Effect of a Task-Oriented Rehabilitation Program on Upper Extremity Recovery Following Motor Stroke: The ICARE Randomized Clinical Trial

Importance  Clinical trials suggest that higher doses of task-oriented training are superior to current clinical practice for patients with stroke with upper extremity motor deficits.

Objective  To compare the efficacy of a structured, task-oriented motor training program vs usual and customary occupational therapy (UCC) during stroke rehabilitation.

Design, Setting, and Participants  Phase 3, pragmatic, single-blind randomized trial among 361 participants with moderate motor impairment recruited from 7 US hospitals over 44 months, treated in the outpatient setting from June 2009 to March 2014.

Interventions  Structured, task-oriented upper extremity training (Accelerated Skill Acquisition Program [ASAP]; n = 119); dose-equivalent occupational therapy (DEUCC; n = 120); or monitoring-only occupational therapy (UCC; n = 122). The DEUCC group was prescribed 30 one-hour sessions over 10 weeks; the UCC group was only monitored, without specification of dose.

Main Outcomes and Measures  The primary outcome was 12-month change in log-transformed Wolf Motor Function Test time score (WMFT, consisting of a mean of 15 timed arm movements and hand dexterity tasks). Secondary outcomes were change in WMFT time score (minimal clinically important difference [MCID] = 19 seconds) and proportion of patients improving ≥25 points on the Stroke Impact Scale (SIS) hand function score (MCID = 17.8 points).

Results  Among the 361 randomized patients (mean age, 60.7 years; 56% men; 42% African American; mean time since stroke onset, 46 days), 304 (84%) completed the 12-month primary outcome assessment; in intention-to-treat analysis, mean group change scores (log WMFT, baseline to 12 months) were, for the ASAP group, 2.2 to 1.4 (difference, 0.82); DEUCC group, 2.0 to 1.2 (difference, 0.84); and UCC group, 2.1 to 1.4 (difference, 0.75), with no significant between-group differences (ASAP vs DEUCC: 0.14; 95% CI, −0.05 to 0.33; P = .16; ASAP vs UCC: −0.01; 95% CI, −0.22 to 0.21; P = .94; and DEUCC vs UCC: −0.14; 95% CI, −0.32 to 0.05; P = .15). Secondary outcomes for the ASAP group were WMFT change score, −8.8 seconds, and improved SIS, 73%; DEUCC group, WMFT, −8.1 seconds, and SIS, 72%; and UCC group, WMFT, −7.2 seconds, and SIS, 69%, with no significant pairwise between-group differences (ASAP vs DEUCC: WMFT, 1.8 seconds; 95% CI, −0.8 to 4.5 seconds; P = .18; improved SIS, 1%; 95% CI, −12% to 13%; P = .54; ASAP vs UCC: WMFT, −0.6 seconds, 95% CI, −3.8 to 2.6 seconds; P = .72; improved SIS, 4%; 95% CI, −9% to 16%; P = .48; and DEUCC vs UCC: WMFT, −2.1 seconds; 95% CI, −4.5 to 0.3 seconds; P = .08; improved SIS, 3%; 95% CI, −9% to 15%; P = .22). A total of 168 serious adverse events occurred in 109 participants, resulting in 8 patients withdrawing from the study.

Conclusions and Relevance  Among patients with motor stroke and primarily moderate upper extremity impairment, use of a structured, task-oriented rehabilitation program did not significantly improve motor function or recovery beyond either an equivalent or a lower dose of UCC upper extremity rehabilitation. These findings do not support superiority of this program among patients with motor stroke and primarily moderate upper extremity impairment.

Trial Registration  clinicaltrials.gov Identifier: NCT00871715

Clinicians providing care for patients with stroke lack evidence for determining the best type and amount of motor therapy during outpatient rehabilitation. Notwithstanding the considerable resources devoted to stroke rehabilitation care, a recent Cochrane review of interventions for improving upper limb function after stroke concluded that high-quality evidence for the superiority of any current routinely practiced intervention is absent, including the amount and content of motor training.¹

