Background Interventional Management of Stroke III did not show that combining IV recombinant tissue plasminogen activator (rt-PA) with endovascular therapies (EVTs) is better than IV rt-PA alone.
Objective To report efficacy and safety results for EVT of intracranial internal carotid artery (ICA) and middle cerebral artery trunk (M1) occlusion.
Methods Five revascularization methods for persistent occlusions after IV rt-PA treatment were evaluated for prespecified primary and secondary endpoints, after accounting for differences in key baselines variables using propensity scores. Revascularization was scored using the arterial occlusive lesion (AOL) and the modified Thrombolysis in Cerebral Ischemia (mTICI) scores.
Results EVT of 200 subjects with intracranial ICA or M1 occlusion resulted in 81.5% AOL 2–3 recanalization, in addition to 76% mTICI 2–3 and 42.5% mTICI 2b–3 reperfusion. Adverse events included symptomatic intracranial hemorrhage (SICH) (8.0%), vessel perforations (1.5%), and new emboli (14.9%). EVT techniques used were standard microcatheter n=51; EKOS n=14; Merci n=77; Penumbra n=39; Solitaire n=4; multiple n=15. Good clinical outcome was associated with both TICI 2–3 and TICI 2b–3 reperfusion. Neither modified Rankin scale (mRS) 0–2 (28.5%), nor 90-day mortality (28.5%), nor asymptomatic ICH (36.0%) differed among revascularization methods after propensity score adjustment for subjects with intracranial ICA or M1 occlusion.
Conclusions Good clinical outcome was associated with good reperfusion for ICA and M1 occlusion. No significant differences in efficacy or safety among revascularization methods were demonstrated after adjustment. Lack of high-quality reperfusion, adverse events, and prolonged time to treatment contributed to lower-than-expected mRS 0–2 outcomes and study futility compared with IV rt-PA.
Trial registration number NCT00359424.
Statistics from Altmetric.com
The Interventional Management of Stroke (IMS) III Study failed to demonstrate that the approach of combining IV recombinant tissue plasminogen activator (rt-PA) with endovascular therapy (EVT) improves clinical outcomes as compared with IV rt-PA alone.1–3 Possible reasons for failure include technical access problems delaying or prohibiting IA treatment, inability to recanalize occluded arteries, serious adverse events caused by EVT, and poor patient selection, leading to late, futile revascularization. IMS III started enrolling patients in 2005 after first experiences with mechanical thrombectomy for ischemic stroke had been published.4 ,5 IMS III added new devices as they became cleared by the Food and Drug Administration but was not designed to compare methods or devices. In addition, because subjects were not randomized to the various EVT approaches, choice of a given approach by IMS III investigators could lead to major imbalances in baseline variables associated with clinical outcome and revascularization. These include location of the arterial occlusion, age, time to treatment, etc. In addition, the small numbers of patients in the various device subgroups yield low power for conclusive statistical comparisons. Nevertheless, the overall result of the combined IV rt-PA and EV paradigm may not be fully understood without exploring potential differences in performance of individual approaches and devices within the trial, and without analyzing the impact of other variables on clinical outcomes such as adverse effects and futile recanalization in comparison with IA thrombolysis as applied in the Emergency Management of Stroke Study,6 a pre-IMS registry,7 and the IMS I and II cohort studies.8–10
Study eligibility/exclusion criteria and randomization methods have been previously reported.1 ,11 Nine hundred subjects were to be randomized, but the trial was stopped early owing to futility after it was determined that the primary endpoint (a 10% difference in the proportions of modified Rankin scale (mRS) 0–2 outcome) was unlikely to be achieved.12 No imaging method to identify target vessel occlusion was required for entry into the trial. CT angiography (CTA) or magnetic resonance angiography was performed in centers where they were established as a local standard of evaluation and care. Prespecified analyses comparing revascularization results and clinical outcomes based on available baseline CTA and 24 h CTA occlusion data for the IV and EVT treatment groups have been reported separately.13
Primary target vessels were defined as the arteriographic occlusion treated by the operator. M1 occlusion was defined as blockage of the middle cerebral artery (MCA) trunk with 100% of MCA cortical distribution at risk, minus flow to the anterior temporal artery. Angioplasty/stenting for cervical internal carotid artery (ICA) stenosis or occlusion in conjunction with intracranial target occlusions (tandem occlusions) was a protocol violation. Five endovascular revascularization methods were approved for use during the trial (thrombolysis via standard microcatheter/guidewire rt-PA infusion or ultrasound-enhanced lysis (EKOS Micro-Sonic SV infusion system)9 ,10; clot removal via the MERCI System,4 ,5 the Penumbra Aspiration System,14 or the ev3/Covidien Solitaire FR Revascularization Device15). Only one approved study device method was to be used in each patient. IA rt-PA infusion was also allowed as an adjuvant to mechanical thrombectomy. A 2000 U bolus of heparin was required for each protocol for EV treatment procedures, followed by 500 U/h. IV infusion.
