Perfusion/Diffusion Mismatch Is Valid and Should Be Used for Selecting Delayed Interventions
- First Online:
- Cite this article as:
- Davis, S., Campbell, B., Christensen, S. et al. Transl. Stroke Res. (2012) 3: 188. doi:10.1007/s12975-012-0167-8
The mismatch between a larger perfusion lesion and smaller diffusion lesion on magnetic resonance imaging is a validated signal of the ischemic penumbra, namely the region at risk in acute ischemic stroke that is critically hypoperfused and the target of reperfusion therapies. Clinical trials have shown strong correlations between reperfusion in mismatch patients and improved clinical outcomes. Attenuation of infarct growth is associated with reperfusion and corresponding clinical gains. Using computed tomography perfusion, the mismatch between relative cerebral blood flow or cerebral blood volume and perfusion delay is a comparable penumbral marker. Automated techniques allow rapid quantitative assessment of mismatch with thresholding to exclude benign oligemia. The penumbra is often present beyond the current 4.5-h time window, defined for the use of intravenous tPA. Treatment beyond this time point remains investigational. Although the efficacy of thrombolysis in mismatch patients requires further validation in randomized trials, there is now sufficient evidence to recommend that advanced neuroimaging of mismatch should be used for selection of delayed therapies in phase 3 trials.
Despite a plethora of clinical trials over several decades, there remains one licensed drug therapy for acute ischemic stroke, namely intravenous tissue plasminogen activator (tPA) . This therapy is clinically proven, with a time window of 4.5 h, based on a series of phase 3 trials and meta-analyses [1, 2]. Currently, clinical guidelines for patients treated with IV tPA within 4.5 h use similar clinical eligibility to the initial NINDS trial in 1995, imaging with non-contrast computed tomography (CT) to exclude patients with hemorrhage, non-vascular pathologies, and (arguably) major early ischemic change. The effectiveness of the treatment substantially diminishes with time. The number of patients needed to treat to achieve an excellent functional outcome ranges from 4.5 within 90 min to 14 for 180–270 min. In the recent meta-analysis, there was no significant association between time and the risk of symptomatic intracerebral hemorrhage (sICH), but the lack of efficacy above 4.5 h was associated with evidence of increased mortality in thrombolysed patients . Substantial additional data are shortly expected from the third International Stroke Trial .
Delayed interventions for acute stroke can be defined as therapies beyond the current 4.5-h time window and are based on the principle of arterial recanalization and rapid reperfusion of the ischemic penumbra. The ischemic penumbra is broadly defined as the region of critically hypoperfused tissue that is at risk of irreversible infarction in the ischemic process but can be potentially rescued by rapid reperfusion [4, 5]. Rapid identification of the penumbra should identify those patients who would most benefit from delayed stroke intervention while excluding those who have already reperfused, those where the infarct has grown into the hypoperfused region and those who have very large areas of established infarction who might be at significant risk of sICH. The penumbra is still present in a substantial proportion of patients at delayed time windows . The argument mounted here is that perfusion/diffusion mismatch with MRI has been sufficiently validated as a measure of the ischemic penumbra to have a role in selecting delayed interventions while awaiting the more definitive evidence produced from phase 3 trials.
Current Status of Delayed Interventions
Delayed interventions may be considered in patients treated initially within 4.5 h with IV tPA who have a poor treatment response, those who reach beyond 4.5 h, patients with wake-up stroke, or an unknown time of onset. Delayed interventions include intravenous tPA and the newer thrombolytic drugs selected with advanced neuroimaging and treated outside the current time window [7–10], intra-arterial (catheter-based) thrombolytic treatment with urokinase or tPA, and increasingly, the use of clot retrieval devices [11, 12]. Of these approaches, there has been only one small positive phase 3 trial beyond 4.5 h, prolyse in acute cerebral thromboembolism-II, which showed efficacy for intra-arterial prourokinase in middle cerebral artery occlusion up to 6 h after onset in 180 patients .
