88 - 94: Rehabilitation and Plasticity

Neurorehabilitation is a quickly developing field accelerated by advances in neurosciences and rehabilitation engineering. In contrast to acute care medicine with its goal to cure, neurorehabilitation aims at minimizing disability and accelerating the return to an independent life. Naturally, it has to focus on all the deficits and dysfunctions of a patient that stand against these goals. Typically, physical (movement), communicational and cognitive deficits occur and interfere with each other. Therefore, the classical neurorehabilitative approach combines physio- and occupational therapy with speech and language and cognitive training, the latter typically provided by neuropsychologists.

Training, i.e. repetitive engagement in challenging exercises, is still the mainstay of neurorehabilitative treatment protocols. Training shares certain neural mechanisms with healthy learning, i.e., what is commonly described as neuroplasticity, although it is unclear in how far these processes overlap or share common mechanisms.

Whether training can be effectively enhanced by plasticity-encouraging interventions such as brain stimulation or certain drugs remains to be proven by clinical research.

What is known from experience as well as science is that higher intensity is more effective than lower intensity [1,2], although this may not hold true for early time points after injury (in this case, stroke) [3]. High-intensity training requires excellent patient compliance and motivation, the major limiting factors for successful training.

Like motor learning in healthy people, rehabilitative training after a lesion to the motor cortex leads to modifications in motor cortices adjacent to the lesion. Motor representations change in size and shape [4,5] and new connections form within and between sensory and motor cortical networks [6]. The representations of complex movement patterns that can be evoked by long high-frequency pulses to the injured hemisphere - a possible reflection of a memory trace for complex movements - are lost after a cortical lesion. With rehabilitative training they recover; the degree of restoration of complex movement patterns evoked by brain stimulation correlates with behavioral recovery [7].

Plasticity also occurs in the uninjured hemisphere contralateral to the lesion. Biernaski and Corbett [8] reported enlarged dendritic trees in the uninjured hemisphere in animals that were trained in an enriched environment after a focal lesion to the motor cortex as compared with control animals that received standard therapy. Indirect evidence for plastic modification was also found in other brain areas that are involved in motor control. In the cerebellum of rats that were subjected to repeated treadmill training after a middle cerebral artery occlusion, the 25-kDa synaptosomal-associated protein and glial fibrillary acidic protein - markers of synaptic plasticity - were upregulated as compared with untrained controls [9]. Similarly in the thalamus, synaptic plasticity indexed by synaptophysin expression was increased in the damaged hemisphere in animals that were trained on a rotarod (skilled locomotor training) as compared to a treadmill (locomotor training of lower difficulty) or no training [10]. In the striatum, expression levels of glutamate, brain-derived neurotrophic factor, and p-synapsin I were increased by treadmill training after middle cerebral artery occlusion [11]. However, the type of motor training certainly plays a role in the pattern of plastic modification across the rat brain: whereas forelimb precision movement training (typically single pellet reaching tasks) depends more on cortical plasticity, locomotor training likely induces reorganization in larger motor control networks including the cerebellum, basal ganglia and thalamus.

In humans, evidence for plastic modification of neural circuits comes from imaging and electrophysiology. Task-related activation in motor cortices or, in case of cortical lesions, the peri-infarct cortex is modified during the course of recovery. After corticospinal tract infarction, initial overactivation of the contralesional motor cortex (ipsilateral to the moving hand) reverts to ipsilesional (contralateral) dominance as recovery progresses [12,13]. The presence of motor potentials evoked from the lesioned hemisphere using transcranial magnetic stimulation (TMS) and corticospinal tract integrity in diffusion tensor magnetic resonance imaging predicted the response to arm training in chronic stroke survivors [14]. Interfering with the contralesional hemisphere by TMS did not affect reaction times with the paretic hand in stroke survivors with varying degrees of recovery at least 2 years after stroke, but interfering with the ipsilesional motor cortex did [15]. This emphasizes the role of the ipsilesional hemisphere (contralateral to the moving limb) for successful recovery.

In predicting therapy gains, the degree of injury to descending tracts from primary motor and premotor cortices is an important surrogate marker. It correlated with gains achieved after hand robotic therapy [16]. Bilateral arm training improved the arm impairment in chronic stroke survivors. This improvement was paralleled by an increase in activation of bilateral premotor cortices after as compared to before training [17]. As compared with conventional physiotherapy according to neurodevelopmental principles, bilateral arm training was as effective but showed a different neural response: where conventional physical therapy had no effect on brain activation, bilateral training led to an increase in bilateral premotor cortex activation [18]. This indicated that different therapies operate through different mechanisms even if they induce a comparable behavioral response.

When to start rehabilitation after stroke or trauma is a matter of debate. It seems that forcing an animal to train too early increases the infarct volume [19]. On the other hand, stroke induces the expression of specific proplastic genes defining an optimal time period in which training should be most effective [20].

In humans, using high-intensity constraint-induced movement therapy (CIMT; 3 h of training based on shaping principles plus immobilizing the intact limb for 90% of the waking hours for 10 days) starting around 10 days after stroke results in adverse outcomes as compared with standard occupational therapy and standard CIMT (2 h of shaping plus 6 h of immobilization per day) [3]. In contrast, Bernhardt et al. [21] suggested that early - within 24 h - mobilization, i.e., getting the patient out of bed, improves the outcome 3 months after the stroke. This finding awaits confirmation in a larger sample.