The primary outcome measure was a mRS score of 0–2 at 90 days. Secondary efficacy endpoints included revascularization, as measured by the modified Thrombolysis in Cerebral Infarction (mTICI) 2–3, mTICI 2b–3, and the arterial occlusive lesion (AOL) score,8 as ascribed by consensus of the angiography core laboratory members (TAT, DSL). Primary safety endpoints were mortality and symptomatic intracranial hemorrhage (SICH), the latter defined as an intracranial hemorrhage temporally related to a decline in neurologic status which, in the judgment of the clinical investigator, warranted medical intervention. Additional safety endpoints included CT-identified parenchymal hematoma (PH) types 1 and 2, asymptomatic ICH (ASICH), subarachnoid hemorrhage (SAH), intraventricular hemorrhage, emboli into a new arterial territory (defined as a new occlusion not present before treatment in the anterior cerebral artery (ACA) beyond the A1 segment, the posterior cerebral, superior cerebellar, or posterior inferior cerebellar arteries, or of an uninvolved M2 division during treatment of an MCA occlusion), and vessel dissection or angiographically identified perforation.
Owing to the small numbers treated with the EKOS and Solitaire devices, statistical comparisons were restricted to subjects treated with a standard microcatheter, the Merci retriever, or the Penumbra system. Although subjects were randomized to the EVT arm, the choice of device for these subjects was not randomized, resulting in a potential selection bias. Propensity score analysis16 was used to compare the various EVT modalities with regard to the specified outcomes (mRS 0–2, mTICI 2–3, AOL 2–3, mortality at 90 days, asymptomatic and symptomatic ICH within 30 h of IV tPA initiation) to account for potential biases and achieve better balance in the comparison of devices. Comparability of the device groups across relevant demographic and clinical characteristics was assessed. Variables for which an imbalance was demonstrated were included in the propensity score model, as were variables for which an association with outcome could be established below a 0.10 level of significance. Propensity scores were estimated via multinomial regression. Logistic regression was used to test for an association between device and outcome, after adjustment for tertiles of the propensity score, in order to balance the devices for comparison. Success of the propensity score approach was assessed by considering the balance of the devices with respect to baseline characteristics after adjustment for the tertiles.
Six hundred and fifty-six subjects were randomized 1:2 to either 0.9 mg/kg IV rt-PA, 10% bolus with remainder over 1 h (n=222), or reduced-dose IV rt-PA (0.6 mg/kg, 15% bolus), followed by arteriography and EV therapy (n=434). Our study population comprised 200 of 328 patients with intracranial occlusions who had the intracranial ICA (n=65) or M1 (n=135) as target vessel.
Of the 434 subjects randomized to the EVT arm, 11 (2.5%) subjects did not have angiograms owing to either worsening (n=3) or improvement (n=2) of their clinical condition, enrollment errors (n=2), protocol violations (n=3), or inaccessible occlusion (n=1). Of 423 receiving arteriography, 89 (21.0%) did not receive EVT, including 80 without treatable occlusion who achieved 62.5% mRS 0–2 outcomes. Of these, 34 were determined to have M3 or M4 occlusion by the angiographic core laboratory. Baseline CTA had been performed in 15 of these 34 subjects, with 14 demonstrating occlusions of the terminal ICA (n=1), M1 (n=8), M2 (n=4), or M3 (n=1) vessels.
