Outcomes assessment

The validity of our results rests firmly on the quality of our data collection. In particular, we must be rigorous in how we assess clinical outcomes. Too often, studies extract modified Rankin scale (mRS) scores from retrospective chart review. This method of deriving mRS scores is inaccurate and suffers from poor interobserver agreement.5 Ideally, outcomes should be determined in a prospective fashion using a standardized formal interview.6

The mRS assessed after 3 months is widely acknowledged as the optimal primary outcome measure. However, it is heavily weighted towards physical disability after stroke and, as such, is fraught with significant limitations. Ideally, stroke outcome scales should reflect the impact of the disease and the intervention on the patient's overall quality of life rather than the impact perceived by the physician who arbitrarily decides that a certain score represents a good versus poor outcome. The patient's perception of outcome is different across individuals with similar neurological deficit and is dependent on multiple factors such as cultural background, degree of education, and familial circumstances. Therefore, development of more comprehensive outcome scores that take into account factors beyond physical disability such as mood and cognition, quality of life and the patient's perception of their own condition may be necessary to accurately determine overall outcomes. In addition, as treatment for acute stroke becomes increasingly more costly, it is imperative that economic analyses measure the cost-effectiveness of different treatment approaches.

Biomarker validation

Stroke is a heterogeneous condition affecting diverse patient populations and, as such, there are a multitude of factors determining eventual clinical outcomes including age, premorbid functional status, comorbid conditions, post-stroke care in the subacute phase (eg, access to rehabilitation), access to medical care and social factors (eg, family support). This represents an enormous challenge for acute stroke reperfusion trials because the effect of any intervention has to be of sufficient magnitude to override all of these factors, which are typically not controlled for in current trial designs. Future RCTs need to take into account the influence of these confounding factors in order to increase the likelihood that the questions investigated by the trial are adequately answered. This will be challenging as post-stroke care and social factors are difficult to standardize. It is therefore important to develop surrogate outcome models specific to the intervention studied (eg, infarct volume for reperfusion therapy) that can be used in conjunction with clinical endpoints and are less prone to influence by outside factors. Akin to the coronary experience where enzymatic leaks are accepted as evidence of myocardial infarction (in conjunction with clinical parameters), outcome measures can be constructed that combine imaging evidence of brain infarction with clinical outcome scales and/or mortality. Early phase studies, in particular, should take advantage of established biomarkers for use as surrogate endpoints.

A surrogate biomarker must be specific to the population, treatment and outcome of interest.7 For example, final infarct volume measured within the first week is an excellent biomarker for 90-day functional independence among patients with anterior circulation stroke treated with IAT.8 ,9 Final infarct volume does not perform as well in general stroke populations largely due to heterogeneity in vessel occlusion site.10 Moreover, a biomarker should be easily measured and highly reliable between observers in addition to being cost-effective (ie, by reducing sample size and shortening follow-up). Validation criteria need to be adopted by the stroke research community, which will yield a standard set of biomarkers for use in clinical trials. A pressing need is to standardize and further refine our angiographic biomarkers of revascularization. In recent device trials, numerous operational definitions have been applied to various grading scales,11 which limits comparisons across studies and likely contributes to the heterogeneous clinical response to ‘successful revascularization’. Using a consensus definition based on antegrade tissue-level reperfusion, recent work has shown that the modified Thrombolysis in Cerebral Infarction (m-TICI) scale is a better predictor of outcome than the Thrombolysis in Myocardial Infarction (TIMI) scale, and that m-TICI 2b–3 (ie, reperfusion in ≥50% of the ischemic territory) is the optimal definition of treatment success.12

Database standardization

In addition to improving outcomes assessment, data collection for baseline clinical, imaging and procedural elements must be standardized. The NIH Common Data Elements Project is an excellent starting point for this process13; currently, few recommendations have been made regarding procedure-related variables. A uniform approach to data collection would allow a meaningful comparison between published studies and would permit a seamless merging of datasets across institutions, which would increase power without increasing noise.