The newer intravenous fibrinolytic strategies (desmoteplase and tenecteplase) have been based on the evidence that these agents are more fibrin-specific and might be more potent recanalizing agents, with a lower degree of hemorrhagic transformation, particularly when selected with advanced neuroimaging [10, 14]. Delayed intervention with clot retrieval is an increasingly used strategy, based on an expanding number of phase 2 trials, where recanalization has been correlated by favorable clinical outcome . The early trials used the MERCI device with more recent strategies including the penumbra system and the Solitaire stentriever device [11, 12, 16, 17]. Of these approaches, the Solitaire stentriever device currently appears to be the most effective with a recanalization rate of over 80 % in registries . It has recently been shown to be superior to the MERCI device in a randomized control trial with a clinical endpoint . However, none of these newer IV lytics or clot retrieval devices has yet been confirmed to be superior to standard care in a phase 3 trial, and their use should be considered investigational. One large trial, the Interventional Management of Stroke Trial 3, is comparing standard dose IV tPA versus a lower dose (0.6 versus 0.3 mg/kg) followed by angiography and clot retrieval. However, advanced imaging is not employed, and patients with large infarcts may potentially receive hazardous and futile reperfusion . In this debate, we argue that patients for trials and any open-label use of delayed interventions should be selected with advanced imaging, based on the demonstration of significant penumbral presence, defined by mismatch using MRI or CT perfusion (CTP).
Imaging the Ischemic Penumbra and Mismatch—Monitoring Brain Ischemia in Real Time
An early research on identifying the ischemic penumbra used positron emission tomography (PET) with 15O, measuring cerebral blood flow (CBF) with C15O2,15O2 to measure oxygen metabolism and the oxygen extraction fraction (OEF), calculated as a ratio of these parameters to determine the ability of the tissue to extract oxygen and maintain tissue function and integrity. The concept of mismatch on PET was first described by a number of investigators including Ackerman , Baron , and Wise  who all postulated that an elevated OEF was the hallmark of the penumbra and indicated a “stressed” brain tissue where there was relative preservation of oxygen metabolism despite hypoperfusion. This was elegantly termed “misery perfusion” by Baron . However, PET is clinically impractical and two other techniques are in increasing clinical use around the world. These include MRI with perfusion weighted/diffusion weighted imaging (PWI/DWI) and CT perfusion.
The operational definition of the ischemic penumbra using MRI is the mismatch between the infarct core measured with DWI and the hypoperfused region measured with PWI (PWI > DWI mismatch). Diffusion weighted imaging measures bioenergetically compromised brain tissue associated with cytotoxic edema, is the hallmark of acute ischemic damage, and almost invariably represents tissue that has already infarcted or is destined to infarct over a short period of time [24, 25]. Although examples of DWI reversal have been identified using intra-arterial therapy, some of these changes are transient . There remained uncertainty whether a portion of DWI lesion was indeed “penumbral” and could be reversed with rapid reperfusion. More recent research by our group based on the EchoPlanar Imaging Thrombolytic Evaluation Trial (EPITHET) and pooled EPITHET-Diffusion and Perfusion Imaging Evaluation for Understanding Stroke Evolution (DEFUSE) trials have shown that the DWI lesion is rarely reversible, certainly when reperfusion is employed longer than 3 h after stroke onset and is therefore a valid assessment of the lesion core in most cases. Adjustment for possible reversal rarely alters the PWI > DWI mismatch signature [27, 28].
Perfusion imaging with MRI has been more problematic because of the realization that a significant proportion of hypoperfused brain tissue in acute stroke represents “benign oligemia,” namely mild reduction of CBF in brain regions that are not at risk of being recruited into the final infarct. To overcome this problem, thresholding has been employed to provide a better measure of the more severely hypoperfused tissue at risk in the ischemic penumbra. Several lines of research have indicated that thresholding in the range of 5–6 s (with Tmax) is a reliable approximation of this boundary [29–31].
In the absence of reperfusion, there is typically an expansion of the DWI lesion into the hypoperfused boundary in the mismatch region [32–34]. Mismatch typically (but not always) reflects the presence of large artery occlusion . Penumbral failure can be accelerated by various physiological factors such as hyperglycemia, hyperthermia, elevated hematocrit, systemic hypoxia, and advancing age [35–38].
A research by our group and others has confirmed the validity of MRI mismatch as a useful clinical signature of the ischemic penumbra. For example, measurement of the DWI volume at 24 h  and 3 days  correlates well with the final infarct volume at 90 days, the traditional time for the clinical end point in stroke trials. Reperfusion can be measured using changes in perfusion (median or dichotomized, for example, >30, 50, or 90 %), correlates well with clinical gains and inversely with infarct growth . Reperfusion has been shown to strongly correlate with clinical outcomes using thrombolytic therapy.
