Injury to the CNS activates neurobiologic recovery processes. Elucidating the underlying neurobiology of spontaneous recovery, particularly after focal CNS lesions, is a rapidly evolving field. Animal models of CNS injury have helped build our understanding of the complex molecular and cellular aspects of recovery in the brain and spinal cord. Rapid, significant changes in gene expression and protein production which occur after injury, help or hinder plasticity in perilesional and remote brain and spinal cord areas.

At the cellular level, processes that promote plasticity include neuronogenesis and angiogenesis, axonal sprouting, dendritic arborization, synaptogenesis and silent synapse unmasking, changes in synaptic strength, and changes in glutamatergic and GABAergic tone. Processes that limit plasticity include excitotoxicity, free radical formation, delayed neuronal death, edema, inflammation, expression of growth inhibitors, and glial scar formation. Importantly, these processes peak and wane within the first weeks to early months after injury, underscoring the limited time window for optimized combination with rehabilitation training.

Research in targeting these plasticity modifiers, either by upregulating or downregulating their effects, is actively underway. For example, in animal models, selective serotonin reuptake inhibitors (SSRIs) increase expression of BDNF and synaptogenesis. A recent multicenter, double-blinded, randomized controlled trial compared 3 months of fluoxetine versus placebo in severely impaired stroke patients, started within 5 to 10 days after stroke. Both groups received conventional rehabilitation. At 3 months, patients who received fluoxetine had significantly improved motor impairment and disability scores than those who received placebo [Level 1].2 There appears to be a modest class-wide effect on recovery [Level 1].3

Using training to capitalize on neuroplasticity after brain injury is the objective of neurorehabilitation. The optimal type, timing, and dose of training required to maximize activity-dependent plasticity after CNS injury is an area of active basic and clinical research. Some direction about best practices can be gained from animal stroke studies in the motor system. One clear principle that has emerged from animal models is that injury-induced plasticity requires adjunctive paretic limb training in order to promote recovery. In animal studies, intracortical microstimulation techniques may be used to track in vivo neuroplastic cortical map changes. In healthy animals, skilled training results in the expansion of cortical representations of the training effectors, whereas rote, unskilled movements lead to no change. These cortical map changes are associated with increased synaptogenesis and dendritic arborization. In a nonhuman primate model of motor stroke recovery, lost cortical motor representations do not spontaneously reappear if the paretic arm is not trained and the nonparetic arm is used to compensate. In animals trained to reach and grasp with their paretic arm, perilesional distal forearm representations expand into areas that previously subserved the proximal arm. These results suggest that training directed at reducing impairment promotes neuroplasticity.

How early to begin therapy is a point of debate, given concerns about glutamate excitotoxicity after CNS injury. In rodent models of stroke, paretic limb training started 5 days after stroke results in greater functional gains than the same amount of training started 30 days after stroke. In a recent phase II randomized controlled trial, a faster return to unassisted ambulation was found in patients who were mobilized earlier (< 24 hour after stroke) and who received more intensive gait training compared to those receiving standard of care; a large phase III clinical trial testing this early intervention is currently underway.

The dose of therapy is also an issue of exploration. From animal and human motor learning studies, it is clear that high doses of repetitive, goal-directed practice are required for long-term skill gains. A recent observational study of inpatient and outpatient rehabilitation sessions found the number of functional repetitions to be an order of magnitude less than would be needed for conventional skill gains. Meta-analyses of dose response to training modestly support improved motor outcomes at higher doses of exercise-based therapy.

However, it is possible that too much therapy, particularly initiated early after injury, may be suboptimal. For example, a recent randomized, single-blind phase II trial investigated the long-term motor benefits of different doses of constraint-induced movement therapy (CIMT) initiated within 2 weeks post stroke. CIMT forces the use of the paretic arm through restraint of the unaffected arm. CIMT performed equally well to dose-matched traditional therapy. However, high-dose CIMT, which restrained the nonparetic arm for 90% of waking hours, resulted in less motor improvement. From these results, it appears that moderately intensive therapy delivered early after an acute CNS insult may be optimal for impairment reduction.