Summary of findings
Description of the condition
Asthma is a chronic inflammatory disorder of the lungs that can lead to structural and functional changes, resulting in bronchial hyperresponsiveness and airflow obstruction (Taylor 2008; Holgate 2009; Zhang 2010; Allen 2012; Brightling 2012). Symptoms of asthma include recurrent episodes of wheeze, cough, breathlessness and chest tightness, together with episodes of marked worsening of symptoms, known as exacerbations (Bateman 2008; Zhang 2010; Brightling 2012).
The diagnosis of asthma is based on the individual's medical history, physical examination findings and lung function and laboratory test results (Sveum 2012). Measurement of lung function provides an assessment of the severity of airflow limitation. These measures yield complementary information about different aspects of asthma control and are obtained by spirometry and by peak expiratory flow measurement (GINA 2011). Assessment of airway responsiveness to factors that can cause asthma symptoms, evaluation of airway inflammation and measurement of allergic status may facilitate the diagnosis of patients with asthma (GINA 2011).
Asthma is a serious public health problem that is a major cause of disability and health resource utilisation for those affected (Bateman 2008; Eisner 2012; To 2012). Around 300 million individuals of all ages worldwide are affected by asthma (Bateman 2008; Bousquet 2010; Brightling 2012). Increases in morbidity, mortality and economic costs are associated with severe or difficult to treat asthma, particularly in industrialised countries (Zhang 2010; Eisner 2012).
Asthma has been associated with symptomatic hyperventilation, which decreases carbon dioxide (CO
Description of the intervention
Although no cure for asthma is known, its symptoms are controllable in most patients (Taylor 2008). Asthma treatment can be pharmacological or non-pharmacological or a combination of the two associated with strategies of symptom control (environmental triggers and asthma education) (Wolf 2008; Burgess 2011; GINA 2011; Welsh 2011).
Medications to treat asthma can be broadly divided into long-term controllers and short-term relievers (Arun 2012). Controller medications are taken daily on a long-term basis, and the relievers are used to rapidly decrease bronchoconstriction and relieve its symptoms (GINA 2011). Such treatment can be administered in different ways (by inhalation, orally or parenterally) (GINA 2011).
Non-pharmacological interventions have gained attention in the treatment of asthma. Complementary and alternative medicine includes breathing exercises, homeopathy, acupuncture, aromatherapy, reflexology, massage, inspiratory muscle training and the Alexander technique (Blanc 2001; Ram 2009; Dennis 2012; McCarney 2012). Breathing exercises have been used routinely by physiotherapists and other professionals to control the hyperventilation symptoms of asthma (Bruton 2005b) and can be performed as the Papworth method, Buteyko breathing technique, yoga or any other similar intervention that manipulates the breathing pattern (Ram 2003).
How the intervention might work
Work undertaken at the Papworth Hospital, in Cambridge, UK, has changed the techniques used for treatment of asthma and hyperventilation (Cluff 1984; Innocenti 1993; Holloway 1994; Lum 1994). The Papworth method focuses on the use of an appropriate breathing pattern to reduce hyperventilation and hyperinflation, therefore increasing CO
The Buteyko method was developed in the 1950s by Konstantin Buteyko; the rationale behind its use is similar to that of the Papworth method for people who experience hypocapnia as a major contributor to their asthma symptoms. This method aims to develop a more efficient pattern of respiration through the use of controlled breathing and respiratory pauses. As a result, the method intends to increase alveolar and arterial CO
Yoga is an ancient discipline from India that has been shown to be an alternative technique for the management of asthma to help reduce anxiety associated with asthma symptoms. Some mechanisms may explain the rationale for yoga, such as reduction in psychological overactivity and emotional instability, and thereby reduction in efferent vagal discharge; increased autonomic control; decreased vagal outflow to the lung causing bronchodilatation and decreased bronchial reactivity (Singh 1990a). Yoga breathing techniques include deep breathing exercises (pranayama), which deal explicitly with control of breathing, postures (asanas), cleansing techniques (kriyas), meditation, prayer and often dietary changes to reduce asthma symptoms (Burgess 2011).
Why it is important to do this review
The worldwide high prevalence of asthma has become a public health problem because of the high healthcare costs resulting from hospitalisation and medication (Giavina-Bianchi 2010).
Breathing exercises have been used widely as an adjunct therapy in the treatment of asthmatic patients, generating considerable interest among researchers to develop studies that seek to show evidence of the effectiveness of this intervention.
