|Home | About | Journals | Submit | Contact Us | Français|
Constraint-induced movement therapy (CIMT) is among the most developed training approaches for motor restoration of the upper extremity (UE).
Very Early Constraint-Induced Movement during Stroke Rehabilitation (VECTORS) was a single-blind phase II trial of CIMT during acute inpatient rehabilitation comparing traditional UE therapy with dose-matched and high-intensity CIMT protocols. Participants were adaptively randomized on rehabilitation admission, and received 2 weeks of study-related treatments. The primary endpoint was the total Action Research Arm Test (ARAT) score on the more affected side at 90 days after stroke onset. A mixed model analysis was performed.
A total of 52 participants (mean age 63.9 ± 14 years) were randomized 9.65 ± 4.5 days after onset. Mean NIHSS was 5.3 ± 1.8; mean total ARAT score was 22.5 ± 15.6; 77% had ischemic stroke. Groups were equivalent at baseline on all randomization variables. As expected, all groups improved with time on the total ARAT score. There was a significant time x group interaction (F = 3.1, p < 0.01), such that the high intensity CIT group had significantly less improvement at day 90. No significant differences were found between the dose-matched CIMT and control groups at day 90. MRI of a subsample showed no evidence of activity-dependent lesion enlargement.
Constraint-induced movement therapy (CIMT) was equally as effective but not superior to an equal dose of traditional therapy during inpatient stroke rehabilitation. Higher intensity CIMT resulted in less motor improvement at 90 days, indicating an inverse dose-response relationship. Motor intervention trials should control for dose, and higher doses of motor training cannot be assumed to be more beneficial, particularly early after stroke.
Stroke is a leading cause of disability, primarily from motor impairments. Experimental restorative treatments include structured therapeutic activity and pharmacologic augmentation.1 For the upper extremity (UE), constraint-induced movement therapy (CIMT) is among the most developed approaches. The basic principles of CIMT were first explored long ago2 but interest has surged recently.3,4 A phase III randomized controlled trial of CIMT initiated in the 3- to 9-month period after stroke onset found it superior to usual and customary care,5 with persistent treatment response.6
The use of CIMT during the inpatient rehabilitation phase, if proven superior to conventional rehabilitation, has several rationales. Substantial rehabilitation resources are expended during this phase of care; more effective retraining might improve outcome without increasing cost. Inpatient rehabilitation is the earliest point where CIMT is practical; early application might preempt learned non-use,7 and exploit any early window for best response to motor training.8 Our pilot trial provided a basis9 for further study.
Acute stroke admissions at Barnes-Jewish Hospital in St. Louis, MO, were screened for study eligibility (figure 1). Three additional subjects were referred by outside neurologists. Subjects were included if they had 1) unilateral ischemic or hemorrhagic stroke (with confirmatory neuroimaging) within 28 days of admission to inpatient rehabilitation; 2) persistent hemiparesis, generally indicated by a score of 1–3 on the motor arm item of the NIH Stroke Scale (NIHSS)10; 3) proximal UE voluntary activity indicated by a score of ≥3 on the upper arm item of the Motor Assessment Scale,11 but wrist and finger movement was not required; 4) preserved cognitive function as indicated by a) a score ≤1 on the consciousness and communication items of the NIHSS, b) the ability to perform two-step commands, and c) a score of <8 on the Short Blessed Memory Orientation and Concentration Scale12; and 5) no UE injury or conditions that limited use prior to the stroke. Exclusion criteria included 1) inability to give informed consent; 2) clinically significant fluctuations in mental status within 3 days of enrollment; 3) not independent prior to the index stroke (determined by scores <95 on the Barthel Index13 or >1 on the Modified Rankin Scale14; 3) hemispatial neglect as determined by asymmetry >3 errors on the Star Cancellation Test15; 4) sensory loss ≥2 on the sensory item of the NIHSS; 5) not expected to survive 1 year due to other illnesses.