Two large rehabilitation trials performed in the long-term phase of stroke, after initial rehabilitation had been completed, suggested that intensive, high-repetition, task-oriented training was superior to usual care for improving upper extremity motor outcomes.^2,3 With rehabilitation training implemented after spontaneous recovery, improvements can be attributed more directly to the training. Even though the rehabilitation interventions differed in these studies (constraint-induced movement therapy³ and robot-assisted training²), both incorporated the same principles of high movement repetitions and structured task-oriented practice. Despite this evidence and expert opinion that more practice enhances recovery, these findings have not been incorporated into clinical practice when patients with stroke actually receive rehabilitation therapy. Typical outpatient treatment sessions last 36 minutes, during which patients engage in an average of only 12 purposeful actions within an otherwise unstructured treatment session.⁴

The goal of the Interdisciplinary Comprehensive Arm Rehabilitation Evaluation (ICARE) was to test the efficacy of the same task-oriented approaches during the outpatient phase of rehabilitation, which typically begins within a month after stroke occurrence. ICARE compared 2 different treatment strategies (intensive, high-repetition, task-oriented training vs unstructured occupational therapy at 2 different doses: 30 hours vs a lower-dose, observation-only control) among patients undergoing outpatient rehabilitation for moderate arm motor impairment after stroke.

A motor rehabilitation trial precludes double-blinding; participants were aware of their group assignment. The protocol and design for this phase 3, parallel, 3-group, single-blind, randomized clinical trial are provided in Supplement 1 and were detailed in the ICARE methods article.⁵ Following institutional review board approvals at each site, participants provided written informed consent. A data and safety monitoring board and medical monitor provided independent oversight. Study biostatisticians with full access to deidentified data managed the statistical analyses with consultation from a blinded statistician. The statistical analysis plan (Supplement 1) was developed and sample size estimations were performed during proposal development and approved by the data and safety monitoring board and the sponsor.

Patients with moderate upper extremity motor impairment who demonstrated at least minimal initiation of hand and finger extension and who could participate in an intense but distributed rehabilitation program were recruited. To preclude severe cognitive and sensory impairments, the National Institutes of Health Stroke Scale subscale scores for neglect and sensory impairment were used, thus allowing focus on upper extremity motor recovery. Per sponsor mandate, race and ethnicity were self-selected from categories determined in 1997 by the US Office of Management and Budget.

Participants were recruited from 7 sites, predominantly during inpatient rehabilitation. Early in recruitment, the randomization window was amended from 30 to 90 days to 14 to 106 days after stroke, enabling randomization 16 days earlier and necessitating a 16-day extension to retain the stratification midpoint. Full eligibility criteria are provided in eTable 1 in Supplement 2. An abbreviated list of inclusion criteria includes age older than 21 years; ischemic or hemorrhagic stroke meeting World Health Organization criteria⁶; upper extremity hemiparesis; voluntary finger extension; no more than 6 outpatient occupational therapy sessions; and absence of traumatic or nonvascular brain injury and subarachnoid or primary intraventricular hemorrhage. Participants were randomized into 1 of 3 treatment groups. Intervention was completed within 16 weeks of randomization. A stratified block randomization scheme within sites balanced assignment by motor severity and time from stroke onset. The biostatistician performed randomization using hybrid block sizes of 3 or 6, depending on anticipated sample size within each site. Once a participant provided informed consent and the baseline assessment was completed, the study site requested randomization; the data manager confirmed eligibility and the site team leader was notified of the assignment.

An overview of the interventions in each treatment group is shown in eTable 2 in Supplement 2. The investigational intervention, a structured, task-oriented upper extremity motor training program, termed the Accelerated Skill Acquisition Program (ASAP), is a best-practice synthesis implementing neuroscientific evidence regarding motor training approaches and schedules. This program is principle based, impairment focused,⁷ task specific,³ intense,⁴ engaging, collaborative, self-directed,^8,9 and patient centered¹⁰; it has been previously described and feasibility tested.^5,8,9,11-14 The ASAP intervention included an initial evaluation and 30 one-hour treatment sessions (3 times per week for 10 weeks). A constraint or mitt worn on the less affected hand was available but not mandated. Purposeful and skilled movement execution was emphasized. Support for patients’ control or autonomy was provided by choices of specific tasks to be practiced, collaborative problem solving to identify and address movement needs, and encouragement of self-direction in extending practice to community contexts.^5,8

The monitoring-only usual and customary care (UCC) and dose-equivalent usual and customary care (DEUCC) therapy groups received outpatient occupational therapy based on usual and customary practice as determined by the therapist(s), local practices, payer guidelines, and participant preferences. These groups differed only by the number of sessions of outpatient therapy. For the DEUCC group, study-related treatment prescribed 30 hours of outpatient therapy. Treatment adherence for the dose-equivalent groups was set at 27 hours or more. The UCC group was only monitored, without specification of dose.

