Research Articles

Constraint-Induced Movement Therapy

During Early Stroke Rehabilitation
Corwin Boake, PhD, Elizabeth A. Noser, MD, Tony Ro, PhD, Sarah Baraniuk, PhD,
Mary Gaber, OTR, Ruth Johnson, MA, Eva T. Salmeron, MD, Thao M. Tran, MD,
Jenny M. Lai, MD, Edward Taub, PhD, Lemuel A. Moye, PhD,
James C. Grotta, MD, and Harvey S. Levin, PhD

Background. Limited data are available about the effectiveness of months after stroke is associated with increased motor
early rehabilitation after stroke. Objective. This is the 1st ran- excitability of the affected cerebral hemisphere.
domized controlled trial of constraint-induced movement ther-
apy (CIMT) in subacute stroke to investigate neurophysiologic Key Words: Stroke—Cerebrovascular disorders—Hemiplegia—
mechanisms and long-term outcome. Methods. Within 2 weeks Randomized controlled trial—Rehabilitation—Transcranial mag-
after stroke, 23 patients with upper extremity (UE) weakness netic stimulation.
were randomized to 2 weeks of CIMT or traditional therapy at
an equal frequency of up to 3 h/day. Motor function of the

C
onstraint-induced movement therapy (CIMT) is
affected UE was blindly assessed before treatment, after treat- a rehabilitation technique for hemiparesis devel-
ment, and 3 months after stroke. Transcranial magnetic stimu- oped in the laboratory of one of the authors (ET)
lation (TMS) measured the cortical area evoking movement of that consists of restraining the unaffected upper extrem-
the affected hand. Results. Long-term improvement in motor
ity (UE) while intensively training the affected arm and
function of the affected UE did not differ significantly between
patients who received CIMT versus intensive traditional ther-
hand to improve performance on functional motor
apy. All outcome comparisons showed trends favoring CIMT tasks.1,2 Although the effectiveness of CIMT in chronic
over intensive traditional therapy, but none was statistically sig- stroke is well established,3-6 support for CIMT early after
nificant except for improvements in the Fugl-Meyer (FM) UE stroke is limited to a single clinical trial involving hospi-
motor scale immediately following treatment and in reported talized patients7 in which outcome was measured imme-
quality of hand function at 3 months. Improvement in UE diately after completion of therapy without including a
motor function on the FM was associated with a greater follow-up assessment. Apart from preliminary reports
number of sites on the affected cerebral hemisphere where from the current study,8,9 no data are available about the
responses of the affected hand were evoked by TMS. Conclusions. long-term maintenance of therapeutic gains or about
Future trials of CIMT during early stroke rehabilitation need the neurophysiologic mechanisms that mediate the
greater statistical power, more inclusive eligibility criteria, and
effects of early therapy. This article reports a clinical trial
improved experimental control over treatment intensity. The
relationship between changes in motor function and in evoked
of CIMT during early stroke rehabilitation that is the 1st
motor responses suggests that motor recovery during the 1st 3 controlled study to evaluate whether improved motor
function is maintained after treatment and to investigate
From the Departments of Physical Medicine and Rehabilitation, how neurophysiologic changes resulting from therapy
Baylor College of Medicine/University of Texas-Houston Medical are related to therapeutic gains.
School, Houston, TX (CB, ETS, TMT, JML, HSL); the Department of The major aim of this single-blind randomized clini-
Neurology Stroke Program, University of Texas-Houston Medical cal trial was to evaluate whether providing 2 weeks of
School, Houston, TX (EAN, JCG); the Department of Psychology,
Rice University, Houston, TX (TR, RJ); University of Texas-Houston
CIMT initiated within 2 weeks of unilateral stroke was
School of Public Health, Houston, TX (SB, LAM); Department of feasible and induced lasting beneficial changes in motor
Physical Medicine and Rehabilitation, Memorial Hermann Hospital, function of the affected upper limb. The study used a
Houston, TX (MG); Department of Psychology, University of Alabama parallel-groups design comparing a group of patients
at Birmingham, Birmingham, AL (ET). who received CIMT with a control group receiving ther-
Address correspondence to Corwin Boake, PhD, TIRR, 1333 Moursund, apy consisting of traditional techniques. Frequency and
Houston, TX 77030-3405. E-mail: [email protected]. duration of therapy were equated between the treat-
Boake C, Noser EA, Ro T, Baraniuk S, Gaber M, Johnson R, Salmeron ment groups to avoid confounding type of therapy with
ET, Tran TM, Lai JM, Taub E, Moye LA, Grotta JC, Levin HS. the intensity of intervention. Motor function was mea-
Constraint-induced movement therapy during early stroke rehabilita- sured by behavioral motor performance tests and by
tion. Neurorehabil Neural Repair 2007;21:14–24. self-report. A follow-up assessment at 3 to 4 months
DOI: 10.1177/1545968306291858 after stroke evaluated whether therapeutic gains had

14 Copyright © 2007 The American Society of Neurorehabilitation

 CIMT During Early Stroke Rehabilitation

been maintained. A secondary aim was to investigate length, 5 mm diameter) into a grid of vertical holes in a
neurophysiologic mechanisms that mediated effects of horizontal 10 cm square surface. The test was discontin-
therapy, using transcranial magnetic stimulation (TMS) ued at 150 s if the patient was unable to insert any pegs.
to map the cortical hand area of the stroke hemisphere. The GPT measure for each hand was the number of
TMS was coadministered on the same occasions as pegs placed per second. The worst possible value of 0
behavioral testing in order to explore correlations was earned by patients who did not place any pegs.
between motor recovery and neurophysiologic changes. Performance by the unaffected hand was used to mea-
sure adverse effects of wearing the restraint.

