Introduction
Chronic obstructive pulmonary disease (COPD) is a progressive respiratory disorder characterized by airflow limitation, dyspnea, and significant extrapulmonary manifestations, including cognitive and emotional impairments (Cleutjens et al., 2016; Dodd et al., 2010). The Official American Thoracic Society Statement defines dyspnea as a subjective experience of respiratory discomfort characterized by sensations that vary in intensity, unpleasantness, and emotional-behavioral significance (Parshall et al., 2012). Its mechanisms are complex, involving physiological, psychological, and emotional factors (Parshall et al., 2012).
Notably, there is often a disconnect between the severity of dyspnea and the clinical severity of the underlying disease; some patients report intense dyspnea despite mild disease, while others with severe pathology remain minimally symptomatic (Banzett et al., 2000; Jack et al., 2004; Teeter & Bleecker, 1998). For this reason. dyspnea in COPD is thought to arise from two interconnected mechanisms: a discriminative process that brings respiratory sensations into awareness and a cognitive-affective process shaped by attention, expectation, mood, and interpretation (De Peuter et al., 2004). As dyspnea worsens, it contributes to a cycle of physical inactivity, muscle deconditioning, and kinesiophobia, reducing adherence to rehabilitation and diminishing quality of life (Roche, 2009).
Pulmonary rehabilitation improves exercise capacity, reduces symptoms, and enhances quality of life in chronic respiratory disease (Donner et al., 2020; Spruit et al., 2013), but some patients continue to experience debilitating dyspnea despite optimal treatment.
Although breathing is primarily automatic, it can be consciously modulated. Biofeedback facilitates this by providing real-time physiological feedback, enabling voluntary control over breathing (Khazan, 2013; Schwartz & Andrasik, 2017). The most recent AAPB/ISNR Evidence-Based Practice document (AAPB, 2023) also provides updated recommendations and efficacy ratings for respiratory and autonomic biofeedback applications, which further support the rationale for implementing such interventions in chronic respiratory conditions.
Biofeedback has shown benefits in conditions including hypertension, anxiety, sleep disturbances, and chronic pain (Fournié et al., 2021), with emerging evidence supporting its role in respiratory diseases (Estève et al., 1996; Gevirtz, 2013; Lehrer & Moritz, 2023). In asthma, heart rate variability biofeedback has reduced medication use and improved symptom severity, though pulmonary function changes may include placebo effects (Lehrer & Moritz, 2023; Lehrer et al., 2004). Despite positive outcomes, broader adoption is hindered by incomplete understanding of the intervention’s key components (Lehrer & Moritz, 2023).
In COPD, devices like Flutter have improved lung function and sputum clearance (Kaja et al., 2020a, 2020b), with biofeedback training also linked to improved function and quality of life (Giardino et al., 2004), while earlier summaries of biofeedback applications (Yucha & Montgomery, 2008) provide historical context rather than current evidence. Similar results have been reported in cystic fibrosis (Delk et al., 1994).
This study investigates the efficacy of respiratory biofeedback training in hospitalized COPD patients, focusing on dyspnea reduction, cognitive-affective outcomes, functional capacity, and quality of life.
Materials and Methods
The study included 35 inpatients with COPD, diagnosed according to GOLD guidelines, admitted to the Pulmonary Rehabilitation Unit at ICS Maugeri in Bari, Italy. Inclusion criteria required FEV1 post brodilatation < 30% (GOLD stage 4) and an mMRC dyspnea score of ≥ 3 despite maximal pharmacological therapy (Bestall et al., 1999). Exclusion criteria included acute exacerbations, non-respiratory conditions, neurological or psychiatric disorders, cognitive deficits, or contraindications to rehabilitation.
Participants were randomly assigned to a control or RBT group using Research Randomizer software (Urbaniak & Plous, 2013) (see Fig. 1). Patients were not allowed to choose their treatment allocation.
Assignment to the Biofeedback or Control group followed computer-generated randomization to avoid selection bias. Comorbidities were extracted from medical records and coded individually (see Table 1). As is typical in late-stage COPD, many patients presented with multiple concurrent medical conditions; therefore, each diagnosis was reported independently rather than categorized as mutually exclusive.Fig.