Establishing a standardized approach to treatment decision-making

A recent survey of IAT practice patterns has revealed a marked heterogeneity in treatment decision-making criteria,14 reflecting the lack of evidence concerning which patients will benefit from treatment. The absence of a standardized approach to IAT poses a major obstacle to both research and quality improvement (QI). Recently, multicenter retrospective studies have examined important questions such as the effect of anesthesia choice on IAT outcomes. The strength of these exploratory analyses is large patient numbers, but this advantage is counterbalanced by confounding from variable selection paradigms, treatment approaches and outcomes assessment. Even within institutions, IAT decisions depend heavily on which neurologist and neurointerventionalist are on call. A standardized algorithm would yield several advantages. The most obvious is rapid and uniform decision-making. In addition, it would facilitate QI initiatives. Because the effect of other variables is minimized, one can confidently attribute improvements in clinical outcomes to the QI measure in question. For research purposes, multicenter analyses can be adjusted for key variables prescribed within institutional protocols. Establishing a standardized algorithm for IAT delivery requires the support of all key stakeholders including emergency medicine, stroke neurology, neurointervention and anesthesiology. This multidisciplinary team can establish an algorithm based on internal review of best available evidence and expert opinion. Table 1 provides an example of the algorithm adopted at Massachusetts General Hospital.

Decision-making criteria

Because treatment decisions are binary, criterion thresholds for decision-making are critical. Numerous studies have identified important predictors of IAT outcome including age, neurological deficit, infarct size and time to treatment,15 but relatively few have identified thresholds relevant for decision-making. This effort is especially important for defining target populations for future clinical trials.

An important principle in patient selection is that patients who are either ‘too good’ or ‘too bad’ to treat should be excluded. Using this approach, are there clinically meaningful thresholds from existing data to guide treatment selection? With respect to neurologic deficit, patients presenting with NIH Strok Scale (NIHSS) scores <10 appear ‘too good’ to treat, with PROACT II demonstrating that such patients do not derive added benefit from IAT, at least by the definition of favorable outcome used in the trial.16 Imaging is helpful for excluding patients where the infarct is already too large (ie, futile recanalization).8 ,9 A pretreatment diffusion-weighted imaging (DWI) volume >70 cm³ or non-contrast CT Alberta Stroke Program Early CT Score (ASPECTS) <5 is highly specific for poor outcomes and identifies a population that does not appear to benefit from endovascular reperfusion.17–19

Age is a strong determinant of treatment response, and patients aged >80 years have higher mortality and often higher rates of dependency.20 In order to improve IAT outcomes among these patients, it is possible that more stringent imaging criteria (ie, smaller infarct size) than for younger patients may be necessary. Recent data suggest that, among octogenarians whose infarct volume is <20 cm³ following intervention, favorable outcomes are seen in over 50%.21 The selection paradigm for IAT at the University of Pittsburgh Medical Center takes into account patient age. When selection is based on volumetric analysis, the infarct volume (in cm³) that qualifies patients for intervention is a maximum of 100 minus the age in years. Alternatively, when selection is based on ASPECTS, the minimum allowed score corresponds to the first digit of the patient's age. This approach requires validation through prospective trials. However, the principle of adjusting allowed preprocedure infarct size by age for selection purposes is supported by recent literature demonstrating that the interaction between final infarct volume and age is the strongest determinant of clinical outcome.8

The optimal time window is uncertain, and some advocate the notion that, beyond the ultra-early period (3 h), time windows should be replaced altogether by a ‘tissue window’ obtained through imaging. This approach is controversial and its validity remains to be proven. Nonetheless, in patients selected solely on the basis of a non-contrast CT whose main purpose is to rule out hemorrhage, the best evidence supports IAT efficacy within the 6–7 h time window.16 ,22

These results provide a good starting point for a standardized framework, but more work needs to be done to clarify stroke populations who are unlikely to benefit from IAT. Moreover, identification of thresholds should be tied to explicit goals of treatment. In the clinical setting it is often desirable to set a threshold that maximizes sensitivity for patients who will potentially benefit from IAT. In the trial setting the threshold should maximize both sensitivity and specificity.

Imaging selection

The fundamental principle underlying imaging-based selection for acute stroke treatment is that the information obtained from the imaging study must be of sufficient value to offset the additional time necessary for performance and interpretation of the study. Because time to reperfusion is increasingly being recognized as a major determinant of outcome in patients with large vessel occlusion, imaging-based trials should incorporate time metrics and guidelines as a key element into their designs. In addition, the added value of advanced imaging for patient selection must be proved in randomized controlled trials.

