Article Text


Original research
Pre-intervention triage incorporating perfusion imaging improves outcomes in patients undergoing endovascular stroke therapy: a comparison with the device trials
  1. Ansaar T Rai,
  2. Karthikram Raghuram,
  3. Jennifer Domico,
  4. Gerald Hobbs,
  5. Jeffrey Carpenter
  1. Interventional Neuroradiology, West Virginia University Hospital, Morgantown, West Virginia, USA
  1. Correspondence to Ansaar T Rai, Department of Radiology/Neuroradiology, West Virginia University Hospital, 2002 Ices Ferry Drive, Morgantown, WV 26508, USA; ansaar.rai{at}


Objectives Endovascular therapy of acute ischemic stroke is evolving towards thrombectomy devices for vessel recanalization. High rates of revascularization have been reported in stroke device trials. However, the discrepancy between recanalization and outcomes raises the question whether patients with irreversible ischemic injury are being exposed to these interventions. This study evaluated a triage methodology that incorporates perfusion imaging against previous device trials that treated all patients within a certain time frame.

Methods 99 consecutive patients were identified with anterior circulation strokes who had undergone endovascular therapy. All patients had a baseline NIHSS score ≥8 and had undergone pre-intervention CT perfusion. Rates of recanalization and functional outcomes were compared with the MERCI, Multi-MERCI and Penumbra trials.

Results This study's recanalization rate of 55.6% is not significantly different from the 46% for MERCI (p=0.15) and 68% for Multi-MERCI (p=0.08) but was significantly lower than the 82% for the Penumbra trial (p<0.0001). Successfully recanalized patients had a significantly higher good outcome of 67% in this cohort versus 46% in MERCI, 49% in Multi-MERCI and 29% in Penumbra. The rate of futile recanalization was 33% compared with 54% for MERCI, 51% for Multi-MERCI and 71% for Penumbra. A small cerebral blood volume (CBV) abnormality (p<0.0001) and large mean transit time–CBV mismatch (p<0.0001) were strong predictors of a good outcome.

Conclusion Despite similar or lower recanalization rates, there was a significantly higher rate of good outcomes in the recanalized population and thus a significantly lower rate of futile recanalization in this study versus the device trials, suggesting a role for pre-intervention perfusion imaging for patient selection.

Statistics from

The current trend in endovascular stroke therapy towards newer thrombectomy devices is predicated on the US Food and Drug Administration approval of previous clot retrievers. The primary endpoint for success in these and ongoing device trials is vessel recanalization, a predictor of good outcome.1–3 However, high rates of recanalization have not been matched by proportionally high rates of good functional outcome.

In order to test the hypothesis that a patient selection methodology including the assessment of preprocedure cerebral perfusion leads to better outcomes, especially in recanalized patients, we compared data from our center that relied on pre-intervention perfusion imaging with the mechanical thrombectomy device trials that treated all comers.


This is a retrospective review of patients with acute ischemic stroke in the anterior circulation that underwent endovascular therapy. The study was performed after approval by the institutional review board. The data from our center were compared with the device trials for acute ischemic stroke as discussed below.

Patient selection

At our institution patient selection for endovascular therapy of stroke incorporates preprocedure perfusion imaging analysis, with the benefit of any doubt always given to the patient. Patients are primarily excluded from an intervention if the non-contrast head CT (NCCT) shows signs of early ischemia affecting at least one third of the occluded vascular territory. Secondary exclusion can be based on the perfusion imaging showing a large cerebral blood volume (CBV) abnormality, for example, 50–75% of the affected vascular territory. The rationale is that an irreversible ischemic injury could be identified using CBV before it becomes abnormal on NCCT. Endovascular therapy in patients with minimal or no NCCT abnormality but with large CBV abnormality may not only be futile but expose the patient to unnecessary risks of the procedure. As mentioned initially, in cases of any doubt we favor treating the patient over no treatment. For example, patients presenting within 3 h but ineligible for intravenous thrombolytics are generally treated unless there is a large NCCT abnormality.

We do not have a cut-off time limit for no treatment. Patients presenting beyond 8 h or with wake-up strokes were treated if they had favorable imaging profiles.

In this stroke treatment triage setting, we applied the following inclusion criteria to all consecutive patients who underwent endovascular stroke therapy over an 8-year period:

  1. Anterior circulation symptoms at presentation with a baseline National Institutes of Health stroke scale (NIHSS) score of 8 or greater.

