Modified Constraint-Induced Movement Therapy for Elderly Clients With Subacute Stroke

Design

This study used an interrupted time series (ITS) design to determine whether CIMT had an effect greater than that of natural recovery on the participation, activity, and impairment of the hemiplegic arm and hand in elderly people after a stroke. ITS designs are characterized by the collection of multiple observations over time that are “interrupted” by an intervention or treatment (Cook & Campbell, 1979). An ITS design enables the researcher to quantify a stable baseline level of function or a growth trend before treatment and then project that trend into the future. The effect of treatment is manifested as the difference between the actual and projected outcomes (Johnston, Ottenbacher, & Reichardt, 1995). This study followed the Effective Practice and Organization of Care Cochrane definition of an ITS design, which requires that there be at least three time points before and after the intervention, that the intervention occur at a clearly defined point in time, and that the study use objective performance or outcome measures (Ramsay, Matowe, Grilli, Grimshaw, & Thomas, 2003).

Participants

Participants were recruited from inpatient and outpatient rehabilitation units at two Toronto-based rehabilitation hospitals. They were eligible to participate if they were ≥65 yr old, had functional written and spoken English, had ischemic or hemorrhagic stroke confirmed by computed tomography or magnetic resonance imaging, and were able to complete at least one of the three upper-extremity tasks at Stage 3 (with 7 being normal movement) on the Chedoke–McMaster Stroke Assessment Impairment Inventory (CMSA–II; Gowland et al., 1995). In addition, they needed to be willing and able to participate in 2 hr of daily therapy (Monday to Friday) for 2 wk and to wear a mitt on the unaffected hand for 6 hr per day. Participants were excluded if they exhibited rapidly improving motor function as identified by the unit physician or practicing clinicians; hemispatial neglect as determined by >10 errors on the left side of the Bells Assessment (Gauthier, Dehaut, & Joanette, 1969); successful completion of Stage 7 on the CMSA–II, upper-extremity injury; conditions that significantly limited use of the affected arm before the stroke, such as osteoarthritis, rheumatoid arthritis, amputation, or previous stroke; or participation in other treatment intervention studies.

Measures

Evaluation at the participation level used real-life activities that were relevant to the participant and his or her daily life, elicited using the Canadian Occupational Performance Measure (COPM; Law et al., 1998; McColl, Paterson, Davies, Doubt, & Law, 2000). The COPM’s test–retest reliability in an adult rehabilitation setting is .80 for performance and .89 for satisfaction (Bosch, 1995; Cup, Scholte op Reimer, Thijssen, & Van Kuyk-Minis, 2003), and it is highly responsive to changes in client outcomes over time (Carpenter, Baker, & Tyldesley, 2001; Chen, Rodger, & Polatajko, 2002).

To measure activity limitations, the self-report version of the FIM™ (FIM–SR; Granger, Hamilton, Linacre, Heinemann, & Wright, 1993; Keith, Granger, Hamilton, & Sherwin, 1987) and the Chedoke Arm and Hand Activity Inventory (CAHAI; Barreca, Stratford, Lambert, Masters, & Streiner, 2005) were used. The internal consistency of the FIM–SR has been reported as .94–.96 in neurological populations (Jensen, Abresch, & Carter, 2005; Masedo, Hanley, Jensen, Ehde, & Cardenas, 2005), and a test–retest reliability of .89 has been reported in patients with spinal cord injury (Masedo et al., 2005). The test–retest reliability of the CAHAI is .98 (Barreca et al., 2005). The CAHAI detects clinically important change in the affected upper extremity’s progression from stabilizer to manipulator when performing tasks of daily living (Barreca et al., 2004; Barreca, Stratford, Dobell, & O’Flaherty, 2001).

To measure impairment, we used the Action Research Arm Test (ARAT; Lyle, 1981). The ARAT is a 19-item assessment with four subscales: Grasp, Grip, Pinch, and Gross Movement. Test–retest reliability is .98–.99 for the subscales (DeWeerdt, 1985; Lyle, 1981). Participants were asked to keep a daily activity log for two purposes: (1) to record their functional use of the impaired upper extremity outside of treatment and (2) to assess adherence to constraint device (mitt) wear.

