Chapter 23 – Evidence-based Motor Rehabilitation after Stroke


Worldwide, stroke is a most common disabling disorder that requires rehabilitation services if curative and preventive treatments fail. There is growing evidence that intensive rehabilitation offered by a multidisciplinary team is effective to improve outcome in terms of independent daily living and health–related quality of life. This conclusion is based on systematic reviews and recent pragmatic phase III and IV trials. Although intensity of practice is an important part of effective stroke care, very early mobilization should be restricted and applied in small doses within 24 hours post-stroke. Systematic review shows that evidence-based therapies for the upper limb are constraint–induced movement therapy and upper limb robotics, whereas interventions that could be beneficial to gait include fitness training and high-intensive, task-specific training. A number of novel therapies, such as combining exercise therapy with transcranial direct current stimulation, repetitive transcranial magnetic stimulation or pharmacological interventions, and virtual reality are under way. However, the evidence for most of these therapies is still unclear and in its infancy.

Chapter 23 Evidence-based Motor Rehabilitation after Stroke

Gert Kwakkel

Janne M. Veerbeek


About 10.3 million people worldwide have a new stroke every year and about 25.7 million stroke survivors and 6.5 million stroke-related deaths are reported (Feigin et al., 2015), making stroke the second most common cause of death and one of the main causes of acquired adult disability (Langhorne et al., 2011; Feigin et al., 2015). The main goal of stroke rehabilitation is to restore patients’ independence in their activities of daily living (ADL) and with that health-related quality of life (HR-QoL). Thrombolysis applied within the first 4.5 hours post-stroke (Kohrmann et al., 2006; Prabhakaran et al., 2015) and endovascular stroke treatment such as thrombectomy (Berkhemer et al., 2015; Saver et al., 2016) are both aimed to restrict the consequences of irreversible brain damage very early post-stroke. Evidence-based stroke rehabilitation is seen as the next most effective option to minimize the disabling consequences of stroke (Langhorne et al., 2011). Rehabilitation has a rather non-specific definition: ‘A problem solving process aiming at reducing the disability and handicap resulting from a disease’ (Langhorne and Legg, 2003). The interventions delivered have different levels of complexity and may be classified at level of services involved (i.e. service level), type of discipline that delivered the therapy (i.e. operator level), or more specific type of intervention applied (i.e. treatment level) (Langhorne and Legg, 2003). These different levels in which therapists need to be educated to implement new protocols are assumed to be better than the usual care, but make the experimental intervention a black box and designing trials to evaluate the effectiveness of rehabilitation interventions rather complex. Effective stroke rehabilitation allows patients to cope with the existing neurological deficits and entails a cyclic process of: (1) assessing to identify and quantify patients’ needs and wishes; (2) defining realistic and attainable goals for improvement; (3) applying the most effective intervention for achieving set goals; and (4) monitoring the improvements in light of set goals (Langhorne et al., 2011).

Classification of Stroke Outcome

Disabling disorders such as stroke can be classified within the World Health Organization’s (WHO) International Classification of Functioning, Disability, and Health model (ICF) (WHO, 2001) (, which provides a framework for the effect of stroke on the individual (Figure 23.1) in terms of pathology (disease or diagnosis), impairment (symptoms and signs), activity limitations (disability), and participation restriction (handicap) (Langhorne et al., 2011). This commonly accepted framework of the ICF is used across all medical specialties involved in stroke care for assessment of impairments such as high blood pressure, glycaemia, hemiplegia, pain, visuospatial neglect, memory deficits, fatigue, and depression. Assessing its second and third domains (activity and participation) is less common in medicine, but fundamental in stroke rehabilitation (Langhorne et al., 2011). Indeed, the final purpose of rehabilitation is to improve patients’ functioning, i.e. the activities that patients perform in their home environment, and their participation in social life. The different domains of ICF should be separately and specifically measured, in order to obtain a comprehensive assessment of the patient (Lejeune and Stoquart, 2015).

Figure 23.1 The International Classification of Functioning, Disability, and Health (ICF) framework for the effect of stroke on an individual. This figure summarizes key features of WHO’s ICF model and includes the most relevant categories affected after stroke and examples of measurement scales used in those categories. ADL = activities of daily living, ECG = electrocardiograph. With permission from the Lancet (Langhorne et al., 2011).

