Introduction
Stroke remains a leading cause of long-term disability worldwide.
1 Deficits in function of the upper extremity represent a significant contributor to decreased function and quality of life poststroke.
2 Interventions such as constraint-induced movement therapy
3 and robot-assisted therapy
4,5 have emerged as promising approaches to promote return of function beyond that which is regained during spontaneous recovery and from conventional rehabilitation. However, the degree to which specific interventions affect disability and improve function after stroke is still poorly understood.
6,7
Numerous outcome measures are available that assess the effect of stroke across all dimensions of the World Health Organization (WHO) International Classification of Function, Disability, and Health (ICF),
8,9 complicating selection of appropriate measurement tools when assessing novel rehabilitation-related interventions.
10,11 Assessments of stroke-related
loss of body function/structure (function/structure, previously referred to as impairment) are often preferred given that these measurements are more objective and easier to define.
1,12 In contrast, measures of
activity limitations (activity) and
participation restriction (participation), while often higher patient priorities,
13,14 are frequently qualitative and rely on patient self-reporting, and are therefore less commonly used when assessing novel interventions.
In general, a limited relationship has been found across WHO ICF dimensions such as loss of function/structure and activity,
12,15,16 and data are conflicting regarding the degree of correlation that exists. For example, a meta-analysis of electromechanical and robot-assisted arm training described high-quality evidence that such interventions improve outcomes in both function/structure and activity, but there was limited evaluation of effects on participation.
17 The Fugl-Meyer (FM) motor scale,
18,19 a commonly employed measure of function/structure, was shown to demonstrate robust correlation with the Action Research Arm Test (ARAT), a measure of activity, when performed by expert raters.
19 On the other hand, studies of constraint-induced movement therapy report discrepancies between treatment-related changes in function/structure when compared with activity.
20 Another report found that a majority of patients demonstrating no measurable upper extremity motor impairment in an assessment of function/structure continued to report deficits as measured by assessments of activity and participation.
21 This inconsistent relationship may be due to numerous factors having greater influence on outcome as one moves from measurements of function/structure to activity or participation.
16 For example, multiple studies have shown improvements in patients’ functional independence measure (FIM) scores (FIM is a measure of activity) over the course of inpatient rehabilitation despite minimal change in FM total score
22 or National Institutes of Health Stroke Scale (NIHSS)
23 (both assessments of function/structure), potentially related to use of assistive devices or learning compensatory skills.
In the setting of a clinical trial on intensive robot-assisted therapy, the current study had two primary aims. First, the study aimed to provide a detailed assessment of the effect of robot-assisted therapy across all ICF dimensions. Second, the study also aimed to assess the degree to which therapy-related improvements in function/structure were related to improvements in activity and participation. We hypothesized that robot-assisted therapy would result in clinically and statistically significant gains in function/structure with smaller gains in activity and participation.
Discussion
Stroke-related deficits in upper extremity function remain a significant contributor to disability worldwide, with persistent deficits directly linked to activity limitations, participation restriction, poorer quality of life, and decreased subjective well-being.
1,21,33 Novel interventions, including robot-assisted therapy, have been developed to reduce stroke-related disability. However, available data provide a limited understanding of the degree to which such interventions produce changes across ICF dimensions. Furthermore, there have been limited studies that directly examine the degree to which treatment-related gains in one ICF dimension are associated with gains across other ICF dimensions.
The current study found that subjects who completed a course of robot-based therapy demonstrated statistically significant improvements in stroke-related loss of body function/structure and activity limitations, but not in participation restriction. Specifically, robot-assisted therapy was associated with statistically significant gains across all measures of function/structure and activity, including both primary endpoints: FM (a measure of function/structure) and ARAT (a measure of activity). Estimates of the minimal clinically important difference (MCID) for the FM and ARAT have been estimated to be 10% of the scale maximum,
18,34 or 6.6 and 5.7 points, respectively. Using this estimate, 17.5% of the study patients achieved MCID on the FM, and 32.5% on the ARAT, after robot-assisted therapy. Lo et al
4 have suggested that for some subjects with chronic stroke, gains below MCID might be clinically meaningful. SIS-hand, which is a patient-reported measure of activity, also showed a statistically significant improvement after therapy, though effect size was notably smaller. Improvements in measures of function/structure were robustly correlated with measures of activity, consistent with prior studies that found measures of function/structure to be significantly correlated with measures of activity, including Barthel Index, FIM, and SIS-16.
35,36 Together, the current results provide an in-depth examination of the degree to which clinical improvements after robot-based therapy extend across ICF dimensions.
Current findings are similar to previously published large-scale clinical trials in which an experimental intervention produced significant gains in measures of function/structure and activity, but not in measures of participation.