METHODS Motor Activity Log.1,13 The Motor Activity Log (MAL), a

rating scale, evaluates how the affected hand is used to per-
Patients form 30 daily living activities (e.g., feeding, turning a door
handle). For each activity, the patient rates how much the
Adult stroke patients were recruited from March 2001 affected hand is used (amount of use) and how well the
to October 2004 from admissions to the in-patient Stroke activity is performed (quality of movement). Ratings are
Service of Memorial Hermann Hospital, a teaching on a scale from 0 to 5, with higher scores representing
hospital in Houston, TX, which is affiliated with the better function. The MAL measures were the average rat-
University of Texas-Houston Medical School. Criteria for ings for amount of use and for quality of use. The standard
enrollment were ischemic or hemorrhagic stroke within MAL was modified for this study by omitting 8 activities
14 days of entering the trial; stroke lesion visualized on that could not be performed during in-patient hospital-
computed tomography or magnetic resonance imaging ization (e.g., turning an ignition key). The MAL was
scan of the brain performed before enrollment; score of administered only at baseline and at the 2 outcome assess-
1 to 3 on item 5 (arm motor) of the National Institutes ments, and not during the treatment period.
of Health (NIH) Stroke Scale (NIHSS);10 at least 10 deg
of active movement in the thumb and 2 or more fingers Transcranial magnetic stimulation. The hand motor area
of the affected hand; total NIHSS score ≤14 if right- and of each cerebral hemisphere was mapped using TMS
≤19 if left-sided stroke; ability to provide informed according to procedures previously validated14,15 and
consent; no previous stroke that would interfere with detailed in our preliminary article.9 TMS was delivered by
interpretation of the results; no neglect or speech com- a Cadwell Laboratories MES-10 polyphasic stimulator
prehension impairment that would prevent participation (Kennewick, WA). Electromyographic (EMG) activity of
in the study assessments and treatment; no pacemaker or the abductor pollicis brevis (APB) of the resting hand
other metallic implant; no UE orthopedic limitation that contralateral to stimulation was recorded using silver-
would affect the results; and readiness to participate in silver chloride surface electrodes attached to a Grass-
standard rehabilitation at the time of enrollment. The Astromed (model IP511, West Warwick, RI) amplifier
NIHSS arm motor score criterion excluded patients with and digitized by a data acquisition card (model CYDAS 8,
no arm movement at all, or no detectable shoulder weak- CyberResearch, Inc., Branford, CT). The EMG signal was
ness. The requirement of active movement in the thumb amplified and filtered for frequencies between 30 Hz and
and fingers ensured that motor function would have been 1000 Hz and was sampled at 1 kHz. During each TMS
sufficient to perform the CIMT training tasks. session, both hemispheres were mapped and the order of
the hemispheres was counterbalanced across patients and
sessions. The hand area and TMS motor threshold for
Outcome Measures each hemisphere were established at the beginning of the
session. The motor threshold was determined at each ses-
Fugl-Meyer Assessment of Motor Recovery.11 The Fugl- sion, and TMS intensity at each session was determined
Meyer Assessment of Motor Recovery (FM), a motor relative to the current threshold. For each hemisphere,
performance test consisting of 33 tasks performed by the TMS intensity was set at 10% above the motor threshold,
affected UE, evaluates the ability to make movements defined as the lowest intensity that induced activation of
outside of a synergistic pattern. Performance on each the contralateral hand on at least 3 of 5 trials. The thresh-
task is rated 0, 1, or 2, with higher ratings representing old definition of 3 out of 5 trials was justified by the need
better performance. The FM measure used in this study to minimize fatigue caused by the long sitting time
was the sum of the 33 ratings (possible range 0 to 66). required of the subacute stroke patients during extensive
TMS and motor testing sessions. Once the cortical hand
Grooved Pegboard Test.12 The Grooved Pegboard Test area had been localized, TMS was applied systematically
(GPT), a test of manual dexterity, evaluates the speed to locations marked directly on the scalp at steps of 1 cm
with which the patient grasps and inserts 25 pegs (3 cm from the initial hand localization site. Activation of a site

Neurorehabilitation and Neural Repair 21(1); 2007 15

Boake et al

was defined as a TMS-evoked motor response detected patients in that the number of hours of therapy per day
by EMG or visually observed movement by 2 observers and the number of therapy days were increased to
on any area of the hand on at least 3 of 5 trials. Visual approximate the frequency and duration of CIMT. No
observation was used as the sole criterion when EMG restraint was used, and patients were free to use either
was unavailable (e.g., technical problems). Both EMG hand for daily activities.
and visual inspection were recorded during the major-
ity of TMS sessions, and both were used as activation
criteria. Thus, during TMS mapping, a given scalp posi- Procedure
tion was classified as an activation site if either criterion
had been met on 3 of 5 trials. Correspondence between The study was approved by the Committee for the
EMG and visual inspection was evaluated in our pre- Protection of Human Subjects at the University of Texas-
liminary study9 and was found to be high. Stimulation Houston Medical School. Consecutive patients admitted
was continued until demarcating an area of active sites to the stroke unit of Memorial Hermann Hospital were
surrounded by a border of sites at which the evoked screened for eligibility. After giving informed consent, eli-
motor responses did not meet either the EMG or visual gible patients underwent baseline testing and were ran-
observation criteria (cf. reference 9). In some cases, domly allocated to either CIMT or traditional therapy.
when TMS of the border regions activated some hand Randomization was stratified for age (≤60 years or
movements, but not enough to meet the criterion of >60 years) and NIHSS arm motor score (1 or 2–3). The
activation on at least 3 out of 5 trials, locations outside NIHSS arm motor score was selected for stratification
this border were also tested to ensure that no more acti- because of its prognostic value, because the score levels are
vation sites could be detected. The dependent measures appropriate to classify patients for our study, and because
for each hemisphere were the number of TMS activa- the score distribution in our stroke admissions was
tion sites and the motor threshold. If fewer than 3 motor already known. Therapy began on the day of baseline test-
responses were evoked at any position on the affected ing or the following day, at a median of 11 days after
hemisphere, then the number of stroke-hemisphere stroke (range 5 to 19 days). All patients received either
TMS activation sites was defined as zero and the motor CIMT or traditional UE therapy at an equal frequency
threshold was defined as greater than 100%. and duration of up to 3 h per day, for 14 to 15 days at a
frequency of 6 days per week excluding Sundays. This pro-
cedure controlled the number of hours of therapy time,
Treatment but not energy expended during therapy. Intervention
was provided by licensed therapists and therapy assistants,
Constraint-induced movement therapy. Therapy sessions including an occupational therapist (MG) and a physical
consisted of performing tasks only with the affected UE. therapist who had undergone training in Dr. Taub’s labo-
Task movements included reaching, grasping, lifting, ratory at the University of Alabama at Birmingham and
and placing. Tasks were individually selected according who provided or supervised the UE therapy for all
to motor ability, to ensure successful experience and pre- patients. Most patients began the study therapy on the
vent frustration leading to learned nonuse.1,16 Task diffi- in-patient rehabilitation unit, were discharged from
culty was progressively increased using behavioral the hospital during the therapy period, and completed
techniques of shaping and successive approximation.17 the remaining therapy sessions in the out patient rehabil-
In addition to individual therapy sessions, patients were itation clinic. Motor tests and TMS mapping were
asked to wear a mitten restraint (Sammons Preston repeated 2 times after treatment. The 1st occasion was a
#6727 “Padded Safety Mitt,” Sammons Preston, Inc., posttreatment evaluation immediately after completing
Bolingbrook, IL) on the unaffected hand during 90% of therapy, and the 2nd was a follow-up evaluation 3 to 4
waking hours, excluding activities when risk of injury months after stroke. Motor testing and TMS mapping of
might increase. The mitten allowed the unaffected UE to a given patient were generally performed on the same day,
assist in transfers and ambulation, but it prevented use of but the order of these procedures was not controlled.
the unaffected fingers to manipulate objects and necessi- Outcome evaluations were performed by personnel from
tated use of the affected hand to perform daily activities. outside of Memorial Hermann Hospital who were blind
to treatment assignment.
Intensive traditional therapy. Therapy sessions consisted
of performing daily living tasks with either hand and
therapeutic activities with the affected UE that were Analyses
intended to improve strength, muscle tone, and range of
motion. This treatment condition differed from the This was a randomized clinical study with 2 treat-
standard therapy regimen provided to non-study- ment groups (CIMT and intensive traditional therapy)