1Study protocolTable 1Demographic and medical characteristics of participants by groupControl group (N = 15)Biofeedback group (N = 15)p valueM (SD)M (SD)Age72.67 (7.45)71.07 (7.27)0.604Years of education9.00 (3.40)10.33 (3.58)0.263BMI27.89 (6.04)25.86 (4.88)0.239N (%)N (%)Sex1.000 Male14 (93.33)14 (93.33) Female1 (6.67)1 (6.67)Medical condition Cardiomyopathy6 (40.00)7 (46.67)1.000 Ischemic cardiopathy1 (6.67)1 (6.67)1.000 Dilated cardiomyopathy0 (0.00)0 (0.00)– Hypertensive heart disease5 (33.33)4 (26.67)1.000 High blood pressure5 (33.33)13 (86.67)0.008 Atrial fibrillation2 (13.33)3 (20.00)1.000 Dyslipidemia6 (40.00)8 (53.33)0.715 Vasculopathy TSA/Aorta4 (26.67)4 (26.67)1.000 Sarcopenia/Muscle Atrophy2 (13.33)4 (26.67)0.651 Diabetes Mellitus4 (26.67)3 (20.00)1.000Respiratory failure Latent exertional respiratory failure3 (20.00)7 (46.67)0.245 Chronic Hypoxemic1 (6.67)2 (13.33)1.000 Chronic Hypoxemic-Hypercapnic7 (46.67)3 (20.00)0.245 COPD Emphysematous6 (40.00)2 (13.33)0.700 Overlap9 (60.00)13 (86.67)0.700Received therapy Oxygen therapy13 (86.67)9 (60.00)0.215 CPAP/NIV7 (46.67)9 (60.00)0.715Endurance training Cycle Ergometer 0 Watt6 (40.00)5 (33.33)1.000 Cyclette/Treadmill9 (60.00)10 (66.67)1.000 Airway clearance3 (20.00)4 (26.67)1.000p values refer to Mann–Whitney U tests (for continuous variables) and Fisher’s exact test (for categorical variables), used to examine potential statistical differences between groups.
All participants underwent a standard pulmonary rehabilitation program: two weekly 30-min strength training sessions with elastic bands or weights (target Borg score 4–5), five weekly 36-min endurance sessions on a cycle ergometer beginning at 60% of peak 6MWT performance and adjusted per Maltais et al. (1996), and one weekly 45-min educational session addressing lifestyle, dyspnea management, and COPD therapies. The RBT group also received 15 respiratory training sessions (35 min/day, five days/week for three weeks), modelled on van Gestel et al.
(2012). Training focused on correcting dysfunctional breathing patterns—rapid shallow breathing, irregular rhythm, and thoracic over-reliance—through diaphragmatic techniques guided by real-time visual and auditory feedback. Each respiratory biofeedback session followed a standardized structure and was delivered individually.
The session began with a brief baseline recording of the patient’s spontaneous breathing pattern, which allowed the clinician to observe respiratory rhythm, depth, thoracic versus abdominal contribution, and irregularities associated with air trapping or accessory muscle recruitment. Following the baseline, patients engaged in diaphragmatic breathing exercises supported by real-time visual feedback displayed on a monitor. The abdominal strain gauges and the BVP sensor provided continuous information on respiratory excursion and autonomic activity.
Visual feedback consisted of animated elements that rose and fell in synchrony with the patient’s respiratory cycle, enabling them to perceive and gradually correct patterns such as thoracic dominance, shallow rapid breathing, or irregular inspiratory–expiratory ratios. When necessary, paced breathing templates were introduced to guide patients toward slower, more regular diaphragmatic cycles, while auditory cues were used to reinforce stable breathing rhythms and prevent hyperventilation. Throughout the session, the therapist continuously adjusted pacing and provided verbal instructions to consolidate effective breathing strategies.
The therapeutic goal of the protocol was to reduce dysfunctional breathing patterns associated with dyspnea, improve respiratory efficiency through enhanced diaphragmatic engagement, and promote greater autonomic stability. All corrections were made dynamically based on real-time physiological signals, ensuring that each session was tailored to the patient’s current respiratory performance and tolerance.