The initial application of MRI perfusion/diffusion mismatch was based on volumetric measurements “offline,” but the clinical utility of this approach has been greatly enhanced by the institution of automated techniques. Eyeball assessment of mismatch at the MRI console is inaccurate . A number these software automations are now in use. The RApid processing of PerfusIon and Diffusion (RAPID) methodology allows rapid estimation online of thresholded PWI, DWI, and PWI/DWI mismatch . It is being employed in clinical trials, and automated strategies are likely to be utilized in commercial MRI software and to have routine clinical application in the near future. The disadvantages of MRI and identification of mismatch regions in acute stroke patients include the time required for data acquisition and the minority of patients (around 10–15 %) who cannot tolerate the procedures because of claustrophobia, contraindication to intravenous contrast, presence of pacemakers, or other metal objects.
Although less validated in clinical trials, the mismatch concept in patient selection for delayed interventions is being increasingly used with CTP-based technology. CT is generally more available, data acquisition is more rapid, and there are less contraindications. Using CTP, a mismatch between the hypoperfused region (for example, measured with mean transit time) and the infarct core (initially defined as the region cerebral blood volume (CBV) <2 g/100 ml) was shown to be a useful measure of the ischemic penumbra, correlating reasonably well with MRI [43, 44]. More recently, it has been shown that relative CBF, with thresholding to about 30 % of the contralateral hemisphere, represents the infarct core better than the CBV [45, 46]. Limited brain coverage has been a limitation with CTP, but newer CT scanners allow whole brain acquisition. Some thrombolytic trials suggest that the use of CTP in patient selection can identify treatment responders [9, 14].
The Evidence for Mismatch Selection with MRI in Clinical Trials of Delayed Interventions
Trials of thrombolysis with penumbral imaging
Time window (h)
No. of patients
EPITHET, Davis SM/Donnan GA et al.
Randomized controlled trial tPA vs placebo; patients treated without reference to MRI findings.
Infarct growth; reperfusion in mismatch patients
Strong trends to limitation of growth; increased reperfusion with tPA
DEFUSE, Albers G et al.
Open label study; all patients received tPA regardless of MRI results
Reperfusion on PWI; clinical outcomes
Early reperfusion in mismatch associated with good clinical outcomes
DIAS, Hacke W et al.
Dose escalation trial; mismatch used for selection
Reperfusion on PWI
Identification of a safe, effective dose that enhances reperfusion and improves outcomes
DEDA, Furlan AJ et al.
Dose escalation trial, similar design to DIAS
Reperfusion on PWI
Similar results to DIAS
DIAS-2, Hacke W, et al
Phase 3 randomized, double blinded, trial; patients selected with penumbra using MR and CT techniques
Clinical improvement score
Negative study with increased mortality in 125 ug per kg group, but due to non-neurological causes
DEFUSE, 2 Albers GW, et al
Some had initial IV tPA. IA-intervention < 12 h
Open label study; patients treated without MRI selection; pre-specified MRI parameters refined
Reperfusion; clinical gains
Favorable clinical outcomes with intervention in mismatch; similar findings to DEFUSE
TENECTEPLASE, Parsons M, et al
Tenecteplase vs tPA
PROBE design; patients selected with dual target of CTP mismatch and large vessel occlusion
Co-primary 24-h reperfusion and change in NIHSS
Superiority of tenecteplase selected with mismatch, including 90-day functional outcome
DEFUSE and EPITHET Trials
The EchoPlanar Imaging Thrombolytic Evaluation Trial trial was a randomized, controlled phase 2 trial in which 100 acute stroke patients were randomized to tPA versus placebo in the 3–6-h time window, imaging patients with multimodal MRI before treatment at 3–5 and 90 days after stroke onset. Again, the patients were not selected according to the presence of mismatch, and the primary hypothesis was that infarct growth would be attenuated by tPA in patients with mismatch, due to increased reperfusion. Like DEFUSE, a perfusion parameter of Tmax ≥2 s was utilized and 20 % designated as representing significant mismatch. The mismatch signature was seen in 86 % of patients, larger than in DEFUSE, and there was a strong trend towards attenuation of infarct volumes with tPA using a suite of growth definitions (p values ranging between 0.24–0.054, depending on measure of infarct growth) and a highly significant increase in reperfusion . Further analyses of infarct growth in EPITHET showed a positive result using baseline PWI/DWI co-registration (for example, geometric mean ratio 0.58, CI 0.33–0.99, p < 0.05) . In addition, reperfusion and infarct growth were strongly correlated with clinical outcomes, further supporting the biological validity of these biomarkers . The EPITHET trial was not powered for a clinical outcome with only 100 patients but did show a non-significant 15 % absolute risk reduction for mRS 0–1 in the mismatch group treated with tPA. We have therefore been able to power our current phase 3 EXtending Time for Thrombolysis in Emergency Neurological Deficits (EXTEND) trial for a two-sided α at 0.05 and power of 90 %, with 200 patients in each arm .
