Biofeedback Training to Increase PCO₂ in Asthma with Elevated Anxiety: A One-Stop Treatment for Both Conditions?

Abstract

Objective:

Anxiety is highly prevalent in individuals with asthma. Asthma symptoms and medication can exacerbate anxiety, and vice versa. Unfortunately, treatments for comorbid anxiety and asthma are largely lacking. A problematic feature common to both conditions is hyperventilation. It adversely affects lung function and symptoms in asthma and anxiety. We examined whether a treatment to reduce hyperventilation, shown to improve asthma symptoms, also improves anxiety in asthma patients with high anxiety.

Method:

One-hundred-twenty English- or Spanish-speaking adult patients with asthma were randomly assigned to either capnometry-assisted respiratory training (CART) to raise PCO₂ or feedback to slow respiratory rate (SLOW). Although anxiety was not an inclusion criterion, 21.7% met clinically-relevant anxiety levels on the Hospital Anxiety and Depression scale. Anxiety (HADS-A) and depression (HADS-D) scales, anxiety sensitivity (ASI), and negative affect (PANAS-N) were assessed at baseline, posttreatment,1-month follow-up, and 6-month follow-up.

Results:

In this secondary analysis, asthma patients with high baseline anxiety showed greater reductions in ASI and PANAS-N in CART than in SLOW (ps≤.005, Cohen’s ds≥.58). Further, at 6-month follow-up, these patients also had lower ASI, PANAS-N, and HADS-D in CART than in SLOW (ps≤.012, Cohen’s ds≥.54). Patients with low baseline anxiety did not have differential outcomes in CART than in SLOW.

Conclusions:

For asthma patients with high anxiety, our brief training designed to raise PCO₂ resulted in significant and sustained reductions in anxiety sensitivity and negative affect compared to slow-breathing training. The findings lend support for PCO₂ as a potential physiological target for anxiety reduction in asthma.

Trial Registration:

clinicaltrials.gov Identifier: NCT00975273.

INTRODUCTION

With over 339 million people worldwide and over 30 million in the United States alone, asthma is one of the most common chronic diseases (1). Despite significant advances in pharmacological treatments for asthma, morbidity and death remain high, underlining the need for better control and prevention methods (2). Mental illness, with co-occurrence ranging from 22–64% (3–6), has been identified as one of the main factors contributing to poor asthma control. Research has found high levels of mental illness in asthma (e.g., 20), which is most consistent for anxiety disorders and asthma (6–19), particularly in severe asthmatics who experience frequent life-threatening asthma episodes (21).

Comorbid anxiety is a risk factor for more significant asthma morbidity (22,23). Symptom overlap may result in errors in diagnosis and treatment, leading to added costs to the health care system (18,24–27). Adverse effects include more prolonged and more frequent hospital stays (28), emergency room visits (29), higher prescription rates, lengthier episodes of oral corticosteroid use (30,31), and greater health care provider use (27). Asthma with comorbid or clinically elevated anxiety has been associated with poorer quality of life, poorer perceived health status (32–35), more frequent reports of psychological asthma triggers (36), and notably greater suicidal ideation, attempts, or completed suicides (11,37). Asthma symptoms experienced in childhood are associated with greater levels of anxiety and shyness (38) and predict the future development of anxiety disorders (39). Anxiety and stress have also been shown to contribute directly to the pathophysiology of asthma by increasing bronchoconstriction (40,41) airway inflammation (42–44), and are prospectively associated with poorer asthma control (45).

Unexpectedly, treatment of comorbid asthma and anxiety can adversely affect both conditions. For example, pharmacological treatments for asthma (e.g., oral corticosteroids, β-agonists, leukotriene inhibitors, theophylline) can induce and worsen panic or anxiety symptoms [46,47]. On the other hand, psychopharmacological treatment can create respiratory side effects that may complicate asthma symptoms. For instance, high doses of benzodiazepines or hypnotics can induce respiratory depression (46–48) and are related to a greater risk of death from asthma (49). Antidepressant, anxiolytic, and hypnotic agent use, which people with asthma use at rates twice as high as the general population, correlated negatively with asthma control (50). Accordingly, psychosocial interventions of comparable efficacy to psychotropic medication for anxiety are preferable. Despite repeated calls for interventions for comorbid asthma and anxiety (5,11,22), trials are largely lacking, and strategies to improve self-management behaviors (47,51), or asthma pathophysiology, show limited efficacy (52–54), or do not target anxiety directly. Two behavioral pilot studies (55,56) offer promising effects but were limited by small sample size or high drop-out rates.