  2. Intracerebral vascular occlusion on admission CT angiography correlating with the neurological deficit.

The endovascular procedures were performed through a transfemoral approach in all patients. General endotracheal anesthesia was used in 62 (62.6%) patients while conscious sedation was used in 37 (37.4%) patients. The type of endovascular therapy involved intra-arterial thrombolytics in only 33 (33.3%) patients, mechanical device in only 24 (24.2%) patients and both thrombolytics and mechanical thrombectomy in 42 (42.4%) patients. All patients went to the medical intensive care unit after the procedure.

Image analysis

In order to ensure consistency of the imaging analysis the source dicom data utilized for the original perfusion study was processed utilizing Vitrea-4 perfusion software by vital images. The software uses a deconvolution method based on singular value decomposition. The image acquisition time is 60 s to ensure inclusion of the complete tissue time-density curve. A neuroradiologist blinded to the procedure or the clinical outcome identified two levels predicting the largest perfusion abnormality on cerebral blood flow, CBV and meant transit time (MTT). Our CT perfusion coverage for the anterior circulation is 40 mm. We believe this is sufficient to cover the internal carotid artery (ICA) or the middle cerebral artery (MCA) territories, which were the vessels that were analyzed. It is possible to miss smaller lesions that are outside the coverage but these will typically involve third order branches, have a low baseline NIHSS and overall carry a favorable prognosis. While it is possible to have part of the lesion outside the field of coverage, we believe for proximal vessel occlusions, such as in this study, enough relevant ‘clinical’ information can be obtained on which to base a treatment decision. A region of interest (ROI) was manually drawn around the perceived perfusion abnormality yielding an area in square millimeters for each level and an average of the two levels showing the largest abnormality was obtained. A mirror image ROI was automatically generated in the contralateral unaffected hemisphere. In addition, the relative perfusion value in the abnormal ROI was measured as a percentage of the normal ROI. The MTT–CBV mismatch was defined as (1 − CBV/MTT) ×100.

Successful recanalization was defined as a post-intervention thrombolysis in myocardial ischemia score of 2 or 3. An independent observer blinded to the outcomes graded the post-procedure recanalization status. A good outcome was defined as a 90-day modified Rankin score of 2 or less.


We compared our procedural and clinical outcomes with the MERCI (Mechanical Embolus Removal in Cerebral Ischemia), Multi-MERCI and the Penumbra Pivotal trials.2–4 We only compared the anterior circulation strokes. This was done because we did not have reliable perfusion imaging data on the posterior circulation, a smaller number of posterior circulation strokes were treated in the device trials as well as our own cohort and basilar artery occlusions carry a very high mortality despite intervention.

Statistical analysis

The significance of simple bivariate associations was assessed using χ2 tests or logistic regression as appropriate. Multiple logistic regression was used when several factors were assessed simultaneously. We used the t-test to compare our sample means against the device trials. All data analysis was performed using JMP statistical software.


A total of 99 patients satisfied the inclusion criteria. A good outcome was seen in 41 patients (41.4%) while successful recanalization was achieved in 55 patients (55.6%). Twenty-five per cent of ICA, 41% of M1 and 69% of M2 occlusion patients had a good outcome (p=0.02). The mortality was 62% in patients with an ICA occlusion, 29% in patients with an M1 occlusion and 12% in those with an M2 occlusion (p=0.001). The baseline characteristics of the patients in comparison with the device trial are listed in table 1. The significant clinical and imaging predictors of outcome and mortality are shown in table 2. Younger age and lower NIHSS score significantly correlated with a good outcome while older age and a higher baseline NIHSS score predicted mortality. Patients with successful recanalization had significantly higher odds of a good outcome and lower mortality than those with no recanalization. Among the 41 patients with a good outcome, successful recanalization was achieved in 37 patients or 90.2%. In contrast, in the 58 patients with a poor outcome, successful recanalization was seen in only 18 patients or 31% (p<0.0001). Likewise, in the 55 successfully recanalized patients, a good outcome was seen in 37 patients or 67.3%, whereas only four (9.1%) of the 44 non-recanalized patients achieved a good outcome (p<0.0001).

Table 1

Baseline characteristics compared with the device trials

Table 2

Predictors of outcome and mortality

Analysis of the perfusion parameters showed that a small area of CBV abnormality and large MTT–CBV mismatch were significant predictors of a good outcome and lower mortality (table 2). We defined futile or ineffective recanalization as patients ending with a poor outcome despite successful recanalization. Among the patients who were successfully recanalized (n=55), 18 (33%) had a poor outcome or futile recanalization.