This is an update of a review last published in 2003, in which the review authors concluded that there was insufficient evidence on the clinical benefits of breathing exercises in patients with asthma. Recently, new studies have been conducted to evaluate the effects of breathing exercises on quality of life, symptom control and lung function in asthmatic patients. Thus, within this review update, we aim to summarise and assess evidence from randomised controlled trials showing the efficacy of breathing exercises in the treatment of patients with asthma.
To evaluate the evidence for the efficacy of breathing exercises in the management of patients with asthma.
Criteria for considering studies for this review
Types of studies
Randomised controlled trials of breathing exercises in adults with asthma.
Types of participants
Adult patients with physician diagnosed asthma and/or diagnosis by internationally established criteria: American Thoracic Society (ATS), European Respiratory Society (ERS) or British Thoracic Society (BTS). Patients may be either community or hospital based with their treatment supervised by a general practitioner or respiratory specialist.
Types of interventions
Intervention: Patients with asthma who have received at least one course of treatment comprising breathing retraining.
Comparison: Control group receiving asthma education or, alternatively, no active control group (e.g. waiting list control).
Types of outcome measures
- Quality of life.
- Asthma symptoms (e.g. measures of dyspnoea or breathlessness with Borg score or visual analogue scale).
- Number of acute exacerbations (mean number and number of participants experiencing one or more exacerbations).
- Inpatient hospitalisation episodes.
- Physiological measures—lung function (especially low flow rates) and functional capacity.
- General practitioner (GP) or hospital outpatient appointments or both.
- Days off work.
- Patient's subjective evaluation of the intervention.
Search methods for identification of studies
Trials were identified from the Cochrane Airways Group Specialised Register of Trials (CAGR), which is derived from systematic searches of bibliographic databases including the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, EMBASE, CINAHL, AMED and PsycINFO, and from handsearching of respiratory journals and meeting abstracts (see Appendix 1 for further details). All records in the CAGR coded as 'asthma' were searched using the following terms:
((breath*) and (technique* or exercise* or re-train* or train* or re-educat* or educat* or physiotherap* or "physical therap*" or "respiratory therapy")) or (buteyko or "qigong yangsheng" or pranayama* OR yoga*) or "breathing control"
For the previous version of this review, searches were conducted up to April 2003. For this version, the literature search has been updated to February 2012. A repeat search was undertaken in January 2013.
Searching other resources
Reference lists of relevant articles found by the above methods were consulted to look for additional studies, and a search in clinical trial registries (clinicaltrials.gov and the WHO trial portal) was undertaken to look for planned, ongoing and unpublished trials.
Data collection and analysis
Selection of studies
Two review authors (DAF and GSSC) independently assessed studies for the possibility of inclusion in this review. We retrieved full text articles and reviewed them to determine eligibility. Final decisions and disagreements were resolved by consultation with a third review author (KMPPM).
Data extraction and management
Two review authors (DAF and GSSC) independently extracted data into RevMan (RevMan 2011) by using a standard data collection form. According to methods described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a), we collected information from the studies, including the following.
- Methods (design, method of randomisation, method of allocation concealment, outcome assessor blinding, withdrawal and dropouts).
- Participants (country, health status, mean age, gender, total sample and exclusion criteria).
- Interventions (methods and types of intervention, including number and duration of sessions and methods used for control group comparisons).
- Outcomes (improvement in quality of life indices, asthma symptoms, number of acute exacerbations, inpatient hospitalisation episodes, etc).
We resolved disagreements by discussion and consensus.
Assessment of risk of bias in included studies
Two review authors (DAF and GSSC) independently assessed the risk of bias using The Cochrane Collaboration’s tool for assessing risk of bias, which includes the following items: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting and other sources of bias. The risk of bias was classified as high, low or unclear, according to the methods described in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b). Disagreements were resolved by discussion and consensus.
Measures of treatment effect
Continuous outcomes were expressed as mean difference (MD) with 95% confidence interval (CI) when outcome measurements were performed on the same scale, or as standardised mean difference (SMD) with 95% CI when studies assessed an outcome by using different methods.
Unit of analysis issues
Trials with a cross-over and cluster-randomised design were not included in the review.
Dealing with missing data
We wrote to authors of included trials to request additional data as required.
Assessment of heterogeneity
We assessed heterogeneity by inspecting the forest plots to detect non-overlapping CIs, while applying the Chi
Assessment of reporting biases
If we had been able to meta-analyse sufficient data (10 studies or more), we planned to assess reporting bias among the studies using the funnel plot method discussed in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011d). If asymmetry was noted, we planned to explore possible causes, including publication bias, poor methodological quality and true heterogeneity.