Mechanisms Underlying Stroke Recovery

Motor recovery is a heterogeneous and complex process that probably occurs through a combination of spontaneous neurological recovery and learning-dependent processes, including restitution (i.e. restoring the functionality of damaged neural tissue), substitution (i.e. reorganization of partly spared neural pathways to relearn lost functions), and compensation (i.e. improvement of the disparity between the impaired skills of a patient and the demands of their environment) (Langhorne et al., 2011; Buma et al., 2013). Several longitudinal studies with repeated measurements in time have shown that about 70–80% of most observed neurological improvements can be explained by spontaneous neurological recovery in the first 3 months (Kwakkel et al., 2006b). Due to the logistic pattern of neurological recovery, meaningful activities return, such as walking ability, upper limb capacity, and performing ADL. However, next to these underlying mechanisms, there is growing evidence that intensive, context- and task-specific motor training has a surplus value beyond expected levels of spontaneous recovery. Estimates from several meta-analyses show that the contribution of rehabilitation on top of this non-linear improvement ranges from 5 to 15% of the variance of outcome (Kwakkel et al., 2004; Veerbeek et al., 2014; Kwakkel et al., 2015). This finding shows that the added value of motor rehabilitation interventions should be seen as ways to enhance the pattern of (spontaneous) neurological recovery in the first 3 months post-stroke, making patients earlier independent in their environment than when therapies are postponed or inadequate (Kwakkel et al., 1999).

Despite growing evidence from a number of animal studies (Murphy and Corbett, 2009), the underlying mechanisms that drive motor recovery early post-stroke are still poorly understood (Cramer 2008a, 2008b). The evidence that motor learning is able to interact with early mechanisms of spontaneous neurological recovery is limited (Buma et al., 2013). Findings from several longitudinal kinematic studies suggest that improvements are mainly based on learning adaptation strategies in which patients learn to optimize the still intact or spontaneously recovered end-effectors after stroke (Buma et al., 2013; van Kordelaar et al., 2013; Kwakkel et al., 2015b; Cortes et al., 2017). Recently, a number of studies suggest that this amount of spontaneous neurological recovery is highly predictable within the first 72 hours post-stroke, which follows a proportional amount of change in motor outcome (Prabhakaran et al., 2008; Winters et al., 2015; Veerbeek et al., 2018; van der Vliet et al., 2020, in press), neglect (Nijboer et al., 2013), and aphasia (Lazar et al., 2010).

Predictability of Stroke Recovery

Due to spontaneous mechanisms of neurological recovery, the time course after stroke follows a logistic pattern, which is characterized by larger improvements in neurological impairments such as synergistic motor control, strength, and visuospatial attention during the first 10 weeks post-stroke (Kwakkel et al., 2006b; Kwakkel and Kollen, 2013). This subacute phase is followed by a gradual levelling off in observed improvements in the months that follow. After 3 to 6 months, recovery of most activities such as dexterity and walking ability plateaus (Kwakkel and Kollen, 2013).

Several cohort studies have shown that final outcomes of the upper (Kwakkel and Kollen, 2007) and lower limb (Kwakkel et al. 2006b; Smith et al., 2017), as well as independency in ADL (Veerbeek et al., 2011b), are highly predictable within the first days post-stroke. In particular, the initial severity of neurological deficits, often assessed with the National Institutes of Health Stroke Scale (NIHSS), and, with that, initial disability (Veerbeek et al., 2011b; Kwakkel and Kollen, 2013) are important indicators of final outcome 6 months after stroke onset (Kwakkel and Kollen, 2013; Scrutinio et al., 2017). In addition, patients of older age have poorer outcomes in ADL when compared with younger subjects. However, it remains unclear to what extent this determinant is modified by the pre-morbidities before stroke (Kwakkel and Kollen, 2013). Important determinants for motor recovery of the upper limb are the voluntary control of finger extension and shoulder abduction (i.e., SAFE model) (Nijland et al., 2010), reflecting the integrity of the corticospinal tract (CST), and the presence of motor evoked potentials (MEPs) (Stinear et al., 2017, Stinear, 2017). However, testing the intactness of CST by transcranial magnetic stimulation (TMS) of the abductor digiti minimi (ADM) (TMS-ADM) within the first 48 hours and 11 days post-stroke showed no added prognostic value when compared with the clinical SAFE model alone for the upper limb (Hoonhorst et al., 2018). For outcome of walking ability, patients’ ability to sit independently without any support for at least 30 seconds and severity of lower limb paresis within the first 72 hours post-stroke are seen as important prognostic indicators (Veerbeek et al., 2011c).