16 Neurorehabilitation and Neural Repair 21(1); 2007

 CIMT During Early Stroke Rehabilitation

Table 1. Demographic and Clinical Features of Patients However, they did not have their type I error rates cor-
Receiving Study Therapy rected for multiple comparisons and are secondary end-
points, with P value assessment at the nominal 0.05
CIMT Traditional Therapy
level. The remaining analyses were exploratory. We used
N 10 13 Pearson correlations to examine the association between
Age, M (SD) 63.1 (14.3) 58.9 (14.0) UE motor performance and number of TMS active sites
Sex, n in the affected cerebral hemisphere. We used the
Female 3 5 Wilcoxon 2-sample test to compare stroke-hemisphere
Male 7 8 motor thresholds between the 2 treatment groups at
Affected side, n
each of the 3 TMS assessments. The nonparametric test
Left 6 6
was used to include patients who had no stroke-hemi-
Right 4 7
NIHSS total score, 4.9 (1.8) 5.3 (3.4) sphere TMS activation sites.
M (SD)
NIHSS arm motor 1.6 (0.84) 1.46 (0.77)
score, M (SD) RESULTS
Type of stroke, n
Infarct 9 9 Feasibility and Safety
Hemorrhage 1 4
Stroke location, n Table 1 presents the demographic and clinical features
Cortical 2 3 of the patients who met entry criteria and were ran-
Subcortical 5 8
domized. The treatment groups did not differ sig-
Brainstem 3 2
nificantly in baseline characteristics, including side of
CIMT = constraint-induced movement therapy; NIHSS = National stroke. All randomized patients underwent the planned
Institutes of Health Stroke Scale.
series of therapy sessions. Time from stroke to the
follow-up evaluation averaged 104.2 days in the CIMT
in a relatively small number of individuals. Baseline group (range 91–137 days) and 108.4 days in the control
measures of demographics and comorbidity were group (range 92–143 days). No significant adverse
assessed using Fisher’s exact test for dichotomous vari- events occurred during the therapy period. No seizures
ables and student’s t test for continuous variables. were induced by TMS. No patient had a recurrent stroke
The prospectively declared endpoint for this study was during participation in the study.
the impact of the intervention on the change of the FM At baseline, the FM was not completed by 1 control
score over time, from the baseline to the follow-up patient. One CIMT patient was not administered the
assessment. Thus, the change in this measure in the MAL at baseline and posttreatment owing to aphasia. At
CIMT group was compared to the change in this mea- follow-up, 1 control patient was not administered the
sure in the control group. The estimates necessary FM and GPT because of UE injury unrelated to the
to complete the sample size calculation for this trial study, and 1 control patient had missing GPT data. Two
were determined from analyses performed on the 1st 8 control patients did not complete any measures at
patients of this study.9 From this evaluation, the follow-up because of refusal to participate and change
observed FM effect size (i.e., change in active group of residence to a different city. Three patients (2 control,
minus change in the control group) was 14.5, with a 1 CIMT) had incomplete FM administrations at base-
standard deviation of 10.76 in the CIMT group and line or follow-up because of omission of up to 3 items
11.36 in the control group. A standard sample size com- due to examiner error. Because of these difficulties
putation determined that a total of 24 patients (12 in obtaining complete data, FM results are reported sepa-
each of the groups) were required to have 90% power to rately for the 16 patients (9 CIMT, 7 control) who had
detect the full effect size for a 2-sided type I error of 0.05. complete FM data at both the baseline and follow-up
We used descriptive statistics (means, medians, and assessments (complete cohort). Figure 1 shows for the
standard deviations) to summarize the data. Any patient complete cohort the flow of patients through the study
with missing or incomplete FM data at baseline or and collection of the FM data. FM results are also
follow-up was excluded from the principal analysis of reported for the 22 patients (10 CIMT, 12 control)
the primary endpoint. A 2-sided test of significance was including those who had incomplete FM data caused by
carried out for the primary outcome, with an alpha level omission of up to 3 items, or who had missing FM data
of 0.045, consistent with an alpha spending function at follow-up (total cohort). For the total cohort, FM
that permitted the interim analysis. The evaluation of scores for the 3 patients with omitted items were obtained
the impact of the intervention on 2 other functional by assigning a rating of zero (worst possible score) for
measures (GPT and MAL) was prospectively declared. those items. For the 3 patients with no follow-up FM

