Predicting the Response to Non-invasive Brain Stimulation in Stroke

摘要

Neuromodulatory non-invasive brain stimulation (NIBS) techniques are experimental therapies for improving motor function after stroke. The aim of neuromodulation is to enhance adaptive or suppress maladaptive processes of post-stroke reorganization. However, results on the effectiveness of these methods, which include transcranial magnetic stimulation (TMS) and transcranial direct current stimulation (tDCS), are mixed. The results of recent large clinical trials and meta-analyses range from no improvement in motor function ( 1 , 2 ) to moderate improvement ( 1 – 6 ) at the group level. Though evidence supporting efficacy is better for TMS ( 7 ) than for tDCS ( 6 ), individual stroke patients' response to NIBS is nevertheless extremely variable ( 8 – 11 ). This is reminiscent of the development of other stroke therapies, such as thrombolysis and mechanical thrombectomy, where early studies were largely mixed before patient selection was refined ( 12 , 13 ). NIBS in stroke faces a similar challenge of refining patient selection and individualizing protocols to determine its therapeutic potential. The variable response to NIBS in stroke patients is a byproduct of multiple factors that influence response to NIBS in healthy controls ( 14 , 15 ), as well as factors that influence the response specifically in stroke patients ( 8 ). The former include factors such as age, gender, anatomical variability, intake of stimulant substances, and baseline neurophysiological state but also technical factors such as stimulation intensity, TMS coil orientation, and stimulation duration ( 16 – 18 ). Specifically in stroke patients, symptom severity, size and location of lesions, stroke etiology, and time from symptom onset to intervention influence the response to NIBS as well. Importantly, these different variability-causing factors interact to affect the response to NIBS, such as the potential amplification of inter-individual differences in brain anatomy ( 19 , 20 ) by stroke lesions ( 21 , 22 ). Such interactions make understanding the causes of NIBS response variability in stroke challenging. Although the need for individualized stimulation protocols in stroke patients is widely accepted, it is still unclear exactly how this will be achieved. At the very least, the factors influencing variability in healthy subjects should be controlled as much as possible through appropriate and careful study design ( 23 ) and checklist-based reporting of factors during data collection ( 24 ). To address the specific factors for stroke, patient selection for NIBS should be informed by pathophysiological processes. This requires that we know which processes are relevant, that we are capable of measuring them, and that we know the optimum timing and patient-related characteristics for treatment administration. Models of Reorganization as a Basis for Stimulation Protocols Until recently, NIBS protocols have mostly been based on the interhemispheric competition model ( 25 , 26 ), which postulates that the unaffected hemisphere overly inhibits the affected hemisphere. Despite NIBS strategies based on this model being largely ineffective at the group level ( 27 – 30 ), it is still a popular approach used by several recent ( 9 ) and ongoing clinical trials. In severely affected patients in particular, the validity of this model has been questioned ( 31 , 32 ) and an alternative, the vicariation model, suggested ( 33 ). The vicariation model postulates that the function of the unaffected hemisphere compensates for the impairment of the affected hemisphere, thereby presenting an adaptive, rather than maladaptive, process ( 32 , 34 – 37 ). These contradictory models have been unified in the bimodal-balance recovery model, taking us a step further to individualized therapy ( 25 ). This uses a metric, the “structural reserve,” defined as the integrity of the white matter motor pathways, to determine whether the inter-hemispheric competition or vicariation model is applicable in a given patient. According to the model, in patients with high structural reserve, the over-activation of the unaffected hemisphere is maladaptive, while in patients with low structural reserve, this over-activation is compensatory. Supporting this model, severely affected patients, with presumably low structural reserve, have poorer outcomes when inhibitory NIBS protocols are applied to their unaffected hemispheres ( 28 , 37 ), emphasizing the need to modify “one-size-fit-all” NIBS protocols. However, it is yet to be resolved which clinical and imaging characteristics are appropriate proxies for structural reserve. Most evidence thus far comes from studies investigating the ability of these characteristics to predict stroke outcome. White matter integrity, quantified with the fractional anisotropy of white matter tracts on diffusion tensor imaging, is commonly used ( 38 – 42 ). However, a good predictor of stroke outcome (prognostic biomarker) is not necessarily useful for