Evidence-based Practice in Stroke Motor Rehabilitation

In the present chapter, we will review the evidence of the most used rehabilitation interventions aimed at improving meaningful activities such as upper limb capacity, balance control, gait, and other (basic) ADL post-stroke. For this purpose, we restricted what we investigated to the pooled and summarized evidence from randomized clinical trials (RCTs) as presented in well-conducted phase III trials that obey the ONSORT statement (Schulz et al., 2010) (, Cochrane reviews of trials, and meta-analyses that obey key criteria of the PRISMA (Schulz et al., 2010) (

In the next paragraphs, we will look mainly at interventions that have shown strong or moderate evidence in terms of body functions and activities. Therapies with strong evidence for improving body functions and meaningful activities (i.e., Level 1) were based on generally consistent findings in multiple, relevant, high-quality RCTs (PEDRO scores >3) (Veerbeek et al., 2014) and analysed in meta-analyses (preferably Cochrane reviews) that after pooling of individual phase II trials of high quality showed significant overall effects favouring the experimental intervention. Alternatively, a single phase III or IV trial of sufficient methodological quality was also considered as strong evidence. Moderate evidence (i.e., Level 2) was based on evidence that was provided in one relevant trial of high quality.

The evidence of (neuro)pharmacological interventions combined with motor rehabilitation as well as studies that used transcranial direct current stimulation (tDCS) or repetitive transcranial magnetic stimulation (rTMS) are beyond the scope of the present chapter and are discussed elsewhere in this book.

Evidence for Very Early Mobilization within 24 Hours Post-stroke

Very early rehabilitation, in which patients are mobilized out-of-bed in the first 48 hours after stroke and during which they are stimulated to be physically active, was strongly believed to contribute to the effectiveness of stroke-unit care. This belief was built on four neutral, phase II trials (N = 218), making the evidence for very early mobilization post-stroke rather weak (Bernhardt et al., 2015a).

The recently conducted phase III trial (Bernhardt et al., 2015b), involving 2104 stroke patients, showed that fewer patients in the very early mobilization group (6.5 times 30 minutes within 24–48 hours post-stroke) had a favourable outcome on the modified Rankin Scale (mRS) when compared with those in the usual care group (who were mobilized 3 times for 10 minutes within the first 24 hours) (N = 480 [46%] vs N = 525 [50%]). Eighty-eight (8%) of all patients died in the very early mobilization group compared with 72 (7%) patients in the usual care group. No differences were found with respect to non-fatal serious adverse events and immobility-related complications in the AVERT study (Bernhardt et al., 2015b; Langhorne et al., 2018). In other words, the intervention tested was not simply earlier, but was also more frequent and of a higher dose than usual care. Indeed, the difference achieved in frequency and dose was greater than the difference in timing of first mobilization (median 18 hours vs 22 hours after stroke onset).

The AVERT study suggests that a higher dose of early mobilization (201 minutes versus 31 minutes) is accompanied by a reduction in the odds of a favourable outcome at 3 months following the mRS. Although not significant, subgroup analysis showed a trend disfavouring early mobilization of severe strokes (NIHSS ≥16 points) and those with a haemorrhagic stroke, suggesting higher risk for poor outcome according to the mRS (mRS >2) when compared with usual care (Bernhardt et al., 2015b). In fact, the findings of the AVERT study suggest that very early mobilization of stroke patients should be restricted to a few times in the first 24 hours and limited to small doses of 10 minutes at most. The AVERT study raises several essential questions to be addressed in future research. First, why is applying a higher dose of out-of-bed therapy in the first 2 days after onset more harmful than a lower dose of shorter duration in patients with severe stroke? Second, is the impaired regional cerebral blood flow in penumbral areas sensitive to orthostatic variation? One may assume that, especially in severe and haemorrhagic strokes, the cerebrovascular autoregulation needed to sustain sufficient regional cerebral blood flow is impaired. High doses of long-duration mobilization very early after stroke, which often result in tired and drowsy patients slumping in their chair, may further reduce the regional cerebral blood flow in critical penumbral and oligaemic brain areas and increase neurological damage. Further research is now needed into the longitudinal association between body and head position and cardiac output on the one hand, and impaired cerebral haemodynamics and reduced cerebral perfusion on the other, in patients with (hyper)acute stroke (Kwakkel, 2015a; Krakauer and Carmichael, 2018). Although the AVERT study suffers from some minor methodological problems such as contamination, as the usual care group started mobilization earlier each year, the indirect message of this groundbreaking trial is that sufficiently powered phase III trials are possible in complex interventions such as stroke rehabilitation (Bernhardt et al., 2015b; Kwakkel, 2015a).