We used The Cochrane Collaboration's statistical package, Review Manager, to combine outcomes when possible (RevMan 2011). We used a fixed-effect model unless substantial heterogeneity (a value of I
We created a 'Summary of findings' table that included the following outcomes, according to the methods described in Chapter 11 of the Cochrane Handbook for Systematic Reviews of Interventions: change in Asthma Quality of Life Questionnaire (AQLQ), change in St George's Respiratory Questionnaire (SGRQ), asthma symptoms, number of acute exacerbations, inpatient hospitalisation episodes, lung function (forced expiratory volume in 1 second (FEV
Subgroup analysis and investigation of heterogeneity
If we were able to combine sufficient data and identify substantial heterogeneity (a value of I
- Degree of asthma severity.
- Age groups (adult versus elderly).
- Duration of treatment.
If we had been able to combine sufficient data, sensitivity analysis would have been performed to explore the influence on the results of the following factors.
- Trial quality (randomised controlled trials with poor methodology).
- Trial size (stratified by sample size).
- Allocation concealment (high risk of bias versus low risk of bias versus unclear risk of bias).
- Assessor blinding (high risk of bias versus low risk of bias versus unclear risk of bias).
Description of studies
Results of the search
For the initial version of the review (1998), full texts of 42 potentially relevant studies were obtained after 182 abstracts and titles revealed by the searches were screened; five studies were included (Nagarathna 1985; Girodo 1992; Fluge 1994; Bowler 1998; Vedanthan 1998). In the 2003 update, two more studies were included (Opat 2000; Thomas 2003). The search of the Airways Group Register for the 2012 update returned 147 references. Of these, twelve were identified as potentially relevant, and the full text articles were retrieved for closer inspection, of which five were new additions in the 2012 update (Holloway 2007; Thomas 2009; Sodhi 2009; Vempati 2009; Grammatopoulou 2011). A repeat search was undertaken from February 2012 to January 2013, and 12 new references were identified. Of these, three were considered eligible and thus were included in the review (Bidwell 2012; Singh 2012; Prem 2013).
We excluded two studies that were included in earlier versions of this review (see Excluded studies).
In total, 13 studies are now included in this review: Nagarathna 1985; Girodo 1992; Fluge 1994; Vedanthan 1998; Thomas 2003; Holloway 2007; Thomas 2009; Sodhi 2009; Vempati 2009; Grammatopoulou 2011; Bidwell 2012; Singh 2012; Prem 2013. See 'Characteristics of included studies' for full details on each study.
Setting and populations
Five trials were conducted in India (Nagarathna 1985; Sodhi 2009; Vempati 2009; Singh 2012; Prem 2013), one in Canada (Girodo 1992), one in Germany (Fluge 1994), three in the UK (Thomas 2003; Holloway 2007; Thomas 2009), two in the USA (Vedanthan 1998; Bidwell 2012) and one in Greece (Grammatopoulou 2011). All papers were written in English with the exception of Fluge 1994, which was written in German. Nine studies were conducted between 2003 and 2013, three were conducted between 1992 and 1998 and one was conducted in 1985. The studies varied in size from 17 to 183 participants. Participants in the included studies were older than 18 years of age, with the exception of Nagarathna 1985 (aged 9 to 47), Thomas 2003 (aged 17 to 65) and Holloway 2007 (aged 16 to 70). We included all studies as the mean age was over 18.
Interventions and control groups
In seven studies (Nagarathna 1985; Fluge 1994; Vedanthan 1998; Sodhi 2009; Vempati 2009; Bidwell 2012; Singh 2012), participants undertook yoga training that involved breathing exercises as the major component, and the control groups did not undergo yoga training but continued taking their usual medication. In Nagarathna 1985, participants in the intervention group underwent training for two weeks and were told to practise these exercises for 65 minutes daily. In Fluge 1994, participants underwent 15 yoga sessions over 3 weeks. In Vedanthan 1998, yoga sessions were performed 3 times a week over 16 weeks. In Sodhi 2009, each yoga training session was of 45 minutes' duration per week with a trained instructor for a period of 8 weeks. In Vempati 2009, the intervention consisted of 2-week supervised training in lifestyle modification and stress management based on yoga followed by closely monitored continuation of these practises at home for 6 weeks. In Bidwell 2012, yoga training consisted of two 1-hour supervised yoga sessions per week for 10 weeks. In Singh 2012, participants attended yoga training provided by a yoga expert for 5 to 6 days. Thereafter, participants were told to practise yoga for an average of 40 to 50 minutes daily at home for 2 months. Participants were called to the yoga centre regularly (about every 7 days) so investigators could see whether they were doing the yoga exercises properly.