VECTORS was a three-arm, single-blind, randomized controlled trial approved by the Washington University Human Studies Committee. The primary endpoint was the 90-day Action Research Arm Test (ARAT). Trained raters performed all blinded evaluations. Using baseline data collected on rehabilitation admission, we adaptively randomized22 the groups balancing for age, total NIHSS score, pretest ARAT score on the affected side, and days from stroke onset to randomization.
Study treatments preempted occupational therapy and occurred 5 days per week for 2 consecutive weeks. Prespecified treatment protocols were developed based on our prior study and incorporated features of the EXCITE trial5 manual. Subjects otherwise received routine inpatient interdisciplinary stroke rehabilitation. Research clinicians trained in study procedures provided all study-related treatments; all were licensed occupational therapists or supervised assistants. The study clinical team met weekly to assure adherence to protocols. Individual sessions and circuit-training techniques were used in all three study groups. The same group of therapists delivered both the experimental and control treatments. Prespecified protocols, developed for both the control and CIMT treatments, were focused primarily on basic activities of daily living (ADL) tasks required for discharge to the community.
The control treatment mimicked traditional occupational therapy, involving compensatory techniques for ADL, range of motion, and strengthening. This treatment consists of 1 hour of ADL retraining and 1 hour of UE bilateral training activities. Adaptive equipment and positioning devices were used; a prespecified cueing strategy neither encouraged nor discouraged the use of the hemiparetic UE. Massed practice, shaping, and constraint were prohibited.
The standard CIMT group received 2 hours of shaping therapy per day and wore a padded constraint mitten for 6 hours per day. The high-intensity CIMT group underwent 3 hours per day of shaping and wore the mitten for 90% of waking hours. This therapy consisted of performance of basic ADL together with supervised massed practice of skilled functional activities. These activities were graded per the study protocol by the treating occupational therapist to match the subject’s motor performance. Increasing levels of demand were based on increases in strength and coordination of the involved joints. As participants achieved successful completion of self-care activities and 80% of the trials for each specific motor activity, the complexity or difficulty of these activities were increased to the next level. Subjects received extensive verbal and written feedback about their performance, including a review of the prior day’s achievements, the day’s proposed goals, and reinforcement of new gains and maintenance of prior gains in graphic and verbal form appropriate to the activities selected for the treatment session.
In rodent brain infarct models, immediate immobilization of the unaffected forelimb causes lesion enlargement by an excitotoxic mechanism.30,31 We therefore carried out an MRI safety substudy in nine subjects with ischemic stroke (two in the control group, four in the standard CIMT group, and three in the high-intensity CIMT group) to ascertain whether very early CIMT as employed in this study resulted in expansion of ischemic lesions. Since both acute ischemia and direct excitotoxic damage result in restricted diffusion by MRI, we determined if any new areas with abnormalities on diffusion-weighted imaging (DWI) and apparent diffusion coefficient (ADC) measurements developed during the course of CIMT.38 An initial set of MRI scans was obtained between 1 and 22 days following stroke immediately prior to beginning the VECTORS treatments and a second set of scans 7–9 days later. MRI was performed with a 1.5-T Magnetom Sonata using an eight-channel head array and Syngo software. Following scout and auto alignment sequences, each subject had fluid-attenuated inversion recovery (repetition time [TR]/echo time [TE]/inversion time [TI] = 9,390/96/2,500, flip angle = 180, matrix = 192 × 256, 64 2-mm-thick slices, voxels = 1 × 1 × 2 mm), T2* (TR/TE = 2,700/61, flip angle = 90, acquisition matrix = 80 × 128, 60 2-mm-thick slices, voxels = 2 × 2 × 2 mm), and single-shot spin-echo echoplanar DWI (TR/TE = 10,000/80, flip angle = 90, acquisition matrix = 80 × 128, 72 2-mm-thick slices, voxels = 2 × 2 × 2 mm, b = 0 and b = 800 mm2/s) for DWI and ADC measurements. Images for each subject were coregistered using Automated Image Registration software (AIR 3.08, Roger Woods, UCLA). Subtraction images of coregistered DWI and ADC image pairs for each subject were visually examined for new areas of DWI or ADC change. In addition, any voxels within the area of infarction for which DWI intensities increased by at least 40% and ADC values decreased by at least 10% between the first and second scans were identified. These are conservative cutoffs based on a study that found mean DWI intensities were at least 90% greater and mean ADC intensities were at least 17% lower than normal within the first 7 days following cerebral infarction.23 Scans from four subjects with intracerebral hemorrhages were excluded from the analysis because of the large changes in DWI/ADC signals for clotted blood that occur over time.
