Historic Basis of Constraint-Induced Therapy
The historic basis behind CI therapy emerged from more than seven decades of research. Tower (27
) noted in 1940 that after unilateral lesions of the pyramidal tracts, monkeys would fail to use the affected limb; however, this use could be improved again with restraint of the unaffected limb. Taub (28
) summarized several studies that explored the effects of the surgical abolition of somatic sensation in the forelimb of monkeys. In these studies, monkeys often quit using the insensate limb after multiple unsuccessful attempts. Consequently, to accomplish basic functional tasks, the monkeys learned compensatory techniques with the nonaffected arm, a phenomenon Taub and colleagues (28
) referred to as learned nonuse. Taub and colleagues observed that successfully developing these compensatory strategies reinforced the nonuse of the affected limb. Thus, animals never realized the possible functional potential of that limb. However, they later discovered that learned nonuse could actually be reversed when the uninvolved UE of deafferented, adolescent monkeys was restrained for 3 days. Furthermore, when the restraint was maintained for a period of 1 to 2 weeks, this return in function could be long lasting.
These fundamental observations have led to numerous studies and reviews that either describe or evaluate the application of CI therapy in humans. These studies are summarized in and include the number and characteristics of participants, intervention description, follow-up time period, outcome measures, and commentary. Ostendorf and Wolf (30
) demonstrated the first application of this approach to humans in a single-subject pilot study. Their approach of exclusively implementing forced use of the hemiplegic UE was then applied in a larger study (21
) involving patients with chronic strokes and traumatic brain injury. The unaffected UE was restrained during waking hours over a 2-week period. Patients performed the protocol at home, with supervision from caregivers. UE improvements in timed or force-generating activities were noted in 19 of 21 functional tasks, and improvements persisted up to 1 year after the intervention.
Summary of Constraint-Induced Therapy Articles
Taub et al (24
) continued this work, adding 6-hour supervised practice sessions for 10 of the 14 days while patients wore a restraint. Four patients underwent unaffected UE restraint, and five subjects were assigned to the comparison group. The restraint was worn 90% of waking hours and only removed during water-based activities (i.e., showering and toileting), naps, and activities where balance might be compromised. The comparison group was given two sessions of passive movement of the affected extremity and instructions for self-range of motion exercises and told to focus attention on using the affected extremity. An earlier version of the Wolf Motor Function Test (WMFT) called the Emory Test and the Arm Motor Activity Test (AMAT) were administered to objectively evaluate UE motor function. Mean performance times decreased significantly on the two motor ability tests for the restraint group compared with the control group. Improvements in strength measures within the Emory Test were noted within the experimental group as well as quality of movement. The Motor Activity Log (MAL) was administered to assess the patients’ subjective impression of how well and how often movement is observed in the affected arm during basic activities of daily living (ADL) before, during, and after the intervention. On this 6-point scale, patients rate their own performance from 0 (cannot use the limb for that activity) to 5 (using the limb as well/as much as before the stroke). A score of 3 is the minimum value that indicates the patient is functionally able to use the limb to complete the task without use of the stronger limb. Statistically significant improvements were made only in the experimental group and were maintained 2 years after intervention. These subjects did exhibit a pronounced improvement in actual attempts to use the limb in all of the activities, a mean increase of 97% compared with pretreatment scores. In unstructured post-interviews, the restraint patients indicated a greatly expanded range of activities in their real-life situation, with two patients gaining the ability to write with their affected UE, an activity that was not previously attempted. Minor adverse events from training, described as typical overuse symptoms of stiffness and discomfort in the affected extremity, occurred in three of the four patients.
Expanding on previous work, Taub and coworkers published a series of case studies (23
) involving a total of 27 patients. Six different modified treatment protocols were described, which included the following: (1) use of a resting splint/sling to restrain the less-affected arm and supervised task practice with the more-affected arm; (2) no restraint (control)—attention group; (3 sling restraint of the less-affected arm and shaping (operant conditioning) of movements of the more-affected arm; (4) half-glove on the less-affected hand as device to remind the subject to avoid using this hand while performing shaping of movements of more-affected arm; (5) no restraint, only shaping of the more-affected arm; and (6) intensive training 6 hours per day for 10 days (no glove or shaping but intensive physical therapy, including aquatic therapy, neurophysiological facilitation, and task practice).