Neurorehabilitation and Neural Repair 21(1); 2007 17

Boake et al

Figure 1. Flowchart of the flow of patients through the study and collection of the Fugl-Meyer (FM) upper-extremity motor scale.
The complete cohort includes only patients with complete FM data at both the baseline and follow-up assessments. CIMT =
constraint-induced movement therapy.

data, FM results at the posttreatment evaluation were P = 0.137, 95% confidence interval [CI] of mean differ-
carried forward. These conventions were determined ence = –3.81, 24.95; total cohort: t = 1.47, df = 20,
prospectively, before unblinding of results. P = 0.157, 95% CI of mean difference = –3.28, 18.98).
The CI of the difference between treatment group
means is coded so that positive values indicate an
Motor Skills advantage for the group receiving CIMT.
Results for the secondary endpoints were as follows. For
Tables 2 and 3 show the results for both treatment the FM, improvement from baseline to posttreatment was
groups on measures of motor performance, self-report, significantly greater in the CIMT group for the 16 patients
and TMS assessed at baseline, immediately after treat- with complete data (t = 2.15, df = 14, P = 0.0497, 95% CI
ment, and at follow-up 3 to 4 months after stroke. of mean difference = 0.011, 16.88) but not in the total
Improvement was seen in both treatment groups on all cohort of 22 patients (t = 1.21, df = 20, P = 0.237, 95% CI
measures of motor function in the affected UE. of mean difference = –3.28, 12.51). Results for the GPT
All of the comparisons of motor performance between and MAL were analyzed for patients with complete data at
the CIMT and intensive traditional therapy group were baseline and follow-up. For the GPT, performance by the
in the predicted direction (i.e., greater improvement affected hand improved from baseline to follow-up in all
in affected UE function in the CIMT group), but the patients, without the CIMT and intensive traditional ther-
primary outcome was not statistically significant. apy groups differing significantly (t = 1.49, df = 17,
Change in FM from baseline to 3 months, which was P = 0.153, 95% CI of mean difference = –0.02, 0.12). For
the prospectively declared primary endpoint, showed the MAL, improvement in the CIMT group from baseline
improvement in both treatment groups, although the to follow-up was significantly better for quality (t = 2.30,
apparent advantage of the CIMT group was not statisti- df = 18, P = 0.033, 95% CI of mean difference = 0.11, 2.32)
cally significant (complete cohort: t = 1.43, df = 14, but not for amount (t = 1.86, df = 18, P = 0.080, 95% CI of

18 Neurorehabilitation and Neural Repair 21(1); 2007

 CIMT During Early Stroke Rehabilitation

Table 2. Fugl-Meyer Assessment Upper-Extremity Motor Scale: Complete and Total Cohort

Baseline Posttreatment Follow-up Difference Follow-up-Baseline

Complete cohort, M (SD) n = 16

CIMT 29.44 (12.11) 47.89 (6.71) 53.44 (8.63) 24.00 (14.60)
Traditional therapy 36.71 (14.22) 46.71 (14.88) 50.14 (10.88) 13.43 (11.34)
Total cohort, M (SD) n = 22
CIMT 30.8 (12.20) 49.0 (7.24) 54.4 (8.68) 23.6 (13.82)
Traditional therapy 29.5 (18.35) 43.6 (14.02) 45.5 (14.4) 15.7 (11.2)
CIMT = constraint-induced movement therapy.

Table 3. Grooved Pegboard Test and Motor Activity Log

Baseline Posttreatment Follow-up

Grooved Pegboard affected side, M (SD)

CIMT 0.01 (0.027) 0.078 (0.063) 0.163 (0.062)
Traditional therapy 0.0046 (0.015) 0.041 (0.075) 0.108 (0.101)
Grooved Pegboard unaffected side, M (SD)
CIMT 0.179 (0.051) 0.256 (0.107) 0.243 (0.059)
Traditional therapy 0.183 (0.073) 0.229 (0.078) 0.268 (0.097)
Motor Activity Log–Amount of Use, M (SD)
CIMT 0.479 (0.426) 2.21 (1.41) 3.10 (1.40)
Traditional therapy 0.690 (0.770) 1.94 (1.49) 2.23 (1.39)
Motor Activity Log–Quality of Use, M (SD)
CIMT 0.474 (0.314) 2.07 (1.26) 3.27 (1.28)
Traditional therapy 0.764 (0.805) 1.82 (1.26) 2.29 (1.38)
CIMT = constraint-induced movement therapy.

mean difference = –0.13, 2.16). GPT performance by the the 2 patients (both controls) who were lost to follow-
unaffected hand showed no worsening over time and no up. As seen in Tables 4 and 5, at the baseline evaluation,
difference between groups. most patients in both treatment groups had no TMS
In view of the discrepancy in motor results from the activation sites in the affected hemisphere, and therefore
preliminary study, which had revealed a large advantage the median motor threshold could not be estimated.
of CIMT relative to the control condition,8,9 we carried Over the study period, there was a dramatic increase in
out exploratory analyses of differences in baseline char- the proportion of patients in whom movements of the
acteristics between patients in the preliminary study affected hand were evoked by TMS, as is consistent with
and those treated afterward. These analyses revealed a the expected course of recovery after stroke. However,
difference in UE motor impairment as measured by the there was no statistically significant difference in motor
NIHSS arm motor score at baseline, with the 8 patients threshold between the 2 treatment groups at baseline
in the preliminary study exhibiting greater shoulder (Wilcoxon statistic = 134.5, Z = 1.02, P = 0.318), the
weakness (mean = 2.25, SD = 0.89) relative to the final evaluation immediately after treatment (Wilcoxon sta-
15 patients (mean = 1.20, SD = 0.56) (t = 3.49, df = 21, tistic = 96.0, Z = 0.08, P = 0.940), or the follow-up evalua-
P = 0.002, 95% CI of mean difference = 0.42, 1.68). We tion at 3 months poststroke (Wilcoxon statistic = 110.0,
attempted to identify patient characteristics associated Z = 0.0, P = 1.0).
with greater response to CIMT relative to the control Data for the number of active TMS sites in the stroke
condition, but these subgroup analyses were not viable and nonstroke hemisphere are summarized in Table 4,
because of small sample sizes. and individual patient data are presented in Table 5.
In the stroke hemisphere, no specific effect of CIMT
was apparent in the number of TMS activation sites
Transcranial Magnetic Stimulation detected at the immediate posttreatment evaluation.
The mean number of stroke-hemisphere TMS activa-
Tables 4 and 5 show the TMS results at the same 3 tion sites at this evaluation did not differ between
occasions. TMS was not completed in the 1st 3 enrolled patients who had been evaluated on the final day of
patients (2 control, 1 CIMT) at posttreatment and in treatment versus those tested later (means 3.55 and 4.67,

