Acute Stroke Imaging Research Roadmap III Imaging Selection and Outcomes in Acute Stroke Reperfusion Clinical Trials

Introduction

During the past 2 decades, an accumulated body of evidence from the stroke research community has led to incremental advances in the standardization of clinical trial methodologies and to the emergence of a central role for imaging in new treatment evaluations. The recent series of positive endovascular trials owe much of their success to the lessons learned from the many previous trials that failed to establish therapeutic efficacy.^1–5 These previous stroke trials have led to an understanding of the roles of vascular, core, penumbral, and collateral imaging and their relationships to treatment response and clinical outcome. The goal of this article is to report on neuroimaging biomarkers for treatment selection and for outcome.

It is beyond question that time from onset of focal cerebral ischemia to reperfusion is fundamental in determining therapeutic efficacy for reperfusion therapies.⁶ The effect of early treatment of stroke with intravenous alteplase demonstrated in the hallmark National Institute of Neurological Disorders and Stroke (NINDS) trial⁷ illustrates this principle; a robust and reliable benefit compared with placebo is related to time from onset to treatment.⁸

However, when time and brain imaging by standard noncontrast computed tomography (NCCT) imaging are insufficient to accurately test a therapeutic hypothesis, selection based on imaging of a biological target for treatment is a logical alternative (Table 1). Examples may be clinical trials in which the anticipated effect size is small (eg, comparing 2 thrombolytic medications or testing of a neuroprotective drug) or in which the treatment is relevant only for a subset of stroke types (eg, large-vessel occlusion). The Stroke Imaging Research (STIR) consortium has recommended the term treatment-related acute imaging target (TRAIT) to describe patient selection based on the biological target of a treatment. The responses of these biological targets to treatment may depend on time.⁹ The series of positive endovascular trials confirmed the value of TRAIT selection and enrichment for endovascular reperfusion strategies (Table 1). The trials demonstrated that patient recruitment limited to an imaging-defined subset of stroke led to positive trials with smaller samples completed within reasonable periods of time. Extending the Time for Thrombolysis in Emergency Neurological Deficits-Intra-Arterial (EXTEND-IA) illustrates how a greater enrichment results into a smaller sample and greater effect size, but potentially also decreased generalizability and excluded patients who may have benefited from treatment.

Imaging Selection in Recent Positive Acute Stroke Endovascular Clinical Trials

After 3 neutral endovascular trials in 2013 (Interventional Management of Stroke [IMS]-III, Mechanical Retrieval and Recanalization of Stroke Clots Using Embolectomy [MR RESCUE], and Intra-Arterial Versus Systemic Thrombolysis for Acute Ischemic Stroke [SYNTHESIS EXP]),^10–13 the years 2014 to 2015 were marked by a historic series of positive acute stroke clinical trials (Table 2). The use of advanced imaging-based selection for patient recruitment in these recent trials is one of the most important factors in the success of these trials (Table 3). The imaging modalities required for each trial were different (Table 4). There is no evidence that the different imaging modalities resulted in different times from symptom onset to treatment (Table 5).

In the Multicenter Randomized Clinical Trial of Endovascular Treatment for Acute Ischemic Stroke in the Netherlands (MR CLEAN) trial,¹ the key imaging findings included a clear benefit of endovascular therapy for NCCT Alberta Stroke Program Early CT score (ASPECTS) scores of 5 to 10, but less certainty for ASPECTS score of 0 to 4. A post hoc analysis demonstrated that a good and moderate collateral score was also associated with a large benefit of endovascular therapy. However, although perfusion computed tomography (PCT) mismatch (cerebral blood volume and mean transit time thresholds) predicted functional outcome, the relative treatment effect in patients with and without mismatch was similar. The use of an ischemic core volume >70 mL on PCT criterion did identify a group of patients with low rates of independent outcome (1/13 [8%] endovascular-treated patients achieved modified Rankin Scale [mRS] score, 0–2) but there were relatively few patients and the interaction test was not significant.¹⁴

The EXTEND-IA trial² showed a robust effect of endovascular therapy over alteplase alone in patients with PCT-defined mismatch and core volume <70 mL. In this group of patients, near complete reperfusion (>90%) in target mismatch patients was strongly tied to favorable clinical outcome (regardless of the treatment strategy), and lack of reperfusion was associated with death or dependence in 70% of patients.