Our objective was to examine whether CIMT was superior to an equivalent amount of traditional Occupational Therapy, and whether CIMT treatment effects would be dose dependent.
The NIH Stroke Scale (NIHSS)10 assesses cognitive, sensory, and motor impairments as an indicator of overall stroke severity.
The total ARAT16 score in the more affected UE was the primary dependent measure. The ARAT assesses UE functional limitations with four ordinal subscales: grasp, grip, pinch, and gross movement. Item scores are summed to create subtest and full-scale scores; the maximum score of 57 indicates normal performance.
The Functional Independence Measure (FIM) is an 18-item disability measure.17 All evaluators in this study were certified by the Uniform Data System. During the inpatient stay, the FIM was scored based on performance; after discharge, we used the Fone Fim18 format developed for telephone administration of the standard FIM items. We created a FIM Upper Extremity Score by summing the 5 FIM items requiring significant hand and arm use.
The five-item hand function subscale of the Stroke Impact Scale19 (SIS) assesses hand strength, dexterity, and fine and gross motor ability associated with the ability to complete functional tasks. Higher scores indicate greater function and life satisfaction.
The Wong-Baker Faces Scale was used to assess pain in the affected shoulder at days 0, 14, and 9020 and during treatment.
The Geriatric Depression–15 Scale21 was used to assess depression at days 0, 14, and 90. Scores below 5 are normal.
SAS Version 9 for Windows was used. A modified intent to treat approach was used, with all available data used for all analyses. One-way analyses of variance were computed to verify the effects of the adapted randomization procedure and to ensure that age, stroke severity, and initial arm function were equivalent across the groups. We hypothesized that subjects randomized to 14 days of standard intensity or high-intensity CIM therapy would have greater improvement in motor function from baseline to 90 days after stroke than subjects randomized to traditional therapy and that they would be more independent in activities of daily living at 90 days after stroke. The total ARAT on the more affected side was the primary dependent measure. Group comparisons were performed with Proc Mixed, a generalization of the standard linear model (GLM) procedure to determine if there were significant main effects for groups and time and a group x time interaction. Random coefficient regression is used to examine the response functions of each treatment group over time. Repeated measures analysis of variance models were fit for each dependent measure with group assignment (control, standard CIMT, and high-intensity CIMT) as the between-subject factor and time of assessment (baseline, day 14, and day 90) as the within-subject factor. Least-square means adjusted for randomly missing values were computed by group for each time of testing. Type 3 tests of fixed effects were used for hypothesis testing. If the group x time interactions were significant (p < 0.05) preplanned comparisons were computed to examine pairwise differences among the three treatment groups. A futility analysis was conducted after 3 years of active subject recruitment. The analysis, completed by the project statistician, confirmed the study was adequately powered with 52 subjects to test the hypothesis.
Figure 1 shows the CONSORT diagram of participant recruitment. Fifty-two individuals were randomized and table 1 presents their baseline characteristics. There were no significant differences on these variables across the treatment groups. There were no significant differences between groups in medical comorbidities: 10% had prior stroke, 65% had hypertension, 23% had diabetes, 6% had atrial fibrillation, and 10% had a history of myocardial infarction. None was on renal dialysis. Rehabilitation length of stay was 22.3 ± 7.1 days.