Evidence for Intensive, Task-oriented Practice beyond the Acute Phase Post-stroke

There is no clear definition of the term ‘intensity’ of practice. In the literature, ‘dose’ or ‘intensity’ of practice has often been described as ‘frequency of repetitions of desired movement’, ‘amount of external work’, or ‘amount of time that is dedicated to practice’ (Kwakkel, 2006a). Preferable measures of dose would be active time in therapy or number of repetitions of an exercise (Lohse et al., 2014). However, actual valid measures of intensity, besides behavioural mapping techniques (Bernhardt et al., 2004), are still lacking. Future trials should use portable activity monitors and/or robotics to offer better estimates of the actual amount of practice applied.

One Cochrane review of 33 trials (N = 1853) showed that repetitive practice of functional mobility-related and upper limb tasks, or parts of these tasks, is favourable for patients after stroke when compared with usual care (French et al., 2016). These findings were most pronounced for improvement of gait and regardless of timing post-stroke. The post-intervention effects of intensive practice did sustain up until a 6-month follow-up (French et al., 2016).

Time dedicated to practice may be a simple surrogate for the actual intensity of practice applied in trials (Kwakkel et al., 1999; Kwakkel et al., 2004; Veerbeek et al., 2011a; Lohse et al., 2014). In the present chapter, we defined intensity as ‘the number of hours spent in exercise therapy’ (Kwakkel et al., 2004; Lohse et al., 2014;Veerbeek et al., 2014). Various systematic reviews and meta-analyses have shown that more time spent in exercise therapy has beneficial effects for patients after stroke (Kwakkel et al., 2004; Veerbeek et al., 2011a; Lohse et al., 2014; Veerbeek et al., 2014). This finding is regardless of timing post-stroke and type of motor rehabilitation intervention. However, the optimal dose of exercise therapy is still unknown after stroke. Systematic reviews with meta-analysis suggest that additional therapy time of ≈17 hours over 10 weeks results in favourable outcomes on all three domains of the ICF (Veerbeek et al., 2014). In the same vein, Lohse and colleagues found after pooling 30 trials (N = 1750) there was a positive relationship between the time scheduled for therapy and therapy outcomes such as ADL and gait. These data suggest that larger doses of therapy may lead to better outcomes post-stroke. Based on cumulative meta-analyses of trials and as a rule of thumb, it is recommended that patients should exercise at least 45 minutes on working days as long as there are rehabilitation goals (Veerbeek et al., 2014).

It is, however, unclear when exactly intensive rehabilitation should be started post-stroke. At least, the AVERT study indicates that the paradigm ‘more is better’ is not the case when starting within the first 24 hours. Generally, there is consensus that rehabilitation services should be start within the first days post-stroke and be based on the patient’s ability. The intensity should be progressive when the first vulnerable days have passed and target meaningful activities such as making transfers, gait, and dressing.

Evidence for Fitness Training and Sports Post-stroke

There is growing interest in fitness training after stroke, although this type of training is still not common in rehabilitation medicine, when compared with other cardiovascular diseases such as myocardial infarction (Saunders et al., 2016). Physical fitness training can be defined as ‘a subset of physical activity which is planned, structured, repetitive, and deliberately performed to train (improve) one or more components of physical fitness’ (Saunders et al., 2016). These components include cardiorespiratory fitness, muscle strength, muscle power, flexibility, balance, and body composition (Saunders et al., 2016). As patients after stroke often are less physically active and frequently have a reduced physical fitness, physical fitness training has the potential to improve these aspects. In general, three types of physical fitness training can be distinguished: (1) cardiorespiratory training; (2) resistance training; and (3) mixed training forms in which cardiorespiratory and resistance training are combined.

