Review of 'Refining the Deep Brain Stimulation Target within the Limbic Globus Pallidus Internus for Tourette Syndrome'

The authors combine clinical outcome data from 15 Tourette patients who had DBS in the ventral globus pallidus pars interna (GPi) with finite element modeling of the expected likelihood of activating axons passing through each voxel of space in and around the active contacts on the DBS leads. This creates a t image mapping clinical improvement in a given measure to location of stimulation. The authors corrected for multiple comparisons based using FSL Randomise. The report also includes key information on clinical benefit and side effects, but this 3D statistical map is of greatest interest. 

The study has two key limitations, one of which can be improved by reanalysis. 

The primary clinical variables tested are change from before to after surgery. This approach, while clinically interesting, confounds actual effects of stimulation with change in disease state over time, expectation bias, placebo benefit, and any microlesion effect of lead placement. Blinded on- and off-stimulation assessments after surgery are more relevant to DBS effects per se. However, the lag between stimulation and benefit is not well understood for tic disorders, complicating this ideal study design.

The authors use permutation statistics to create the t distribution at each voxel, and test for multiple comparisons. The voxelwise permutation approach should take care of the concern that (unlike in neuroimaging data generally) the degrees of freedom vary substantially over the image. However, as we have discussed elsewhere (Eisenstein et al 2014), this approach does not correct appropriately for the possibility that stimulation anywhere may be effective (the "sample mean problem"). 

Nevertheless, the overall approach of this report is strong, and the authors are to be congratulated on a very useful report that links clinical symptom improvement in TS to the exact location of stimulation in 3D space. This approach can not only be expected to improve clinical outcomes but also to better understand the pathophysiology of tics and associated symptoms.