In the Endovascular Treatment for Small Core and Proximal Occlusion Ischemic Stroke (ESCAPE) trial,³ an imaging strategy of NCCT ASPECTS scores of 6 to 10, as well as good and moderate collateral scores on computed tomography (CT) angiography, showed a robust effect favoring endovascular therapy. ASPECTS and collateral scores were highly correlated. Patients with higher clot burden assessed using the clot burden score demonstrated more treatment effect.

In the Solitaire With the Intention for Thrombectomy as Primary Endovascular Treatment (SWIFT PRIME) trial,⁴ a target mismatch based on perfusion imaging combined with successful recanalization was associated with a favorable outcome. Final infarct volume (FIV) strongly correlated with clinical outcome in both treatment groups.¹⁵ Baseline ischemic core volume predicted 27-hour infarct volume in patients who reperfused.¹⁶ In target mismatch patients, the combination of baseline core and 27-hour hypoperfusion volume predicted FIV.

The Randomized Trial of Revascularization With Solitaire FR Device Versus Best Medical Therapy in the Treatment of Acute Stroke Due to Anterior Circulation Large Vessel Occlusion Presenting Within 8 Hours of Symptoms Onset (REVASCAT) trial⁵ supported NCCT-based patient selection, only requiring ASPECTS score of ≥6, demonstrating a robust treatment effect. However, significant discrepancies were observed between the centralized core laboratory ASPECTS and the investigators’ ASPECTS, and some benefit with lower ASPECTS scores (0–4) cannot be excluded. A pooled analysis of all patients with ASPECTS score of 0 to 4 across all endovascular trials is needed, but may be too small to draw reliable conclusions about endovascular treatment effects. Interestingly, there were also significant discrepancies between M1 versus M2 occlusions between the core laboratory and the investigators. It is important to note that, if the inclusion criteria were expanded to fully embrace the actual recruited subjects (eg, lower ASPECTS to 3–10 range) that a similar cohort would be enrolled and still show benefit.

Assess the Penumbra System in the Treatment of Acute Stroke (THERAPY) (ClinicalTrials.gov Identifier: NCT01429350), which required hyperdense clot length measurement ≥8 mm on NCCT for trial inclusion, suggested that the benefit of bridging endovascular therapy relative to intravenous thrombolysis alone increased with hyperdense clot length, and large infarcts as measured by final NCCT ASPECTS score of 0 to 4 to be associated with poor outcome providing further support for this threshold as a useful treatment exclusion criterion.

The Trial and Cost Effectiveness Evaluation of Intra-Arterial Thrombectomy in Acute Ischemic Stroke (THRACE) study (ClinicalTrials.gov Identifier: NCT01062698) has not been published to date. This study required demonstration of an arterial occlusion but similar to MR CLEAN, did not use NCCT or other criteria to exclude patients with a large ischemic core.

Opportunities for Standardization

Although the above-listed stroke clinical trials had several elements in common (occlusion location and ischemic core size), they also had significant differences, which represents a unique opportunity for standardization. More specifically, the scoring systems used to characterize ischemic core and collateral circulation varied from trial to trial. The pooling of the imaging data from these trials offers great opportunities to refine the imaging selection of patients for acute reperfusion therapy and trials (last column in Table 4). A statistical analysis plan for the pooled analysis of all the endovascular trials has been published,¹⁷ which will focus on ASPECTS, M1 versus other arterial occlusion sites, and good/moderate versus poor collaterals. The optimal set of imaging biomarkers to select patients with acute stroke may vary depending on the revascularization therapy being considered, the population being studied, and the time window under investigation, in agreement with the concept of TRAITs defined in STIR Roadmap II.¹⁸ Imaging remains essential for phase II trials, and more than 1 imaging method is probably acceptable for patient selection purposes, as long as reasonable cross-modality concordance and within modality standardization and reliability are achieved. The Stroke Treatment Academy Industry Roundtable (STAIR)/STIR imaging workshop recommends imaging-based selection for acute stroke reperfusion clinical trials (not limited to endovascular therapies) as outlined in Table 1.

The specific imaging methods proposed for patient selection using each TRAIT are outlined in Table 1. Table 1 contains the acceptable options for patient selection in clinical trials and are not listed in any order of priority.