Further analysis of the EPITHET results indicated that patients rarely responded if they had an initial DWI lesion greater than 25 ml, while those above 70 ml were at high risk of hemorrhagic transformation . Using MRI, very intense hypoperfusion measured by very low cerebral blood volume has been shown to be an even better predictor of sICH than either DWI or PWI baseline volumes . Neither EPITHET nor DEFUSE were adequately powered to test the mismatch hypothesis, but the results of these trials suggested that an enriched population focusing on target mismatch patients would select responders to delayed thrombolysis. Although underpowered, these and other studies have shown generally a lack of growth in non-mismatch patients [32, 34, 47]. A pooling collaboration between the EPITHET and DEFUSE investigators provided further evidence in support of the mismatch hypothesis and refined the criteria for future trials . Patients with benign oligemia could be excluded by the use of Tmax 6 rather than the Tmax 2 used in EPITHET and DEFUSE. Patients with large clot volumes (particularly those with carotid T occlusion) rarely respond to IV tPA in this time window, generating the hypothesis that this group should ideally be triaged for clot retrieval .
The DEFUSE II trial, recently reported, explored MRI predictors of therapeutic response in an open-label study of 138 patients treated with delayed intervention using various IA-therapies . Mismatch criteria used were Tmax >6 s PWI lesion, mismatch ratio >1.8, absolute mismatch >15 ml with automated RAPID processing but were not employed in selection. Similar to DEFUSE, this trial showed that patients with the target mismatch pattern with reperfusion were clinical responders. Patients without target mismatch and particularly those with the malignant profile did not benefit, and indeed, there was a trend towards worse outcomes with reperfusion in these patients.
A number of trials have been performed with desmoteplase, a more fibrin-specific thrombolytic, with the patients selected on the basis of MR- or CT-based mismatch. In the initial desmoteplase in Acute Ischemic Stroke Trial (DIAS) and Dose Escalation of Desmoteplase for Acute Ischemic Stroke (DEDAS) trials, IV desmoteplase 90 or 125 μg/kg in the 3–9-h window was associated with increased reperfusion and potential clinical benefits, although these were small phase 2a trials [7, 8].
These trials were followed by a relatively small phase 3 trial of 186 patients treated 3–9 h after stroke onset, DIAS-2, which used placebo and desmoteplase doses of either 90 or 120 μg/kg . This trial was negative, and there was an unexpected increase in mortality in the higher dose group at 120 μg/kg, although the majority of these deaths were late and not associated with sICH. Concerns about the trial have included a milder patient population, selected with “on-console” estimations of MRI and CT mismatch by the investigator, without standardized processing, thresholding of perfusion, or automation. A low proportion of the patients had vessel occlusion. Arterial occlusion appeared a better determinant of treatment response than mismatch in this trial, and the ongoing DIAS 3 and 4 trials testing IV desmoteplase are using this parameter in patient selection (NCT00790920, NCT00856661).
Meta-analysis of Mismatch-Based Delayed Thrombolysis
A meta-analysis was based on the use of mismatch imaging in the DIAS, DEDAS, DIAS-2, DEFUSE, and EPITHET trials in 502 patients thrombolysed beyond 3 h . This showed that delayed thrombolysis in patients selected according to mismatch exhibited increased recanalization/reperfusion, which in turn was associated with improved clinical outcomes. However, delayed thrombolysis in mismatch patients was not confirmed to improve clinical outcome. Given the relatively small sample size, a useful clinical benefit remained possible, and the investigators concluded that validation of the mismatch selection paradigm required a phase 3 trial. With the exception of DIAS-2, the other trials supported the physiological basis of selection of patient with mismatch.