Neurorehabilitation and Neural Repair 21(1); 2007 19

Boake et al

Table 4. Transcranial Magnetic Stimulation Motor Threshold, Number of Patients Without Motor-Evoked Potential,
and Number of Active Transcranial Magnetic Stimulation Sites

Baseline Posttreatment Follow-up

Patients without stroke-hemisphere MEP, n without MEP/n tested

CIMT 8/10 2/9 1/10
Traditional therapy 7/13 2/11 1/11
Motor threshold (%MSO) in stroke hemisphere, Mdn
CIMT > 100 83.0 75.0
Traditional therapy > 100 83.0 67.0
Active TMS sites in stroke hemisphere, M (SD)
CIMT 1.00 (2.16) 3.00 (2.12) 6.10 (3.28)
Traditional therapy 1.53 (2.22) 4.09 (3.39) 4.54 (2.84)
Active TMS sites in nonstroke hemisphere, M (SD)
CIMT 7.00 (4.66) 6.56 (2.83) 5.70 (4.14)
Traditional therapy 4.54 (2.07) 4.55 (1.81) 5.36 (1.86)
MEP = motor evoked potential; CIMT = constraint-induced movement therapy; MSO = maximum stimulator output; Mdn = median; TMS =
transcranial magnetic stimulation.

Table 5. Number of Active Transcranial Magnetic Stimulation Sites Detected in Individual Patients

Stroke Hemisphere Nonstroke Hemisphere

Group/Patient # Baseline Posttreatment Follow-up Baseline Posttreatment Follow-up

CIMT
3 0 NT 8 7 NT 2
4 0 7 10 9 5 14
7 0 3 11 6 10 5
8 0 0 7 19 12 12
11 0 3 4 5 5 1
13 0 3 5 6 8 6
14 4 4 6 4 6 4
18 0 0 0 3 4 4
21 6 4 7 3 4 5
22 0 3 3 8 5 4
Traditional therapy
1 0 NT 1 1 NT 2
2 1 NT 9 3 NT 6
5 0 0 0 4 4 3
6 0 0 2 6 4 8
9 0 4 NT 5 2 NT
10 6 11 4 6 8 5
12 2 8 8 5 5 6
15 0 5 7 5 5 5
16 0 2 4 6 6 4
19 6 7 NT 1 5 NT
20 2 4 6 6 2 6
23 3 6 4 3 6 8
24 0 7 5 8 3 6
CIMT = constraint-induced movement therapy; NT = not tested.

respectively; t = 0.83, df = 18, P = 0.418, 95% CI of mean stroke-hemisphere activation sites from baseline to
difference = –3.96, 1.72). The number of active TMS follow-up was greater in the CIMT patients than in the
stroke-hemisphere sites increased in both treatment intensive traditional therapy patients, as in our prelimi-
groups from baseline to follow-up, at which time it nary results,9 but the difference was not statistically sig-
approximated the number of active sites in the non- nificant (t = 1.20, df = 19, P = 0.241, 95% CI of mean
stroke hemisphere. The increase in the number of difference = –1.33, 4.99). In the nonstroke hemisphere,

20 Neurorehabilitation and Neural Repair 21(1); 2007

 CIMT During Early Stroke Rehabilitation

Table 6. Correlation of Motor Performance With Number of Active Transcranial Magnetic Stimulation Sites in the Stroke
Hemisphere

Motor Test Baseline Follow-up Difference Follow-up-Baseline

Fugl-Meyer upper-extremity motor scale, r

All patients 0.45* 0.47* 0.50*
CIMT 0.69* 0.59 0.69*
Traditional therapy 0.35 0.36 0.14
Grooved Pegboard Test, r
All patients 0.15 0.35 0.32
CIMT 0.47 0.70* 0.70*
Traditional therapy –0.20 0.02 –0.07
CIMT = constraint-induced movement therapy.
*P < 0.05, 2-tailed.