Exclusion of patients with large ischemic core was a feature of most of the recent positive acute stroke clinical trials. As the interaction of treatment with this imaging variable cannot be determined reliably because of the small numbers of subjects across all trials, neither safety nor efficacy of reperfusion therapies in this group is established. Future studies investigating the sensitivity and specificity of each method/modality used to define ischemic core is essential.^16,19 Furthermore, studies investigating the relationship between the ischemic core volume and collaterals²⁰ should be pursued. The definitions of ischemic core will need to be revisited in populations of patients with ultrafast reperfusion. The geographic distribution of the ischemic core may need to be considered in addition to its volume to reflect the eloquence of the infarcted region. Finally, future studies will need to determine whether treatment of patients with larger ischemic cores is associated with higher rates of symptomatic intracranial hemorrhage when treated. The research priorities for core and the other TRAITs are outlined in Table 6.

Standardization of the grading of collateral circulation on and between CT and magnetic resonance imaging (MRI) are needed. The importance of collateral circulation must also be more robustly validated in prospective acute ischemic stroke. Future studies comparing single-phase and multiphase CT angiography²¹ for this purpose are warranted, considering that a dichotomous definition of collaterals (absent/poor versus good/moderate) is probably sufficient.

Perfusion-derived entities, such as the core and penumbra, are the imaging biomarkers that will require the largest effort in terms of standardization considering the number of existing definitions and the differences between imaging modalities. Core is defined generally as the irreversible ischemic area that is injured beyond therapy benefit. Penumbra is defined generally as the at risk hypoperfused area surrounding the core that is the target for therapy to be salvaged. There are now data sets available to benchmark and compare processing of acute PCT against a concurrent diffusion-weighted imaging scan.¹⁹ Also, much of the previous study to define optimal thresholding did not involve patients with ultra early reperfusion, and repeat study should be undertaken using the imaging data collected in these patients.

These efforts to refine and standardize imaging selection must also inform the concept of futility in stroke reperfusion therapy. A futile imaging profile should identify groups of patients in whom a therapy offers little to no clinical benefit particularly if an increased risk of harm is greater than any predicted benefit. A futile profile will depend on many considerations, including time from onset window, anatomic location of existing core infarction, type of treatment, and other clinical variables, such as patient age, NIH Stroke Scale (NIHSS) score, and patient preferences.²² One commonly used definition of unfavorable outcome, mRS score of 3 to 6, ignores potentially meaningful shifts from severe to moderate disability. The dichotomous approach has been modified to classify mRS score of 4 to 6 as poor clinical outcome (eg, hemicraniectomy for space occupying cerebral edema). However, an ordinal analysis approach using the full scale of the mRS to generate numbers needed to treat to achieve an improvement of at least 1 level on the mRS (perhaps combining 5 and 6 if that transition is not deemed meaningful) is an alternative approach that avoids arbitrary dichotomies. Similarly, patient-oriented outcomes, such as the NeuroQol or Patient-Reported Outcomes Measurement Information System (PROMIS), may also be considered. Recent small studies have shown that they correlate well with the mRS but have greater capacity to discriminate smaller but still meaningful change.^23,24 To address the issue of futility, future research efforts should use pooled analysis of data from recent trials as well as large imaging-based observational studies that enroll either patients without the TRAITs or all comers with a subsequent analysis of outcome by imaging profile to derive futility thresholds for current reperfusion therapy.²⁵

Two ongoing trials, Promoting Acute Thrombolysis for Ischemic Stroke (PRACTISE) (ClinicalTrials.gov Identifier: NCT02360670) and Pragmatic Ischaemic Stroke Thrombectomy Evaluation (PISTE)-2, have been designed to better understand imaging selection strategy and the impact on treatment, rather than to test a specific treatment. PRACTISE is currently testing CT-based advanced imaging selection in intravenous thrombolysis decisions. PISTE-2 will have 2 arms, 1 with advanced imaging and 1 without advanced imaging selection, and it is hoped that these will provide information on the added value of advanced imaging.