In the Girodo 1992 study, participants undertook a 16-week programme of deep diaphragmatic breathing exercises and were compared against a group of controls that were on a waiting list. Thomas 2003 compared participants who completed three short breathing retraining sessions (total contact time 75 minutes), taught by a physiotherapist, with a control group that received asthma education from a nurse. In the Holloway 2007 study, the intervention group completed five 60-minute individual sessions on the Papworth method provided by a respiratory physiotherapist. The control group received no additional treatment. In the Thomas 2009 study, the breathing training group attended three sessions (one small group session and two individual sessions) that provided an explanation of normal breathing and possible effects of abnormal 'dysfunctional breathing'. During individual sessions, participants were taught diaphragmatic and nasal breathing techniques and were encouraged to practise these exercises for at least 10 minutes per day. The control group received three sessions of nurse-provided asthma education. The intervention group in the Grammatopoulou 2011 study received twelve individual breathing retraining sessions, and the control group received usual asthma care. In the study of Prem 2013, 120 participants were divided into three groups: Buteyko, yoga and control. Participants assigned to Buteyko or yoga groups received 3 to 5 days of sessions totalling 60 minutes each day. Participants in the control group followed routine physician care involving pharmacological management.
The primary outcome in Holloway 2007, Thomas 2009, Bidwell 2012 and Prem 2013 was quality of life, although different instruments (SGRQ in Holloway 2007 and Bidwell 2012, and AQLQ in Thomas 2009 and Prem 2013) were used. Asthma symptoms as measured by the Asthma Control Test score were the main outcome in Grammatopoulou 2011. In Vempati 2009, pulmonary function was the primary outcome.
Secondary outcomes were asthma symptom and lung function in Holloway 2007 and Thomas 2009. Asthma symptoms were measured by the Nijmegen questionnaire in Holloway 2007 and by the Asthma Control Questionnaire in Thomas 2009. In Grammatopoulou 2011, secondary outcomes were quality of life (as measured by the Short Form (SF)-36 v2 Health Survey) and lung function. In Vempati 2009, the secondary outcome was quality of life (as measured by the AQLQ).
In the other included trials, primary and secondary outcomes were not specified, but the authors reported several main outcome measures, including pulmonary function (Fluge 1994; Sodhi 2009), asthma symptoms (Girodo 1992), number of acute exacerbations and pulmonary function (Nagarathna 1985), quality of life and asthma symptoms (Thomas 2003), asthma symptom and lung function (Vedanthan 1998), and lung function and quality of life (Singh 2012).
After the full text of potentially eligible trials was retrieved, a total of 43 studies were excluded from the review. Two studies previously included were excluded in the 2012 update (Bowler 1998; Opat 2000). Reasons for exclusion are described in the Characteristics of excluded studies.
Risk of bias in included studies
|Figure 1. Risk of bias summary: review authors' judgements about each risk of bias item for each included study.|
Four studies reported adequate sequence generation and were judged to have low risk of bias (Thomas 2003; Holloway 2007; Grammatopoulou 2011; Prem 2013). Nine studies were reported as randomised but gave no description of the methods used and were therefore judged to be at unclear risk of bias (Nagarathna 1985; Girodo 1992; Fluge 1994; Vedanthan 1998; Sodhi 2009; Thomas 2009; Vempati 2009; Bidwell 2012; Singh 2012).
Thomas 2003 recruited individuals with high Nijmegen scores who were currently being treated for asthma at a general practice. Volunteers were randomly assigned by numbering them alphabetically and using random number tables to assign them to trial groups. In Holloway 2007, randomisation was undertaken by a computer-generated number sequence that assigned consecutive subject ID numbers a 1 or a 2 to denote intervention or a control condition. Random allocation was undertaken with sealed envelopes in Grammatopoulou 2011. In Prem 2013, participants were assigned to three groups (Buteyko, yoga and control) through block randomisation.