To prevent contamination of treatment across groups, at each site, the DEUCC and UCC therapies were delivered by different clinicians situated in separate locations from the investigational intervention. The ASAP Manual of Procedures was embargoed during the trial⁵; it included a formal standardization process. All ASAP therapists signed nondisclosure agreements.

Participants were assessed at baseline (prerandomization), postintervention (end of therapy), 6 months, and 12 months (end of study) after randomization by clinicians masked to treatment assignment and standardized in administration of primary and secondary outcome assessments.⁵ A 12-month change in the log-transformed Wolf Motor Function Test (WMFT) time score was the primary outcome. The WMFT time score is a mean of 15 hierarchically arranged timed arm movements and hand dexterity tasks.^15-18 If a task could not be completed in 2 minutes (ie, 120 seconds), a time score of 121 seconds was assigned. A log transformation is used to accommodate for the nonnormality in the raw time score.

Secondary outcomes included the 12-month change in WMFT time score (minimum clinically important difference, 19 seconds faster, 16% of the range on the dominant side and unknown on the nondominant side¹⁹) and improvement in participant-reported Stroke Impact Scale (SIS), version 3.0, hand subscale score (percentage with improvement ≥25 points). Because the SIS hand function range is 0 to 100 points, participants with SIS hand function scores of 75 or higher at baseline (n = 15) were excluded from this analysis. The minimal detectable change and minimal clinically important difference for the SIS hand function subscale are 25.9 points and 17.8 points, respectively.²⁰ The primary end point was the end of study (12 months after randomization; mean, 13.5 months after stroke), while end of therapy (16 weeks after randomization; mean, approximately 22 weeks after stroke) was a secondary end point.

Patients (or their designated representatives or caregivers) were contacted by monthly telephone calls to review health status, health care utilization, medications, and adverse events. Adverse events were reported regarding 3 independent characteristics: expectedness, relation to study, and severity.⁵

Baseline characteristics and outcomes, attrition, adherence, and adverse events were compared across groups. Data distribution assumptions were assessed and data transformations made as necessary. All analyses were performed using the intention-to-treat principle, comparing outcomes by assigned group. Data were analyzed for patterns of missing data; because no pattern was found, Markov chain Monte Carlo imputation was used to account for missing data from loss to follow-up or missed evaluations at end of study. Complete case analyses were also performed, although there were no differences in results, so we report the planned intention-to-treat results with multiple imputation herein. Changes from baseline in continuous outcomes were analyzed using analysis-of-covariance models adjusting for baseline values and stratification factors (site, severity, time since onset). For the primary outcome, a 2-sided P < .05 was used to indicate a difference between the ASAP intervention and DEUCC therapy. To adjust for multiple comparisons in secondary aims (DEUCC vs UCC and ASAP vs UCC), a 2-sided P < .025 was used. Per the protocol, the proportion of participants who improved 25 points or more on the SIS hand subscale, that which exceeds a clinically important change²⁰ and would represent at least a full category of difficulty improvement (eg, from “very difficult” to “somewhat difficult”), was also examined. This outcome was evaluated using logistic regression models.

Analyses were performed using SAS software, version 9.3 (SAS Institute Inc). In addition to data checks before analyses, residuals for the total group and by group were examined to ensure that model assumptions were met. Besides means and confidence intervals, effect size (Cohen d) was computed to determine the magnitude of effects. This effect size was calculated as the between-group difference in outcome variable mean divided by the common standard deviation of the outcome variable²¹; confidence intervals for d were estimated as described by computing the 95% confidence intervals for the noncentrality distribution from the t test of group differences.²²

Sample size calculations were performed during ICARE proposal development to ensure sufficient power to detect a difference in log-transformed WMFT at 12 months between the ASAP and DEUCC groups at a significance level of .05 and 80% power. Given the population being recruited and the relatively short duration of time from stroke occurrence to enrollment in ICARE, the primary outcome group differences were expected to be between those found in EXCITE³ and VECTORS.²³ EXCITE resulted in d = 0.50 for log-transformed WMFT between the high-functioning and the delayed intervention (control) groups and d = 0.63 between the low-functioning and the delayed intervention groups. For EXCITE, there was a 26% change in the control group (from baseline to 1 year) for the primary outcome. VECTORS resulted in a smaller effect between the high-dose and control groups (d = 0.20) for the Action Research Arm Test,²⁴ although the variance of the effect was higher for VECTORS (inpatient acute) than EXCITE (outpatient subacute). For VECTORS, there was a 77% change in the control group (from baseline to 90 days) for the Action Research Arm Test.

