In the Netherlands, each year more than 32,000 patients sustain a stroke [1, 2], and the incidence is expected to have increased by 30-45% by 2015 . About 80% of stroke survivors suffer from an upper limb paresis immediately after stroke , hampering movement of the paretic arm and bimanual coordination .
Spontaneous recovery after stroke is limited, and knowledge about which mechanisms lead to spontaneous recovery is incomplete . Restitution of non-infarcted penumbral tissue (i.e., reestablishment of metabolism in the tissues surrounding the infarcted area)  and resolution of diaschisis (i.e., relief of suppression of anatomically related brain areas) , together with recovery of neurotransmission in spared tissue near and remote from the infarct [7, 8], are held mainly responsible for the nonlinear recovery pattern observed in the first weeks post-stroke .
In addition to these early post-stroke developments, functional recovery of the upper extremity is promoted by plastic changes in the functioning of the brain, which, in general, also occur in learning . These experience-induced changes are brought about by a combination of neural repair and neuro-anatomic reorganization, and include greater excitability and recruitment of the neurons in both hemispheres, sprouting of dendrites, and strengthening of synaptic connections [7, 10–16].
Although the aforementioned processes may suggest an optimistic view on post-stroke recovery, only one third of all stroke patients regain some dexterity within six months using conventional treatment programs . This means that 60-70% of all stroke survivors will continue to experience major functional limitations of the upper extremity , which are associated with diminished health-related quality of life after stroke [7, 18].
In light of this grim prospect, it is encouraging that recent studies, capitalizing on the concept of experience-induced neuroplasticity, have produced promising results using specific interventions aimed at arm-function improvement. One such intervention is bilateral arm training with rhythmic auditory cueing (BATRAC), which has been shown to have beneficial effects on the paretic arm in chronic stroke patients , possibly as a result of changes in contralesional cortical networks . This suggests that motor function in the impaired paretic arm may be regained by exploiting interhemispheric interactions . In particular, based on the principle of interhemispheric recruitment from the non-affected hemisphere (i.e., exercise-induced neuroplasticity by means of "neural cross-talk"), BATRAC may serve as an effective therapy for patients in whom the corticospinal tract (CST) system is seriously affected  - a group of patients for which effective therapies are urgently lacking and prospects of arm function recovery are particularly poor [8, 17]. Furthermore, a recent meta-analysis on upper limb robotics suggests that distally oriented repetitive bilateral arm training is more effective than a more proximally oriented approach . In addition, longitudinal studies with repeated measurements in time suggest that an early return of wrist and finger extension is a pre-requisite for regaining some dexterity [24–26]. These findings support the hypothesis that the effectiveness of BATRAC may be enhanced by performing repetitive flexion and extension movements of wrist and fingers, rather than rhythmic movements of more proximal parts of the arm.
In contrast, various controlled trials have suggested that intensive unilateral training by constraining movements of the non-paretic arm (constrained-induced movement therapy, CIMT) is an effective method for improving upper limb function in chronic stroke patients [18, 27–29]. This suggests that training may also induce beneficial changes in the affected rather than the non-affected hemisphere and raises the question whether the improved functionality of the paretic arm with BATRAC indeed results from exploiting interhemispheric interactions, or merely from training with the affected arm .
The ULTRA-stroke program entails a randomized clinical trial (RCT) in which the merits of both BATRAC and CIMT are compared with each other and those of an equally intensive (i.e., dose-matched) conventional treatment program. To this end, participants will be divided over three intervention groups and the effects of the interventions will be assessed prior to training (t0), after 6 weeks of training (t1), and 6 weeks after training (t2). The primary aim of the ULTRA-stroke program is to assess the relative effectiveness of the three interventions on a group level and as a function of patient characteristics. In addition, the program aims for delineating the functional and neurophysiological changes that are induced by those interventions. This led to the following research questions:
▪ Which of the three interventions - modified BATRAC, modified CIMT, or a dose-matched conventional treatment (DMCT) - is more effective in terms of recovery of (unimanual and bimanual) hand and arm function in subacute stroke patients?
▪ How are the observed changes in functionality related to changes in peripheral stiffness, interlimb interactions, and cortical inter- and intrahemispheric networks?
The effectiveness will be assessed by a range of functional outcome measures pertaining to motor ability of the paretic arm, activities of daily living (ADLs), bimanual coordination, and peripheral motor functioning. Besides further elucidating the merits of bilateral versus unilateral upper limb training in general, the study will generate specific insights into the effectiveness of distally oriented (modified) BATRAC, specifically aimed at improving wrist and finger extension [24–26], and into the effectiveness of (modified) CIMT as applied in a thus far hardly studied stage after stroke .