A maladaptive behavior common to both asthma and anxiety is hyperventilation. Decades of research demonstrate the mutually adverse effects of hyperventilation (hypercapnia; below normal levels of PCO₂) on health, anxiety, or asthma patients (57–62). Hypocapnia in both asthma and anxiety is associated with similar symptoms (shortness of breath, palpitations, and faintness (63), negative affect/trait anxiety (64,65), intense stress, and lower quality of life (66). Acute decreases in PCO₂ have been found in anxious patients when subjected to threatening stimuli (67–69), and is linked to increased airway obstruction and hyperreactivity (60,70) in asthmatics. Panic-induced hyperventilation has been shown to precede asthma exacerbation (71), and marked bronchoconstriction was found during exposure to phobic stimuli (41). Psychosocial interventions for anxiety also involve potential risks for patients with asthma, where interoceptive exposure training might induce hyperventilation, which will worsen asthma symptoms (72,73). Reducing hyperventilation may therefore be a key strategy to improve both conditions.

This study aimed to examine the effects of capnometry-assisted respiratory training (CART) on reducing anxiety in asthma. In a randomized control trial (RCT), 120 asthma patients received either 4-weeks of CART to improve asthma pathophysiology by reducing hyperventilation, or a slow breathing control training (SLOW). One in five (21.6%) patients met clinically elevated anxiety. The primary treatment outcome analysis of this data demonstrated sustained elevations in PCO₂ six months posttreatment, accompanied by significant improvements in asthma control (74). Additionally, mechanical lung function (as measured by forced oscillation) during training improved over the slow-breathing control group, which also deteriorated in their tolerance of airway obstruction by methacholine.

Six prior RCTs have tested the efficacy of CART. Four demonstrated the efficacy of CART in reducing panic and anxiety (75–78). Two more trials tested the efficacy of CART on improvements in asthma symptomatology (74,79). The degree to which CART can improve anxiety symptoms in individuals with asthma was not tested. In the present trial (74), the treatment rationale exclusively focused on improvements in asthma symptomatology, and at no point in training were potential effects on anxiety made salient to patients. Nevertheless, anxiety was monitored throughout treatment, enabling us to study anxiety-reducing effects of PCO₂ elevations without creating expectancy effects (and thus biased reporting of anxiety) in patients.

Moreover, the control group in this study was tightly matched in aspects of attention, instrumentation, and credibility. Individuals assigned to CART focused on steadily reducing respiration rate (RR) by paced breathing while at the same time breathing shallower than normal to elevate their PCO₂ using RR and PCO₂ feedback from a portable capnometer. SLOW trained only the slow breathing component using RR, but no PCO₂ feedback, which resulted in only small and transient improvements in PCO₂ (74). This allowed us to isolate the relation between PCO₂ training and anxiety levels, free from patients’ expectancy and nonspecific intervention characteristics. There is general skepticism about whether psychological treatments for chronic disease work beyond nonspecific effects such as expectancy or therapeutic alliance (80).

Based on the previous CART studies (which only included participants with elevated levels of anxiety or panic symptoms), we anticipated that CART, compared to SLOW, would reduce anxiety and negative affect in participants who had clinical levels of anxiety. However, since participants with lower anxiety levels would have little room for anxiety improvement, we did not expect that low anxiety participants would have greater reductions in anxiety in CART than in SLOW. Since our sample was not recruited for high levels of anxiety, our participants had a wide range of anxiety scores (range: 0–16 on the 0–21 Hospital Anxiety and Depression Scale-Anxiety (HADS-A) (81)), with 21.7% of the sample (n=26) having clinically elevated baseline HADS-A scores (HADS-A ≥ 8). Thus, we examined baseline anxiety levels as a dimensional (continuous) moderator of the effect of treatment on improvement in outcome over time, rather than treating baseline HADS-A as dichotomous moderator. We hypothesized a triple interaction between baseline level of anxiety, treatment group, and time, such that participants with higher baseline anxiety would improve more from baseline to 6MFU in CART than SLOW. Positive results would demonstrate the potential of CART training to address comorbid asthma and anxiety unbiased by expectancy effects. Seven covariates were included in our analyses since they are known to affect mood or lung function or airway inflammation parameters: sex, age, BMI, cold and flu symptoms, beclomethasone equivalent doses, frequency of ICS use, and frequency of leukotriene modifier use. Of these, the latter three were particularly important because modification of medication intake could have powerful effects on asthma outcomes and confound biobehavioral treatment effects.

