Portal de Periódicos da CAPES

The Utility of Domain-Specific End Points in Acute Stroke Trials

2021; Lippincott Williams & Wilkins; Volume: 52; Issue: 3 Linguagem: Inglês

10.1161/strokeaha.120.031939

1524-4628

Steven C. Cramer, Steven L. Wolf, Jeffrey L. Saver, Karen C. Johnston, J Mocco, Maarten G. Lansberg, Sean I. Savitz, David S. Liebeskind, Wade S. Smith, Max Wintermark, Jordan Elm, Pooja Khatri, Joseph P. Broderick, Scott Janis, Daofen Chen, Alexander W. Dromerick, Kari Dunning, Andrew W. Grande, Randolph S. Marshall, Caitlyn Meinzer, Stephen J. Page, Aimee Reiss, Lorie Richards, Lawrence R. Wechsler, Carolee J. Winstein, Cheryl Bushnell, Dorothy F. Edwards, Warren Lo, Jin‐Moo Lee, Catherine Amlie‐Lefond, Gregory W. Albers, Robert F. Dempsey, Toby Gropen, Edward C. Jauch, Enrique C. Leira, Renee Martin, Brett C. Meyer, Kiva Schindler, Phillip Scott, Aneesh B. Singhal, Claudia S. Moy,

Stroke Rehabilitation and Recovery

HomeStrokeVol. 52, No. 3The Utility of Domain-Specific End Points in Acute Stroke Trials Free AccessArticle CommentaryPDF/EPUBAboutView PDFView EPUBSections ToolsAdd to favoritesDownload citationsTrack citationsPermissions ShareShare onFacebookTwitterLinked InMendeleyRedditDiggEmail Jump toFree AccessArticle CommentaryPDF/EPUBThe Utility of Domain-Specific End Points in Acute Stroke Trials Steven C. Cramer, MD Steven L. Wolf, PhD Jeffrey L. Saver, MD Karen C. Johnston, MD J Mocco, MD Maarten G. Lansberg, MD Sean I. Savitz, MD David S. Liebeskind, MD Wade Smith, MD, PhD Max Wintermark, MD Jordan J. Elm, PhD Pooja Khatri, MD Joseph P. Broderick, MD Scott Janis, PhD Steven C. CramerSteven C. Cramer Correspondence to: Steven C. Cramer, MD, Department of Neurology, UCLA, 710 Westwood Plaza, Reed C239, LA, CA 90095-1769. Email E-mail Address: [email protected] https://orcid.org/0000-0002-6214-6211 Department of Neurology, David Geffen School of Medicine at the University of California, Los Angeles (S.C.C., J.L.S., D.S.L.). California Rehabilitation Institute, Los Angeles (S.C.C.). , Steven L. WolfSteven L. Wolf https://orcid.org/0000-0002-9446-8995 Division of Physical Therapy, Department of Rehabilitation Medicine, Emory University School of Medicine, Atlanta, GA (S.L.W.). , Jeffrey L. SaverJeffrey L. Saver https://orcid.org/0000-0001-9141-2251 Department of Neurology, David Geffen School of Medicine at the University of California, Los Angeles (S.C.C., J.L.S., D.S.L.). , Karen C. JohnstonKaren C. Johnston Department of Neurology, University of Virginia, Charlottesville (K.C.J.). , J MoccoJ Mocco Department of Neurosurgery, Mt. Sinai, New York (J.M.). , Maarten G. LansbergMaarten G. Lansberg Department of Neurology, Stanford University, CA (M.G.L.). , Sean I. SavitzSean I. Savitz Institute for Stroke and Cerebrovascular Disease, University of Texas Health Science Center, Houston (S.I.S.). , David S. LiebeskindDavid S. Liebeskind https://orcid.org/0000-0002-5109-8736 Department of Neurology, David Geffen School of Medicine at the University of California, Los Angeles (S.C.C., J.L.S., D.S.L.). , Wade SmithWade Smith Department Neurology, University of California, San Francisco (W.S.). , Max WintermarkMax Wintermark https://orcid.org/0000-0002-6726-3951 Department of Radiology, Stanford University, CA (M.W.). , Jordan J. ElmJordan J. Elm Department of Public Health Sciences, Medical University of South Carolina, Charleston (J.J.E.). , Pooja KhatriPooja Khatri https://orcid.org/0000-0002-7344-8266 Department of Neurology, University of Cincinnati (P.K.). , Joseph P. BroderickJoseph P. Broderick https://orcid.org/0000-0002-7323-7709 Department of Neurology and Rehabilitation Medicine, University of Cincinnati Gardner Neuroscience Institute, University of Cincinnati Academic Health Center, OH (J.P.B.). and Scott JanisScott Janis Division of Clinical Research, The National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD (S.J.). and on behalf of the NIH StrokeNet Recovery and Rehabilitation Group and the Acute Stroke Group* Originally published10 Feb 2021https://doi.org/10.1161/STROKEAHA.120.031939Stroke. 2021;52:1154–1161Domain-specific end points are assessments that correspond to the output of individual neural systems and are useful for capturing treatment effects on specific behaviors. By contrast, global end points combine several attributes into a single score and are useful for capturing broad treatment effects in a summary way. While global end points have become the de facto mechanism required to define benefit in stroke trials, they also have important limitations, some of which might be addressed by simultaneously measuring domain-specific end points. Substantial opportunity remains to identify quantifiable patient benefit that would otherwise not be captured by global end points. Potential advantages of incorporating domain-specific end points in acute stroke trials are discussed, such as increased granularity of measurement, improved understanding of how therapies affect the brain between acute treatment and day 90, and optimized therapeutic translation. Potential disadvantages are also considered, including time