In this context, the term shaping
refers to a “specific behavioral training technique in which a desired motor or behavioral objective is approached in small steps, by successive approximations” (32, p. 40). Specific values for outcome measures on these additional patients were not provided. However, descriptively, the results interpreted from the MAL and WMFT changes suggested that the sling and shaping group and the half-glove and shaping group improved as well as the original group of restraint and supervised task practice. The shaping-only group improved as well as these groups, but the “transfer of improvement to the life situation, while substantial, was not as great as in the constraint groups (32, p. 51).” The intensive therapy group, receiving conventional therapy for 6 hours per day over a period of 10 days, improved just as much as the patients in the sling plus shaping group. From these results, the relative importance of restraint (in combination with intensive training) as a basis for improved function is unclear, but these data suggest that constraint of the less-affected limb assists in improving subsequent use of that limb during daily activities. Only the nine original participants from the first article in the series (24
) were randomized. Although in the original study (24
) and again in a later article (23
), Taub and colleagues noted that patients underwent 14 days of restraint and 10 days of training, the other two studies in the series (31
) suggest that patients received only 12 days of restraint and 8 days of training. From these conflicting reports, it is unclear which treatment interval was used for these groups.
Miltner et al (25
) evaluated the effects of restraint with a resting hand splint for 12 days while using shaping techniques 7 hours on each of 8 days in 15 patients. Improvements were expressed as effect sizes rather than as percentages. The effect sizes on the WMFT and MAL were described as “extremely large by standards of the field (p. 25).” However, as in previous studies (24
), the quality of movement portion of a primary outcome measure, the MAL, was administered each day of the treatment phase as well as before and after intervention. Familiarizing patients with the MAL questions on a daily basis may permit recollection of previous responses, thereby narrowing the choice points on the MAL, thus effectively reducing the variability and enhancing the effect size. Additionally, the possible confounding influences of the Hawthorne effect (i.e., effects from repeated testing) cannot be dismissed using this method of outcome measurement administration. The WMFT analysis showed significant improvements in each measure from pretreatment to posttreatment but no significant differences from baseline to pretreatment or posttreatment to follow-up. Follow-up data represented only 12 of the 15 patients for the MAL and 9 patients for the WMFT.
Kunkel et al (26
) studied five patients who wore a restraint for 14 days while performing shaping procedures for 10 of those days. Improvements were seen in the MAL, the Actual Amount of Use Test (AAUT), the Arm Motor Ability Test (AMAT), and WMFT outcome measures. The MAL score improvements indicate that patients progressed from using the affected limb as merely an assist to using the limb to complete a task independently. Although the AAUT represents an objective and unobtrusive measure of motor function of the UE, actual reliability and validity levels of this test were not provided. For the WMFT and AMAT test results, the average improvement scores were 42% and 19%, respectively. Large effect sizes were noted for each of these four tests as well. Small sample size and absence of a comparison group limit the conclusiveness of these results. No data are offered regarding baseline measurement comparisons so that patient homogeneity could be determined; consequently, potential effects of intersubject variability are not addressed.
Vander Lee et al (22
) have performed the largest controlled CI therapy trial to date, comparing CI therapy to treatment emphasizing bimanual activities. Their results showed that 1 week after the last treatment session, the change within the CI therapy group was significantly different compared with the group undergoing bimanual training for Action Research Arm test (ARA) and the MAL amount of use score. Other outcome measures (MAL quality of movement score, Rehabilitation Activities Profile, and the UE section of the Fugl-Myer Assessment scale) did not improve significantly. Only the ARA scores showed persistent change at 1-year follow-up. The greater treatment effect measured by the ARA was clinically relevant for CI therapy patients with sensory disorders, and the greater treatment effect measured by the MAL amount of use scores was relevant for CI therapy patients with hemi-neglect. Differences in kind and intensity of CI therapy intervention in this study compared with previous studies (23
) may be responsible for these results. The experimental patients were treated in groups of four by one or two therapists, and no intensive shaping techniques were used. Except for the MAL, different outcome measures were used, also limiting comparisons to previous studies.
Preliminary studies using CI therapy have also addressed patients with acute and subacute stroke. Dromerick et al (33
) evaluated the effects of the CI therapy protocol in a randomized controlled trial of 20 patients with stroke less than 3 months from the ictus. Their results showed the CI treatment group (CI therapy complementing regular treatment) attained significantly higher scores on total ARA and pinch subscale scores compared with the control group who received regular treatment only, but with control over total treatment duration for both groups. The differences in the mean ARA grip, grasp, and gross movement subscale scores did not reach statistical significance. No difference was noted between groups for UE ADL performance measures (Barthel and Functional Independence Measure). Most disability measures do not distinguish unilateral from bilateral ADL tasks; consequently, a lack of significance for the ADL measures may be attributable to the insensitivity of these measures to detect change in impaired UE function. This study illustrates the feasibility of applying a modified CI therapy protocol to the acute stroke population. Different outcome measures evaluating change and function compromise comparisons to other studies in the chronic stroke population.
