Electroencephalographia as Predictor of Effectiveness HD-tDCS in Neuropathic Pain: Machine Learning Approach

EEG as Predictor of Effectiveness of HD-tDCS in Treatment of Neuropathic Pain After Brachial Plexus Injury: Machine Learning Approach

Contextualization: Neuropathic pain is a complication present in the clinical picture of patients with traumatic Brachial Plexus injury (BPI). It is characterized by high intensity, severity and refractoriness to clinical treatments, resulting in high disability and loss of quality of life. Due to loss of afferent entry, it causes cortical and subcortical alterations and changes in somatotopic representation, from inadequate plastic adaptations in the Central and Peripheral Nervous System, one of the therapies with potential benefit in this population is the Transcranial High Definition Continuous Current Stimulation (HD-tDCS). Thus, by using connectivity-based response prediction and machine learning, it will allow greater assurance of efficiency and optimization of the application of this therapy, being directed to patients with greater potential to benefit from the application of this approach. Objective: Using connectivity-based prediction and machine learning, this study aims to assess whether baseline EEG related characteristics predict the response of patients with neuropathic pain after BPI to the effectiveness of HD-tDCS treatment. Materials and methods: A quantitative, applied, exploratory, open-label response prediction study will be conducted from data acquired from a pilot, triple-blind, cross-over, placebo-controlled, randomized clinical trial investigating the efficacy of applying HD-tDCS to patients with neuropathic brachial plexus trauma pain. Participants will be evaluated for eligibility and then randomly allocated into two groups to receive the active HD-tDCS or simulated HD-tDCS. The primary outcome will be pain intensity as measured by the numerical pain scale. Participants will be invited to participate in an EEG study before starting treatment. Clinical improvement labels used for machine learning classification will be determined based on data obtained from the clinical trial (baseline and post-treatment evaluations). The hypothesis adopted in this study is that the response prediction model constructed from EEG frequency band pattern data collected at baseline will be able to identify responders and non-responders to HD-tDCS treatment.

Using connectivity-based prediction and machine learning, the objective is to assess whether characteristics related to baseline EEG predict the response of patients with neuropathic pain after BPI to the effectiveness of HD-tDCS treatment. An observational, retrospective cohort study will be carried out, of predictive response with a quantitative approach, of an applied nature, of an exploratory and open-label type, related to the efficacy of HD-tDCS4x1 in patients with neuropathic pain due to BPI, from an analysis of data obtained from a pilot, placebo-controlled, triple-blind, randomized, crossover type clinical trial, in accordance with the CONSORT guidelines, which will investigate the effectiveness of treatment with HD-tDCS.

Inclusion Criteria:

1. Age over 18 years;
2. Moderate to severe pain score according to the Numerical Pain Scale (4-10);
3. Persistent pain and refractory to clinical treatment for at least 3 months;
4. Appropriate pharmacological treatment for pain for at least 1 month before the start of the study;
5. Not presenting contraindications for Non-Invasive Brain Stimulation;
6. Absence of concomitant diseases of the Central Nervous Sistem or Peripheral Nervous Sistem.

Exclusion Criteria:

1. Failure to sign the informed consent form;
2. Missing two consecutive or three alternate sessions during treatment;
3. Developing a disabling condition that prevents further participation in the study