Mindfulness Development via Biofeedback for Anxiety in Hospitalized Patients: A Randomized Controlled Trial

Background: Anxiety disorders are prevalent and debilitating. Given challenges in maintaining mindfulness at one-month remission, integrative approaches targeting anxiety and psychological resources (especially mindfulness) have gained interest. This study aimed to assess the effectiveness of adding biofeedback to the treatment of anxiety disorders. Methods: Inpatients with anxiety disorder (ICD10 F40/F41) were randomized to three groups: (1) biofeedback for enhancing mindfulness (BFB, n = 76); (2) medication therapy (MT, escitalopram 10 mg/day, n = 46); and (3) combined therapy (BFB + MT, CT, n = 66). Per-protocol analysis: 180 completers; ITT: all 188 randomized patients. The primary outcomes were changes in the Hamilton Anxiety Rating Scale (HAM-A) total score and the Five Facet Mindfulness Questionnaire total score, assessed at baseline, after 10 days of treatment, and at one-month follow-up. The HAM-A assessor was blinded to group allocation. Results: In the per-protocol analysis, the BFB and CT groups showed a significantly greater reduction in the HAM-A score than the MT group did. For example, post-treatment, the mean difference between BFB and MT groups was -6.30 points (95% CI: -9.12 to -3.48, p = 0.00046). The BFB and CT groups did not differ significantly. For mindfulness, the BFB and CT groups showed an increase, whereas the MT group showed a decrease (BFB vs. MT: mean difference 30.28 points, 95% CI: 23.54-37.02, p < 0.000001). Intention-to-treat (ITT) analysis (n = 188) confirmed the robustness of these findings. Conclusions: BFB appears to be a promising approach for reducing anxiety and fostering mindfulness skills; these findings suggest it may be worth exploring further as a potential component of comprehensive therapy for anxiety disorders. Clinical Trial Registration: ClinicalTrials.gov NCT07628153 (retrospectively registered). The trial was retrospectively registered, and the primary analysis was conducted on a per-protocol population.