Methods

Participants

This study was a secondary analysis of data collected initially from a randomized controlled trial with 120 adult patients with asthma of all severity grades, aged 18–65, English- or Spanish-speaking (74). Enrollment period was between October 22, 2008 and December 30, 2011. The patients, who were recruited through physician referrals and advertisements, had a wide range of scores on our baseline anxiety measure, HADS-A (range: 0–16 on the 0–21 HADS-A scale), with 21.7% (n=26) of the patients reporting clinically-elevated anxiety symptoms (HADS-A ≥ 8). Inclusion criteria for the study were: a verified asthma diagnosis (20% drop in forced expiratory volume in the 1^st second [FEV₁] following methacholine challenge or a 12% increase in FEV₁ following nebulizer with 0.083% albuterol, if initial FEV₁ was at least 60% of predicted), agreement to remain on asthma treatment (standard of care) until 1-month follow-up. Short-acting bronchodilators (albuterol) with dose counters embedded in Ventolin-HFA MDI devices (GlaxoSmithKline) were provided throughout the study period. Exclusion criteria included requiring oral corticosteroids in the past three months, present smoking or past smoking (>10 pack-years), uncontrolled medical or psychiatric comorbidity, pregnancy, and patients who experienced stronger exacerbations (peak expiratory flow [PEF]<60% of predicted, together with the development of strong symptoms). Patients were randomly assigned to either 4-weeks of (i) capnometer-assisted respiratory training (CART) or (ii) slow breathing and awareness training (SLOW), using a computer-generated block randomization by initial PCO₂-levels to ensure equal distribution of patients with PCO₂-levels<35 mmHg in both groups. The institutional review boards approved the study at the participating sites (Southern Methodist University [#KS08-051] and Baylor University Medical Center [#008-180]). All patients provided written informed consent.

Interventions

Treatments were identical regarding duration, dose, homework, patient-therapist contact, and biofeedback device use, except feedback of PCO₂ and respiratory rate (RR) provided in CART vs. feedback of only RR provided in SLOW. Comprehensive hand-outs were provided, detailing aspects of pulmonary physiology and the effect of breathing on asthma symptoms. Patients were also given instructions for daily exercises and a portable capnometer with additional pulse oximetry capability (TIDALWAVE^®, Respironics, USA) for in-session and at-home training. PCO₂ and RR were displayed continuously and stored with time and date stamps for compliance checks. All prior-week exercises were downloaded for in-session discussion and further training. Treatments were offered in English or Spanish by doctoral or postdoctoral students.

Capnometry-Assisted Respiratory Training (CART).

The goal of CART was to reach or maintain a PCO₂ of around 40–42mmHg by reducing ventilation primarily through shallow breathing aided by continuous feedback of PCO₂ and RR. Slow, regular, nasal, and abdominal breathing was also emphasized. Patients were educated about the mechanisms triggering asthma symptoms through hyperventilation and hyperpnea. They were prescribed to do homework twice-daily, 17-min, audio-directed, exercises with (a) 2-minute quiet sitting, eyes-closed baseline, (b) 10-minute paced breathing following tones, and (c) 5-minute without pacing. Rising tones indicated inspirations; falling tones, expirations; and silence indicated the pause between expiration and inspiration. The fractional inspiratory ratio (inspiratory time/total time of a breath) was 0.4. The tone pattern was set to correspond to a respiratory rate of 13 breaths per minute in the first week and rates of 10, 8, and 6 breaths per minute for weeks 2–4, respectively. Additional observation of oxygen saturation served to educate patients about the stability of oxygen saturation despite possible symptoms of dyspnea.

Slow breathing and awareness training (SLOW)

Slow breathing and awareness training (SLOW) followed the same structure and dose as CART, but only used the capnometer to display RR for feedback (PCO₂ display deactivated) to achieve reductions in RR with the help of the same weekly homework assignment and schedule, and the same pacing tones (13, 11, 9, and 6 breaths/min) as CART. As in CART, the patients’ goal was to breathe regularly with the tones and maintain their respiration rate as close as possible by using the display of respiration rate from the capnometer. Slow, abdominal, nasal, and regular breathing was also trained, with the rationale focused on irregular and fast breathing (‘chaotic breathing’) as a source of symptoms. Maintenance of a “basic breathing rhythm” was encouraged by inner counting and directing attention to sensations of breathing in and out. The awareness-of-breathing exercises were adapted from an autogenic training protocol (82); however, instructions were restricted to breathing and bodily awareness, not relaxation. Patients were instructed to adjust their inspiratory volume to a “comfortable amount of air.”