In a Cochrane review by Saunders and colleagues (2016), a total of 58 trials (N = 2797) was identified, in which mainly ambulatory stroke patients in the chronic phase were included. Physical fitness training appeared to be safe (Saunders et al., 2016). Meta-analyses showed clear evidence that fitness training including cardiorespiratory training involving functional tasks such as walking is beneficial for patients after stroke in terms of walking performance. There was some evidence that cardiorespiratory training has the ability to improve cardiorespiratory fitness, and that mixed training may improve measures of walking performance. In line with another Cochrane review, no evidence was found to support resistance training or strengthening exercises alone (Saunders et al., 2016). In addition, most benefits disappeared at follow-up, except for walking distance and maximal gait speed. The effects of cardiorespiratory training and mixed training on quality of life, mood, death, and dependency are still unclear (Saunders et al., 2016).

Evidence-based Interventions to Improve the Upper Limb

Around 80% of the stroke survivors have motor impairments of the upper limb that affect their ability to perform, and participate in, activities of daily living (ADL). The severity of upper limb paresis and in particular the patients’ ability to extend wrist and fingers are the strongest determinants that define upper limb capacity at 3 and 6 months post-stroke (Nijland et al., 2010; Stinear, 2010; Coupar et al., 2013; Kwakkel and Kollen, 2013). Therapies that are targeting these patients with an initial favourable prognosis at baseline of their trial have also shown beneficial effects on outcome of upper limb capacity. In contrast, rehabilitation interventions trials that selected patients without voluntary motor control of wrist and finger extension have shown that their beneficial effects are restricted to improvements in motor impairments alone. This finding further supports the notion that evidence-based therapies for the upper limb strongly depend on ability of the patients to voluntarily control the hand (i.e. have a favourable prognosis) (Langhorne et al., 2011). With that, next to the therapy, selecting patients for a specific intervention is an important part of adequate rehabilitation care post-stroke.

Significant effects are found for: (1) constraint-induced movement therapy (CIMT) and its modified versions (mCIMT); (2) electromyography-triggered neuromuscular stimulation (EMG-NMS) of wrist and finger extensors; (3) mental practice; and (4) virtual reality training of the upper limb (Veerbeek et al., 2014). Next to these interventions, positive effects have also been found for the body functions and/or activities level for (5) bilateral arm training (BAT) with rhythmic auditory cueing (BATRAC); (6) mirror therapy; and (7) robot-assisted therapy for the upper limb. The evidence for these interventions will be discussed below.

(Modified) Constraint-Induced Movement Therapy

Constraint-induced movement therapy was developed to overcome upper limb impairments after stroke and is the most investigated intervention for the rehabilitation of patients with upper limb limitations (Figure 23.2).

Figure 23.2 Task-oriented practices with the paretic limb in constraint-induced movement therapy (CIMT). With permission from the Lancet Neurology (Kwakkel et al., 2015).

Practices include: (A) cutting bread, (B) pouring water, (C) picking up and replacing money, and (D) playing a game. Use of the unaffected limb is restricted by a padded mitt.

The signature protocol for the original form of CIMT contains (1) intensive, graded practice of the paretic upper limb to enhance task-specific use of the affected limb for up to 6 hours a day for 2 weeks; (2) constraint or forced use therapy, wearing a mitt on the non-paretic hand to promote the use of the impaired limb during 90% of the total hours awake; and (3) adherence-enhancing behavioural methods designed to transfer the gains obtained in the clinical setting or the laboratory to patients’ real-world environment (i.e. a transfer package) (Kwakkel et al., 2015b). Modified versions of CIMT often apply only the first two elements of CIMT with a limited intensity. CIMT or modified versions of this treatment use operant training techniques to enhance upper limb capacity. However, for this behavioural therapy some preservation of finger and wrist extension of the paretic upper limb is needed.

A recent meta-analysis of 51 trials (N = 1784) showed beneficial effects on motor function, arm–hand activities, and self-reported arm–hand functioning in daily life of both types (i.e. CIMT or mCIMT) immediately after treatment and at follow-up (20 weeks), whereas no evidence was found for the efficacy of constraint alone (as used in forced use therapy) (Kwakkel et al., 2015b). Figures 23.3 and 23.4 show forest plots of overall effect sizes of CIMT, modified versions of CIMT (mCIMT), and forced use therapy.