Procedures

The study was approved by research ethics boards at two Toronto hospitals; informed, written consent was obtained from all participants. Participants were assessed 4 or 5 times before the intervention over approximately 3 wk. After a 2-wk modified CIMT program, they were reassessed another 4 or 5 times, again over approximately 3 wk. The assessments and treatment were delivered at a place convenient to participants, either in their home or at one of the hospitals.

Intervention

In this study, we used a modified CIMT protocol that was based on suggestions made by Dromerick, Edwards, and Hahn (2000). Modifications were reductions in the duration of mitt wear and massed practice compared with the traditional protocol (Taub, Crago, & Uswatte, 1998). Massed sessions were 2 hr, 5 days per wk for 2 wk. During the treatments, the participant wore the padded mitt on his or her unaffected side and focused on activities that required use of his or her affected side. Participants were also instructed to wear the mitt outside of the treatment sessions for an additional 4 hr per day, resulting in a total of 6 hr/day of mitt wear.

The training consisted primarily of shaping. The treatment protocol was individually customized; specific task selection depended on the participant’s functional goals identified on the COPM. Each participant identified three to five functional goals. For example, Participant 1 had as one of her goals, “To hold a glass of water in my left hand and drink from it.” Each treatment session consisted of three main components: (1) motor tasks related to the participant’s functional goals (e.g., grasping and releasing plastic cones); (2) goal-specific functional tasks (e.g., picking up a glass and drinking from it), and (3) other tasks of daily living (e.g., using a telephone). Techniques focused on motivating the participants to improve their performance and providing them with feedback. Examples of the techniques include the occupational therapist’s recording the time it took to complete a task, the number of repetitions completed within a certain time frame, the percentage of assistance required to complete a task, and the participant’s perceived exertion in completing a task. If the goal-specific functional tasks exceeded the participant’s motor ability, a task analysis was performed, and only components of the goal were addressed (e.g., if the participant’s goal was to lift a glass of water and drink from it, he or she might work on grasp and release before attempting to lift the glass). Over the 2 wk of treatment, components of the goal were addressed, and each treatment was shaped toward performing the final goal by the end of the 2 wk.

Throughout all treatment sessions, the therapist redirected participants’ attention and effort to the use of the affected upper extremity. Each task was repeated approximately 6–10 times with the affected extremity, even if the less affected extremity would normally be used for that function. This approach was also true for tasks that were bilateral in nature. Instead of removing the mitt and performing the tasks with both upper extremities, the participants performed the task in a modified fashion, using the more affected upper extremity or enlisting the assistance of the researcher to serve as a “second” upper extremity (Morris & Taub, 2001). This approach assisted in achieving the necessary massed practice.

Data Analysis

Data analysis included descriptive analysis of mean differences and trend line analysis to test whether this modified CIMT had a larger positive effect on impairment, activity, and participation than natural recovery. For each participant, the individual preintervention data points were used to determine a mean preintervention score on each measure; the same procedure was followed to obtain mean postintervention scores. Preintervention scores were also used to create trend lines for each participant. Trend lines were created using the Excel 2003 scatterplot function, and the linear regression was calculated to predict the postintervention results if no change was expected. Ninety-five percent confidence intervals (CIs) were calculated for each data point. The predicted trend lines were then compared with the actual postintervention results in terms of slope and in terms of the overlap in 95% CIs for each data point. A significant change was considered to have occurred if the CIs did not overlap (Borg & Gall, 1989).

