Mindfulness and Biofeedback for Anxiety Disorders

This was an open-label, randomized, parallel-group study conducted at the Department of Psychotherapy of [Blinded for Review] between July 2022 and February 2025. Adult inpatients (age 19-74 years) with a first-time diagnosis of anxiety disorder according to ICD-10 codes F40 and F41 were eligible. Exclusion criteria included photosensitive epilepsy, severe cognitive impairment, exacerbation of mental illness, lack of motivation, and inability to provide informed consent.

Participants were randomly assigned (simple unrestricted randomization) to one of three groups:

BFB group (n=76): 10 daily 60-minute individual sessions of mindfulness training using the "Reacor" adaptive biofeedback system, which records physiological parameters (heart rate, EEG alpha rhythm) and converts them into real-time auditory and visual signals.

Medication therapy (MT) group (n=46): escitalopram 10 mg once daily from day 1, continued through follow-up.

Combined therapy (CT) group (n=58): both BFB training and escitalopram 10 mg/day as above.

All patients received standard inpatient medical care. No concomitant psychotherapy, physiotherapy, or additional psychotropic medications were allowed.

Assessments were conducted at three time points: baseline (T0), immediately after the 10-day treatment (T1), and one month after treatment completion (T2). The primary outcomes were the change from baseline in the Hamilton Anxiety Rating Scale (HARS) total score (assessed by an independent blinded psychiatrist) and the Five Facet Mindfulness Questionnaire (FFMQ) total score (self-reported). Secondary outcomes included HARS subscales (somatic and psychic anxiety), FFMQ subscales (observing, describing, acting with awareness, non-judging, non-reactivity), heart rate, and EEG alpha-rhythm power.

Adherence was monitored through daily rounds and pill counts. A per-protocol analysis was conducted on 180 completers; an intention-to-treat analysis (worst-case imputation) was performed on all 188 randomized patients. Repeated-measures MANOVA and ANOVA with Scheffé post-hoc tests were used. A priori sample size calculation (GPower 3.1.9.7) based on a medium effect size (f=0.30), α=0.05, power=0.80, and correlation among repeated measures r=0.5 indicated a required total sample size of 138. Allowing for 15% dropout, we aimed to recruit at least 159 patients; the final per-protocol sample of 180 exceeded this requirement.