For ICARE, the DEUCC group was estimated to demonstrate a proportional recovery of approximately 50% on the primary outcome (midway between the 26% in EXCITE and the 77% in VECTORS); thus, the effect size for power was set at d = 0.42. With this effect, using a 2-sided t test with a type I error of .05 and 80% power, at least 90 patients per group would need to be evaluable. With an expected attrition rate of 25%, 120 patients per group was the planned sample size. The secondary outcome measure was the success rate in the SIS hand subscale score, defined as the proportion of patients with a change of 25 points or higher out of 100 on the normalized SIS hand at end of study. Given the estimates for SIS hand function subscale change from acute (VECTORS, 46% [n = 13]), postacute (Kansas City home-based exercise study,²⁵ 35% [n = 71]), and subacute (EXCITE, 24% [n = 86]) control data, if the DEUCC group success rate was greater than in EXCITE, less than in VECTORS, and comparable with that of the Kansas City study, then ICARE was powered sufficiently to detect a minimum of 21% difference between groups.

From June 2009 through February 2013, 11 051 patients were screened. Given the need for participants to remain in the trial, the most frequent reason for exclusion at chart screening was the clinician’s best judgment about the likelihood of the patient completing the study (44%), which included consideration of social instability, inability to travel for study-related procedures, and concomitant illness. The second most frequent reason for exclusion was no diagnosis of stroke (19.5%). During clinical screening, other exclusions included possible dementia (11.4%), upper extremity Fugl-Meyer (UEFM) motor score too high (11.3%), and clinician’s best judgment (10.9%). See Figure 1 for participant flow through the study.

Table 1 summarizes baseline characteristics. Mean age of the participants was 60.7 (SD, 12.5) years; 56% were men and 42% were African American. The 2 largest sites (Atlanta, Georgia, and Washington, DC) were urban Stroke Belt hospitals.²⁶ At randomization, the mean number of days since stroke was 45.8 (SD, 22.4). Eighty-three percent experienced an ischemic stroke without hemorrhagic conversion; 43% affected the left hemisphere; 48% affected the right hemisphere; and 7% affected the brain stem. The stroke affected the dominant upper extremity in 49% of the sample. Sixty-eight percent of participants underwent inpatient rehabilitation and 18% were referred directly from acute care. Fifty-seven percent were in the less severe (UEFM score ≥36), early-onset (<60 days) stratum, with 19% in the next most common stratum of more severe (UEFM score <36), early onset. A median National Institutes of Health Stroke Scale score of 4 (interquartile range, 2-5) was consistent with a study population with primarily motor hemiparesis without substantial cognitive or sensory impairment. The overall mean UEFM motor impairment score of 41.6 (95% CI, 40.7-42.6) for the entire sample was in the moderately severe range of 21 to 50; 78% were within this range at baseline and 21% were classified with mild impairment (>50). The baseline mean WMFT time score was 14.9 (95% CI, 12.9-16.8) seconds. Twenty-nine percent of participants had received some outpatient occupational therapy prior to randomization, though none received more than 6 hours.

A total of 304 patients (84%) completed the 12-month evaluation (87% in the ASAP, 88% in the DEUCC, and 79% in the UCC groups; P = .09 for difference in primary end-point attrition).

Treatment adherence for the dose-equivalent groups, defined as completion of at least 27 treatment hours, was comparable across the groups at 79% for the ASAP intervention and 74% for DEUCC participants. The mean number of treatment visits per week in the ASAP group was 2.4 (95% CI, 2.3-2.5) vs 2.1 (95% CI, 2.0-2.2) in the DEUCC group. The mean number of hours of treatment received was 28.3 (95% CI, 27.1-29.5) and 26.7 (95% CI, 25.3-28.1) for the ASAP and DEUCC groups, respectively. In the UCC group, 81% (99/122) reported receiving some upper extremity therapy; the mean treatment time per week was 1.7 hours (95% CI, 1.5-1.9 hours). The UCC group completed a mean of 11.2 (95% CI, 9.5-12.9) total hours of treatment that varied widely (range, 0-46 hours; see the eFigure in Supplement 2 for details). Monitoring of health and adverse events via monthly telephone calls was comparable across groups. There was an overall 54% improvement in mean motor performance as measured by the WMFT time score, and 71% (215/302) of participants met or exceeded a meaningful change in the SIS hand subscale score (≥25 points at the primary end point). In 7 (0.8%) of the 898 follow-up evaluation sessions, the assessor was unblinded to group assignment.