Sessions could be paused or rescheduled when patients experienced fatigue, dyspnea escalation, oxygen desaturation, or anxiety. This flexibility ensured safety and individualized tolerance of the intervention.
All biofeedback sessions were continuously supervised by a trained clinical psychologist to ensure safety and correct breathing execution. Control rehabilitation sessions were also supervised by physiotherapists and nursing staff according to unit protocols.
Physiological signals were monitored via abdominal sensors and BVP on the index finger, transmitted to the ProComp5 Infiniti system and visualized through Biograph Infiniti software (Thought Technology, Montreal, Canada). At baseline assessments included the mMRC (Bestall et al., 1999), Barthel Dyspnea Index (Vitacca et al., 2016), CAT (Jones et al., 2009), six-minute walk test (6MWT) (Holland et al., 2014; ATS, 2002), ABG, spirometry, and quality of life questionnaires (EUROQOL-5D, SGRQ) (Jones et al., 1992; Balestroni et al., 2012). Psychological and cognitive assessments comprised the MoCA and HADS (Iani et al., 2014; Santangelo et al., 2015).
Medication use, oxygen therapy, and ventilation were tracked during the intervention. At the end of the training period, a re-test was conducted employing psychological, cognitive, respiratory, and quality of life questionnaires.
Data analysis used Bayesian repeated-measures ANOVA (JASP, Version 0.19.3) to evaluate main effects of Time, Group, and their interaction, with model selection guided by Bayes factors (BF10, BFincl) (Lee et al., 2014). Pre-post changes were normalized using MCIDs to assess clinical relevance (Jaeschke et al., 1989; Norman et al., 2003), with values ≥ 1 reflecting meaningful improvement.
Results
A total of 30 participants (Biofeedback group = 15; Control group = 15) completed the study. Table 1 presents demographic and clinical characteristics. Five participants did not complete the intervention: two withdrew, one was discharged early, one was transferred due to complications, and one deceased.
On average, participants completed 13.53 ± 1.68 of the 15 planned sessions.
Results are reported across three domains: (1) respiratory symptoms and functional performance, (2) quality of life and disease impact, and (3) cognitive and emotional functioning. Bayesian repeated-measures ANOVA was applied to examine Time (pre vs post), Group (Biofeedback vs Control), and Time × Group interaction. Model comparison used Bayes factors (BF10), and effect inclusion was assessed using BFincl values.
MCID-normalized changes support clinical interpretation. Full statistics are in Appendix Table 2.
At baseline, no significant differences were found between groups (p > .05), except on the MoCA Attention subscale, where the Biofeedback group scored higher (p = .032).
Respiratory Symptoms and Functional Outcomes
Across the CAT, mMRC, and Barthel Dyspnea scales, the model including only the effect of Time was consistently best supported (BF10 = 1). There was extreme evidence for improvement over time: CAT (BFincl = 1549.29), mMRC (BFincl = 28,464.29), and Barthel Dyspnea (BFincl = 81,913.38). Evidence for a Time × Group interaction was weak to moderate (BFincl range: 0.46–0.60).
Functional measures followed a similar pattern. For the SPPB, Time was strongly supported (BFincl = 810.63), and interaction effects were minimal (BFincl = 0.55). The 6MWT also showed significant improvements over time (BFincl = 320.38), with weak support for interaction (BFincl = 0.78).
These results indicate that both groups benefitted significantly from standard rehabilitation, with no additional gains from biofeedback in these domains.
Quality of Life and Disease Impact
The SGRQ total score improved significantly over time (BF10 = 1; BFincl = 1276.50), with minimal support for interaction (BFincl = 1.31). Similarly, the EQ-5D showed moderate evidence for Time (BFincl = 2.26) and limited evidence against the interaction effect (BFincl = 0.54). While both groups improved, the Biofeedback group showed slightly greater improvements in the EQ-5D “Usual Activities” dimension, though overall group differences remained inconclusive.