The clinical utility of an imaging modality presupposes that it has an acceptable degree of technical reliability. This is questionable for perfusion imaging.23 Despite calls for improved standardization, most centers use local protocols based on vendor software and institutional experience. The generalizability of any perfusion imaging study is therefore limited to the investigating center.24 Moreover, several studies have raised concerns regarding the degree of error in perfusion quantification,25–27 which may render thresholded perfusion maps too noisy for accurate lesion detection and volume quantification.28 Future studies should identify patient- and protocol-related factors that influence perfusion values and seek ways to improve measurement reliability in patients with proximal artery occlusions. The same holds true for CT angiography source imaging.29

Until then, imaging selection studies should use validated techniques such as DWI lesion volume or the ASPECTS score on non-contrast CT. DWI is the reference standard technique for delineating the extent of early ‘hyperacute’ infarction and, in reported series, may be acquired rapidly and in up to 85% of stroke patients.30 ,31 In routine clinical practice, however, MRI may produce significant delays, as evidenced by the recently completed Mechanical Retrieval and Recanalization of Stroke Clots Using Embolectomy (MR RESCUE) trial where the mean time from imaging to groin puncture was approximately 2 h. Improving rapid access to MRI and reducing delays associated with MRI performance are important QI initiatives that can yield significant time gains.32 Moreover, future work should identify methods to improve the accuracy of CT techniques. A recently reported malignant collateral profile on CT angiography is highly specific for detecting large DWI infarcts.33 Potentially, this imaging finding may improve the sensitivity of CT for detecting large infarcts when added to non-contrast CT evaluation.

Time window

The importance of time to revascularization is clear,22 ,34 but is one factor in shaping tissue viability. The interdependence of time and collaterals in determining infarct progression provides an opportunity to use imaging to extend the time window or include patients with uncertain times of onset.35 In extended window populations it will be important to identify which time intervals (eg, procedural duration36) impact on outcome and subsequently the optimal time thresholds to guide clinical practice. It should also be examined whether a good imaging profile at later time points predicts a benign natural history regardless of treatment.

Predicting recanalization efficiency

The link between better patient selection and better outcomes assumes an effective treatment (ie, a high rate of recanalization without procedural complications). Clot lengths longer than 8 mm may identify patients who are unlikely to recanalize after intravenous tissue plasminogen activator and who may benefit from added IAT.37 Future studies should investigate other clot imaging characteristics such as clot density that may further improve prediction of who may benefit from a bridging approach.38 ,39 In addition, preprocedure collateral status in the affected hemisphere has been shown to play a major role in predicting likelihood of recanalization with IAT.40

Prehospital delay

The vast majority (99%) of Americans live within 6 h of a neurointerventional facility.42 Despite such access, treatment delivery is hampered largely by delays in seeking medical care.43 Public health campaigns should emphasize acute stroke treatment options to motivate timely patient presentation.44 In addition, first-responders must be trained in stroke recognition,45 using validated prehospital clinical scales46 to transfer patients likely to have large vessel occlusions directly to comprehensive stroke centers (CSCs) with endovascular capability.47 Pre-arrival hospital notification for early neurointerventional activation must become routine. Telestroke systems have proved effective, but interfacility transfers to CSCs require significant improvement.48 Transfer time should become a key metric in maintaining primary stroke center certification, with expectations calibrated by distance.

The economics of IAT resource utilization require focused research. Is a rotating call system between regional CSCs more economical than multiple neurointerventional teams in the same city on IAT call simultaneously? Are economic incentives aligned for the best, fastest and most cost-effective care? The answers to these and similar questions may yield a profound rearrangement of current paradigms and will be increasingly topical with heightened national attention to health economics.

Intrahospital delay

Current CSC recommendations call for door-to-IAT times of 2 h or less.49 At first glance this time frame appears too long, but such delays have been reported at well-established centers.32 ,50–52 Miley et al52 reported a mean time from non-contrast head CT to microcatheter placed in the cerebral circulation of 174±60 min across three academic centers performing endovascular acute stroke treatment in the USA. This report does not contain a detailed breakdown of times required for each step from CT to microcatheter. Costalat et al50 presented a more detailed analysis of these steps in a group of acute stroke patients selected with MRI and treated with IAT at a single center in Europe. In this report the mean time to MRI was 59 min, the mean MRI duration was 22 min, the mean time from MRI to groin puncture was 81 min, and the mean time from groin puncture to reperfusion was 54 min. Adding these times yields a mean of approximately 160 min from door to groin puncture and 215 min from door to reperfusion. Clearly, a radical rethinking in our approach to delivering intra-arterial acute stroke care is necessary in order to dramatically improve outcomes with this procedure.