At the end of the study, the mean times for lnWMFT were as follows: for ASAP, 1.4; DEUCC, 1.2; and UCC, 1.3 (Table 2 and Figure 2). The differences in lnWMFT change between groups at the end of the study (12 months) were small (eTable 3 in Supplement 2): for ASAP vs DEUCC, the mean difference was 0.14 (95% CI, −0.05 to 0.33; P = .16); for ASAP vs UCC,−0.01 (95% CI, −0.22 to 0.21; P = .94); and for DEUCC vs UCC, −0.14 (95% CI, −0.32 to 0.05; P = .15); in raw seconds, these differences ranged from 0.5 to 2 seconds. Associated effect sizes of group pair differences were small. The Cohen d was 0.15 (95% CI, −0.30 to 0.21) for the primary pairwise comparison of ASAP vs DEUCC; for the secondary pairwise comparisons, d = −0.01 (95% CI, −0.26 to 0.24) for DEUCC vs UCC and d = −0.10 (95% CI, −0.15 to 0.36) for ASAP vs UCC (see eTable 3 in Supplement 2 for end-of-therapy and end-of study pairwise group comparison effect sizes).

For all participants, the baseline mean WMFT time score was 14.9 (95% CI, 12.9-16.8) seconds. At the end of the study, the mean time score had improved to 6.8 (95% CI, 5.3-8.3) seconds but was still slower than the age-matched normative mean time of 1.3 seconds,¹⁸ indicating persistent motor impairment. All groups showed comparable improvements in WMFT times of 7 to 8 seconds, adjusted for baseline and covariates; by group, WMFT mean times improved; ie, for ASAP, 8.8 seconds, for DEUCC, 8.1 seconds, and for UCC, 7.2 seconds. Between-group changes were not significantly different: for ASAP vs DEUCC, the mean difference was 1.81 (95% CI, −0.83 to 4.45) seconds (P = .18); for ASAP vs UCC, −0.59 (95% CI, −3.77 to 2.60) seconds (P = .72); and for DEUCC vs UCC, −2.12 (95% CI, −4.52 to 0.27) seconds (P = .08).

Hand function improvement of more than 25 points on the SIS subscale at the end of the study was 73% (72/99) in the ASAP group, 72% (75/104) in the DEUCC group, and 69% (68/99) in the UCC group; none of the group pairwise comparisons were significant (P>.21) (Table 2). Differences in mean changes were 0.6 (95% CI, −6.4 to 6.5; P = .99) for ASAP vs DEUCC, 2.6 (−3.8 to 9.0; P = .42) for ASAP vs UCC, and 2.1 (95% CI, −4.0 to 8.3; P = .49) for DEUCC vs UCC. Thirty-six percent of participants had less than complete recovery (<100 points), despite improvement beyond a clinically meaningful amount (mean, 36 points; 95% CI, 34-40 points) by the end of the study.

There were no significant differences across groups in number of serious adverse events whether on or off site, expected, or study related, when either treating each event as separate or accounting for repeated observations within individuals (all P>.40). There were 168 serious adverse events involving 109 participants; the most common were hospitalization (n = 143; 85% of adverse events; 25% of randomized participants) and recurrent stroke (n = 42; 25% of adverse events; 9% of randomized participants). Two of the serious events were deemed related to the intervention, one from hypertension and the other from a wrist fracture.