Cognitive and Emotional Outcomes
Cognitive outcomes revealed more substantial differences. The MoCA total score favored the full model with Time, Group, and interaction effects (BF10 = 1). Evidence was strong for Time (BFincl = 690.62), Group (BFincl = 15.10), and the Time × Group interaction (BFincl = 20.41).
Subdomain analysis showed meaningful interaction effects in Memory Recall (BFincl = 4.71) and Naming (BFincl = 5.62), with greater gains in the Biofeedback group. Other domains, like Executive Function and Attention, showed higher baseline scores in the Biofeedback group but no significant interactions.
For anxiety (HADS-A), the model including Time and Group was best supported (BF10 = 1), with moderate evidence for both (BFincl = 2.65 and 3.01), and weak evidence against interaction (BFincl = 0.77). In contrast, depression scores (HADS-D) supported the full model (BF10 = 1), with strong evidence for Time (BFincl = 5.40), Group (BFincl = 10.68), and interaction (BFincl = 8.92). These results indicate greater emotional improvements in the Biofeedback group, particularly for depression.
A summary of changes in all main outcomes is presented in Fig. 2.Fig. 2Pre- post-interventions changes in all outcome measures
Normalized MCID.
To aid clinical interpretation, Fig. 3 presents pre-post changes normalized by MCID. Negative values reflect symptom improvement (e.g., dyspnea, depression), and positive values indicate functional or cognitive enhancement.
Dashed lines at ± 1 represent clinical significance thresholds.Fig. 3Mean pre-post change normalized by minimal clinically important difference (MCID). Note: Bars represent 95% confidence intervals.
Values < 0 indicate symptom improvement; values > 0 indicate functional or cognitive enhancement. Dashed lines at ± 1 represent the threshold for clinical significance
Both groups experienced meaningful improvements in respiratory and functional domains. However, the Biofeedback group showed notably greater gains in cognitive and emotional outcomes. MoCA improvements exceeded four times the MCID, while changes in the Control group did not reach clinical relevance.
Depression scores (HADS-D) improved beyond the MCID in the Biofeedback group only. Quality of life (SGRQ) improved in both groups, with slightly larger—but not clinically significant—gains in the Biofeedback group.
In summary, while standard rehabilitation effectively improved physical and respiratory outcomes across groups, the biofeedback intervention was associated with greater benefits in cognitive and emotional functioning, suggesting a potentially unique therapeutic role for respiratory biofeedback in advanced COPD.
Discussion
In contrast to prior studies reporting reductions in dyspnea following biofeedback interventions in COPD (De Souto Barbosa et al., 2023; Giardino et al., 2004), our Bayesian analysis did not support a clear advantage of biofeedback-enhanced physiotherapy over standard rehabilitation for reducing breathlessness. It is important to consider that many of these investigations employed uncontrolled or single-arm designs (e.g., De Souto Barbosa et al., 2023; Giardino et al., 2004). Trials including a control condition are relatively scarce and tend to show more heterogeneous effects, which may partly explain the discrepancies between their findings and those observed in our randomized protocol.
For instance, De Souto Barbosa et al. (2023) observed improvements in dyspnea and 6MWT performance, highlighting the synergistic potential of combining respiratory training with biofeedback. Similarly, Giardino et al.
(2004) and Wu et al. (2024) reported broad benefits, including improved autonomic regulation and subjective respiratory relief, suggesting biofeedback exerts both physiological and psychological effects.
Our study did not reveal significant between-group differences in standard clinical outcomes such as CAT, SPPB, or 6MWT. This divergence likely reflects differences in patient characteristics and treatment settings. Unlike prior studies involving patients with mild-to-moderate disease, our cohort consisted exclusively of individuals with very severe COPD (GOLD 4) in maximal pharmacological therapy, all undergoing comprehensive inpatient rehabilitation.
These patients had already received maximal conventional therapy. In this context, biofeedback was implemented as a supplementary, last-resort intervention—potentially limiting its observable additive impact due to ceiling effects in physical performance.
Although respiratory biofeedback may be more effective in earlier stages of the disease, when residual autonomic plasticity is preserved, our results demonstrated clinically meaningful within-group improvements in cognition (MoCA) and depressive symptoms (HADS-D) among patients in the biofeedback group. This suggests that biofeedback may provide unique benefits in neuropsychological domains often under-addressed by traditional pulmonary rehabilitation programs.