Only two trials described adequate allocation concealment and were then judged to have low risk of bias (Grammatopoulou 2011; Prem 2013). In Grammatopoulou 2011, allocation concealment was undertaken with sealed envelopes, and in Prem 2013, the method of allocation was concealed in sequentially numbered, sealed, opaque envelopes. The other eleven studies gave no description of the methods of allocation concealment used and were therefore judged to have unclear risk of bias (Nagarathna 1985; Girodo 1992; Fluge 1994; Vedanthan 1998; Thomas 2003; Holloway 2007; Thomas 2009; Vempati 2009; Sodhi 2009; Bidwell 2012; Singh 2012).
A study is classed as double-blinded if neither the investigator nor the participant involved knows the identity of the intervention. Blinding reduces bias in a trial. Double-blinding is not possible or practical in these studies. Participants in these trials must know whether or not they are undertaking breathing training or asthma education, as compliance is critical to the study. However, it is possible to blind the assessor who is analysing the results of the trial.
Five trials described that the blinding of participants and personnel was not possible; these studies were judged to have a high risk of bias, as it was determined that the outcomes may be influenced by the lack of blinding (Thomas 2003; Holloway 2007; Thomas 2009; Vempati 2009; Grammatopoulou 2011). Eight studies did not describe blinding of participants and personnel, so they were judged to have an unclear risk of bias (Nagarathna 1985; Girodo 1992; Fluge 1994; Vedanthan 1998; Sodhi 2009; Bidwell 2012; Singh 2012; Prem 2013).
Blinding of outcome assessors was described in five studies (Vedanthan 1998; Thomas 2003; Holloway 2007; Grammatopoulou 2011; Prem 2013). In Grammatopoulou 2011, participants were assessed by the same trained assessor, who was blinded to the participants' treatment allocation. In Thomas 2003, questionnaires completed by the participants were scored by the blinded investigator. In Vedanthan 1998, records of the yoga and control groups were coded during the study period, and the decoded data were unavailable to the principal investigators. In Prem 2013, the parameters were recorded before and after training by a person blinded to the allocation of groups. Eight studies did not describe blinding of outcome assessment, so they were judged to have an unclear risk of bias (Nagarathna 1985; Girodo 1992; Fluge 1994; Thomas 2003; Vempati 2009; Sodhi 2009; Bidwell 2012; Singh 2012).
Incomplete outcome data
Two studies did not describe the occurrence of withdrawals and dropouts and were judged to be at unclear risk of bias (Girodo 1992; Sodhi 2009). The study of Nagarathna 1985 affirmed that in total 25 participants dropped out of the study. However, this study was judged to have an unclear risk of bias because it did not describe how many participants dropped out of the study in each group (intervention and control). In three studies, no withdrawals or dropouts were reported; these studies were judged to have a low risk of bias (Vedanthan 1998; Grammatopoulou 2011; Bidwell 2012). Seven studies described withdrawals and dropouts and were also judged to have a low risk of bias because the missing outcome data were balanced in numbers across intervention groups (Thomas 2003; Holloway 2007; Thomas 2009; Vempati 2009) or because the reasons for missing outcome data were unlikely to be related to true outcomes (Fluge 1994; Singh 2012; Prem 2013).
Two studies were registered on clinicaltrials.gov, and all of the prespecified primary and secondary outcomes were reported in a prespecified way (Holloway 2007; Vempati 2009). These studies were judged to have a low risk of bias. Six studies adequately reported outcome data for all primary and secondary outcomes as listed in the methods, although the protocol for each study is not available ( Nagarathna 1985; Thomas 2003; Thomas 2009; Sodhi 2009; Grammatopoulou 2011; Singh 2012). Five studies were judged to be at high risk of bias because one or more outcomes of interest in the review were reported incompletely, so that they cannot be entered into a meta-analysis (Girodo 1992; Fluge 1994; Vedanthan 1998; Bidwell 2012; Prem 2013).
Other potential sources of bias
We were unable to identify any other potential biases in four studies (Thomas 2003; Holloway 2007; Grammatopoulou 2011; Prem 2013). Nine studies were judged to be at unclear risk of bias, as they did not provide sufficient information to allow assessment of whether an important risk of bias is present (Nagarathna 1985; Girodo 1992; Fluge 1994; Vedanthan 1998; Thomas 2009; Sodhi 2009; Vempati 2009; Bidwell 2012; Singh 2012).