Details of immediate post-procedure and revascularization and 90 day mRS 0–2 outcome results for 328 evaluable subjects are given in table 1. Revascularization rates for all anterior circulation AOLs (n=311) were: AOL 2–3 recanalization 78.8%, mTICI 2–3 reperfusion 73.3%, mTICI 2b–3 reperfusion 39.9%, with 35.1% achieving mRS 0–2 outcome.
Intracranial ICA or M1 occlusions were treated in 200 subjects and subjected to propensity score analysis, with baseline clinical characteristics given in table 2. In this group, high AOL recanalization rates (82%) and mTICI 2–3 reperfusion rates (76%) allowed only 29% of EV-treated subjects to achieve good outcome. Figure 1 shows the revascularization results and clinical outcome results (mRS 0–2) for each device/method. More than 1 study device was used in 12 subjects, and non-study devices in 3, as protocol violations. mTICI ≥2 reperfusion was achieved in 70.8% of thrombolysis-only treatments via ultrasound or standard microcatheter, and 76.7% of all thrombectomy-device procedures. Four M1 occlusions were treated solely by the Solitaire device. Table 3 outlines mRS 0–2 outcomes for intracranial ICA and M1 occlusion for each mTICI reperfusion level, and for reperfusion >mTICI 2a, (the prespecified trial revascularization efficacy endpoint), evidence that the likelihood of good outcome increases as reperfusion improves. Figure 2 shows differences in levels of AOL recanalization and mTICI reperfusion by device, indicating good recanalization does not necessarily indicate good reperfusion.
Figure 3 summarizes 90 day mRS 0–2 outcomes by device for those reaching mTICI 2–3 reperfusion. The Merci device performed less well numerically, and thrombolysis performed similarly to other thrombectomy devices by this measure.⇓
Adverse events for intracranial ICA and M1 occlusions in each device group are summarized in table 4. Comparisons of devices with regard to baseline characteristics indicated some level of imbalance for atrial fibrillation, age, and systolic blood pressure, as well as the time from onset to IV rt-PA initiation and IV rt-PA initiation to groin puncture. These variables were included in the propensity score model for all outcomes.17 Additional variables included in the model were baseline characteristics associated with clinical outcome (α=0.1) and therefore varied for each outcome modeled. Despite suggested differences in performance noted in figures 1⇑–3, when adjusted for key clinical variables via propensity score tertiles, there was insufficient evidence to conclude that the revascularization methods differed with respect to the proportion of subjects with mRS 0–2 (p=0.80), mortality (p=0.32), mTICI 2–3 reperfusion (p=0.16), AOL 2–3 recanalization (p=0.86), or ASICH (p=0.17). Too few cases of SICH or emboli into new arterial territories (ENT) precluded propensity-adjusted analysis. SAH and PH-2 were numerically more common with the Merci devices.
ENT were identified in 9/40 (22.5%) evaluable ICA occlusions by the core laboratory. Twenty-five of 65 intracranial ICA occlusions were not evaluable because collateral vessels were not examined at arteriography to distinguish pre-existing emboli versus ENT. ENT were identified in 17/135 (12.6%) M1 occlusions. mRS 0–2 outcomes were numerically less frequent in subjects with an ENT (19.2%) versus subjects with no ENT (32.2%).
Of 172 evaluable target intracranial ICA and M1 occlusion cases, forty-seven cases had tandem ICA stenosis or occlusions.
IMS III clinical outcome varied according to recanalization and reperfusion results for subjects with intracranial ICA and M1 occlusion. However, high AOL recanalization rates (82%) and high mTICI 2–3 reperfusion rates (76%) allowed only 29% of EV-treated subjects to achieve good outcome. Although mTICI 2a reperfusion led to a greater proportion with a good outcome than for those with little or no reperfusion, mTICI 2b was better than mTICI 2a, supporting mTICI 2b–3 as the goal of reperfusion therapy for intracranial ICA and M1 occlusions.18 Both %AOL recanalization and %mTICI reperfusion are an improvement over IMS I and II, with %mTICI 2–3 reperfusion more closely approaching %AOL recanalization as measurable endpoints with the techniques used in this study.19 It must be acknowledged that thrombectomy devices and methods used in the study, with the exception of the few uses of the Solitaire device, are now obsolete, with a variety of stentrievers and improved suction thrombectomy in use or under investigation.
