Probabilistic mapping of tremor control and gait ataxia risk in deep brain stimulation

Deep brain stimulation (DBS) is an effective therapy for medication-refractory essential tremor, yet clinical outcomes remain highly variable and stimulation-induced side effects, particularly gait ataxia, frequently limit long-term benefit. In routine clinical practice, it is difficult to predict which stimulation settings will provide optimal tremor control while avoiding disabling motor complications. We analysed long-term tremor and gait outcomes in 73 patients from two DBS centres with chronic bilateral (sub)thalamic DBS, assessed before surgery and at follow-up after 12 months using blinded ratings of standardized video recordings. We evaluated the influence of demographic, clinical, and stimulation parameters and applied voxel-wise probabilistic stimulation mapping to identify anatomical regions associated with tremor suppression and stimulation-induced gait ataxia. DBS produced a marked reduction in tremor (46.3% ± 4.3), with 77% of patients free of functionally disabling tremor (> 2 cm amplitude) at follow-up, but 44% exhibiting clinically relevant gait ataxia. No demographic, clinical, or stimulation parameter was independently associated with tremor improvement. However, probabilistic mapping revealed a tremor "sweet spot" just superior to the medial subthalamic nucleus, and stimulation field-based modelling predicted individual tremor outcome in leave-one-out cross-validation, explaining 21% of variance. In contrast, higher stimulation amplitudes and higher age at surgery were associated with greater stimulation-induced gait worsening, which localized to posterior-inferior stimulation in the subthalamic area. Integrating these factors enabled leave-one-out cross-validated prediction of gait outcome, explaining 49% of variance. Together, these results show that the clinical variability of DBS in essential tremor reflects both spatial targeting and patient-specific susceptibility. Individualized prediction models based on probabilistic stimulation mapping can integrate these factors to estimate stimulation outcome.