Effects of interventions
Breathing exercises versus inactive control
Primary outcome: quality of life
Six studies involving 381 participants reported improvement in quality of life in the groups that submitted to breathing exercises (Holloway 2007; Vempati 2009; Grammatopoulou 2011; Bidwell 2012; Singh 2012; Prem 2013). One study (Holloway 2007) reported data at baseline and at 6 and 12 months after baseline; one (Grammatopoulou 2011) at baseline and at 1, 2 and 6 months; and one (Vempati 2009) at baseline and at 2, 4 and 8 weeks. In Bidwell 2012, Singh 2012 and Prem 2013, data were reported at baseline and at post-treatment. Three of these studies (Holloway 2007; Vempati 2009; Prem 2013) were included in the meta-analysis. However, the studies of Vempati 2009 and Prem 2013 assessed quality of life by the AQLQ, and the study of Holloway 2007 assessed this outcome by the SGRQ. When the AQLQ showed the opposite direction of effect to the SGRQ, these questionnaires were analysed separately.
For the outcome 'Change in AQLQ' ( Analysis 1.1), which included the studies of Vempati 2009 and Prem 2013, meta-analysis showed significant differences favouring the breathing exercises group (MD 0.79, 95% CI 0.50 to 1.08). The postintervention values for the AQLQ in the study of Prem 2013 were provided by the author as means and standard deviations.
The statistical analysis for the SGRQ in the study of Holloway 2007 has a P value that has been adjusted for a baseline covariate. However, the adjusted mean difference was not given in this study. Thus, after the adjusted mean difference was calculated, the analysis for the outcome ‘Change in SGRQ’ ( Analysis 1.2) yielded a P value smaller than the 0.186 reported in the paper (6 months post-baseline), whereas no difference was seen in the final scores analysis (12 months post-baseline).
We were not able to include the other three studies in the meta-analysis (Grammatopoulou 2011; Bidwell 2012; Singh 2012). Of these, Grammatopoulou 2011 showed that the group that performed breathing exercises showed improvement in the physical component of the SF-36 quality of life questionnaire compared with controls in all assessments (2, 3 and 6 months after intervention, with P values of 0.003, 0.0002 and 0.066, respectively). Bidwell 2012 found significant improvement in the three aspects of the SGRQ (symptoms, activity and impact) in the yoga group compared with the control group (P < 0.05). Singh 2012 observed a significant difference favouring the group submitted to the intervention, with a P value ˂ 0.001 for all four domains of the AQLQ.
Holloway 2007 also assessed the Hospital Anxiety and Depression Score (HADS). This study found significantly lower HADS scores in the intervention group than in the control group, with a P value of 0.002 for HADS Anxiety score and of 0.075 for HDAS Depression score at 6 months post-baseline assessment. At 12 months post-baseline assessment, significant differences favoured the intervention group, with P = 0.772 for the HADS Anxiety score and P < 0.001 for the HADS Depression score.
Secondary outcome: asthma symptoms
Five studies involving 331 participants reported asthma symptoms (Girodo 1992; Vedanthan 1998; Holloway 2007; Grammatopoulou 2011; Prem 2013). Meta-analysis was possible for two studies for this outcome (Holloway 2007; Grammatopoulou 2011) ( Analysis 1.3; Figure 2). Assessment of heterogeneity revealed no significant difference between these two studies (I
|Figure 2. Forest plot of comparison: 1 Breathing exercises versus inactive control, outcome: 1.3 Asthma symptoms.|
Secondary outcome: number of acute exacerbations
Only one study reported this outcome (Nagarathna 1985). Over two weeks, the intervention group involving 53 participants attended a daily yoga programme lasting two and a half hours. Comparison between the two groups (yoga and control) showed significant improvement (P < 0.005) in the number of exacerbations in the group that received the intervention (Nagarathna 1985).
Secondary outcome: physiological measures
Six studies involving 462 participants reported improvement in spirometry in the groups that performed the intervention (Nagarathna 1985; Sodhi 2009; Vempati 2009; Grammatopoulou 2011; Bidwell 2012; Prem 2013). Four studies did not show significant differences in this outcome (Fluge 1994; Vedanthan 1998; Holloway 2007; Singh 2012). Only two of the ten studies (Sodhi 2009; Vempati 2009) were included in the meta-analysis ( Analysis 1.4, Analysis 1.5, Analysis 1.6, Analysis 1.7; Analysis 1.8). However, because of the substantial heterogeneity indicated by an I
Two studies also assessed capnography (Holloway 2007; Grammatopoulou 2011). The study of Holloway 2007 did not find significant differences between intervention and control groups regarding end-tidal carbon dioxide. However, values for relaxed breathing rate over a 10-minute period showed better results in the intervention group than in the control group, with P < 0.001 at 6- and 12-month post-baseline assessments. In the study of Grammatopoulou 2011, the intervention group compared with the control group showed increased end-tidal carbon dioxide (P = 0.002 and 0.003 for 1- and 6-month post-baseline assessments, respectively; P < 0.0001 for 2-month post-baseline assessment). The intervention group showed a decreased respiratory rate compared with the control group (P < 0.0001).