As in IMS I and II, subjects randomized to the IMS III EVT arm with no major AOL at the time of angiography contributed to a near doubling of the proportion of subjects with good outcome, despite reduced-dose IV rt-PA: 62.5% of those with no major AOL versus 35.6% for all EV-treated IMS III subjects. However, absence of a treatable occlusion at arteriography did not indicate the absence of AOL at baseline. Of EVT subjects with pretreatment CTA, 89.6% exhibited CTA intracranial occlusions. Early partial or complete recanalization of CTA-demonstrated AOL after IV rt-PA was identified at subsequent arteriography in 8/52 (15.4%) ICA and 34/93 (36.6%) M1 occlusions.20 These data point to the effectiveness of rapid recanalization with IV rt-PA in some patients with large artery occlusions, and also to the limit or ceiling of good outcomes that may be achieved with even the earliest revascularization with any IV or EV therapy in a population selected based on clinical and limited imaging criteria.
We identified no clear difference in clinical or revascularization results according to the revascularization method of intracranial ICA and M1 occlusions, after adjusting for differences in key baseline clinical variables using the propensity score. However, the small number of subjects in the various EVT approaches limits power, and as with most statistical approaches, the propensity score adjustment works best with larger sample sizes. Nevertheless, outside of large randomized trials, in which the treatment groups are balanced for key baseline variables, careful attention needs to be given to any direct comparisons between EVT approaches within or between studies. The TREVO-2 and SWIFT Trials, which compared stent retriever devices with the Merci device, did identify differences in revascularization, safety, and functional clinical outcome between technologies.21 ,22 The Merci device exhibited numerically greater SAH, PH2, ENT in IMS III, while achieving lower mRS 0–2 with mTICI 2–3 reperfusion.
SICH and several other adverse events occurred only infrequently so that they could not be included in the propensity analysis to identify differences among devices. In IMS III, SICH was similar in the EV (6.2%) and IV (5.9%) arms. However, SICH occurred in 7.2% of 334 subjects who actually underwent revascularization.
Emboli into new territories, usually the ACA, occurred during treatment of 1.7% M1 occlusions in IMS I and II, but in 12.6% of subjects in IMS III. With intracranial ICA occlusion in IMS I and II, distal ACA occlusion (A2 or beyond) was confirmed in approximately 15% subjects before treatment with IA rt-PA, and new additional distal emboli were introduced in 15% of subjects during treatment,23 with new ACA emboli interfering with collateral flow to the occluded MCA and contributing to poorer outcome. In IMS III, 9 new emboli were confirmed in 40 evaluable intracranial ICA occlusions (22.5%). New emboli were numerically less frequent with thrombolytic treatment of intracranial ICA and M1 occlusions than with thrombectomy devices, although not statistically different. The 13% difference in mRS 0–2 outcomes between those subjects with and without new emboli, predominantly occurring with clot-grasping devices, is notable as a potential contributor to study futility.
Catheterization-related dissection (2.1%) was predominantly identified in the cervical ICA as a complication of guide catheter placement. It is intuitive that it should occur more frequently in treatment methods requiring placement of large catheters (8 or 9 F), as opposed to thrombolytic methods allowing smaller (5 or 6 F) catheter use.