These findings support emerging frameworks that emphasize the interconnectedness of physiological regulation and cognitive-emotional functioning in COPD (Cleutjens et al., 2016; Dodd et al., 2010). Cognitive impairments and affective disturbances are highly prevalent in late-stage COPD and often resist pharmacological and exercise-based interventions. The observed improvements in executive function, attention, and memory may reflect enhanced autonomic balance and cerebral perfusion, consistent with prior studies linking HRV biofeedback to modulation of fronto-limbic networks (Giardino et al., 2004; Wu et al., 2024).
Furthermore, the significant reduction in depressive symptoms in the biofeedback group highlights its psychological utility. Mechanisms likely include increased self-efficacy, emotional self-regulation, and improved interoceptive awareness—factors known to enhance emotional resilience. These effects are consistent with prior research suggesting that biofeedback enhances perceived control and reduces psychological distress (Chen & Guo, 2016).
Although physical performance did not significantly differ between groups, the biofeedback group reported larger improvements in health-related quality of life (SGRQ), nearing the threshold of clinical relevance. This suggests that biofeedback may indirectly influence functional outcomes by boosting emotional well-being and motivation. Given that poor adherence is a major barrier in COPD rehabilitation, improvements in engagement and mood could have long-term functional implications (Chen et al., 2015) that deserve further investigation.
However, because outcomes were collected only during the inpatient stay, we were unable to determine whether the cognitive and emotional benefits observed in the biofeedback group persisted after discharge, as no post-discharge follow-up was performed.
Overall, our findings emphasize the importance of tailoring biofeedback interventions based on disease stage and therapeutic setting. While early-stage COPD patients may gain more from biofeedback in terms of physical outcomes, those with advanced disease may benefit primarily in cognitive and emotional domains. The use of MCID-normalized outcomes in our analysis underscores the clinical significance of these changes, even in the absence of between-group differences in conventional respiratory endpoints.
Importantly, the observed gains in psychological and cognitive functioning—domains critical for long-term rehabilitation success—support the inclusion of biofeedback in multidimensional COPD care models. These results align with current calls for broader frameworks that incorporate neuropsychological and behavioral metrics alongside traditional pulmonary indices (Bonini et al., 2020; Demeyer et al., 2014).
Our study also presents several methodological strengths. Conducted in a high-intensity, real-world rehabilitation setting, it provides strong translational relevance. By targeting a severely impaired clinical subgroup with limited treatment options, we address a significant gap in the literature.
The protocol was rigorously standardized, delivered by trained personnel using validated, multimodal biofeedback tools, ensuring high fidelity and reproducibility. We also used a robust analytic framework, combining Bayesian statistics with MCID-based interpretation, offering nuanced insight into treatment effects in small, heterogeneous samples.
Adherence to the intervention was high despite the complexity of the patient population, supporting its feasibility and acceptability in clinical practice.
Future research should explore the neurobiological mechanisms underlying biofeedback’s cognitive and emotional effects, potentially using fMRI or near-infrared spectroscopy to examine brain–heart interactions. Large-scale, multicenter trials with extended follow-up are needed to determine the durability of these effects and their impact on outcomes such as hospital readmissions and mortality. Combining biofeedback with behavioral therapies (e.g., CBT or mindfulness) may also enhance its psychological impact, particularly in patients with high emotional comorbidity.
Implementation science will be essential to identify barriers in resource-limited or home-care environments.
In conclusion, this pilot study contributes to the growing body of evidence supporting the use of respiratory biofeedback in advanced COPD. Although biofeedback did not yield superior outcomes in physical performance or dyspnea, it produced clinically meaningful improvements in cognitive function, depressive symptoms, and quality of life. These findings suggest that biofeedback may be a valuable adjunctive tool for addressing psychological and cognitive challenges in late-stage COPD, especially when conventional options have been exhausted.
While not a replacement for standard rehabilitation, biofeedback offers a complementary approach that supports holistic, multidimensional care.