and cost of administering domain-specific end points, as well as statistical implications. Domain-specific end points and global end points are not mutually exclusive, and both capture clinical benefits to patients. Incorporating a broader set of outcome assessments in stroke trials, including both global and domain-specific end points, is warranted.Acute stroke trials generally enroll patients within 24 hours after stroke onset, with the aim being to salvage threatened tissue, for example, through reperfusion. Because large volumes of brain are often prevented from infarction, clinical benefits can be substantial when patients are assessed 90 days later. Acute stroke trials have generally relied on global end points to capture treatment efficacy.Global end points can be defined as measures that combine several attributes into a single score, often with an emphasis on functional outcomes.1,2 The global end point most commonly used in acute stroke trials is the modified Rankin Scale (mRS) assessed at day-90, which has been the primary3–8 or secondary9 outcome measure in positive reperfusion trials. The mRS is often endorsed by the FDA as the primary end point for acute stroke trials10 and is included with the definition of stroke disability co-developed by the FDA.11 The mRS is a 7-level functional outcome scale that ranges from no functional deficits (score of 0) to death (score of 6).10Global end points are useful for capturing broad treatment effects in a summary way. For example, the parsimony, simplicity, and clinical relevance of mRS scoring have enabled measurement of treatment effects across numerous studies. However, global end points also have important limitations, as discussed below, and some of these might be addressed by simultaneously measuring domain-specific end points. While global end points have become the de facto mechanism required to define benefit in stroke trials, substantial opportunity remains to identify quantifiable patient benefit that would otherwise not be captured by global end points, particularly as novel concomitant therapies or expanded indications for treatment are evaluated.Domain-specific end points, also known as modality-specific end points,12 are behavioral measures that correspond to the output of individual neural systems. The brain is composed of dozens of interconnected neural systems, each defined as a circuit of neurons that gives rise to a specific behavior.13 For most neural systems, the behavioral output can thus be measured using a domain-specific measure, either once as a day-90 cross-sectional end point, or serially to capture stroke recovery; domain-general circuits support diverse processes14 and are not further considered here. Examples of domain-specific end points appear in Table 1; more detailed lists appear elsewhere.15–18 Abnormalities in domain-specific end points, such as hemiparesis and aphasia, are often useful for diagnosing acute stroke,19 but detailed assessment is not commonly performed. For many domain-specific end points, validity and reliability have been documented, sensitivity to change has been established, and the minimal clinically important difference has been defined. Domain-specific end points have been successfully used in neurological clinical trials outside of stroke (eg, a common end point in multiple sclerosis trials is the MS Functional Composite, which measures gait via a 25-foot walk test, hand dexterity via the 9-hole peg test, and memory/attention via the paced auditory serial addition task) and in trials focused on stroke recovery therapeutics, but to date have been uncommon in acute stroke trials and in stroke prevention trials. Domain-specific end points have been the basis for FDA approval of neurological therapies, such as 4-aminopyridine, which was approved for walking ability in patients with multiple sclerosis.20,21Table 1. Examples of Domain-Specific End PointsDomain-specific end pointBehavioral domain assessedFugl-Meyer arm motor scaleUpper extremity motor deficitsGait velocityFunctional walking abilityWestern aphasia battery-revised (bedside)AphasiaLine cancellation testHemineglectFunctional Oral Intake ScaleDysphagiaPatient Health Questionnaire-9DepressionThe use of domain-specific end points has several potential advantages for acute stroke trials in terms of understanding and measuring the recovery processes that occur between acute treatment and outcome assessment 90 days later.Increased Resolution of MeasurementThe most commonly used global end points in acute stroke trials, such as the mRS or the Barthel Index, are ordinal scales with a limited number of scoring levels.10 Fewer scoring levels means lower resolution and less biological information to detect smaller but still clinically meaningful changes. In addition, lower resolution scales sometimes increase the study's required sample size.22 With continuous variables, there is the same difference across successive numerical intervals, for example, the difference in volume of cerebral infarction is precisely the same when going from 25 to 30 cc as compared with going from 55 to 60 