Many studies have compared different training therapies but have shown similar therapeutic responses. Experience suggests that the intensity of training is a major determinant of the training effect. Additional leg training resulted in better walking ability and arm training in better dexterity as compared with immobilization in subacute stroke survivors enrolled within 14 days of stroke onset indicating that the intensity of training counts [10]. Similarly, a community-based physical activity program improved walking velocity, balance and quality of life measures as compared with standard care also emphasizing training intensity [22]. On the other hand, as mentioned above, high intensity early after stroke may result in adverse functional outcomes [3].

The main obstacles that prevent high-intensity training are fatigue and motivation. Fatigue can be counteracted by sufficient interruptions and rest periods. Motivation requires a motivating environment, positive feedback from the therapist or a training device that specifically addresses motivational aspects, e.g. by providing rewarding schemes in a gaming environment. How to optimally use motivational therapy to improve training outcomes is still unclear.

Little is known about which training principles are most effective and for whom. Active training is better than passive movement, but also mental imagery of movement has a training effect in combination with active training [23]. Bilateral training is similarly effective as unilateral training when looking at the population mean; individual patients may respond better to one or the other [18]. It remains unknown which factors predispose a patient to respond to a specific treatment.

One of the largest randomized controlled trials (n = 222 participants) in neurorehabilitation showed that CIMT is superior to conventional care in patients between 3 and 9 months after a stroke [24]. CIMT was developed on the basis of the finding that disuse of an extremity leads to functional deterioration. The constraint element enforces active movement of the affected limb by immobilizing the unaffected. While this trial clearly demonstrated the lasting benefits of CIMT, the controlled comparison depends on the nature of conventional care. In most countries and medical systems, conventional care in the time frame of 3-9 months is limited to low-frequency outpatient therapy, e.g. physiotherapy and occupational therapy once a week. It seems expected that such low intensity is less effective than CIMT which engages patients in training for several hours per day. Hence, it is unknown whether it is simply the intensity that renders CIMT more effective, or it is the strictly unilateral nature of the training, or any other element of therapy.

Progression is another therapeutic principle that may be a key to success. Adding to the complexity, the loading or the velocity of training provides a constant stimulus for learning as the patient continues to improve [25].

Robots were initially developed as assistants to the therapist enabling highly repetitive, uniform passive movements. In patients with spinal cord injury, robotic gait training may be superior to standard overground training [26], but further studies are needed to confirm these differences. For patients with brain lesions, various studies using different upper or lower extremity robotic training devices were performed and showed inconsistent results ranging from inferiority to equality as compared with conventional physiotherapy [27,28].

Robots, however, enable a form of standardized precision training that differs from training under the direction or with the assistance of a therapist. Robots may complement therapist-based training. Precision movement training is often implemented in a computer game. Gaming environments may also increase the motivation for training by introducing reward or competition with others. Finally, robots may partially support the patient's movement (assistance mode) or may perturb the patient's movement, e.g. by applying forces that counteract the patient. Whether assistance or perturbation per se are effective training elements remains to be investigated.

If training is performed in a setting in which plastic reorganization of the brain is facilitated, its effects may be higher, faster and/or longer lasting. Repetitive TMS (rTMS) or transcranial direct current stimulation provide such a setting. They likely work by modulating cortical excitability (in the case of rTMS by increasing or decreasing excitability depending on the frequency of the repetitive stimulation) [29]. As compared with sham stimulation, 10 sessions of rTMS as an adjunct to conventional physical and occupational therapy starting 5-10 days after stroke improved disability (as measured by the Barthel index) and impairment (NIHSS) [30]. In a small randomized trial, rTMS in combination with task-oriented walking exercises improved gait symmetry more than walking exercises alone (combined with sham stimulation [31]. Transcranial direct current stimulation in combination with conventional physical therapy improved arm impairment and activity of daily living function more than physical therapy alone (combined with sham stimulation) [32]. Positive effects of rTMS and transcranial direct current stimulation were also reported for the treatment of aphasia [33,34].

Drugs potentially improve the effects of rehabilitative training interventions. One larger trial in 118 acute ischemic stroke patients found that adding fluoxetine to standard rehabilitation improves motor impairment at 3 months [35]. The mechanisms by which this selective serotonin reuptake inhibitor exerts this effect are unknown.

Neurorehabilitative interventions are multidimensional often combining different treatments such as exercise and brain stimulation. Finding optimal therapy protocols is much more complex than in pharmacological therapy because many parameters need to be optimized, such as timing, intensity, duration, and rest periods. The design of a clinical trial in neurorehabilitation is further complicated by the fact that valid control interventions or groups are difficult to define. Because there is no placebo pill to a training therapy, it is difficult and often impossible to double/single-blind the trial. This leaves the investigation vulnerable to placebo effects, e.g. seeing an expensive robot equipment may introduce a placebo effect over the conventional physical therapist. In addition, the outcome measures and scales for movement, speech language and cognitive function all have their shortages. There is little agreement which scale best reflects a clinically relevant treatment effect. Because training interventions require large amounts of therapist time, trials in neurorehabilitation are expensive.

These factors all contribute to the fact that little evidence exists in neurorehabilitation today. Large and expensive trials are risky as long as optimal treatment protocols are not known. Small studies with less than 100 participants are common. Clinical guidelines therefore remain brief and leave much room for personal experience as well as treatments that can still be considered experimental.