the number of activation sites decreased slightly from DISCUSSION

baseline to follow-up in the CIMT patients but the
observed change did not differ significantly between The main results of this study are, 1st, that long-term
treatment groups (t = 1.43, df = 19, P = 0.170, 95% CI of improvement in motor function of the affected UE did
mean difference = –4.56, 0.86). not differ significantly between patients who received
Table 6 presents exploratory correlation analyses CIMT and those who received traditional therapy at the
to examine the association between UE motor perfor- same frequency and duration, and 2nd, that motor
mance on the FM and GPT and number of stroke- improvement was related to enlargement of the area of
hemisphere sites where motor responses of the affected cortical excitability. On all measures of motor function
hand were evoked. Correlations are presented at baseline of the affected arm and hand, patients who received
and follow-up for all patients (i.e., combined over treat- CIMT showed an apparent advantageous trend over
ment groups) and for each treatment group separately. patients who received intensive traditional therapy.
As in the preliminary study,9 a positive correlation was Relative to the control group, the CIMT group reported
found in the combined sample between improvement in significantly greater improvement in quality of per-
UE motor function on the FM and increase in the forming daily activities using the affected hand. In addi-
number of active TMS sites in the affected hemisphere. tion, on one test of motor performance by the affected
FM scores correlated significantly with the number of UE (FM), they demonstrated significantly better
sites on the stroke hemisphere at which responses of the improvement over the 2-week treatment period.
affected hand were evoked by TMS before treatment and However, no other comparisons showed a statistically
at follow-up. The right-hand column of Table 6 presents significant between-group difference in motor function
the longitudinal correlations between changes in each of the affected UE, including improvement on the FM
variable, measured as the difference from baseline to from baseline to follow-up, which was the primary end-
follow-up. A significant positive longitudinal correlation point. Therefore, as an evaluation of the relative effec-
was found in the combined sample between improve- tiveness of CIMT in early stroke rehabilitation, as
ment in the FM score and increase in the number of compared to traditional interventions provided at the
stroke-hemisphere TMS activation sites. same frequency and duration, the results are overall
The same correlations were computed within each neutral but contain some encouraging trends.
treatment group to evaluate if the association between Rehabilitation therapy plays an important role after
changes in TMS activations and motor function differed stroke, yet evidence for the effectiveness of early rehabil-
as a function of treatment. It should be noted that these itation interventions including CIMT is limited. In the
results are exploratory and based on small samples. As one previous randomized controlled trial of CIMT dur-
seen in Table 6, all the within-group correlations with ing early stroke rehabilitation, Dromerick et al7 reported
the FM were in the predicted direction but were that patients who received CIMT demonstrated better
stronger in the patients who had received CIMT. Table 6 motor performance relative to a traditional therapy
also presents correlations between number of TMS group when assessed immediately after treatment.
active sites and motor performance on the GPT. No sig- Similar to our findings, they reported greater improve-
nificant correlation was found in the combined sample. ment by the CIMT group on all motor function mea-
However, the within-group results suggest that correla- sures, including motor performance and activities of
tions between GPT performance and TMS activation daily living, but the advantage was statistically significant
sites were stronger among the CIMT patients. on only 1 subscale of the Action Research Arm Test.18

Neurorehabilitation and Neural Repair 21(1); 2007 21

Boake et al

The consistency of our motor performance results and the relationship between spontaneous motor recovery
their similarity with previous findings support the inter- and ability to elicit motor-evoked potentials (MEPs). An
pretation that the present study was underpowered to additional and unexpected finding was that the associa-
detect a relative therapeutic advantage of CIMT as com- tion between increased TMS activations and motor
pared to a control intervention provided at equal fre- recovery tended to be stronger in patients who had
quency and duration. On the basis of the unusually received CIMT, relative to the traditional therapy group.
positive results observed in the preliminary phase of this Although this finding might suggest that the association
study,8,9 we probably overestimated the effect size of was facilitated by CIMT, this interpretation is prelimi-
CIMT relative to traditional therapy and projected a nary because the analyses were exploratory and based
sample size for the present study that was too small. In on small samples. In addition, the significant correla-
retrospect, it appears that the projected between-group tion between FM performance and TMS activation sites
difference of 14.5 points on the FM was too optimistic at baseline suggests that the association may have been
and that the study was not adequately powered to detect preexisting rather than a result of therapy.
a smaller but more plausible treatment effect. The differ- Our TMS results are consistent with those of previous
ential effect of CIMT on motor improvement observed longitudinal studies19,20 of motor recovery in chronic
in the preliminary study, with statistically significant patients who were assessed beginning 2 or more months
advantages for the CIMT group on most measures, was after stroke. Recently, Platz et al,21 in a trial of stroke reha-
not confirmed in the completed study. Although greater bilitation, demonstrated that medial shift of the motor
proximal weakness was observed in the preliminary map was associated with motor improvement. A positive
study sample, this discrepancy could have been due to association between poststroke motor function and TMS
random differences between small groups. activation area has also been reported in cross-sectional
Unfortunately, recruitment of a sample size adequate studies.22 These longitudinal results indicate that TMS
to detect a small to medium effect size was not possible map-size enlargement signals at least one mechanism
at our single center. The main limitation on recruitment through which cerebral reorganization participates in
was that our entry criteria excluded most patients admit- motor recovery. Interpretation of map-size enlargement
ted to our stroke unit. Ro et al,9 in the preliminary phase is problematic because changes in map size reflect alter-
of this study, reported that fewer than 5% of stroke ation in motor cortical output excitability and can be
admissions had qualified for randomization and deter- explained by different mechanisms. It has been suggested
mined that the major exclusion was for motor impair- that TMS motor map enlargement could be due either to
ment less severe than allowed by the NIHSS arm motor participation of additional neurons in muscle activation
score criterion. Our exclusion of patients without or to increased excitability of the motor representation by
detectable shoulder weakness (i.e., NIHSS arm motor stimulation of the surrounding cortical area.23 Our
score = 0) was based on the expectation that such results do not clearly support either explanation, and it
patients would recover full UE motor function without remains unclear whether stroke rehabilitation studies can
intensive rehabilitation. However, the UE motor perfor- distinguish between these mechanisms, for example, by
mance test results at 3 months after stroke (Table 2) administering TMS before and after therapy sessions.20
demonstrate that most study patients achieved less than Explanations of TMS map enlargement during stroke
full recovery and suggest that less impaired patients recovery need to account for use-dependent plasticity, as
could also have benefited from therapy. shown by the fact that map enlargement can be produced
This study extends previous findings demonstrating by rehabilitation interventions that are instituted more
an association between TMS motor map changes and than 6 months after stroke.24-27
motor recovery during the initial months after stroke. The experimental design of our study is relevant to
The longitudinal study design, with assessment of methodological issues that are increasingly recognized in
motor performance and TMS activation area at the rehabilitation research. First, our study and that of
same time-points before and after treatment, revealed a Dromerick et al7 were designed so that treatment groups
positive relationship between extent of hand motor were balanced for frequency and duration of interven-
recovery, as measured by the FM, and the number of tion. The rationale for this design was to prevent con-
scalp positions over the stroke hemisphere at which founding type of intervention with time spent in
motor responses of the affected hand could be evoked. therapy, as would have occurred if intervention had been
An important consideration in interpreting this finding provided to the control patients for fewer hours or days
is that because the proportion of patients exhibiting than to the CIMT group. In that case, an advantage of
evoked motor responses in the stroke hemisphere CIMT over traditional therapy could have been attrib-
increased dramatically over the study period, the appar- uted either to a different type of therapy or to greater
ent association between improved motor function and time spent in intervention. As a consequence of balanc-
increased map size could have come about because of ing frequency and duration of intervention between