Comparison with the device trials

The MERCI trial

The MERCI trial was a prospective, single-arm, multicenter trial to test whether the device could safely achieve recanalization at a rate greater than a prespecified rate of spontaneous recanalization. The rate of spontaneous recanalization used was that reported in PROACT-II.5 Of the 151 patients enrolled, angiography results were available in 141 patients, and of these 90-day follow-up was available in 138 patients. Of these, 47 patients had an ICA (37%) and 80 patients (63%) had an MCA occlusion, these 127 anterior circulation strokes formed the cohort for comparison. The mean age and gender distribution was similar to our study (table 1). A comparison of recanalization rates, good outcomes and mortality is shown in table 3A. Our recanalization rate was not significantly different from the MERCI trial, and while the percentage of good outcomes in our series was higher, the most notable difference was the significantly increased odds of achieving a good outcome in recanalized patients in our series as opposed to the MERCI trial. To account for the lower baseline NIHSS score in our study compared with the MERCI trial (table 1) we used the trial's mean baseline NIHSS score of 20, which predicts a good outcome of 57% for recanalized patients. This is still significantly higher than the 46% good outcome in recanalized patients in the MERCI trial (p=0.008). Our time to treatment from symptom onset was significantly higher than the MERCI trial (table 1), which could potentially have an adverse effect on the outcomes (table 3).

Table 3

A comparison with the device trials

The mortality in recanalized patients was lower in the current study compared with the MERCI trial; however, there was no difference in outcomes or mortality in the non-recanalized patients and no difference in overall outcomes or mortality based on site of occlusion.

The rate of futile anterior circulation recanalization in the MERCI trial was 54%, which is significantly higher than the 33% rate in the current series (OR 2.3, 95% CI 1.3 to 4.2; p=0.002).

The Multi-MERCI trial

The Multi-MERCI trial was similar in design to the MERCI trial, but tested a newer version of the thrombectomy device. The mean age and gender distribution and composition of the occlusion site was similar to our study (table 1). While the overall baseline NIHSS score was slightly lower in our study, the baseline NIHSS score for patients with ICA and MCA occlusions was not significantly different (table 1). As shown in table 3B, the overall recanalization rate was higher in the Multi-MERCI trial. The overall percentage of good outcomes and mortality was not different; however, the percentage of good outcomes in recanalized patients was significantly higher in our series. Using the mean baseline NIHSS score of 19 (similar to the Multi-MERCI trial) predicts a good outcome of 60% in our recanalized cohort compared with the 49% good outcome in recanalized patients in the Multi-MERCI trial (p=0.007). Again, our time to procedure from symptom onset was significantly higher and should theoretically offset at least partly the effect of a lower NIHSS score. The percentage of futile recanalizations in the Multi-MERCI trial of 51% was significantly higher than the 33% in the present series (OR 2.1, 95% CI 1.2 to 3.7; p=0.009).

The Penumbra Pivotal trial

The Penumbra Pivotal study enrolled a total of 125 patients in a prospective, single-arm, multicenter trial. The baseline sample characteristics compared with our study do not show a difference between the mean age, baseline NIHSS score or distribution of the occlusion site. The mean time to procedure initiation was 4.3 h (±1.5) was, however, significantly lower than the present study. As shown in table 3C, the rate of recanalization in the Pivotal trial was significantly higher than the current series; in fact it was the highest reported recanalization rate of any of the device trials. However, the rate of good outcomes was not only significantly lower than the current series but was also lower when compared with the MERCI trials. In addition, the patients who were recanalized did not have significantly better outcomes than the overall cohort. Most importantly, however, the percentage of patients with successful recanalization who achieved a good outcome was significantly lower when compared with the current series. By contrast, the rate of ineffective recanalization of 71% was the highest of any of the device trials and was significantly higher than the 33% observed in our series (OR 4.97, 95% CI 2.7 to 9.1; p<0.0001). The overall mortality and the mortality in the recanalized and the non-recanalized groups was not different when compared with the current series. The Penumbra trial also allowed comparison of the outcome and mortality rates among the recanalized and non-recanalized cohorts by the site of vascular occlusion, as shown in table 4. Again, the percentage of patients achieving a favorable outcome with successful recanalization was significantly higher in the present series than the Penumbra trial for both the ICA and MCA occlusions. The Penumbra trial had a very high rate of futile recanalization for both the ICA and MCA occlusions. For ICA occlusions, the odds of ineffective recanalization in the Penumbra trial were more than eight times higher than the current series, a rate of 74% in the Penumbra trial versus 25% in the current study (OR 8.5, 95% CI 4.5 to 16.1; p<0.0001). Likewise, for an MCA occlusion, the odds of futile recanalization in the Penumbra trial was four times higher, a rate of 68% futile recanalization compared with 33% observed in the current series (OR 4.1, 95% CI 2.28 to 7.43; p<0.0001).