cc. However, ordinal variables have intervals with marked differences between single steps across the range of the scale. For example, the step between mRS scores of 0 and 1 represents the clinical difference of no symptoms versus symptoms with no limitations, whereas the step between mRS scores of 5 (severely disabled) and 6 (death) is a large clinical difference. Because ordinal variables often treat interval values as homogenous, they are a comparatively weaker form of measurement as compared with continuous variables.23,24 This comparison is not universal: not all global end points are ordinal, although the one used most often in acute stroke trials10 (mRS) is, and not all domain-specific end points are continuous, although even when they are not they tend to have a large number of intervals (Table 1).Achieving finer resolution may increase the ability to detect smaller treatment-related benefits. This level of resolution might not be a priority when salvaging large volumes of brain, but could be important when evaluating a therapy for which benefits might be smaller, such as when (1) targeting specific subpopulations, as seen among the oldest old or patients treated with greater delay or patients with distal vessel occlusion; (2) trying to detect differences related to a variant treatment, such as higher number of catheter passes, different approaches to anesthesia, or tighter blood pressure control; or (3) examining the added value of a combination therapy, such as when comparing endovascular therapy plus neuroprotection versus endovascular therapy alone. Identifying small benefits may not be important if the differences detected are not clinically important.Serial measurement of a domain-specific end point that has good resolution may be useful to understand which patient features (and their change over time) are most associated with a good treatment response, or to understand how an acute stroke therapy interacts with the processes of recovery. Scales with higher granularity might also be useful to identify the relative contributions that specific neural systems make toward improvement.Greater Insight Into Acute Therapy Effects on Individual Brain SystemsOne approach to increasing measurement resolution is to assess the individual behavioral components that together constitute a global end point. Reperfusion therapies target clots and arteries to reverse ischemia in neural systems that support motor, sensory, coordination, cognitive, attention, vision, language, and other behaviors. Differences in infarct size and location determine which neural systems are injured and the severity of their involvement.25,26 Moreover, the extent and rate of recovery after stroke are not the same across these different neural systems.12 For example, a patient might recover language function but not the ability to functionally use the hand. Neglect might resolve but the patient might remain fully hemianopic.An effective acute stroke therapy improves recovery from treatment to the day-90 outcome. This enhanced recovery occurs variably in each of the neural systems involved by stroke, and so using domain-specific end points to measure recovery in key affected neural systems provides greater insight than using a single measure that collapses many neural systems' recovery into one global end point score. Because a global scale such as the National Institutes of Health Stroke Scale combines all of this neurological information into a single score, it may be less sensitive and provide weaker insights compare to domain-specific end points that capture behavioral recovery related to affected neural systems. For example, the overall National Institutes of Health Stroke Scale would be a comparatively weaker choice of end point than one or more domain-specific end points in a thrombectomy trial targeting patients with acute posterior cerebral artery occlusion because much of the National Institutes of Health Stroke Scale does not measure loss of function likely to occur when tissue is salvaged within the visual, memory, and sensory systems subserved by this artery. Using domain-specific end points to measure therapeutic effects on the individual neural systems involved in each patient (Braun R, Heitsch L, Cole J, Lindgren A, de Havenon A, Cramer S, and Worrall B, unpublished data, 2020) might also bolster personalized medicine approaches to care.Foster a Common Language Across All Stroke TrialsInclusion of domain-specific end points in acute stroke trials is rare.27 Similarly, in trials of stroke recovery therapeutics, inclusion of the mRS is uncommon, although doing so has recently been recommended.18 Adoption of a common language across different stages of stroke therapeutic investigations can enable a more cohesive system for understanding the benefit of stroke therapeutics, from the acute phase through recovery to the chronic phase,28 and extending to stroke prevention. Achieving this goal might require revision of educational goals in acute stroke, stroke recovery, and stroke prevention training programs. There are many domain-specific end points that might be used. Choosing