22 Neurorehabilitation and Neural Repair 21(1); 2007

 CIMT During Early Stroke Rehabilitation

treatment groups, our design compared CIMT to a con- used in earlier clinical trials with chronic stroke
trol condition that received more hours of therapy than patients.1,2 Recent clinical trials involving chronic stroke
is typically provided. Therefore, the results of this study patients suggest that CIMT remains effective when pro-
should not be interpreted as a comparison between vided less than the daily regimen of 6 h used in earlier
CIMT and traditional UE therapy when provided at the studies.4,6 The 2-week treatment duration used in this
customary frequency of up to 1 or 2 h per day. Greater study, which is identical to most previous clinical trials
experimental control could have been achieved by inclu- of CIMT, could not be completed during the in-patient
sion of a 2nd control group that received traditional rehabilitation stay and required that patients return to
therapy at the customary frequency, or by a factorial complete the therapy sessions on an out-patient basis.
design investigating the interaction of therapy type and Limitations of our study, in addition to small sample
hours of therapy time. However, the necessary increase size and design issues discussed above, include attrition
in recruitment would not have been feasible at our sin- and incomplete motor data on a subset of patients. This
gle center. A 2nd issue in our study design is the com- speaks to the difficulty in accurately obtaining a large
parison of an experimental treatment such as CIMT to a amount of motor data in patients who are often unco-
control group receiving standard or traditional therapy. operative or unable to carry out many of the functions
If the standard therapy is not provided according to an tested and who are dependent on other family
explicit protocol and does not have its own evidence members, friends, or health care workers for arranging
base, then the treatment condition is less easily repli- their long-term care, transportation, and living arrange-
cated or compared among centers.28 Measures of therapy ments. This makes planning of follow-up visits prob-
process (e.g., time spent performing activities with the lematic. A 2nd limitation, which is especially salient in
affected UE, compliance in mitten wear) could improve the acute stroke population, is the exclusion of a high
specification of treatment conditions and improve com- proportion of patients because their motor function
parability between studies. Finally, a methodological was too severely impaired to participate in training with
issue in trials conducted during the subacute stage after the affected limb or because motor function had recov-
stroke is that effects of therapy are confounded with ered to a level above the range specified for CIMT. A 3rd
spontaneous changes in motor function and brain orga- limitation is that the study could have been powered to
nization. Spontaneous recovery probably occurred to a detect an improvement in motor function that was
greater extent in our patients who were studied and defined as clinically important, although this remains to
treated fairly soon after stroke onset, as compared to be defined for the motor tests used in this study. Fourth,
studies in which patients were treated in the chronic accuracy of TMS motor mapping may have been
stage. Although spontaneous recovery should have decreased by using a polyphasic stimulator, by record-
occurred equally between the randomized groups, such ing from the APB, by defining the motor threshold on
recovery could dilute a treatment effect so that further the basis of only 5 trials, by using visual inspection, and
improvement due to intervention would be limited by not controlling the order of TMS and behavioral
because of ongoing recovery. An additional implication motor testing. Although visual inspection could have
is that our finding of an association between motor been prone to larger errors of measurement, particu-
improvement and increased TMS activation sites may be larly because TMS-evoked responses in stroke patients
largely due to spontaneous recovery and that it may not are likely to be smaller and less visible, our preliminary
be possible to detect a supplementary effect of treatment study9 found that correspondence between visual
on this association. inspection and EMG was high. Moreover, inaccuracy of
Our study and previous ones7,9 demonstrate that TMS maps would not explain the increase in TMS map
implementation of a modified form of CIMT during size over the study period or its association with motor
early stroke rehabilitation is safe and feasible. No signif- improvement. Although sampling at 1 kHz leaves open
icant adverse events occurred during the treatment the possibility of aliasing error, this issue probably did
period, and no loss of motor function caused by not affect the results, because only noise is likely to be
restraint of the unaffected hand could be detected. affected by the aliasing and MEP signal power is mostly
Therefore, available evidence in human stroke patients in lower frequencies. Finally, although there are poten-
undergoing CIMT does not bear out the potential tial limitations to replacing missing data at follow-up by
adverse consequences of early overutilization of the carrying forward the posttreatment observation, this
affected limb as seen in animal studies.29 It is possible procedure allowed an intent-to-treat analysis and would
that earlier restraint placement29 and other procedural not be expected to introduce bias.
differences in animal studies account for the discrepant The completion of this study without adverse events
results. The frequency of 3 h of UE therapy per day as combined with the encouraging trends provides a ratio-
provided in our study is greater than the 2 h per day pro- nale for further evaluation of CIMT for subacute stroke.
vided by Dromerick et al7 but less than the 6 h per day However, the difficulties in carrying out this study argue