Final Infarct Volume

FIV can potentially be a useful biomarker in phase II trials to provide an early signal of efficacy for a new treatment. The rationale is that FIV is a more direct measurement of biological effect of acute treatment compared with clinical outcome at 90 days or later which may depend heavily on infarct location and can be affected by unrelated pathology. However, it is not clear that FIV is an equivalent or more powerful measure of treatment effect than clinical measures of outcome. This is an important research question that has been addressed in earlier treatment trials of tissue-type plasminogen activator (imaging outcomes less powerful than clinical outcome measures to detect treatment effect with tissue-type plasminogen activator) but has yet to be investigated in the current endovascular trials. What is clear is that all FIV imaging approaches are known to correlate with long-term clinical outcome. However, what matters is not the degree of correlation but rather the ability to properly classify patients to predict accurately the long-term outcome.

The best approach and timing for measuring FIV requires further investigation. Measuring FIV early after stroke treatment (within 24–48 hours) has the advantage that the majority of patients remain in hospital, but the disadvantages that the lesion volume and signal intensity may still be changing or may be confounded by edema and by parenchymal hematomas. Early mortality at this time point is uncommon and becomes increasingly problematic with later imaging end points because it inevitably leads to missing data in a biased manner. Measuring FIV later (30–90 days) has the advantage of a more stable true final lesion, but the patient is less likely to be available for follow-up scan, tissue atrophy may underestimate the infarct volume, and distinguishing the index infarct from chronic ischemic damage may be impossible, or at least subjective. At all time points, lesion detection and contrast is superior for MRI than CT, making it the preferred modality for final lesion volume measurement. However, CT may be required when MRI is contraindicated or unavailable. The recommended MRI sequence to determine the FIV is diffusion-weighted imaging at 24 to 48 hours.²⁶ Performing diffusion-weighted imaging earlier than 24 hours risks underestimating lesion volume due to temporary postreperfusion reversal.²⁷ MRI with fluid attenuated inversion recovery imaging performed at 3 to 5 days or just before discharge is an alternative approach that reduces the potential risk of late infarct growth occurring in nonreperfused patients while minimizing loss to follow-up.²⁸ However, differentiating the acute lesion from chronic ischemia can be more challenging and edema is prominent at this time. The optimal timing for CT follow-up (when MRI is not available) needs further investigation (ie, 24–72 hours versus 3–5 days). Research on confounding factors, including edema, hemorrhagic transformation, contrast staining on CT, fogging, etc, are necessary to increase validity of the use of FIV as a biomarker. Adjustment to account for the anatomic location and distribution of the final infarct relevant to clinical outcome whether it affects eloquent regions would clearly be relevant to models aiming to predict functional outcome. However, for assessment of biological treatment effect, removal of this potential confound may be a benefit rather than a pitfall. The research priorities for FIV are outlined in Table 6.

Imaging Technology Issues

Imaging selection for acute stroke could benefit from several technological improvements that would ensure that the requirement for speed does not result in reduced use of advanced imaging, which could impair future pathophysiologic insights and treatment advances.

MRI use could become more widespread with recent advances in rapid stroke imaging protocols but would require an effective fast safety screening process. The risk associated with the administration of gadolinium needs to be addressed, and alternative approaches to assess perfusion such as arterial spin labeling need to be further evaluated.

NCCT could benefit from a focus on improving image acquisition quality and workflow that would improve core detection, including characterization of ASPECTS score. A focus on standardizing optimal acquisition techniques, and the biophysics of image reconstruction algorithms, would be helpful, and should consider a wide range of CT technologies available, including the emerging availability of CT-equipped mobile stroke ambulances.

PCT would benefit greatly from increased signal contrast to noise through improved software and perhaps contrast agent approaches. Faster image reconstruction, transfer, and processing are critical, not just to produce standardized maps but to rapidly generate dynamic angiography. Minimum hardware requirements such as ability to operate at low kilovoltage of 80 kV (or 70 kV when available), volumetric coverage, and safety dose-check features should be considered. Rapid technological advances could open new horizons in terms of imaging selection of patients with acute stroke for treatment.

Conclusions

Recent positive acute stroke endovascular clinical trials have demonstrated the added value of neurovascular imaging. The optimal imaging profile for endovascular treatment includes large vessel occlusion, smaller core, good collaterals, and large penumbra. However, equivalent definitions for the imaging profile parameters across modalities are needed, and a standardization effort is warranted, potentially leveraging the pooled data resulting from the recent positive endovascular trials.

Acknowledgments

We thank Gary Houser for his invaluable efforts in organizing and facilitating the STAIR conference.

Footnotes