3,37,38 This difficulty to generate changes in participation likely reflects both the multitude of factors that contribute to participation in society and the complex interaction of such factors.
1 Previous studies have identified demographic variables, socioeconomic status, injury characteristics, medical and psychiatric comorbidities, social support, and functional status as contributors to deficits in poststroke participation.
2,39 Given that participation can be influenced by so many variables, including those related to both the function/structure and activity dimensions,
40 it may be that detection of meaningful improvements in participation requires a longer time period in order to develop and thus an extended follow-up period, consistent with the observation from longitudinal studies that quality of life after stroke can take months to years to manifest.
41 In line with this, Wolf et al
42 found initial changes in SIS functional domains were only followed by improvements in SIS participation domains after a period of 12 to 24 months beyond when the intervention was completed.
The high number and density of movement repetitions provided by the robot represent several strengths of the current study. First, standardization of treatment by the robot minimizes variability in conventional rehabilitation interventions, which could confound the relationship between improvements after robot-assisted therapy across ICF dimensions.
43 Second, robot-based interventions often deliver a higher intensity of movement repetitions compared with conventional interventions.
43,44 In one study, conventional therapist-administered interventions provided an average of 32 repetitions/session,
45 which is too few movements to induce neural plasticity
46 and an impediment to improving behavior, particularly in higher order ICF dimensions, including participation.
47 In contrast, patients in the current study averaged 2579 thumb movements and 1298 wrist movements per treatment session, exceeding the 600 to 700 repetitions per day
48 in rats and the 924
49 movements per day in primates considered necessary to realize functional benefits with stroke rehabilitation. Third, each hand movement in the robot was coupled to video games. Games promote patient involvement in health care
50,51 and motivate patients to engage in enjoyable play behavior that involves therapeutically relevant movements.
52,53 Use of games alters cognitive context
54 and, compared with rote movement repetition, increases activity in cognitive networks in patients with stroke.
55
After robot-assisted therapy, improvements in function/structure and activity were not associated with improvements in participation. Recently published results from 2 large-scale clinical trials also reported a dissociation between treatment-related motor gains and change in measures of participation.
16,38 Lang et al
38 found that while the majority of subjects reported overall perception of meaningful change with treatment, as a group they did not demonstrate statistically significant improvement in measures of participation. A cross-sectional study
31 comparing SIS with SS-QOL also reported significant dissociation, with SIS subscores being more responsive to treatment as compared to SS-QOL subscores. On the other hand, Roth and Lovell
56 reported a correlation between FIM scores >80 at 1 year poststroke and increased community/home participation as defined by the Frenchay Activities Index. Together, these results highlight an ongoing need for robust, responsive, and specific measures of participation, particularly when evaluating the effect of novel interventions for reducing stroke disability. Strikingly, in a review of the 116 instruments on
www.rehabmeasures.org28 designated for stroke, only 15 instruments were strictly participation measures. The overwhelming majority of those instruments assess only 3 or 4 items to characterize overall quality of life, and thus provide coarse gradation across patients. Under-scoring the limited attention to participation in studies of stroke rehabilitation therapies, a report by Salter et al
57 showed only 25% of randomized controlled trials of stroke rehabilitation in the past 4 decades included an assessment of participation.
57 These prior studies indicate the ongoing interest and the paucity of knowledge regarding the translation between improvements in function/structure and participation, changes that are more clinically meaningful to patients.
Several limitations are associated with the current study. First, subjects were studied 11 to 26 weeks poststroke. As such, it is unknown the degree to which the current findings can be extended to patients who initiate therapy at an earlier or later time poststroke. In addition, while the literature reports 95% of patients with upper extremity deficits reach recovery plateau by 11 weeks poststroke,
58 the improvements demonstrated after robot-assisted therapy (
Table 2) may not be entirely attributable to the robotic intervention and instead may be confounded by a degree of natural recovery. Any contributions of natural recovery to the current results are likely attenuated in this study, however, as serial baseline assessments indicated that subjects had reached stable arm motor status at the time of enrollment. The primary endpoint of the study was at 1 month posttherapy, which was an additional study limitation. It may be that any long-term changes in participation and their relationship over time
59 with structure/function and activities, were not captured as a result of this 1-month focus. Finally, reflecting the paucity of meaningful participation measures reported in the stroke literature, the SS-QOL was the only participation measure used in the current study. Although other measures of participation are available, the SS-QOL has been validated in stroke patients in multiple studies
60,61 and includes measures of quality of life as they relate to upper extremity function, making it especially suited for the current study.
Though the generalizability of these results to robot-assisted therapy beyond the upper extremity is yet unknown, studies showing correlations between function/structure of the lower extremity
62 and activity using the Barthel Index,
35,62 SIS-16,
36 and FIM
22,36 suggest that further studies in other key components of movement are warranted and may show similar relationships.