Participants

Of the 8 potential participants who met the inclusion criteria, 5 agreed to participate and 4 completed the treatment. Two of the 3 who did not agree to participate did not have motor goals, and the daughter of the third did not feel her parent would comply with the protocol. The participant who dropped out of the study reported that the treatment protocol was too rigorous. The characteristics of the remaining participants (2 men, 2 women) are shown in Table 1. The participants had a mean age of 82 yr (range = 71–91), and preintervention CMSA–II motor recovery ranged from Stage 3 to Stage 6 for the arm (mean [M] = 4.8) and from Stage 2 to Stage 6 for the hand (M = 3.8).

Mean Difference Analysis

The before, after, and mean difference scores for all measures are presented in Table 2. Positive change scores were found for all participants on all participation, activity, and impairment measures with two exceptions. Participant 3 showed a small negative change on the FIM–SR, and Participant 2 showed a small negative change on the ARAT pinch. Of note is that 3 of the 4 participants (1, 3, 4) had mean difference scores on the CAHAI that were much greater than the change score of 6.3 deemed to indicate clinical significance by Barreca and colleagues (2005). The remaining participant approached this level (mean difference = 5.7).

Trend Line Analysis

Figures 1 and 2 illustrate the trend line analysis for COPM performance and satisfaction with performance scores. For all 4 participants, the postintervention trend lines for the performance component of the COPM fell above the predicted trend lines. For 3 of the participants (1, 2, 4), there was no overlap in CIs between the predicted and postintervention COPM performance trend lines, indicating improvements associated with treatment rather than natural recovery.

OPEN IN VIEWER

Figure 1.

Trend analysis for the performance scores on the Canadian Occupational Performance Measure (COPM) for all participants. For 3 of the 4 participants (1, 2, 4), there was no overlap in confidence intervals (CIs) between the predicted and postintervention COPM performance trend lines, indicating improvements associated with treatment rather than natural recovery.

OPEN IN VIEWER

Figure 2.

Trend analysis for the satisfaction scores on the Canadian Occupational Performance Measure (COPM) for all participants. For 3 of the 4 participants (1, 3, 4), the postintervention trend lines for the satisfaction component of the COPM fell above the predicted trend line. Participants 1 and 4 had no overlap in the confidence intervals (CIs) between the predicted and postintervention trend lines.

In the COPM satisfaction with performance component (Figure 2), the postintervention trend lines for 3 of the participants (1, 3, 4) fell above the predicted trend line. Participants 1 and 4 had no overlap in the CIs between the predicted and postintervention trend lines.

In the interest of space, we do not present trend line analysis figures for the FIM–SR, CAHAI, and ARAT. In the FIM–SR trend line analysis, 3 of the 4 participants (1, 2, 3) had postintervention trend lines above the predicted trend lines. Of those, only Participant 1 had no overlap in CIs between the predicted and postintervention trend lines.

On the CAHAI, all participants had postintervention trend lines fall above the predicted trend lines. Although all the participants had overlapping CIs, 100% of the postintervention data points exceeded the highest preintervention data point. Also suggestive of improvement is that for Participant 3, the slope of the postintervention trend line was substantially steeper than the slope of the preintervention trend line.

The data also suggest improvement in the grasp subtest of the ARAT, because for 3 of 4 participants (2, 3, 4) the postintervention trend lines are higher than the predicted trend lines. However, the CIs for all 4 participants overlapped, and none of the postintervention points were higher than the preintervention points. For the grip data, only Participant 2 had a postintervention trend line that was higher than the predicted trend line. Only Participant 1 had no overlap in CIs for pinch. On the gross motor subtest of the ARAT, only Participant 4 had a postintervention trend line that fell above the predicted trend line, because there was a ceiling effect for the other participants. The CIs of this participant did not overlap.

Daily Activity Log

Only 1 of 4 participants completed a daily activity log to record the types of activities he performed with the affected arm outside of the treatment time. Types of activities included opening and closing doors, turning lights on and off, turning faucets on and off, peeling a banana, and eating papaya and sausage with a fork. He did not wear the mitt during this time, nor did he record the number of hours he used the affected arm each day. The remaining participants did not wear the mitt outside of the treatment period, nor did they complete an activity log. They reported that it was too difficult to manage around the home when wearing the mitt or that they were frustrated wearing the mitt to complete tasks.