All but two participants were available for assessment at the 90-day primary endpoint. ARAT total and subtest scores for the affected arm are presented in table 2. The primary test of the study hypothesis is the comparison across groups of the slopes of the total ARAT scores from baseline to day 90 (group x time). The overall main effect for treatment groups was not significant (group: F = 0.50, p < 0.61). As expected, total ARAT scores improved from baseline to 90 days for all groups (time: F = 45.15, p < 0.0001). There were no significant differences between the standard CIMT and control patients from baseline to day 14 or baseline to day 90 (table 2, figure 2). Unexpectedly, for high-intensity CIMT, a significant group x time interaction (F = 3.06, p < 0.01) was observed for the total ARAT when all three groups were compared. Contrary to the study hypothesis, high-intensity CIMT patients had significantly lower gains in total ARAT scores from baseline to day 90 than control patients (p < 0.006) or standard CIMT patients (p < 0.01). Group comparisons of baseline vs day 14 were similar; high-intensity CIMT patients performed less well than control patients (p < 0.03) or standard CIMT patients (p < 0.003) (see figure 2A).
The same analyses were computed for the ARAT subtests (table 2). All subtest scores increased significantly from baseline to day 90. A group x time interaction was observed for the ARAT pinch subtest (p < 0.03). Pinch scores were highest in the standard CIMT group at day 14 and day 90 and lowest for the high-intensity CIMT pinch scores at day 90. The group x time interactions for the ARAT grip and gross motor subtests approached but did not reach statistical significance (p < 0.06). High-intensity CIMT patients had the lowest scores on these subtests at day 14 and day 90.
We evaluated the effects of standard or high-intensity CIMT on activities of daily living with the FIM Upper Extremity Score and the SIS hand subscale (figures 2B and 2C). We found no significant differences across the three treatment groups for the FIM Upper Extremity Scale; all groups achieved similar gains in function (p < 0.16). Mean ratings for SIS Hand subscale varied by group over the two times of testing [F(1,50) = 3.88, p = 0.02]. The control group scores were higher at day 14, while the standard CIMT group reported the highest scores at day 90. The high-intensity CIMT group had significantly lower scores at day 90 (p < 0.02). Thus, CIMT was no more effective than the control treatment in ameliorating UE-related disability and the high-intensity CIMT group had significantly less improvement.
Mean pain scores were low (≤3 out of 10 possible points), and shoulder pain increased significantly from baseline to the 90-day endpoint [F(2,49) = 4.01, p < 0.01]. There were no significant differences among the three treatment groups [F(2,49) = 0.22, p < 0.80], and all were similar to the prevalence of shoulder pain in other cohorts.24 Group means for the Geriatric Depression Scale were within the normal range for all three groups at all three time points.
The nine subjects in the MRI substudy were initially studied 1–22 (mean 13.2) days after onset and again 7–9 (mean 7.3) days later. Infarct volumes defined by thresholding the DWI/ADC ratio image ranged from 0.1–13.5 mL. Eight of the nine subjects (one control, four standard CIMT, and three high-intensity CIMT) showed no areas of DWI or ADC change upon visual inspection and had no voxels with changes in DWI or ADC signal intensities above threshold. Another control subject had five scattered voxels above threshold which could be attributed to imaging noise.
During acute inpatient stroke rehabilitation, CIMT was not superior to an equal amount of traditional occupational therapy. Further, we found an inverted dose response effect, in which a higher dose of shaping and constraint led to significantly less UE motor improvement at 90 days. These striking results were consistent across all endpoints.