Among participants with primarily moderate upper extremity motor impairment after stroke, there were no group differences in upper extremity motor performance at 12 months after randomization with a structured, task-oriented motor training program compared with UCC occupational therapy during outpatient rehabilitation. Specifically, the structured, task-oriented motor therapy was not superior to usual outpatient occupational therapy for the same number of hours, showing no benefit for an evidence-based, intensive, restorative therapy program. In addition, there was no advantage to providing more than twice the mean dose (mean, 27 hours) of therapy compared with the average 11 hours received by the observation-only group, showing that substantially more therapy time was not associated with additional motor restoration. With payer pressures on reducing inpatient rehabilitation, outpatient rehabilitation may be of greater importance for patients with stroke. The findings from this study provide important new guidance to clinicians who must choose the best treatment for patients with stroke. The results suggest that usual and customary community-based therapy, provided during the typical outpatient rehabilitation time window by licensed therapists, improves upper extremity motor function and that more than doubling the dose of therapy does not lead to meaningful differences in motor outcomes.

However, these results cannot be assumed to generalize to other outcomes (eg, health-related quality of life, community participation), other stroke-related impairments such as walking,²⁷ other stroke populations, or rehabilitation time points earlier or later following stroke than those studied here. Moreover, the lack of difference between dose groups may be a measurement artifact. Although the mean dose of therapy for the DEUCC group was more than twice that of the UCC group, the latter group had considerable variation in actual dose (ie, a range of 0-46 hours). Thus, mean differences may misrepresent the actual differences between groups. This variability in the observation-only group was not unexpected and was suggested by pretrial survey data (Supplement 1). Post hoc exploratory analysis found no systematic relationship between dose of usual therapy and magnitude of change in WMFT time scores, similar to results from secondary outcomes of the EXCITE trial,²⁸ even though variability in dose across patients was much lower in EXCITE than in ICARE.

The ICARE results support the necessity of well-designed dose-response studies of motor training at key clinical time points after stroke, including the periods of inpatient and outpatient care and the long-term period beyond a year after stroke. Different physiological and psychological responses to rehabilitation training might occur at different times after stroke onset,²⁹ raising the possibility that dosing and timing are not independent factors in stroke rehabilitation intervention trials. This observation could explain why trials conducted early after stroke (eg, AVERT³⁰ and VECTORS²³) found undesirable effects of more intensive interventions, whereas the EXCITE trial,³ conducted later after stroke, found a positive effect. ICARE, timed in between, found no dose effect. This time window hypothesis provides a plausible explanation for why the results from ICARE are consistent with previous studies of therapy dosing that did not find large dose effects^31,32 but also are inconsistent with other findings in which a high dose of task-oriented training, as in EXCITE, was shown to be more effective compared with the control group when the dose of outpatient therapy varied considerably. Future trials in rehabilitation should use designs and methods that consider the natural recovery transpiring during the early period after stroke³³ as well as the degree of neurological severity and disability.

The data pertaining to dose of rehabilitation therapy may be important to policy makers and may be useful to estimate the cost and expected effect of aftercare in the outpatient setting. Future neurorehabilitation trials should include physiological and psychosocial domains and use designs that directly test hypothesized mechanisms of action.³⁴ This consideration may include choosing primary outcomes that are sensitive to participation and quality of life and are consistent with patient-centered aspects of health care reform.³⁵

ICARE has several limitations related to the disadvantages that are typical of pragmatic trials of patient-centered treatment decisions. Several consequences resulted from choosing an experimental treatment in the context of current practice.

First, spontaneous recovery may have been greater than any treatment effect. Within 6 to 10 weeks after stroke, time from stroke occurrence is independently associated with spontaneous recovery of impairments and activities, explaining 16% to 42% of the observed improvements.³³ Previous upper extremity stroke rehabilitation trials (ie, EXCITE and VA Robotics)^2,3 were performed later after stroke occurrence, after the early recovery had plateaued and when rehabilitation is typically no longer prescribed.

Second, differences in the dose of rehabilitation therapy in the ICARE trial may have been insufficient despite being considerably more than what was reported previously⁴ and consistent with reports of modified schedules of constraint-induced movement therapy that were shown to be effective.^36-38 A different treatment schedule (for example, the same number of hours delivered over a shorter period) might have been more effective. Time on task was not measured in this study, and information about the content of usual therapy was limited and dependent on site-specific clinical documentation and procedural billing codes (eg, Current Procedural Terminology codes); these choices were necessary for study completion.

Third, the use of standard care as a control group in this pragmatic trial imposes limitations, particularly when testing a nonpharmacological intervention. Usual outpatient upper extremity therapy may have evolved over time to resemble the investigational intervention, by following evidence-based guidelines. Continuing education, increased attention to practice guidelines,^39,40 awareness of the ongoing ICARE trial,⁴¹ and site (partner or affiliate of nearby academic medical center) may have influenced usual care practices. This limitation is inherent to pragmatic trial designs; for ICARE, both usual therapy groups represented a relatively high standard of outpatient practice.