CTP Trials Using Mismatch
Computed tomography perfusion was used in a small subset of the DIAS-2 trial for patient selection, based on visual mismatch of CBF–CBV at the console. A recent phase 2B randomized trial was conducted to compare the standard dose of tPA with two different doses of tenecteplase within 6 h of the onset of ischemic stroke, using CTP and CTA to select patients most likely to respond to early reperfusion, defined as those with large vessel occlusion and a large perfusion lesion in the absence of a large infarct core . The pooled tenecteplase group had greater reperfusion and clinical improvement at 24 h than the alteplase group. The higher dose of tenecteplase was superior to the lower dose and to tPA for all efficacy outcomes, including a good functional outcome at 90 days .
Ongoing Trials Using Mismatch in Patient Selection for Delayed Interventions
The Mechanical Retrieval and Recanalization of Stroke Clots using Embolectomy (MR RESCUE, NCT00389467) study has enrolled 120 patients, with MRI performed before clot retrieval (MERCI or Penumbra) or standard care. Patients were not selected using mismatch criteria, but randomization was stratified on the basis of the MRI profile. Recruitment has completed and the results are awaited.
Ongoing trials using penumbral selection in acute stroke therapy
Time window (h)
No. of patients
MR RESCUE, Kidwell C, et al
IA-clot retrieval with Merci or Penumbra devices
MRI not used in selection but pre-specified hypotheses
90-day functional outcome
Recruitment complete; results pending.
EXTEND (IV), Donnan GA, Davis SM, et al
4.5–9 h, wake-up stroke
Patients selected with automated RAPID technology on basis of MRI or CTP mismatch
90-day functional outcome
ECASS 4- EXTEND, Hacke W, et al
Patients selected with automated RAPID technology on basis of MRI mismatch
90-day functional outcome
About to commence
EXTEND (IA), Campbell B, Donnan GA, Mitchell PM, Davis SM
Clot retrieval with Solitaire device after IV tPA vs IV tPA alone
tPA < 4.5 h; commence IA < 6 h
Patients selected with automated RAPID technology on basis of MRI or CTP mismatch
Co-primary 24-h reperfusion and 3-day NIHSS
We are about to commence a parallel trial EXTEND-IA (NCT01492725), a phase 2 design in 100 patients treated with standard sub-4.5-h tPA who have evidence of mismatch on CTP or MRI and an occluded vessel, then randomized to clot retrieval with the Solitaire device or current standard of care. We are using a co-primary endpoint of reperfusion at 24 h and favorable neurological response (improvement of ≥8 points NIHSS or recovery to 0 or 1) at 3 days.
Current Status of Mismatch Imaging: Is it Ready for Prime Time?
To summarize the accumulated evidence, thrombolytic therapy, proven to be effective in unselected patients up to 4.5 h with intravenous tPA, is based on the principle of vessel recanalization with tissue perfusion in the critically threatened ischemic penumbra. A large body of evidence has indicated that MR perfusion/diffusion mismatch provides a valid representation of the penumbra, particularly with standardized MRI protocols and optimized thresholding of perfusion to exclude benign oligemia. Multimodal MRI can be performed within an acceptable time frame in most acute stroke patients. There is a strong correlation between recanalization, reperfusion in mismatch patients, and clinical benefits. Furthermore, MRI parameters can effectively exclude those patients at significant risk from acute thrombolytic therapy. CTP mismatch has become increasingly validated and is a useful alternative to MR mismatch in selecting patients most likely to benefit and those most likely to be harmed by thrombolysis.
It is acknowledged that limitations of the mismatch hypothesis are, first, that the benefits of thrombolysis in a mismatch population have yet to be confirmed in an adequately powered phase 3 trial. Second, the lack of benefit in patients without mismatch also remains an unconfirmed hypothesis despite substantial evidence that patients without mismatch do not exhibit significant infarct growth or benefit from reperfusion therapies. Clearly, delayed interventions for acute stroke, particularly clot retrieval, must be considered investigational until the results of randomized phase 3 trials provide level 1 evidence. Our contention is that these approaches should optimally be tested in patients with mismatch. If used open-label, evidence of significant tissue at risk should be a key criterion.
Conflicts of Interest
The authors declare that they have no relevant conflicts of interest.