among these can be based on expert consensus,18 available common data elements,29 matching preclinical end points,22 or relevance to underlying treatment mechanisms.30Better Understanding of Treatment MechanismDomain-specific end points can provide mechanistic insights into what is changing in the brain with treatment. Such end points can improve treatment-related insights into clinical-imaging relationships.31,32 For example, when trying to understand the impact of endovascular therapy targeting eloquent cortex, a language measure, such as naming, might be more sensitive to treatment effects than a global measure.33 Many language-focused end points are available depending on the study's needs, ranging from the language subscore of the NIH Stroke Scale (quick but low granularity) to the Philadelphia Naming Test34 (relatively brief with better granularity) to the bedside Western Aphasia Battery-Revised (longer with substantial granularity). Such insights enable the study of specific brain targets, modulation of which might provide further benefit.35 The initial infarct and the pattern of resulting behavioral deficits vary tremendously across patients. Domain-specific end points enable an improved understanding of the relationship between specific cerebral pathologies and behavioral outcomes.36 A related potential benefit is that domain-specific end points, whether measured serially from acute to chronic or just once in the chronic phase, can add information regarding the natural history of stroke and so allow for more informed planning for future trials.Better Understanding of What Goes on in the Brain From Acute Treatment to Day-90Currently, many studies treat the time period between the acute intervention and the day-90 outcome as a black box, with little or no measurement of events or influences that might affect final status. However, many factors between acute stroke treatment and day-90 outcome scoring might influence patient outcomes,37 such as occupational therapy for arm motor deficits or speech therapy for aphasia. Domain-specific end points therefore provide data that directly correspond to the output of neural systems targeted by standard-of-care interventions and so may provide useful insights into global outcomes. For example, intensive physical therapy enhances recovery of gait and balance after stroke.38 If some patients in an acute stroke trial received intensive physical therapy and have superior improvement in gait, such knowledge could be used to better interpret how the experimental treatment affected a global outcome measure such as the mRS. Measuring stroke recovery, and its influences, through more frequent and more detailed assessments over a 90-day interval, can therefore provide insights useful for understanding the effects of an acute intervention. Furthermore, it may prove evident that domain-specific end points consistently demonstrate benefit sooner than 90 days. Such knowledge could inform trial design and reduce research costs.Optimizing Therapeutic TranslationDomain-specific end points may also be useful for optimizing T1 (from preclinical to initial human) translation by better matching preclinical measures with clinical trial end points.39 Preclinical studies often evaluate the translational potential of a stroke therapy using domain-specific end points, especially those related to motor function, but acute stroke clinical trials often rely on global end points of function, such as the mRS.40,41 Inclusion of domain-specific end points could improve the clarity with which candidate acute stroke therapeutics are translated from animals to humans. For example, let us say that a preclinical study suggests efficacy based on motor testing in animals, but the human translational study fails to show a benefit using a global end point, such as the mRS. The question is whether the study was negative because (1) the treatment improved motor outcomes (as in preclinical studies) without affecting mRS or (2) the treatment simply did not improve motor outcomes.41 In addition, mechanistic insights into how brain function and behavior are enhanced by reperfusion or neuroprotection therapies can inform therapy development as well as T2 to T4 (efficacy trials, implementation investigations, and population studies) translation.Support Therapeutic Targeting of Individual Neural Systems AcutelySeveral therapeutic developments suggest potential utility for understanding acute stroke therapies in relation to the specific neural system affected. Some acute therapies target specific penumbral areas, eg, via brain stimulation.42,43 Endovascular therapy can in some instances preferentially target elegant or eloquent cortex,44 an approach that can be guided by weighting affected brain voxels according to the functional consequences of infarction.45 For these approaches to acute stroke therapy, measuring the behavioral output of target neural system(s) using domain-specific end points, as compared to global end points, would likely improve detection of efficacy. For example, if an endovascular therapy targeted voxels in motor cortex and