Assessments

Assessments took place at baseline and posttreatment, 1- and 6-month follow-up by independent assessors that were blind to the patients’ treatment assignment.

The present report focuses on the psychological effects of the training on the following outcome measures:

HADS (81) measures anxious and depressed mood, with seven questions related to anxiety (HADS-A) and seven related to depression (HADS-D). The scores for HADS-A and HADS-D range from 0–21, and literature supports a cut-off on the HADS-A of 8 points or higher for clinically relevant anxiety (83,84). It is ideally suited for clinical settings because items avoid reference to physical symptoms that overlap with asthma symptoms and have demonstrated sensitivity to change (85). Cronbach’s alpha for HADS-A was .80, .79, .84, and .83, and was .81, .79, .80, and .82 for HADS-D for the four assessments.

Fear of bodily symptoms was assessed with the anxiety sensitivity index (ASI). The 16 items comprising the ASI (86) assess beliefs that unexplained somatic sensations are dangerous and may cause harmful physical, psychological, or social consequences that go beyond any immediate physical discomfort. Items are rated on a scale from 0 (very little) to 4 (very much). Anxiety sensitivity is conceptualized as a dispositional and dimensional construct that determines the tendency to respond fearfully to anxiety symptoms (87,88). Cronbach’s alpha for ASI was .87, .85, .86, and .86 for the four assessments.

Positive and Negative Affect Schedule (PANAS; (89)). Ten items (rated 1=“very slightly/not at all” to 5=“extremely”) of the PANAS measured negative affect (PANAS-N) during the past week. Cronbach’s alpha for PANAS-N was .87, .85, .88, and .88 for the four assessments.

The asthma outcome measures are reported elsewhere (74). Key measures were used here for baseline patient characterization and secondary analysis of the relation between asthma physiology and anxiety. These asthma outcomes included:

The fraction of nitric oxide in exhaled air (FE_NO, in ppb) at a flow of 50ml/s. It was measured with an electrochemical gas analyzer (NIOXmino, Aerocrine, Solna, Sweden), typically interpreted as an indicator of airway inflammation in asthma, as well as mechanical lung function by spirometry (AM2, Jaeger/Toennies, AM2, Höchberg, Germany), with PEF and FEV₁ extracted from the best of three blows.

The Asthma Control Test (ACT; (90,91), a 5-point self-report scale (total score, 5–25; higher scores indicate greater control, with values >=20 indicating well-controlled asthma).

Bronchodilator use was recorded with a dose counter embedded in the shell of the MDI. Patients received refills as needed during the study period. Dosage (converted to beclomethasone equivalent dosage for inhaled corticosteroids (92)) and frequency of intake of maintenance medication including leukotriene modifiers were assessed by self-report at each assessment point, communicating to patients the scientific importance of reporting actual intake, which can be different from prescribed dosages. Patients were instructed to discontinue use of the short-acting bronchodilator for 8 hours before the assessments.

Basal PCO₂ levels, together with basal RR, were assessed during a 2-min quiet sitting in the clinic without feedback. Adverse events were assessed at each visit, with none being reported.

Data Analysis

We performed a multivariate (multiple dependent variables) repeated measures ANCOVA implemented using multivariate multilevel models (MMLM) as an intent-to-treat method to analyze the psychological outcomes (HADS-A, HADS-D, ASI, and PANAS-N). The MMLM approach to multivariate repeated measures ANCOVA is conceptually similar to Multivariate ANCOVA (MANCOVA), but MMLM is an intent-to-treat analysis that includes all participants regardless of missing data or missing assessments (which improves power and generalizability), while MANCOVA only includes participants with full data (all outcome measures at all assessments). Also, MMLM can model the covariance matrix of the repeated measures more flexibly than MANCOVA. Because of these (and other) advantages, MLM is the recommended method for longitudinal data analysis (93). This single multivariate analysis also maintains the study-wise Type I error at p<.05. Our 2 × 4 MMLM implementation of MANCOVA consisted of 2 treatment conditions (CART vs. SLOW) and four assessments over time (baseline, posttreatment, 1MFU, 6MFU).