Using Modified CIMT in an Elderly Population

The mean age of our study participants was 82, much higher than the mean ages in previous CIMT studies. Of the 5 initial participants, only 1 dropped out, suggesting that older people may be able to tolerate an intense treatment such as CIMT. Participants all tolerated 2 hr of intense daily therapy wearing the restraint mitt but, in general, did not wear the restraint outside of the treatment sessions. Only 1 participant reported diligently using his affected arm outside of the treatment session, but he also did not wear the mitt, and his gains were not greater than those of the other participants. Poor mitt-wearing adherence in this study suggests that a modified CIMT protocol with less restraint time may be more appropriate for older adults such as those in this study (i.e., age ≥70).

The poor adherence resulted in a study protocol of 2 hr of mitt wear versus the original study protocol of 6 hr of mitt wear. Nevertheless, substantial improvements were noted, consistent with the other subacute studies using modified protocols, such as those described by Dromerick et al. (2000). A direct comparison cannot be made between the studies described in Dromerick et al. and our study, however, because mitt adherence was much higher in those studies (5–6 hr of mitt wear time). One possible explanation for why we were still able to see a strong effect is that the high-intensity program of massed practice may be more important than motor restriction of the less affected arm, as suggested by Liepert and colleagues (1998) and Taub and colleagues (1998).

Participation, Activity, and Impairment

CIMT modified to 2 hr per day of mitt wear and task-specific practice using the affected upper extremity had a positive effect on participation. This finding supports and extends the findings of the three other studies that have investigated participation as an outcome (Dettmers et al., 2005; Gillot, Holder-Walls, Hurtz, & Varley, 2003; Roberts,Vegher, Gilewski, Bender, & Riggs, 2005).

This study also demonstrated an improvement at the level of activity. Our findings are consistent with those of Dromerick et al. (2000). In their randomized controlled study, mean difference scores on the FIM–SR were higher for the CIMT group than for the control group; the only significant difference occurred on the upper-extremity dressing item.

The effects of modified CIMT on impairment as measured by the ARAT were less evident, with the exception of the results on the Grasp subtest. On the Grasp subtest, all 4 participants had higher mean scores postintervention than preintervention, 3 of the 4 participants had postintervention trend lines above the predicted trend lines, and 100% of their postintervention data points exceeded the highest preintervention data point, suggesting that modified CIMT had a beneficial effect. On the other three subtests, the results were variable, making it difficult to draw conclusions.

The ARAT is one of three primary measures used across CIMT studies for impairment. Dromerick and colleagues (2000) analyzed the ARAT by subtest and found that all mean posttreatment ARAT subtest scores were higher for patients in the CIMT group than in the control group; however, only the Pinch subtest scores achieved statistical significance. It is difficult to explain why Pinch was the only significant change when it was the least sensitive of the four subtests in our study; with the exception of age, the two studies’ participants appeared to be well matched for motor abilities, cognitive status, and CIMT protocol. Perhaps this difference is related to the type of intervention (i.e., a larger focus of the treatment intervention in the study by Dromerick et al., 2000, was on pinch activities). Other researchers, using total ARAT score, have demonstrated improvements (Page, Sisto, Levine, Johnston, & Hughes, 2001; van der Lee et al., 1999).

Limitations

Several considerations limit the conclusions that can be drawn from this work. Only 4 participants were evaluated, and additional study with a larger sample is required. Nonblinded assessment of the outcome measures may have been a source of bias. We attempted to control for this limitation by using standardized measures, by retesting participants without looking at previous scores, and by validating test scores against videotapes.

In some cases, participants’ behavior systematically improved during the baseline period and continued to improve during the treatment phase. This improvement made it difficult to separate treatment effects from naturally occurring recovery or placebo effect. The trend line analysis is reported to be robust in discriminating between the effects of time and intervention (Borg & Gall, 1989), and the intervention is the most likely explanation for improvements in these participants.