Blanton and Wolf (34
) reported one of the first applications of CI therapy for a patient with subacute stroke. This patient improved both MAL and WMFT scores, and these improvements were still present 3 months later. The CI therapy protocol used with this patient is similar to the intervention used in studies conducted by Taub and colleagues (23
) with half-glove restraint and intensive task practice for 6 hours per day for 10 days. Although limited in relevance because of a case study format, the results from this case study support additional investigation of CI therapy for this patient population.
Most recently, Page et al (35
) evaluated a modified CI therapy protocol in a randomized controlled trial of six subacute patients (between 3 and 9 months). Two patients were placed in each of the three groups: CI therapy (restraint of unaffected arm 5 days per week during 5 hours identified as times of frequent use in addition to 30-minute physical and occupational therapy sessions three times a week for 10 weeks), regular therapy (30-minute physical and occupational therapy sessions three times a week for 10 weeks), and no therapy. Substantial improvements were found for the CI therapy group in the Fugl-Myer Assessment of Motor Recovery, ARA, WMFT, and MAL. No specific statistical analysis was done on the small sample size, and comparisons with other studies are limited because of the use of a different, modified CI therapy protocol. These results are encouraging for the application of CI therapy in the subacute stroke population, but an evaluation of a much larger number of patients is needed.
In summary, CI therapy holds substantial potential to assist patients with stroke to improve UE function. Although multiple articles describe the benefits of CI therapy, conclusive evidence is limited because of the lack of any large randomized clinical trial (RCT). Presently, most of the evidence is presented in uncontrolled patient series and four small RCTs. Larger, controlled studies with uniform CI therapy intervention and measurement protocols will help to define more clearly the benefits of this intervention.
Is There an Appropriate Patient Population for Constraint-Induced Therapy?
Despite evidence that CI therapy may provide an avenue for significant functional improvements in the hemiparetic limb, not every patient recovering from a neurologic injury has benefited from this treatment approach. Although Taub et al (36
) have suggested that CI therapy may be applicable to greater than 75% of the stroke population with chronic unilateral motor deficits, the validity of this possibility has yet to be demonstrated. The primary criteria for patient eligibility for a CI therapy intervention include adequate balance and safety while wearing the restraint and the ability to initiate at least 20 degrees of wrist extension and at least 10 degrees of extension at two digits in addition to the first digit of the affected hand. These criteria are derived from electromyographic biofeedback studies performed by Wolf and Binder-Macleod (37
) that indicated voluntary movements of finger and wrist extension were a better predictor of future acquisition of independent limb use than was the ability to reduce the hyperactive responses from stretching UE flexor muscle groups. The number of patients that actually achieve this level of functional recovery in the upper limb after a stroke is estimated to be approximately 20% to 25% of the population (37
). These criteria would seem to constitute important prerequisites to consider for reacquisition of important function using the CI therapy paradigm or any other treatment approach. Because most UE use involves grasping and manipulating objects, the preparatory movements of wrist and finger extension of the involved limb are essential to achieving at least some success within the patient’s ADLs without requiring assistance from the other limb or another person.
With such specific criteria to consider, neurorehabilitationists must realize that the generalizability of the CI therapy literature is limited to patients with the minimal movement criteria. Consideration must also be given to the need for solid family support, good cognitive skills, and emerging movement in the hand as well as other UE joints.
Elements of Intensive Training Associated With the Constraint-Induced Therapy Training
The two approaches used in CI therapy training include general task practice and “shaping” (adapted task practice). General task practice refers simply to the practice of a full functional task that may have multiple steps for completion (i.e., eating a meal, making coffee, or finding and dialing phone numbers from a city directory).
Shaping or adapted task practice is defined as a method in which a motor or behavioral objective is approached in small steps by successive approximations or by making the task more difficult in accordance with the patient’s motoric capabilities. In using a telephone, for example, one would practice grasping the telephone and holding the receiver before practicing pushing the telephone buttons. The task selection is based on the patient’s specific movement deficits and preference. Specific knowledge of results about a patient’s performance is given after each trial or practice session. Generally, this feedback is in terms of number of repetitions per unit of time or time required to perform a set task.