These results are consistent with other reports, particularly with studies in which control groups received equal amounts of therapy time. CIMT was found to be superior to an equal amount of neurodevelopmental treatment in a more clinically heterogeneous sample of chronic patients who were at least 1 year poststroke.25 However, this superiority was limited to the subgroups with sensory loss and hemineglect, two groups excluded from VECTORS. A smaller acute study with a similar design using treatment sessions distributed over 3 weeks, and including subjects 2 weeks poststroke, found no difference in prespecified primary outcomes compared to a dose-matched control.26 Our prior work found a benefit for CIMT compared to a dose-matched control, but the sample was much smaller and the randomized groups were not balanced on important clinical features.9 A phase III trial of CIMT performed in the 3- to 9-month period poststroke demonstrated the ability of a CIMT program to improve UE motor function during that period after stroke when compared to the usual and customary care provided in the United States.5 Because usual and customary care in the United States during this 3- to 9-month period involves far less motor training than that delivered to the CIMT group, the extent to which the benefit was related to the CIMT treatment paradigm, to a nonspecific effect of a higher dose of motor training, or to both27 is unclear. Two other studies using a dose-matched control did find CIMT to be superior, but used a small sample28 or a later time after stroke onset.29
Our finding of an inverted dose-response curve bears further comment, and we have considered several explanations. In rodent stroke recovery models, immediate casting of the unaffected forelimb has been reported to cause lesion enlargement30 that is presumed excitotoxic31 and is associated with a decrement in motor recovery.32 We found no evidence for CIMT-caused lesion enlargement in a small MRI substudy. Shoulder pain and depression have both been reported to hinder stroke recovery.33 However, self-reported pain and depression were not significantly different across the three study groups. Another possible explanation is fatigue or injury related to overtraining. Overtraining effects are best described in the resistance training literature, where effects of resistance loading and training volume have been studied in some detail.34 The possibility of overtraining was recently described in another stroke rehabilitation study using different training modalities to improve gait35; in this case, however, progressive lower extremity strength training was an explicit intervention goal, an approach quite unlike CIMT. Yet another possibility is that increasing the time spent daily in shaping sessions by 50% effectively led to a change in practice schedule, from a “distributed” practice schedule to a more “blocked” schedule. There is a substantial motor learning literature suggesting that distributed practice schedules lead to better task learning and performance.36,37 Thus the extra time spent by the high-intensity CIMT group actually may have interfered with motor learning.
NIHSS scores were relatively low in our sample, a consequence of enrolling cognitively capable persons who had some residual UE motor abilities. Eliminating persons with complete paralysis or severe disorders of consciousness, sensation, aphasia, or neglect must result in NIHSS scores within the range of our study participants. Because all subjects otherwise received usual and customary care after the 14-day study treatment, data about treatments after discharge were not collected; CIMT was prohibited outside the study treatment period. Although this study used a relatively small sample, futility analysis indicated sufficient power to test the hypothesis.
Our results have two important implications. First, our finding that equal time of CIMT and conventional therapy were equally effective in ameliorating UE impairment and disability emphasizes the need for clinical trials with control groups that match therapy doses. Without such a control, the superiority of a specific training paradigm cannot be demonstrated. Second, we found significant dose effects for training, suggesting that motor training interventions, like drug interventions, can have dose-response effects. However, as with drugs, it appears possible that doses higher than an optimal range may be less effective than lower doses, and even detrimental. Establishing optimal therapy dosing may require determination of specific dose-response curves through clinical studies that directly test different doses of therapy. Given the large variety of results of timing and dose studies in rodent models of stroke recovery, we suggest that such studies must address specific time points after stroke onset.
Statistical analysis was performed by J.P.M.
The authors thank Rahsaan J. Holley, MS, OTR/L, for assistance with manuscript preparation. The authors also thank Michele Hahn, MS, OTR for assistance with treatment design.
Dr. Dromerick has received research support from NIH, NINDS, and United States Veterans Affairs, and Intramural support from the United States Army, Department of Defense. Dr. Lang has received research support from NIH, NINDS, and the Missouri Physical Therapy Association. Dr. Miller has served on the Data and Safety Monitoring Board for Ethicon. Dr. Powers has received research support from University of North Carolina, Washington University, University of Iowa, University of Kentucky, Harvard University, Bowdoin College, NIH, Legatus Emergency Services, LLC, Neutral, LLC, EDJ Associates, Hitchcock Foundation, Dartmouth-Hitchcock Clinic, Certus International, Inc., Companion Baking Company, and Union Square Hospitality Group. Dr. Wolf is supported by NIH, Allergan, and AMES.
Address correspondence and reprint requests to Dr. Alexander W. Dromerick, National Rehabilitation Hospital, 102 Irving Street, NW, Washington, DC 20010 Alexander.w.dromerick/at/medstar.net
Editorial, page 170
e-Pub ahead of print on May 20, 2009, at www.neurology.org.
This study was funded by NIH grant 1 RO1 NS41261-01A1 and James S. McDonnell Foundation grant 21002032.
Disclosure: Author disclosures are provided at the end of the article.
Received October 24, 2008. Accepted in final form March 16, 2009.