In light of contrasting results and divergent perspectives regarding underlying mechanisms of current interventions , their potential dependence on the neurological characteristics of stroke survivors will also be a topic of investigation in the ULTRA-stroke program. It has been proposed that the effectiveness of CIMT is dependent on CST integrity [32–34], which is essential for motor control of the distal part of the upper limb. On the other hand, BATRAC may be expected to be less dependent on the integrity of the CST, as it appears to induce reorganizations in cortical inter- and intrahemispheric networks [21, 22]. To cope with this issue, participants will be categorized in terms of their motor ability of the distal part of the arm .
In short, we hypothesize that both modified CIMT and modified BATRAC significantly improve upper limb function when compared to DMCT. Modified CIMT is expected to have a larger impact on those subjects who already showed some dexterity at recruitment than on subjects that were more restricted in this regard, given the proposed importance of CST integrity for motor control of the distal part of the upper limb . Modified BATRAC, on the other hand, is expected to be also effective for the latter group of subjects, thanks to influences stemming from and reorganizations in the contralesional hemisphere (see also [35–37]). The effects of modified BATRAC and modified CIMT are both expected to sustain during the follow-up period of 6 weeks. To uncover the mechanisms associated with intervention-induced functional improvement, three kinds of analysis will be included.
First, endpoint mechanical behavior of the paretic wrist, under both passive and active conditions, will be assessed to determine both paresis and stiffness, the latter described in terms of intrinsic visco-elasticity and reflexive feedback properties . In active posture tasks the negative signs of post-stroke limb dysfunction prevail (viz., paresis and poor adaptation of reflexes; ). Under passive conditions, however, enhanced joint stiffness and hyper-excitability of the reflex loop (viz. enhanced tendency for synchronous and self-sustained firing of the motor neuron pool) are evident [40, 41]. Because the spinal reflex loop is under control of higher brain areas, loss of CST integrity and persistent central nervous system reorganization is anticipated to be related to high joint stiffness, absence of reflex modulation, and signs of hyper-excitability of the reflex loop. Assuming that increased joint stiffness is specifically associated with loss of CST integrity, modified BATRAC is expected to be more effective in reducing these effects than modified CIMT for participants with minimal hand/finger control.
Second, bimanual coordination will be examined in all detail. Bimanual coordination is characterized by interlimb interactions [42, 43] that result in stabilization of specific bimanual coordination patterns [5, 44–49]. The success of bilateral training protocols (such as BATRAC) has been ascribed to the presence of such interlimb interactions , suggesting that influences from the contralesional hemisphere are beneficial for performance of the paretic limb. Bilateral training may also induce adaptations in these interactions, potentially strengthening its advantageous influence on paretic arm performance as well as improving bimanual performance. Therefore, modified BATRAC is expected to induce more improvement in these interactions than both modified CIMT and DMCT.
Third, treatment-induced neuronal reorganization will be identified using magneto-encephalographic recordings (MEG). Given its high temporal resolution, MEG is a very suitable non-invasive tool for studying patterns of correlated neuro-electrical activity within and across hemispheres. MEG recordings of unimanual and bimanual tasks will be conducted prior to and after interventions to investigate treatment induced changes in these patterns.
Functional MRI studies and TMS studies already indicated that, during paretic arm movement, CIMT results in increased metabolic activation in the primary sensorimotor cortex of the affected hemisphere [50–62], whereas BATRAC results in increased metabolic activation in the contralesional cerebrum and ipsilesional cerebellum [20, 21].
Modified CIMT is hence expected to result primarily in changes in ipsilesional hemisphere functioning, i.e., greater activity in the primary sensorimotor cortex of the affected hemisphere and increased phase synchronization between regions surrounding the lesion (we note that assessing the latter requires the high temporal resolution of encephalographic recordings), which may be related to restitution of its former functionality. In contrast, modified BATRAC is expected to primarily induce adaptations in the contralesional hemisphere, enhanced activity in the (ipsilesional) cerebellum (possibly reflecting enhanced timing abilities), and a considerably greater increase in the degree of phase synchronization between the lesioned hemisphere and the contralesional hemisphere than will occur as a result of either modified CIMT or DMCT. This finding would indicate compensatory cortical reorganizations in which the coupling to the nonaffected hemisphere acquires a special role in the motor control of the paretic arm.