) and Schmidt (39
) note that part-task training can be an effective way to retrain some activities if they can be naturally divided into units that reflect their inherent goals. Morris and Taub (40
) note that shaping is unique because patients are given explicit feedback concerning even small improvements in performance. Activities are chosen that can be easily broken down into subtasks that are objectively measured, and then these subtasks are repeated for a specific number of times (i.e., at least 10 trials). Thus, with CI therapy, the amount and frequency of feedback and knowledge of results is increased, as is the typical number of repetitions for a given task. Consequently, the intensity of training for any given task is greater, based on the assumption that repetitively engaging relevant neural substrates will achieve greater plastic changes within the CNS.
Mapping Clinically Relevant Cortical Responses to Constraint-Induced Therapy
For any therapeutic intervention, there is a need to understand the mechanisms contributing to its effectiveness. Several methods have been used to map cortical changes in animals and humans after lesions. The technique most often used in nonhuman primates has been neurophysiological mapping; in humans, positron emission tomography (PET), functional magnetic resonance imaging (fMRI), and trans-cranial magnetic stimulation (TMS) have been used most often.
Changes in primary somatosensory or primary motor cortex have been observed with specific interventions or manipulations (49
). The results in monkeys demonstrating the effects of rehabilitative training after stroke using restraint of the unimpaired arm are very similar to those found after CI therapy (52
). Yet repetitive motor activity alone does not seem to produce functional reorganization of cortical maps (53
), and it seems that motor skill acquisition, or motor learning, is a prerequisite factor in driving representational plasticity in M1 of the adult squirrel monkey (46
Over the last 10 years, fMRI has been used to enhance the understanding of the underlying mechanisms of motor control during motor learning (55
) and motor functional recovery after stroke using fMRI (57
) and PET (62
). However, few functional imaging experiments have investigated the effects of rehabilitative interventions (65
), specifically CI therapy, on cortical reorganization after stroke.
At this time, there is only a single functional imaging study that attempts to evaluate the effect of upper limb CI therapy on cortical motor reorganization (66
). Qualitative assessment of the images showed larger areas of activations bilaterally and in the peri-infarct area after 2 weeks of CI therapy. The study was hampered by design and methodological problems, leaving definitive conclusions of cortical reorganization after CI therapy problematic.
Continued use of these neuroimaging techniques will advance our understanding of the mechanisms of CI therapy. To date, the one approach in humans that has received the most attention in investigating the mechanisms of cortical change after CI therapy is transcranial magnetic stimulation.
Basic Overview of Principles Underlying Transcranial Magnetic Stimulation
Since its introduction by Barker in 1985, TMS has gained recognition as a safe, relatively painless and noninvasive method for mapping cortical motor representation areas (67
). Recently, TMS has been used to investigate the possible mechanisms underlying both spontaneous and therapy-induced poststroke motor recovery.
TMS is based on the principle of electromagnetic induction. Electrical current is directed through a hand-held copper-stimulating coil, with the consequent production of a transient magnetic field. When held over the scalp, the rapidly changing magnetic field is able to induce a small electrical current in underlying brain tissue, without significant attenuation. When performed over the primary motor cortex at low stimulus intensities, TMS is thought to stimulate the corticospinal tract indirectly (trans-synaptically) via horizontal fiber depolarization (67
). The resultant efferent volleys can then be recorded as motor evoked potentials (MEPs) via surface or indwelling electrodes at peripheral target muscles.