The MMLM model had 3 levels: the 4 outcome measures (level-1 of the MMLM) were nested within repeated assessments (level-2), which were nested within individuals (level-3). Thus, the MMLM data file had the 4 outcome measures “stacked” in a single column within each assessment, and each assessment was “stacked” within each participant. This resulted in up to 16 lines of data (4 outcomes at each of 4 assessment timepoints) per participant. The error covariance matrix at level-1 (the error covariance between the 4 outcome variables) was modeled as unstructured because that structure fit the data better (lower AIC and BIC) than simpler structures (Toeplitz, AR(1), or compound symmetry, with heterogeneous or homogeneous variances). Models included a random intercept for individuals (at level 3). Random slopes for time (at level 3) were not included since the model would not converge with these random slopes. Each individual measure was z-scored across all assessments as recommended by Hox et al. (94). Maximum likelihood estimation was used in all analyses.

MMLM also allows the inclusion of continuous moderators. In our case, we hypothesized that the Treatment x Time interaction would be moderated by baseline anxiety level (baseline HADS-A) as a continuous moderator, as evidenced by a Treatment x Time x Baseline Anxiety interaction. The MMLM model included this triple interaction and all of its subcomponents. Covariates were: sex, age, BMI, cold and flu symptoms, beclomethasone equivalent doses, frequency of ICS use, and frequency of leukotriene modifier use. Non-significant covariates were dropped for the final analysis. See Supplemental Digital Content for the MMLM syntax for the final MMLM model and for more details on the MMLM analysis. Significant multivariate effects were followed by univariate MLMs for each outcome to determine which of the individual outcomes were significant. P-level was set at .05, two-sided, for all analyses.

In exploratory analyses, we investigated whether baseline anxiety was also a moderator of the treatment group effects on physiological outcomes also (PCO₂, PEF, FEV₁ percent of predicted, RR). Since these outcomes reflected diverse dimensions that might be affected differently by the treatment groups, each of these physiological outcomes were analyzed separately in univariate MLMs. Like the MMLM, these univariate MLMs were 2 × 4 ANCOVAs with baseline anxiety as a moderator of each ANCOVA effect (Treatment group, Time, and Treatment x Time). The Benjamini-Hochberg correction (95) was used to maintain the false discovery rate at p<.05. All data analyses were performed using SPSS 26.0 mixed models.

Results

Efficacy findings for asthma-related variables were previously reported in Ritz et al. (74). A total of 120 adults with asthma participated. Randomization resulted in n=62 patients in CART and n=58 in SLOW, providing sufficient power (greater than .80) to detect sizes greater than d=.40, between a small (d=.20) and medium (d=.50) effect size. Missing data through the end of treatment was 11.3% for CART and 13.8% for SLOW. Missing data through 6MFU was 17.7% for CART and 22.3% for SLOW. Patients with missing data did not differ from those without missing data on any baseline measure. These data are consistent with data missing at random (MAR), an assumption of MLM.

Mean baseline HADS-A scores was 5.38 (SD=3.6), with a range of 0–16. Baseline HADS-A did not differ between conditions, Mean(CART)=5.61, SD=3.8, Mean(SLOW)=5.02, SD=3.5, p=.39. Although baseline HADS-A was used as a dimensional moderator of treatment differences in change over time, we examined whether treatment groups differed in the proportion of patients with clinically-relevant baseline anxiety. Analyses indicated that the proportion of patients with clinically-relevant levels of HADS-A (HADS-A ≥ 8) at baseline did not differ significantly between treatment groups (p>.11), but the proportion meeting this level was numerically higher in CART (17 out of 62, 27%) than in SLOW (9 out of 58, 16%). Attrition rates, missing data rates at posttreatment, and missing data rates overall did not differ between those with clinically-relevant levels of HADS-A at baseline and those who did not have clinically-relevant levels of HADS-A at baseline, ps>.61 (e.g., attrition by 6MFU for high baseline HADS-A was 23% compared to 20% for low baseline HADS-A). Baseline characteristics (sex, age, BMI, cold and flu symptoms, beclomethasone equivalent doses, frequency of ICS use, and frequency of leukotriene modifier use) did not differ between patients with high baseline HADS-A and those without (ps>.11).

As reported previously (74), treatment groups did not differ significantly on drop-out, completion rates, demographics, anthropometrics, or disease-specific characteristics, except for age. Airway function was somewhat compromised in both groups, with overall FEV₁ at 78.8% of predicted and FE_NO at 46.3ppb. Most patients’ asthma was categorized as not well controlled (overall 38.8%) or very poorly controlled (overall 50.0%). Maintenance medication (including leukotriene modifiers) was taken by 63.9% in CART and 58.2% in SLOW. Rescue inhalers were used as the only medication by 36.1% in CART and 41.8% in SLOW.