Among patients with motor stroke and primarily moderate upper extremity impairment, the use of a structured, task-oriented rehabilitation program, compared with an equivalent dose of customary occupational therapy or with usual and customary occupational therapy, did not significantly improve motor function or recovery after 12 months. These findings do not support superiority of this task-oriented rehabilitation program for patients with motor stroke and moderate upper extremity impairment.

Corresponding Author: Carolee J. Winstein, PhD, University of Southern California, 1540 Alcazar St, CHP 155, Los Angeles, CA 90033 (winstein@usc.edu).

Author Contributions: Dr Winstein had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Winstein, Wolf, Dromerick, Nelsen, Lewthwaite.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Winstein, Wolf, Dromerick, Lane, Nelsen, Lewthwaite, Cen.

Critical revision of the manuscript for important intellectual content: Winstein, Wolf, Dromerick, Lane, Nelsen, Lewthwaite, Azen.

Statistical analysis: Lane, Cen, Azen.

Obtained funding: Winstein, Wolf, Dromerick, Azen.

Administrative, technical, or material support: Winstein, Wolf, Dromerick, Lane, Nelsen, Lewthwaite, Cen.

Study supervision: Winstein, Wolf, Dromerick, Nelsen.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.

Funding/Support: The ICARE trial was funded jointly by the National Institutes of Health, the National Institute of Neurological Disorders and Stroke (primary), and the National Center for Medical Rehabilitation Research of the Eunice Kennedy Shriver National Institute of Child Health and Human Development (grant U01NS056256). Each author received support from this grant during the conduct of the study.

Role of the Funder/Sponsor: The funders provided input and oversight related to the design and conduct of the study, as well as collection, management, analysis, and interpretation of the data. Preparation, review, and approval of the manuscript and decision to submit the manuscript for publication were solely the responsibility of the authors.

Disclaimer: This article does not necessarily represent the official views of the National Institutes of Health.