Secondary outcomes: inpatient hospitalisation episodes, GP appointments, days off work and subjective evaluation of the intervention
These outcomes were not reported in any of the studies.
Breathing exercises versus asthma education
Primary outcome: quality of life
Two studies involving 194 participants assessed this outcome (Thomas 2003; Thomas 2009). Both studies had follow-up periods of 1 and 6 months. The study of Thomas 2003 showed a statistically significant improvement (P = 0.018) in overall AQLQ scores in the intervention group compared with the control group after 1 month. After 6 months, only the improvement in the activities domain of the AQLQ was significantly greater in the intervention group than in the control group (P = 0.018). The study of Thomas 2009 showed no significant between-group differences in the four subdomains of the AQLQ at 1-month assessment. At 6 months, significantly greater improvements were found in the intervention group in terms of symptoms (P = 0.01), activities (P = 0.01) and emotions (P = 0.05) domains but not in the environment domain (P = 0.40) compared with controls, with a significant between-groups difference favouring the intervention group (P = 0.01) for the total score (see Analysis 2.1).
Thomas 2009 also assessed the Hospital Anxiety and Depression Score (HADS). This study found significant reductions in HAD Anxiety and Depression domain scores in both groups 1 month after the intervention, with no significant difference noted between the groups. At the 6-month assessment, significant between-group differences were observed to favour the intervention group for Anxiety score (P = 0.02) and Depression score (P = 0.03).
Secondary outcome: asthma symptoms
Two studies involving 194 participants assessed asthma symptoms (Thomas 2003; Thomas 2009). Both studies carried out assessment of symptoms at baseline and 1 month and 6 months after the intervention (Thomas 2003; Thomas 2009). In Thomas 2003, the between-group difference favouring the intervention was statistically significant only after 6 months (P = 0.01). In Thomas 2009, no between-group difference was noted for the ACQ, whereas a statistically significant difference favoured the intervention group at 6-month assessment for the Nijmegen Questionnaire (P = 0.005).
Secondary outcome: physiological measures
Only one study assessed spirometric values (Thomas 2009). This study assessed FEV
The study of Thomas 2009 also assessed resting end-tidal carbon dioxide concentration, showing that values for this outcome did not change significantly within or between groups.
Secondary outcomes: numbers of acute exacerbations, inpatient hospitalisation episodes, GP appointments and days off work and subjective evaluation of the intervention
These outcomes were not reported in these two studies.
Summary of main results
This systematic review assessed available evidence for the efficacy of breathing exercises in the treatment of patients with asthma. A total of 13 studies involving 906 participants satisfied the inclusion criteria. Although these studies met the inclusion criteria, they differed significantly in terms of intervention characteristics, such as type of breathing exercise, number of participants, number and duration of sessions, reported outcomes and statistical presentation of data. These differences limited meta-analysis.
The included studies reported that breathing exercises were well tolerated by participants, and no adverse effects related to the intervention were described, showing that this is a safe non-pharmacological intervention. All eight studies that assessed quality of life reported improvement in this outcome. Improvement in the number of acute exacerbations was observed by the only study that assessed this outcome. Six of seven included studies showed a significant difference favouring breathing exercises for asthma symptoms. Effects on lung function were more variable, with no difference noted in five of the eleven studies that assessed this outcome, although the other six showed a significant difference for this outcome that favoured breathing exercises.
Because the included studies employed different interventions by using different methodologies, meta-analysis for lung function was not possible because of high heterogeneity. For asthma symptoms and changes in AQLQ, meta-analysis showed improvement favouring the group that submitted to breathing exercises. However, each meta-analysis was performed with only two studies.
Overall completeness and applicability of evidence
The types of breathing exercises that were related to improvements in quality of life, asthma symptoms and numbers of exacerbations were the Papworth method, Buteyko, diaphragmatic breathing and yoga. The ones that improved lung function were Buteyko, yoga and diaphragmatic breathing. However, the effects seen may represent a combination of many characteristics rather than breathing exercises alone. Some of the included studies involved group sessions in which participants were able to talk to each other and share their experiences. This can also be considered a therapeutic procedure that may affect the sensation of well-being (Evans 1993). Awareness of participation in the study, the sensation of increased care and cure and the specialists’ recommendations to continue regular asthma medication are characteristics that must be considered when the findings of an experimental study are interpreted (Grammatopoulou 2011).