in corticospinal tract that meet mismatch criteria, a measure of motor behavior might provide greater insights into treatment effects than a global measure. A similar approach could be adopted for aphasia and neglect. Domain-specific measures, acquired at the time of enrollment, might also serve as stratifying variables.46In addition, certain recovery-based therapies that target individual neural systems are also initiated during the acute stroke admission–the unfolding of acute injury effects and the initiation of neural repair are intertwined.47 As with acute reperfusion therapies, such recovery therapies introduced during the acute phase might also benefit from the information provided by domain-specific end points.36,48–50Capture Improved Outcomes in Additional, Patient-Centered DimensionsTreatment-related gains in outcomes beyond the mRS might also be important. For example, a treatment that substantially reduced depression or increased gait endurance might be perceived as clinically important, independent of whether mRS scores showed a treatment-related difference.51 This is particularly important for patients within the common mRS categories of 2 and 3, where there are very wide ranges of functional ability and quality of life within each score level. As an example, in the BETAS trial of growth factors administered 24 to 48 hours after onset of ischemic stroke,36 domain-specific measures provided improved resolution of treatment-related change over time, eg, one patient had modest improvement in National Institutes of Health Stroke Scale score, going from a baseline score of 10 to a day-90 score of 5, but tremendous gains in arm motor status measured as a gain in Fugl-Meyer score of 40 points on this 66-point scale. In many cases, these likely represent meaningful differences for patients, caregivers, society, and health care payers. Further direct comparisons are needed modeling domain-specific end points in relation to global end points.Domain-specific measures can provide information deemed valuable from a number of perspectives. The World Health Organization International Classification of Function divides end points into 3 constructs: loss of body structure/function (previously referred to as impairment), activities limitations (previously disability), and participation restrictions (previously handicap). Global end points generally capture activities or participation. Domain-specific end points often capture impairment (eg, the Fugl-Meyer score) but can also measure activities limitations (eg, the Wolf Motor Function Test).Analogous Issues With Neuroimaging End Points in Acute Stroke TrialsThe clinical reasoning described above can also be applied to neuroimaging end points. The imaging counterpart of the mRS is the final infarct volume, which has been used in phase I and phase II trials as a surrogate end point for clinical outcome.52,53 The final infarct volume fails to capture many clinically relevant aspects of stroke injury that can be assessed by domain-specific neuroimaging end points. Examples include atlas-based approaches that weigh the infarct volume by the eloquence of stroke-affected brain regions,54,55 measurement of injury to specific neural systems,56,57 and measures of connectivity within and between neural networks that have been correlated with stroke recovery.58,59 Domain-specific neuroimaging end points, as compared with total infarct volume, can increase resolution of measurement, provide greater insights into injury and function of individual neural systems, and help optimize translation from preclinical models to clinical trials.57,60 Many of the same benefits thus exist for domain-specific neuroimaging end points as those listed above for clinical end points.Limitations and ChallengesWhile the above sections outline potential advantages of adding domain-specific end points to acute stroke trials, there are also potential limitations and challenges (Table 2). Some domain-specific end points have properties that can at times limit their application, such as time to administer, need for specific testing equipment, or the requirement for specific training or personnel. For some domain-specific end points, further data are needed, for example, to understand the natural history of scores (eg, across the first 90 days poststroke), as such knowledge is critical to properly powering future trials.12 Moreover, experience remains limited for many end points. Other scales would benefit from further characterization, for example, of psychometric qualities such as validity, tendency toward floor/ceiling effects, and distribution of variance. Domain-specific end points could be advanced as a primary end point, added as secondary outcomes, or in certain cases be considered as a co-primary end point. Some of these approaches might require adjustment for multiplicity, and statistical strategies might need to be developed to deal with multiple streams of information, especially if results across end points are conflicting. Reliance on a domain-specific end point might increase or decrease sample size requirements, depending on