Figures 23.3 & 23.4 Forest plot of overall effect sizes of constraint-induced movement therapy (CIMT), modified CIMT, and forced use therapy post intervention (Figure 23.3) and at follow-up (Figure 23.4). Effects classified in accordance with the International Classification of Functioning, Disability, and Health (ICF; WHO). Diamonds represent the overall effect sizes after pooling the standardized mean differences (SMD). The SMD was based on adjusted Hedges’ g (95% CI) model. If pooling was not possible, the individual SMD is shown based on an adjusted Hedges’ g analysis. The SMD Hedges’ g model is a model calculated on the basis of the difference between the means of the experimental and the control groups divided by the pooled standard deviation of both groups in a trial and multiplied by a correction factor called J for the degrees of freedom. The different ICF categories: body functions (black outline), activities (grey), and participation (dotted outline). ADL = activities of daily living. E = experimental group. C = control group. N/A = no data available. *Sufficient statistical power (1–β ≥0.80). With permission from the Lancet Neurology (Kwakkel et al., 2015).

Figures 23.3 & 23.4 (cont).

Sensitivity analyses showed no significant differences in effect sizes between original CIMT and mCIMT, dose of CIMT (additional time spent in exercise therapy between 5 hours and 60 hours), or timing of CIMT (trials started within or after the first 3 months after stroke onset) (Kwakkel et al., 2015b).

In summary, (m)CIMT is found to be the most effective therapy to enhance upper arm function and arm–hand activities post-stroke (Kwakkel et al., 2015b). In contrast, meta-analyses showed no evidence for grip strength, sensibility, or pain, while for patients beyond 3 months post-stroke, no effects were found for motor function of the paretic arm. These findings strongly suggest that with (m)CIMT patients learn to optimize motor performance (Buma et al., 2013; van Kordelaar et al., 2013; Kwakkel et al., 2015b).

EMG-NMS of Wrist and Finger Extensors

With electromyography-triggered neuromuscular stimulation (EMG-NMS), peripheral nerves and muscles are stimulated with electrodes (Pomeroy et al., 2006). In this chapter, only stimulation with external electrodes is discussed. Electrostimulation can be applied during training of activities (i.e. functional electrical stimulation [FES]) or in a non-functional manner (e.g. performing wrist and finger extension without a functional purpose) (Veerbeek et al., 2014). Stimulation in EMG-NMS is only activated when the patient actively attains an individualized, pre-set threshold value of muscle activity (Veerbeek et al., 2014). It was found that EMG-NMS of the wrist and finger extensors of the paretic arm has significant positive effects on synergy independent motor control of the paretic arm and arm–hand activities (Veerbeek et al., 2014). These findings were based on 25 trials (N = 492) (Veerbeek et al., 2014). Application of EMG-NMS of the wrist and finger extensors should be considered in patients with some voluntary wrist and/or finger extension. In addition, it should be noticed that application of EMG-NMS is mainly investigated in patients who are beyond 1 month and quite often beyond 6 months after stroke onset. The optimal frequency and intensity of the stimulation are unclear (Veerbeek et al., 2014).

Mental Practice with Motor Imagery

Mental practice of motor actions and/or activities for the upper limb aims to improve their performance (Barclay-Goddard et al., 2011; Langhorne et al., 2011a; Veerbeek et al., 2014) and is often combined with physical practice. Based on a meta-analysis of 14 RCTs (N = 424), mental practice combined with physical practice has significant beneficial effects for outcome of arm–hand activities at termination of the intervention period, but the sustainability is unclear (Barclay-Goddard et al., 2011). No effects have been found for synergy independent motor control of the paretic arm, nor for muscle strength or basic ADL (Veerbeek et al., 2014). To participate in this type of training, patients needed to be able to have some active movement abilities in the paretic arm. However, it should be noted that testing the patient’s ability to perform mental imaging is difficult and therefore implementation in daily practice is hampered.

Only gold members can continue reading. Log In or Register to continue

Mar 22, 2021 | Posted by in NEUROLOGY | Comments Off on Chapter 23 – Evidence-based Motor Rehabilitation after Stroke
Premium Wordpress Themes by UFO Themes