Process improvement begins with a detailed assessment of workflow from patient arrival to intra-arterial recanalization. Process mapping should be anchored around critical time points in patient evaluation and treatment, specifically arrival at the emergency department, imaging evaluation, determination of IAT candidacy, suite arrival, anesthesia initiation, groin access, vessel catheterization and recanalization. Measuring these time intervals provides insights into major causes of delay and permits targeted measures for reducing these delays.

Process improvements should be developed through literature review, root cause analysis, flowcharting and team brainstorming, and must be incorporated into a standardized protocol.53 One high-impact measure that has been shown to produce dramatic time gains is the use of parallel workflow in which the interventional and anesthesia teams are activated during patient evaluation.32 Off- versus on-hour approaches may differ, and a tolerable false-positive rate for team activation must be agreed upon. Data collection should be an ongoing effort and provide critical feedback on the success of QI measures. Sustained success requires committed institutional leadership and a collaborative multidisciplinary team.

Procedural duration

For QI measures, neurointerventionalists should record specific time points including suite arrival, anesthesia initiation, groin access, time to base catheter, clot access and recanalization. A ‘stroke pack’ containing all stroke-specific equipment should be available to increase speed.

Anesthesia

Conscious sedation is an attractive option for IAT, hastening times from suite arrival to groin access by eliminating anesthesia induction and intubation times as well as permitting intraprocedural neurological examination. In addition, retrospective data suggest improved neurological outcome with conscious sedation versus general anesthesia and also a lower final infarct burden, lower incidence of pneumonia and shorter stay in the ICU.54 ,55 Conscious sedation should therefore be used if tolerated by the patient.

Alternative vascular access

Although procedure duration comprises both access site to base catheter and base catheter to recanalization, most of the advances in procedural technology have focused on the latter step while the former has largely been neglected. Even under optimal conditions, the time required to access the clot from a transfemoral approach can be considerable. A recent study suggests that, in 50% of patients, groin puncture to carotid catheterization takes >20 min while, in 25%, this procedural step is >30 min.56 This can be much longer in the presence of atherosclerotic disease or vessel tortuosity. In difficult cases, alternative access routes should be employed. Transradial access has been described57 but is rarely used. Direct carotid puncture provides rapid access and short working distances, and may be performed using ultrasound-guided Seldinger access techniques to allow placement of a 5 Fr or 6 Fr sheath into the carotid artery.58 ,59 Puncture should occur in the lower cervical region to ensure sheath insertion proximal to the bifurcation, and short sheaths (eg, 5 cm) should be used to avoid injury to the internal carotid artery.58 Reduced distance to the target allows more efficient suction and traction to be applied to the thrombus, although this would require modifications to existing equipment. Securing the arteriotomy is a particular concern. A major potential complication is hematoma at the access site, with risk of airway compromise, especially in patients who have received a thrombolytic drug.

Distal revascularization

Proximal vessel recanalization may be complicated by distal emboli, which are poorly responsive to current treatments. Novel thrombolytic agents with improved revascularization and safety profiles may better target these distal occlusions.60 In addition, smaller devices such as the Mindframe (Covidien), Bonnet (Phenox GmbH, Bochum, Germany) and Catch devices (Balt, Montmorency, France) are available. None of these are yet available in the USA. Aspiration through a pump (Penumbra) or manually through a syringe61 is another approach that holds promise in this circumstance as aspiration catheters are available to accommodate these distal vessels. However, caution must be exercised when considering revascularization of smaller branches. Due to their increased fragility, safety concerns become significant and thus risks may outweigh benefits.

Device characteristics and evaluation

With stent retrievers, TIMI 2–3 revascularization rates are approaching a ‘ceiling’. To further stratify device effectiveness it will be necessary to change our definition of treatment success to TICI 2b–3. Furthermore, time metrics should be incorporated into our revascularization endpoints (eg, TICI 2b–3 achieved within 30 min of base catheter). Despite the success of stent retrievers, improvements remain to be made in support systems, catheter trackability and device delivery. Simplicity of device preparation and deployment will further aid in time gains. Device designs will evolve with a better understanding of clot/device and vessel/device interactions. Advancing medical technology requires close interaction between clinicians and industry, and it will be critical to manage conflicts of interest proactively and pursue innovation under a unified ethical code.62