Baseline Anxiety Moderating the Effects of Treatment on Psychological Outcomes

None of the covariates in the MMLM analysis were significant, so they were dropped, and the MMLM analysis recomputed (analyses including all of the covariates yielded identical significant results with similar p-values). Results indicated that the effect of Treatment over Time on our multivariate outcome (consisting of HADS-A, ASI, PANAS-N, and HADS-D) was moderated by baseline anxiety, as indicated by a significant Baseline anxiety x Treatment Condition x Time interaction, F(3,278)=4.80, p=.003 from the overall MMLM ANOVA (see Figure 1a–d).

Figure 1a-d:

Treatment effects of Capnometry-Assisted Breathing Training (CART) compared to slow breathing and awareness training (SLOW) on the Anxiety subscale of the Hospital Anxiety and Depression Scale (HADS-A), Anxiety Sensitivity Index (ASI), and the Negative Affect scale of the Positive Affect Negative Affect Schedule (PANAS-N) across baseline, posttreatment, 1-month follow up (1MFU), and 6-month follow up (6MFU)

We used simple slopes analyses to estimate treatment group effects for patients at different levels of baseline anxiety (see, for example, (96)). This approach includes the full sample in the analyses and centers the moderator variable alternately at low, and then at high, levels of the moderator, typically 1 SD below the mean and 1 SD above the mean (96), to obtain estimates of effects of treatment for those low and high levels of the moderator. These simple slopes analyses indicated that, for those with low baseline anxiety (1SD below the mean on the HADS-A [i.e., HADS-A=1.78]), treatment group did not affect improvement from baseline to 6MFU (p=.50 for the Treatment group x Baseline-to-6MFU contrast). But for patients with high baseline anxiety (1SD above the mean on the HADS-A [i.e., HADS-A=8.98]), those in CART improved significantly more from baseline to 6MFU than those in SLOW, b=.62, 95%CI: [.27, .97], t(284)=3.53, p<.001, d=.42. Follow-up, univariate analyses of each individual outcome showed that the triple interactions from the univariate ANOVAs were significant for HADS-A, F(3, 152)=3.50, p=.017; for ASI, F(3, 92)=3.41, p=.021; and for PANAS-N, F(3, 92)=3.44, p=.020, but not for HADS-D (p=.39) (Figure 1a–d). For these individual outcomes, the difference between treatments in the change from baseline to 6MFU was not significant for any outcome for patients with low baseline anxiety. However, for patients with high baseline anxiety, the decrease from baseline to 6MFU was greater in CART than in SLOW on ASI, p=.005, d=.58, and on PANAS-N, p=.002, d=.62; and was marginally greater on HADS-D, p=.099, d=.34, but was not significant for HADS-A (p=.25).

This MMLM triple interaction also resulted in treatment group differences at the 6MFU. For patients with low baseline anxiety, there were no significant differences between CART and SLOW on the multivariate outcome at the 6MFU (p=.94). But for patients with high baseline anxiety, those in CART had lower symptoms at the 6MFU than those in SLOW, b=.56, 95%CI: [.24, .88], t(308)=3.47, p<.001, d=.40. Univariate analyses showed that this pattern was evident for ASI, PANAS-N, HADS-D, and marginally significant for HADS-A. Specifically, CART and SLOW did not differ on any outcome at the 6MFU (ps=.28-.86) for patients with low baseline anxiety. However, for patients with high baseline anxiety, those in CART had lower symptoms than those in SLOW on ASI, PANAS-N, and HADS-D, and marginally lower symptoms on HADS-A, p=.007, d=.55; p=.012, d=.54; p=.001, d=.70; p=.090, d=.19, respectively.

Examining the effect of treatment group at more moderate levels of baseline HADS-A (e.g., examining the effect of treatment using HADS-A=3.0 as “low” HADS-A (instead of 1SD below the mean=1.78) and using HADS-A=8.0 as “high” HADS-A (instead of 1SD above the mean=8.98) yielded very similar results to the above-reported analyses. The effect of treatment at low HADS-A (HADS-A=3.0) produced the same non-significant differences between CART and SLOW for low HADS-A as reported above. And using HADS-A=8.0 for “high” HADS-A yielded the same significant differences between CART and SLOW for patients with high HADS-A as reported above.

Exploratory Analyses

Baseline Anxiety as a Moderator of Treatment Effects on Respiration and Asthma Outcomes.