Members of the ICARE Investigative Team: University of Southern California, Administrative Coordinating Center, Los Angeles, California: Carolee J. Winstein, principal investigator; Monica A. Nelsen, program director, SIS standardization assessor; Jennifer Bandich, financial manager; Shannon Massimo, project assistant; Michelle Haines, project assistant; Patricia Hatchett, project manager, ASAP therapist (pilot and proposal phases); Veronica T. Rowe, WMFT standardization assessor, WMFT FAS rater; Claire McLean, blinded evaluator; Arlene Yang, blinded evaluator, WMFT FAS rater; Rachel Tabak, blinded evaluator; Kristin McNealus, Blinded Evaluator; Veronica Strickland, Blinded Evaluator; Julie Y. Kasayama, WMFT FAS Rater; Rebecca Lewthwaite, Co-Investigator, ASAP standardization assessor; Burl Wagenheim, coinvestigator, study psychologist; Brent Liu, neuroanatomical and imaging database director; Ximing Wang, neuroanatomical and imaging data manager; Kevin Ma, neuroanatomical and imaging data manager; David Russak, WMFT quality assurance; Nhi Dang, WMFT quality assurance; Katie Wongthipkongka, WMFT quality assurance; Yanshu Hou, WMFT quality assurance; Sue Duff, T32 fellow; Richard Nelson, webmaster (public site). Emory University, Atlanta, Georgia: Steven L. Wolf, co–principal investigator; Sarah Blanton, site team leader, ASAP therapist; David Burke, Physician investigator; Susan Murphy, research assistant; Aimee Reiss, blinded evaluator; Marsha Bidgood, blinded evaluator; Lois Wolf, ASAP therapist; Gina Holecek, usual care therapist; Megan Hite, usual care therapist; Melissa Tober, usual care therapist; Sara Zeforino, usual care therapist. National Rehabilitation Hospital and Georgetown University, Washington, DC: Alexander W. Dromerick, co–principal investigator, physician investigator; Matthew A. Edwardson, neuroanatomical and imaging analyst (NINDS-StrokeNet fellow); Kathaleen Brady, site team leader (August 2012-2015), ASAP therapist; Rahsaan Holley, site team leader (August 2009–August 2012), ASAP therapist; Lori Monroe, site team leader (August 2008–August 2009); Matthew Elrod, ASAP therapist (pilot and proposal phases); Kate Burdekin, research assistant; Siena Quitania, research assistant; Sara Loftin, research assistant; Margot Gianetti, research assistant; Rebecca Feldman, research assistant; Deirdra Tiffany, research assistant; Annie Simons, research assistant; Lauro Halstead, physician investigator; Kanan Desai, blinded evaluator; Hansen Chan, blinded evaluator; Jessica Barth, blinded evaluator; Sambit Mohapatra, blinded evaluator; Carrie Pappe, blinded evaluator; Nora E. Barrett, blinded evaluator; Ericka Breceda Tinoco, blinded evaluator; Diane Nichols, ASAP therapist; Alison Lichy, ASAP therapist; Melissa Cross, ASAP therapist; Mara Levy, usual care therapist; Connie Guercin, usual care therapist; Teresa Tracy, usual care therapist; Jutta Brettschneider, usual care therapist; Molly Fieldsend, usual care therapist; Janice Underwood, usual care therapist. University of Southern California Data Management and Analysis Center, Los Angeles: Christianne J. Lane, DMAC director (years 2-7); Stanley P. Azen, blinded statistician; Steven Yong Cen, database director; James Gardner, database programmer; Li Ding, statistician; Christopher Hahn, statistician; Jiaxiu He, statistician; Caron Park, statistician; Ge Wen, statistician; Miwa Takayanagi, statistician; Anny Xiang, DMAC director (year 1). Long Beach Memorial Medical Center, Long Beach, California: Candice Burtman-Regalado, site team leader, ASAP therapist; Charro Scott, site team leader, ASAP therapist; Richard Adams, physician investigator; Diehma Hoang, physician investigator; Audrey Huang, physician investigator; Shannon Massimo, research assistant; Shuywe Jenq, usual care therapist. Casa Colina Hospital and Centers for Rehabiliation, Pomona, California: Stephanie Kaplan, site team leader, ASAP therapist, ASAP standardization assessor; David Patterson, physician investigator; Cathelyn Timple, ASAP therapist; Shelia Mendon, usual care therapist; Jacob Hazen, usual care therapist; Deborah Ouellette, usual care therapist. Rancho Los Amigos National Rehabilitation Center, Downey, California: Oscar Gallardo, site team leader, ASAP therapist (Spanish bilingual); Covey Lazouras, site team leader, ASAP therapist; Xiao-Ling Zhang, physician investigator; Babak Bina, physician investigator; Joaquín Torres, research assistant; Rosemary Tamayo, research assistant; Byanca Rodriguez, research assistant; Chester Lin, research assistant; Alex Villegas, blinded evaluator (Spanish bilingual); Claire Smith, ASAP therapist; Mark Wolfson, usual care therapist; Kelly Love, usual care therapist. Cedars Sinai Medical Center, Los Angeles, California: Richard Riggs, physician investigator; Michelle Demond, site team leader; Pamela Roberts, site team leader; Nuvia Solis, research assistant; Sara Benham, ASAP therapist; Karina Fakheri, ASAP therapist; Aimee Davis, ASAP therapist; B. A. MacCormack, usual care therapist. Huntington Rehabilitation Medical Associates, Pasadena, California: Sunil Hegde, physician investigator; Cynthia Kushi, site team leader, ASAP therapist; Ilin Ohanessians, clinical site coordinator, site team leader; Michael Parkinson, ASAP therapist; Miriam Burch, usual care therapist. Rehabilitation Institute of Chicago, Chicago, Illinois: Richard L. Harvey, medical safety monitor. The following individuals provided oversight during the conduct and analysis of the study: Data Safety and Monitoring Board: Bruce Coull (chair), University of Arizona, Tucson; Stephen Nadeau, University of Florida, Gainesville; Michael Parides, Icahn School of Medicine at Mount Sinai, New York, NY; Sue Ann Sisto, Stony Brook University, Stony Brook, NY. National Institutes of Health, Bethesda, Maryland: Scott Janis, scientific program officer (NINDS); Janice Cordell, program official (NINDS); Louise Ritz, DSMB liaison (NINDS); Mary Ellen Michel, program official (NCMRR). Ancillary Study (ICAREGEN; NIH grant R01 NS058755): Steven C. Cramer, principal investigator; Jill See, project coordinator, University of California, Irvine.