study details and the number of end points employed. Studying the behavioral output of different brain systems does add complexity compared with scoring the single-digit ordinal mRS, but measuring behavioral data is crucial to achieving a better understanding of brain function61 and its modulation by therapeutic interventions.Table 2. Potential Advantages and Disadvantages of Using Domain-Specific End Points in Acute Stroke TrialsAdvantages Increased resolution of measurement Greater insight into acute therapy effects on individual brain systems Foster a common language across all stroke trials Better understanding of treatment mechanism Better understanding of what goes on in the brain from acute treatment to day-90 Optimize therapeutic translation Support therapeutic targeting of individual neural systems acutely Capture improved outcomes in additional, patient-centered dimensionsDisadvantages Can require longer times to administer Some domain-specific end points require specific testing equipment Some domain-specific end points require specially trained personnel Incomplete knowledge exists for the natural history of some domain-specific end points Some domain-specific end points require further study of psychometric qualities such as validity Experience remains limited for some domain-specific end points Incorporating multiple domain-specific end points can increase risk of a type I errorConclusionsDomain-specific end points and global end points are not mutually exclusive, and regulatory authorities are urged to incorporate elements of both as they relate to meaningful clinical benefits to patients,62 for both acute stroke trials and stroke recovery trials. Each type of end point has unique potential to contribute to interpreting research findings. Studies can measure both types of end point in parallel, depending upon a study's goals and desired balance, for example, between granularity and time constraints. Strategies regarding choice of domain-specific end points must consider a number of factors, including allocating resources for patient testing, facilitating comparisons across trials, relevance to treatment mechanism, aligning with preclinical end points, feasibility of testing, cost, trial priorities, and burden of testing on patients and research teams. Domain-specific end points, many of which provide high-resolution measurements that are linked with treatment mechanism and preclinical end points, may be particularly valuable in phase II trials. Global end points, many of which capture broad effects of therapy on function, may be most valuable in phase III trials. However, these distinctions are not absolute, as domain-specific end points have been the basis for FDA approval,20,21 and global end points provide useful insights in phase II trials.Domain-specific end points complement global end points and warrant evaluation in acute stroke trials. In the current environment of endovascular therapy and fibrinolytics demonstrating substantial efficacy, with combination acute therapies or expanded indications likely to be tested with much greater frequency, with the development of numerous promising restorative therapies, and with the growing efficacy of preventative therapies, incorporating a broader set of outcome assessments in stroke trials, including both global and domain-specific end points, is warranted.Sources of FundingThis study is funded by Grant U01 NS086872 from National Institutes of Health.Disclosures Dr Cramer has served as a consultant for Constant Therapeutics, MicroTransponder, Neurolutions, SanBio, Fujifilm Toyama Chemical Co, Medtronic, and TRCare. Dr Saver co-created the Rankin Focused Assessment (RFA) while a University of California employee, and Dr Saver, collaborators, and the University of California Regents have made the RFA freely and permanently available as a no-fee public resource under a Creative Commons, use-freely-with-attribution license. Dr Saver also co-created a written vignette rater certification program for the RFA while a University of California employee. The written vignettes are an optional system for training and certifying raters in the use of the free RFA resource. The University of California Regents, along with Dr Saver and collaborators, hold a copyright for the written vignette rater certification system. Any revenues received under that copyright are used to support the training of Vascular Neurology Fellows at UCLA. Dr Wolf has served as a consultant for SAEBO, Inc, Motus Nova, MicroTransponder, and Fujifilm Toyama Chemical Co. Dr Lansberg has served as a consultant for Biogen, Nektar Therapeutics, and Roche/Genentech. Dr Khatri's department has received funds for her efforts from Cerenovus (Investigator-Initiated Study), Nervive (NIH SBIR co-investigator), Lumosa (consultant), and Diamedica (Scientific Advisory Board). She has also received funds from Bayer (National Trial PI) and UpToDate, Inc (royalties). Dr Broderick's department received funds from Genentech for his role on steering committee of TIMELESS Trial and from Ono Pharmaceuticals for consulting work. Dr Lansberg repor