Baseline anxiety did not moderate the effect of treatment on PCO₂, p=.11 for the Baseline anxiety x Treatment group x Time interaction (Figure 2). There was a significant Treatment group x Time interaction for PCO₂, F(3, 118)=3.11, p=.029, which resulted in PCO₂ being significantly higher (better) in CART than in SLOW at the 6MFU, b=2.85, 95% CI: [.90, 4.81], t(94)=2.90, p=.005. None of the other physiological variables tested (PEF, FEV₁ percent of predicted, nor RR) showed any significant effects of baseline anxiety as a moderator. However, it is important to note that, as expected, treatment groups did not differ on their change in RR over time (p>.30 for the Treatment x Time interaction), but that there was a significant main effect for Time indicating that RR decreased in both groups over time, F(3,92)=73.52, p<.001, attesting to the success in manipulating it in both CART and SLOW (estimated means [breaths per minute] at baseline, posttreatment, 1MFU, and 6MFU were: 14.29, 9.80, 10.38, 10.13 for CART and 15.17, 9.54, 10.21, 10.64 for SLOW). Further, there were no significant triple interactions for the other asthma outcomes or medication variables tested (ACT, bronchodilator use, beclomethasone equivalent doses, frequency of ICS use, and frequency of leukotriene).

Figure 2:

Treatment effects of Capnometry-assisted breathing training (CART) compared to slow breathing and awareness training (SLOW) on basal PCO2 levels across baseline, posttreatment, 1-month follow up (1MFU), and 6-month follow up (6MFU)

Discussion

This study showed that CART, a hypoventilation treatment that elevates basal PCO₂, leads to substantial reductions in anxiety sensitivity, negative affect, and depressive symptoms in asthma for patients that suffer from higher anxiety levels at the beginning of treatment. Slow breathing did not alter anxiety or negative affect.

Prevalence rates of anxiety in asthma are high and associated with more significant mortality and suffering (22,23). The comorbidity is not surprising considering that many symptoms patients with asthma fear are inherent to panic and anxiety sufferers (72,97). Sensations such as shortness of breath, dizziness, and heart racing impending threat. Unlike in individuals with panic or anxiety, for whom the perceived consequences of these symptoms are unlikely to occur (e.g., “I am suffocating,” “I will pass out,” “I am having a heart attack”), the same symptoms can signal true threat in asthmatics (i.e., acute bronchoconstriction), and require adequate action (e.g., using a bronchodilator). However, the very symptom that signals threat in asthma (e.g., dyspnea) can be produced by factors independent of asthma pathology, such as sustained or acute decreases in PCO₂. Dyspnea may consciously or unconsciously trigger a maladaptive compensatory response of further hyperpnea, which can cause an array of short-term complications, such as bronchoconstriction of hyperreactive airways, and problematic long-term consequences, such as suboptimal asthma control and rescue-medication overuse.

A complicating factor in the majority of intervention studies for treating psychological symptoms is the patient’s awareness and desire to reduce the very symptoms that are being targeted (80,98). The degree to which nonspecific mechanisms, such as patients’ desire to improve, self-observed change, or wanting to please the assessor, influence outcomes remains largely unknown (80,98). Unlike in prior CART trials for panic and anxiety, the incentive and rationale of the present trial was to reduce asthma symptoms and improve lung function and at no point was anxiety or stress reduction mentioned or targeted. Furthermore, participants in SLOW were blinded to their PCO₂ values. The highly specific and significant reductions in anxiety sensitivity and negative affect in CART are thus particularly noteworthy. The specificity of PCO₂ as a mediator of anxiety reduction has been observed in prior studies with panic patients. In these studies, CART led to sustained increases in PCO₂ levels and significant reductions in panic and anxiety symptomatology (e.g., (75,76)).

Several explanations for the association between PCO₂ elevations and anxiety reduction are plausible: Hypocapnia, caused by increased minute ventilation in excess of metabolic demand, has been linked to physical symptoms (e.g., dyspnea, dizziness) that are feared by patients with PD. Elevation of low levels of PCO₂ should thus ameliorate bodily symptoms and, consequently, their catastrophic interpretation (i.e., fewer sensations of shortness of breath should lead to less fear engendered by shortness of breath). This temporal pathway is consistent with the idea that successful alteration of one system (i.e., physiology) induces changes in another system (i.e., cognition (99). As treatment progresses, patients may become more successful in counteracting the onset of physical symptoms by breathing more slowly and shallowly. Mediational and temporal analyses from prior panic and anxiety trials with CART support this notion in that the correction of sustained hypocapnia by therapeutically increasing PCO₂ levels mediated improvements in panic symptomatology, symptom appraisal, and perceived control (76,100–102). Trainings such as heart rate variability (HRV) biofeedback have not demonstrated mediational evidence for HRV and PCO₂ (103) or do not assess PCO₂ (104) despite proposing hyperventilation-reducing mechanisms.