Table 4

Comparison of outcome and mortality with the Penumbra Pivotal trial based on recanalization and site of occlusion

A significantly lower mortality was observed in our series for patients with ICA occlusions than the Penumbra trial; however, no such difference was elicited for patients with MCA occlusions. Similarly, no difference in outcome or mortality was observed in patients who were not recanalized. A graphic representation highlighting the differences in futile recanalization and the rates of good outcomes in recanalized and non-recanalized patients is shown in figure 1.

Figure 1

A comparison of effective and futile recanalization between the current study and the device trials. mRS, modified Rankin score.


The ultimate goal of an endovascular stroke intervention is neurological recovery or improvement. Recanalization of an arterial occlusion is central to achieving this goal; however, higher recanalization rates are not being paralleled by equally higher rates of favorable outcomes in recanalized patients. Patient selection thus remains crucial in achieving not only recanalization in the right patient but also in accomplishing a favorable neurological outcome. The extent of irreversible ischemic injury is related to the site of vascular occlusion4 ,6 and the duration of such occlusion. The NIHSS score and the time from symptom onset serve as surrogate markers for the severity and duration of ischemia, respectively. As endovascular stroke therapy is initiated after expiration of the intravenous time window or for cases refractory to intravenous thrombolytics, it is by default selecting patients with ischemia of longer duration and strokes of worse severity. The mean time to initiation of intra-arterial therapy was almost 6 h in our study and is typically reported to be 4 h or longer in previous endovascular stroke trials,1 ,3 ,5 as opposed to 90 min in the NINDS trial.7 In terms of severity, patients with proximal vessel occlusion such as the MCA or intracranial ICA are less responsive to intravenous thrombolytics8 and are increasingly becoming the target population for stroke interventions. The current arbitrary recommendations for endovascular therapy based on a time window of up to 8 h rely on using time as a surrogate for cell death. A problem with this approach of treating all comers within a certain time frame is that endovascular procedures are not benign and carry a fairly significant morbidity and mortality. Subjecting a patient with minimal or no chance of recovery makes the procedure not only futile but also dangerous. By the same token restricting treatment to a time frame may deprive a patient of a procedure that could help regain function. Patients waking up with a stroke are prime examples of the latter and patient selection beyond 8 h based on imaging has shown promising initial results.9 ,10

The preprocedure imaging methodology at our institution evolved towards CT perfusion due to its round-the-clock availability and immediate proximity to the emergency department; a setting that is not unique to our hospital. For stroke imaging and intervention to disseminate beyond large metropolitan centers, such logistic concerns can be important. Furthermore, the utility of CT in determining viable ischemic tissue versus core infarct correlates well with MRI.11–13 Our patient selection based on utilizing MTT and CBV has valid precedence in the literature.14 Similar mismatch can be constructed by using cerebral blood flow and CBV.15 In addition, a visual evaluation of the perfusion maps has been shown to correlate with final infarct volumes,16 ,17 a fact that is important in timely assessment of the perfusion maps, and mitigates the vendor software variability that primarily affects absolute values in quantitative perfusion maps.18 Visual assessment may overestimate the size of the MTT abnormality in some patients by including non-threatened tissue;19 however, overestimating the MTT abnormality will favor treatment as it will overestimate the size of the mismatch or penumbra. An overestimation of MTT is a serious issue if it would result in excluding patients who would otherwise be eligible for therapy, but this is not the case with visual assessment, and so we feel in a real-world scenario when time is of the essence a visual subjective analysis yields enough useful information on which to base a treatment decision. The infarct core determination based on cerebral blood lesion volume is strongly supported by the literature,13 ,20 and CBV predicting clinical outcomes in patients undergoing both intravenous and intra-arterial stroke therapies is also documented.13 ,21–24 We realize that CBV may not exactly match the diffusion-weighted imaging abnormality and that it may even slightly underestimate the infarct compared with diffusion-weighted imaging;12 however, again, underestimation of the CBV size favors endovascular therapy and will result in the patient being treated. Our strategy for triaging patients even though based on imaging hinges on the philosophy that in case of any doubt the benefit is given to the patient and we err on the side of treatment. Another confounding variable is a true characterization and definition of infarct core on imaging, which may be different from its pathological description. The basis of incorporating perfusion imaging in stroke triage is to use this methodology in furnishing a functional and clinical concept of infarct core, ie, the identification of ischemic tissue that is resistant to revascularization. Our data suggest that the CBV lesion size is a surrogate for irreversible ischemic injury and thus is an appropriate parameter to aid in triaging stroke patients. Extensive discussion on appropriate perfusion imaging techniques, factors impacting vendor variation and standardization of imaging methodology is beyond the scope of the current paper and has been discussed in the literature. There is in summary sufficient evidence to support CT perfusion as a safe,25 rapid,26 and reliable imaging methodology.

