which the endovascular interventions were only successful in a minority of the treated patients. Are we really ready to discard penumbral imaging-based patient selection (and EVT) on the basis of a study in which only 64 patients underwent EVT and of whom only 16 were effectively revascularized? Finally, while we applaud the effort of the study investigators in attempting to study the critical question of penumbral-based patient selection, evaluation of this aspect of the study is difficult given that details regarding specific techniques used for CT perfusion analysis were not detailed, rendering assessment of the design and utility of these protocols impossible.

Despite the small number of patients enrolled and the low rates of successful intervention, the MR RESCUE trial did show that successful 7-day revascularization and reperfusion was strongly associated with better patient outcomes regardless of their penumbral pattern or treatment disposition. In fact, no patients in either the penumbral or non-penumbral groups had modified Rankin Scale (mRS) scores of ≤2 at 90 days if they were not revascularized while, if they were successfully revascularized, 35% of the penumbral patients and 21% of the non-penumbral patients had mRS scores of 0–2.

Similar problems are evident in the SYNTHESIS Expansion trial which tested an experimental treatment protocol that was (and is) well outside the current standard of practice in the management of acute ischemic stroke. In this study, 181 stroke patients presenting within 4.5 h were randomized to endovascular treatment and 181 patients were randomized to treatment with intravenous tPA. The outcomes (mRS ≤1 at 90 days) were no different between the groups and the authors again concluded that EVT offered no benefit over standard medical management with intravenous tPA. However, it is critical to emphasize that these trial results only apply to the experimental treatment algorithm employed in this study and are in no way generalizable to the vast majority of patients undergoing intra-arterial therapy for acute stroke.

In the SYNTHESIS Expansion protocol, intravenous tPA was withheld from all patients allocated to the EVT arm.3 ,17 At the same time, patient enrollment was based on basic head CT alone without any confirmation of an accessible LVO. Moreover, almost half of the patients included in the study had a NIH Stroke Scale (NIHSS) score of <11, indicating a low likelihood of an LVO. Thus, patients in the EVT arm who were most likely to benefit from intravenous tPA (ie, those with distal occlusions or angiographically occult occlusions) did not receive intravenous tPA while those in the medical arm did. The very design therefore introduces a tremendous bias in favor of the medical therapy group.

In addition, treatment was not initiated in patients in the EVT arm until, on average, 1 h later than those assigned intravenous tPA and, when treated, most interventions employed antiquated techniques and devices. Specifically, 66% of patients were treated with intra-arterial tPA alone or with microcatheter and microwire clot disruption, probably achieving very low rates of revascularization. If no LVO was encountered at angiography, operators were directed to infuse intra-arterial tPA into the vascular distribution which they felt was most likely to be affected, at their discretion, up to a maximum dose of 0.9 mg/kg. This unique practice represents a novel and exploratory extrapolation from most established intra-arterial treatment protocols.

Even more astounding is the failure of the authors to report the percentage of EVT patients who actually had an LVO, the location of these occlusions or the recanalization rates and TICI scores from the subsequent interventions. It is impossible to ascertain how many patients in the SYNTHESIS Expansion trial had LVOs, so we do not even know what percentage of the patients in the study had the index disease process! However, given that so many enrolled patients had a low NIHSS score, it is likely that many did not.

It should also be stressed that, despite these significant concerns regarding study design which resulted in a powerful bias towards the medical arm, the EVT arm still performed equally well across all outcome measures. The study conclusion might just as easily have been ‘EVT is equally efficacious as traditional intravenous tPA even when applied to a large percentage of patients who are traditionally thought ineligible for EVT’.