Alternatively, repeated exposure to increasing levels of PCO₂ during homework exercises may have led to a desensitization of a hypersensitive suffocation alarm system (105). Klein’s theory proposes that rising PCO₂ falsely heralds impending suffocation. CART would thus reduce panic vulnerability by building tolerance toward feared respiratory-related sensations and the absence of compensatory hyperventilation episodes (see (106) on mechanistic role of dyspnea in CART). Research has demonstrated chemosensitivity of cells in the amygdala (107), with increases in PCO₂ directly leading to an intense fear response. It could be speculated that attenuations in fear responses are linked to alterations in these chemosensory properties. Arguably, although respiratory symptoms are among the most frequent and distressing physical symptoms of anxiety or panic (63), they reflect only a subset of possible symptoms patients could have. However, due to the interconnectedness of autonomic subsystems (e.g., cardiorespiratory integration), ameliorations in respiratory function could influence symptom production in other systems that produce relevant symptoms, such as intercostal muscle tension (108) or heart-racing cardiac activity. Whether the correction of hypocapnia is associated with improved general health outcomes (109) requires further research.

Finally, CART-trained patients were instructed to gradually regulate their RR down to six breaths/min in the fourth week of training, a breathing rate that has been shown to stimulate and exercise the baroreflex (110). However, as we found in a previous investigation (101,102,111), there was little evidence that RR was related to changes in symptom appraisal or panic symptom severity, despite its successful reduction.

Strengths of our study, beyond the tight control of nonspecific treatment factors by our slow respiration biofeedback group, was the diversity of our sample. Minority populations, particularly African-American and Puerto Rican minorities, shoulder a disproportionate burden of poor asthma control. Our exploratory findings hinted at a particular efficacy for anxiety reduction of both device-guided techniques, CART and SLOW. Future studies comparing the appeal, uptake, and effectiveness of device-guided interventions more systematically between racial and ethnic groups are necessary to consolidate such findings.

The limitations of our study were that we did not explicitly recruit patients that were comorbid with anxiety. Still, this secondary analysis relied on the natural variation of anxiety levels we found in this population of asthma patients. The number of participants with elevated anxiety levels was on the lower end of what is reported in the literature. One contributing factor could be that our study required very stringent inclusion and exclusion criteria, including a methacholine-challenge verified asthma diagnosis. This required procedure may have inadvertently resulted in lower numbers due to the significant symptom-inducing nature of those test (e.g., extreme shortness of breath). We also did not directly compare CART protocols targeting both conditions, anxiety, and asthma, to develop an optimized treatment for comorbid patients. However, it can be argued that by testing the efficacy for anxiety reduction of this CART protocol for asthma, we have established the lower limit of the potential for anxiety reduction, as greater anxiety reduction could be achieved by a protocol that explicitly addresses anxiety and panic and one in which the participants expect anxiety reduction. In terms of its effect size, the observed anxiety reduction under the present conditions was at a respectable level (around d=.50, a medium effect size) and clinically relevant.

In sum, the findings of this study support CART to serve as a two-pronged adjunctive treatment that addresses comorbid asthma and anxiety in one protocol. In isolating PCO₂ change as a factor by a carefully designed control condition, we were able to show that changes elicited by this training, over and above pure slow breathing, can have anxiolytic effects through the 6MFU. Future studies will need to examine this effect in larger samples of comorbid asthma patients and dissect active ingredients of CART training, distinguishing acute sensations elicited by this training from resulting change in respiratory gas exchange.

Abbreviations:

CART

capnometry-assisted respiratory training

SLOW

Slow breathing and awareness training

RR

respiration rate

FEV₁

forced expiratory volume in the 1^st second

PEF

peak expiratory flow

FE_NO

fraction of nitric oxide in exhaled air

HADS-A

Hospital Anxiety and Depression Scale

ASI

Anxiety Sensitivity Index

PANAS-N

Negative Affect Scale of the Positive Affect Negative Affect Schedule

1MFU

1-month follow up

6MFU

6-month follow up