Neurorehabilitation and Neural Repair 21(1); 2007 23

Boake et al

for refinement of eligibility criteria and the need for 10. Goldstein LB, Bertels C, Davis JN. Interrater reliability of the
a multicentered approach to increase sample size. NIH Stroke Scale. Arch Neurol 1989;46:660-2.
11. Fugl-Meyer AR, Jaasko L, Leyman I, et al. The post-stroke hemi-
Recruitment and retention of the necessary sample size plegic patient. 1. A method for evaluation of physical perfor-
are likely to be serious obstacles to future therapy trials mance. Scand J Rehabil Med 1975;7:13-31.
for subacute stroke, particularly if the eligibility criteria 12. Kløve H. Clinical neuropsychology. Med Clin North Am 1963;
are restrictive and if enrollment is limited to a single 47:1647-58.
center. Inclusion in the experimental design of an addi- 13. Uswatte G, Taub E, Morris D, et al. Reliability and validity of the
upper-extremity Motor Activity Log-14 for measuring real-
tional control group, given conventional therapy at the world arm use. Stroke 2005;36:2493-6.
typical number of hours per day, is needed to distin- 14. Rossini PM, Barker AT, Berardelli A, et al. Non-invasive electrical
guish the effects of type of intervention from the fre- and magnetic stimulation of the brain, spinal cord and roots: basic
quency and duration of therapy. Potential predictors of principles and procedures for routine clinical application. Report
therapeutic response, including baseline motor status, of an IFCN committee. Electroencephalogr Clin Neurophysiol
1994;91:79-92.
need to be investigated prospectively. Functional brain 15. Ro T, Cheifet S, Ingle H, et al. Localization of the human frontal
imaging of stroke patients in randomized controlled eye fields and motor hand area with transcranial magnetic stim-
trials of rehabilitation therapies could improve under- ulation and magnetic resonance imaging. Neuropsychologia 1999;
standing of neurophysiologic mechanisms underlying 37:225-31.
recovery and therapeutic improvement. 16. Taub E. Somatosensory deafferentation research with monkeys:
implications for rehabilitation medicine. In: Ince LP, ed. Behavioral
psychology in rehabilitation medicine. Baltimore: Williams &
Wilkins; 1980:371-401.
ACKNOWLEDGMENTS 17. Taub E, Crago JE, Burgio LD, et al. An operant approach to reha-
bilitation medicine: overcoming learned nonuse by shaping. J
This work was supported by National Institutes Exp Anal Behav 1994;61:281-93.
18. Lyle RC. A performance test for assessment of upper limb func-
of Health grants R24HD39629 (Western Medical tion in physical rehabilitation treatment and research. Int J
Rehabilitation Research Network; Principal Investigator: Rehabil Res 1981;4:483-92.
Marjorie E. Anderson) and P50NS044227 (University of 19. Traversa R, Cicinelli P, Bassi A, et al. Mapping of motor cortical
Texas Specialized Program in Acute Stroke; Principal reorganization after stroke. A brain stimulation study with focal
Investigator: James C. Grotta). We are grateful to Mary magnetic pulses. Stroke 1997;28:110-7.
20. Koski L, Mernar TJ, Dobkin BH. Immediate and long-term
Nowak for help in manuscript preparation. changes in corticomotor output in response to rehabilitation:
correlation with functional improvements in chronic stroke.
Neurorehabil Neural Repair 2004;18:230-49.
REFERENCES 21. Platz T, van Kaick S, Moller L, et al. Impairment-oriented train-
ing and adaptive motor cortex reorganisation after stroke: a
1. Taub E, Miller NE, Novack TA, et al. Technique to improve fTMS study. J Neurol 2005; 252:1363-1371.
chronic motor deficits after stroke. Arch Phys Med Rehabil 1993; 22. Bastings EP, Greenberg JP, Good DC. Hand motor recovery after
74:347-54. stroke: a transcranial magnetic stimulation mapping study of
2. Taub E, Uswatte G, Pidikiti R. Constraint-induced movement motor output areas and their relation to functional status.
therapy: a new family of techniques with broad application to Neurorehabil Neural Repair 2002;16:275-82.
physical rehabilitation—a clinical review. J Rehabil Res Dev 1999; 23. Wassermann EM, McShane LM, Hallett M, et al. Noninvasive
36:237-51. mapping of muscle representations in human motor cortex.
3. van der Lee JH, Wagenaar RC, Lankhorst GJ, et al. Forced use of Electroencephalogr Clin Neurophysiol 1992;85:1-8.
the upper extremity in chronic stroke patients: results from a sin- 24. Kopp B, Kunkel A, Muhlnickel W, et al. Plasticity in the motor
gle-blind randomized clinical trial. Stroke 1999;30:2369-75. system related to therapy-induced improvement of movement
4. Sterr A, Elbert T, Berthold I, et al. Longer versus shorter daily after stroke. Neuroreport 1999;10:807-10.
constraint-induced movement therapy of chronic hemiparesis: 25. Liepert J, Uhde I, Graf S, et al. Motor cortex plasticity during
an exploratory study. Arch Phys Med Rehabil 2002;83:1374-7. forced-use therapy in stroke patients: a preliminary study. J
5. Alberts JL, Butler AJ, Wolf SL. The effects of constraint-induced Neurol 2001;248:315-21.
therapy on precision grip: a preliminary study. Neurorehabil 26. Schaechter JD, Kraft E, Hilliard TS, et al. Motor recovery and cor-
Neural Repair 2004;18:250-8. tical reorganization after constraint-induced movement therapy
6. Page SJ, Levine P, Leonard AC. Modified constraint-induced in stroke patients: a preliminary study. Neurorehabil Neural
therapy in acute stroke: a randomized controlled pilot study. Repair 2002;16:326-38.
Neurorehabil Neural Repair 2005;19:27-32. 27. Kim YH, Park JW, Ko MH, et al. Plastic changes of motor
7. Dromerick AW, Edwards DF, Hahn M. Does the application of network after constraint-induced movement therapy. Yonsei Med
constraint-induced movement therapy during acute rehabilita- J 2004;45:241-6.
tion reduce arm impairment after ischemic stroke? Stroke 2000; 28. Weinrich M, Good DC, Reding M, et al. Timing, intensity, and
31:2984-8. duration of rehabilitation for hip fracture and stroke: report of a
8. Grotta JC, Noser EA, Ro T, et al. Constraint-induced movement workshop at the National Center for Medical Rehabilitation
therapy. Stroke 2004;35:2699-701. Research. Neurorehabil Neural Repair 2004;18:12-28.
9. Ro T, Noser E, Boake C, et al. Functional reorganization and 29. Bland ST, Schallert T, Strong R, et al. Early exclusive use of the
recovery after constraint induced movement therapy in subacute affected forelimb after moderate transient focal ischemia in rats:
stroke: case reports. Neurocase 2006;12:50-60. functional and anatomic outcome. Stroke 2000;31:1144-52.

24 Neurorehabilitation and Neural Repair 21(1); 2007