Measured TMS variables often include the location of the “hot spot” (the most active scalp position for the target muscle), the excitability threshold (measured at the hot spot), the area of motor output representation, the MEP latency, the amplitude-weighted center of gravity, the MEP amplitudes (at rest and sometimes with facilitation), and the MEP recruitment curves (73
). A thorough review of the physics and measurement characteristics of TMS can been found elsewhere (67
Response to Repetitive Task Practice
Recent work with animal models has suggested that the specificity and difficulty of training may impact the extent of use-dependent cortical plasticity (53
). Similar findings have been reported in motor recovery in human subjects after stroke. Leipert et al (86
) examined the effect of one intensive session of physical therapy in nine subjects 4 to 8 weeks after stroke. Subjects received 1.5 hours of manual dexterity exercises in addition to ongoing standard therapy. TMS mapping of the abductor pollicis brevis (APB) representation was performed 1 week before, immediately before, immediately after, and 1 day after the training session. Measures of motor output area, excitability threshold at the APB hot spot, and center of gravity for the APB muscle (location at which an evoked muscle response greater than 50 μ
V in amplitude is seen at minimal stimulus intensity) of the AH and unaffected hemispheres (UH) did not significantly change between the two pretraining measures, indicating that no significant changes occurred because of spontaneous recovery or nonspecific training. AH area increased significantly immediately after training but then decreased toward baseline after 1 day. Increased AH motor output area was associated with improved dexterity on a clinical measure (the Nine Hole Peg Test) in seven of the subjects, although the amount of clinical improvement did not correlate with the extent of change in area. The excitability threshold at the hot spot and the center of gravity was unchanged after training, possibly signifying that the AH area increases were attributable to increased excitability at the edges of the map. The rapid change detected in the TMS-derived maps after brief training epochs suggests that functional, rather than structural, mechanisms were involved. Potential mechanisms discussed include the modulation of inhibitory GABA-ergic transmission at the borders of the motor map and alteration in glutamate transmission (86
). Classen and coworkers (87
) have suggested that the “motor cortex builds up, and then loses, in a short time, memory traces of movements retaining the subject’s recent history of performance.” They (86
) postulate that either the short duration of a training effect is a function of the short duration of the training (and, hence on a continuum with long-term storage) or, alternatively, that the primary cortex subserves short-term plasticity, with long-term procedural memory requiring involvement of other cortical and subcortical structures.
Transcranial Magnetic Stimulation Mapping in Constraint-Induced Therapy
Recent studies have used TMS motor mapping to investigate the effect of CI therapy for the more-affected UE. Liepert et al (86
) used focal TMS to construct cortical output maps to the APB in six chronic stroke patients before and after 10 days of CI therapy. As noted in prior studies of poststroke subjects, significantly higher motor thresholds, smaller amplitudes, and a smaller area of excitable cortex were observed in the AH. After CI therapy, TMS showed no change in thresholds but significant increases in MEP amplitude and APB motor output area in the AH, possibly indicating increased excitability of surrounding neuronal networks. The UH output areas were smaller after the training period, perhaps because of decreased use of the less-affected UE, normalization of the UH APB representation, or increased transcallosal inhibition of the UH by the AH. Center of gravity shifts were significant (in the mediolateral axis) only for the AH, suggesting possible recruitment of adjacent areas along the motor cortex. All subjects improved significantly in their use of the affected extremity, but scores on the motor activity log (MAL) did not correlate to degree of map change. The Leipert group suggests that “physiotherapy induces use-dependent reorganization which supports recovery-associated plastic changes.” (86, p. 321)
In another study (77
), clinical (MAL) and TMS measures were made at multiple time points before and after CI therapy in 13 chronic stroke patients. Neither baseline measure showed appreciable change at 2 weeks and 1 day before CI therapy, suggesting little spontaneous recovery and good test-retest reliability. Again, the AH showed a smaller APB representation area at baseline, with a near doubling of the area after CI therapy.
Motor activity improvements (MAL) were maintained at the later measurement points. However, a return toward baseline AH APB representation area was seen at the 4-week and 6-month TMS sessions, indicating a possible “normalization after therapy-induced hyperexcitability” via improved synaptic efficiency or the relegation of motor function to TMS-inaccessible regions.
Considerations Regarding Somatotopic Organization
Some physiologists believe that the motor cortex represents movements rather than individual muscles (87
). Mosaic patterns of movement representations have been found via intracortical mapping studies in nonhuman primates (54
). Although many studies involving TMS report motor output maps to a single muscle as a confluent area of excitability, the topography of the human motor cortex also seems to be complex (69
). Wasserman et al (97
) found significant overlap in the motor output areas for four different UE muscles in healthy subjects, although there was a clear somatotopic organization of the hot spots for these muscles. Similarly, Mortifee et al (76
) identified considerable overlap in the motor output areas for the abductor pollicis brevis and abductor digiti minimi in a group of healthy controls. The hot spots for the two muscles were also discretely positioned.
The extent to which TMS map changes and clinical improvements are attributable to changes in the cortical output to the individual target muscle or changes in activity of multiple agonists and antagonist muscles remains to be elucidated. In the future, intrasubject comparison of TMS-derived motor maps with those achieved through intracortical microstimulation may help to clarify this issue. Multichannel needle electromyography of multiple agonist and antagonist muscles may also yield important information. In this regard, Rossini and Pauri (43
) have demonstrated the ability to map up to 12 UE muscles concurrently and display a distributed motor network along the precentral cortical area. This kind of display represents our best alternative to simulating mapping procedures from intracortical stimulation studies in nonhuman primates and may provide the basis for future detailed intrasubject examination of cortical reorganization in response to any intervention.