Two US Food and Drug Administration-approved devices1–3 are currently being utilized for endovascular stroke therapy, with several others undergoing clinical trials and yet more in the development stage. Most if not all are non-inferiority trials designed to show a primary endpoint of similar or better acute vascular recanalization compared with a previously approved device. Recanalization as a predictor of good outcome is well supported by the literature.5 ,6 ,9 ,27 ,28 In the MERCI and the Multi-MERCI trials,1 ,2 recanalized patients had better outcomes; however; the majority of patients in the recanalized cohort, more than half, did not achieve a favorable outcome. The Penumbra Pivotal trial had the highest rates of recanalization but the lowest percentage of good outcomes, and recanalization was not even associated with a clear improvement in outcome.3 This mismatch between recanalization and favorable outcomes is termed ‘futile recanalization’.29 ,30 As shown in figure 1, our rate of futile recanalization is the lowest of all the device trials while the rate of effective recanalization is the highest. One hypothesis to explain futile recanalization could be that subjecting all patients within a certain time window to endovascular therapy disregards their individual physiology thus limiting the impact of revascularization. The fact that such response can differ among patients based on collateral circulation, comorbid conditions and preprocedure physiology is well studied and documented. Our finding of a pre-intervention imaging profile defined by a small CBV abnormality predicting a favorable outcome supports this hypothesis. This is in keeping with multiple previous studies that have determined CBV to be a determinant of functional outcome.15 ,31 ,32 We did not find any difference in mortality or outcomes in non-recanalized patients (table 3) indicating that restoration of blood flow remains vital to a good outcome. However, this restoration is more significant in terms of its impact in patients who still have viable brain tissue.

The intent of pre-intervention imaging is to identify patients who will respond to therapy. Such information, while not the sole determinant for triage, can nonetheless add a layer of knowledge to the clinical analysis and decision process. The reasons why this stated intent has not been fully realized are manifold. First, in contrast to preprocedure variables such as NIHSS score and age, which have been shown to predict outcome in large studies, no comparable, clear or consistent information exists when it comes to parametric perfusion studies. Second, imaging represents a snap shot in time in a complicated and rapidly evolving process. Due to the logistics involved in endovascular therapy an intervention may not be performed for an hour or more after the imaging study and revascularization may not be achieved until even later. The further out the intervention from the imaging, the less relevant the information obtained from it, as given time, most if not all ischemic tissue will succumb to infarction. With the introduction of the next generation of thrombectomy and retrieval devices for stroke, it is important that in order to maximize the effect of these devices, revascularization is linked to neurological recovery. These devices will have a greater impact when used on the right patient and a rapid physiological assessment of cerebral perfusion can support pre-intervention triage in selecting these patients.


A major limitation of our study is the comparison of a retrospective cohort with controlled, prospective device trials. While our data are very similar to these trials in terms of baseline characteristics, and even allowing for a higher initial NIHSS score we demonstrate significantly better outcomes, this limitation cannot be ignored. However, given the scarcity of literature showing a role of imaging in improving outcomes we hope that our data offer evidence supporting perfusion imaging as a possible tool in stroke triage.


The most significant difference between our study and the device trials was a higher rate of good outcomes in patients who were successfully recanalized and thus a much lower rate of futile recanalization. This was despite a longer time to intervention, a similar or lower recanalization rate and accounting for the difference in NIHSS scores. The endpoint of any stroke intervention is not just re-establishing blood flow but to achieve better functional outcomes as a consequence of restoring blood flow. Therefore, adding cerebral perfusion information to other clinical predictors of outcome may increase the confidence with which an intervention is decided upon.


View Abstract


  • Competing interests None.

  • Ethics approval Ethics approval was provided by the local institutional review board.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

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.