Multiple devices were used in 15 subjects (including 4 additional uses of the Solitaire device) and are not included in the propensity score analysis owing to small sample size. Numerically, these subjects had the highest AOL 2–3 recanalization (93.3%) and mTICI 2–3 (93.3%) and mTICI 2b–3 (60%) reperfusion (table 2), as well as the highest mRS 0–2 outcome (40%) rates. However, they also had the highest mortality (40%) and cumulative SAH, SICH, and ASICH as well. Operator-reported dissection was low for the entire EVT group but highest in the multiple-device subgroup. The risk of multiple-device application may remain a safety concern in future studies where more effective identification and exclusion of subjects with little hope of improvement will place a greater premium on safety for those who can improve.19 ,24–27
For all occlusion sites, mTICI 2–3 reperfusion was more frequent in IMS III than in IMS I and II, where only thrombolysis, without or with ultrasound assistance, was used. The reason for differences in mortality and mRS 0–2 outcomes despite better revascularization outcomes in IMS III versus IMS I and II is not easily explained by differences in device performance or baseline clinical variables. However, EVT in IMS III began 37 min later than in IMS I and II (127 vs 90 min), despite IV treatment being 6 min earlier, on average. mRS 0–2 outcomes were lower for all revascularization endpoints in IMS III compared with IMS I–II (figure 4). A longer time to EV treatment in IMS III is a key variable that probably explains differences in outcome results compared with IMS I–II, and probably also contributed to trial futility.28 In IMS III, every 30 min delay in angiographic reperfusion reduced the relative likelihood of a good clinical outcome by 12% in adjusted analysis for intracranial ICA, M1, and M2 occlusion.29
The last-added IMS III device, the Solitaire FR revascularization device, was introduced after the angiographic core laboratory had reviewed the imaging studies and the angiograms from the Solitaire Retrospective Trial and device clearance by the Food and Drug Administration.15 The 85% TICI 2b–3 reperfusion results recorded, using IMS criteria from the same core reader, led to expectation of introduction of the device in the study. The mRS 0–2 outcomes of 56% in the Solitaire study probably reflect more rapid reperfusion after device deployment, more complete recanalization after its retrieval, or patient selection advantages. However, the device was subsequently used only five total times per protocol before IMS III trial termination.
The IMS III primary analysis demonstrated a trend for improved outcome for subjects with National Institutes of Health Stroke Scale score ≥20 in the EVT group at 90 days, compared with IV treatment alone in those subjects treated with IV in <2 h.1 More rapid treatment times comparable to IMS I and II, in combination with earlier introduction of stent retriever technology, might have led to a positive result in this subgroup, and prevented trial futility. Further analysis of relevant treatment time points is the subject of a separate manuscript.30 In the final analysis, an overall positive result with EVT might have been possible only with much more rapid, effective, and safer reperfusion results overall.
In IMS III, clinical outcome correlated to good recanalization and reperfusion results. Despite high AOL recanalization rates (82%) and high mTICI 2–3 reperfusion rates (76%) for EV-treated subjects with intracranial ICA and M1 occlusion in IMS III, late time to treatment in combination with 43% mTICI 2b–3 reperfusion allowed only 29% to achieve a good outcome. No clear difference between EVT methods in clinical and imaging measures related to outcome was seen.
Collaborators The IMS III Study Group.
Contributors The following authors meet the criteria for authorship as noted: TAT, JPB, JS, YYP, JC, MDH, AMD, TJ, PK, DSL, MG, RvK, BY, OOZ, WS, SE, SDY, RM; conception or design of the work: TAT, JPB, JS, YYP, MDH, PK, SDY, AMD; acquisition, analysis, or interpretation of data for the work; TAT, MDH, SDY, DSL, LF, JC, MG, AMD, TJ, RvK, BY, OOZ, WS, SE, PK, JS, YYP, JPB; drafting the work or revising it critically for important intellectual content; TAT, MDH, TJ, PK, DSL, MG, RvK, BY, WS, SE, SDY, RM, PK, YYP, JPB; final approval of the version to be published; TAT, DSL, MDH, TJ, PK, MG, RvK, BY, SDY, JPB; and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
Funding The study was supported by the National Institue of Health and National Institute of Neurologic Disease and Stoke (UC U01NS052220, MUSC U01NS054630, and U01NS077304), and by Genentech, EKOS, Concentric Medical, Cordis Neurovascular, and Boehringer Ingelheim.
Competing interests None.
Patient consent Obtained.
Ethics approval Institutional review boards at all participating institutions.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Multiple concurrent manuscripts have been submitted, or are in preparation, originating from data not included in the this manuscript. A public use dataset form original case report data is in preparation.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.