Moreover, asthma severity of participants from the included studies ranged from mild to moderate, so it was not possible to assess the effects of breathing exercises on participants with severe asthma. The samples from studies consisted solely of outpatients. Besides that, four of the eight outcomes proposed by this review were not addressed: inpatient hospitalisation episodes, reduction in general practice (GP) and hospital outpatient appointments, days off work and participants' subjective evaluation of the intervention.
Quality of the evidence
This systematic review was limited by the quality of existing data. Some points must be taken into consideration when the results of this review are analysed, including small sample size and small number of sessions of some studies coupled with limitations in the design and reporting of studies, leading to risk of bias.
In general, the included studies had a small number of participants. The impact of a small sample size on trial results was already reported in a previous study (Moher 1994), which reviewed 383 randomised controlled trials. This study concluded that most trials with negative results did not have large enough sample sizes to detect relative differences. Furthermore, the description of how sample size was determined is recommended by the CONSORT statement (CONSORT 2010). Three of the included studies performed sample size and power calculations: Thomas 2003 (based on AQLQ), Holloway 2007 (based on SGRQ) and Grammatopoulou 2011 (based on ACT). Moreover, the number of sessions involved in these studies was small, given the longer duration of 6 months.
Of the thirteen studies included in the review, only four described the method of random sequence generation and were classified as “low risk of bias”. In addition, the allocation concealment was described in only two studies, which had a low risk of bias for this item. According to Savović 2012, inadequate reporting of trial methods can severely impede assessment of trial quality and of risk of bias in trial results. According to this study, this is a particular problem for the assessment of sequence generation and allocation concealment, which often are not described in trial publications (Savović 2012). In addition, inadequately reported randomisation has been associated with bias in estimating the effectiveness of interventions (Moher 2001).
In a randomised controlled trial, at least three distinct groups (trial participants, trial personnel and outcome assessors) can potentially be blinded (Savović 2012). When a randomised controlled study that involves breathing exercises is conducted, it is not possible for the participants and the personnel to be blinded to the intervention (Holloway 2007). According to Savović 2012, the lack of or unclear double blinding (participants and personnel) can be associated with marked exaggeration of intervention effect estimates.
Potential biases in the review process
Although an attempt was made to apply a systematic process for including and excluding studies in this review, besides following the criteria prespecified in the protocol with robust methods for data collection and risk of bias assessment, final decisions are open to interpretation or criticism.
Incomplete outcome data may be considered a potential source of bias of this review. This factor has also limited analysis, as the data from these studies could not be entered into a meta-analysis. Also related to meta-analysis, the subgroup and the sensitivity analysis were not possible because of the impossibility of obtaining sufficient data. This could have showed possible differences in degree of asthma severity, age groups and duration of treatment. Moreover, sensitivity analysis could have identified the influence of some factors (such as trial quality and trial size) on the results, thus revealing a source of the substantial heterogeneity found among studies on lung function.
Agreements and disagreements with other studies or reviews
The current review update included eight new randomised controlled trials and removed two trials that were included in the last published version of this review. In addition, this review brings together trials that were not included in previous systematic reviews (Ernst 2000; Ram 2003). These two reviews assessed the effectiveness of breathing exercises in the management of asthma. The outcomes assessed by Ernst 2000 were asthma symptoms and lung function, and those assessed by Ram 2003 were quality of life, asthma symptoms, number of exacerbations and lung function.
The findings of this review show that, even though outcomes reported by individual trials showed improvement in quality of life indices, asthma symptoms, number of exacerbations and lung function of participants who submitted to breathing exercises, evidence supporting the efficacy of breathing exercises in these outcomes is not sufficient. The systematic review performed by Ernst 2000 affirmed that, on the basis of available data, it was not possible to make firm judgments. Also, this review suggested that breathing techniques seem to have some potential and should be tested rigorously in the future. Similarly, the systematic review performed by Ram 2003 concluded that because evidence available from the small randomised controlled trials included in the review is limited, it was not possible to draw any firm conclusions as to the effectiveness of breathing exercises in the treatment and management of asthma.
It is important to emphasise that, even though the results of this review are consistent with results reported by the two previous systematic reviews (Ernst 2000; Ram 2003), some methodological differences have been noted among these studies. The review of Ernst 2000 was published more than one decade ago. Moreover, this review included two cross-over studies and one study that was performed with children admitted to hospital with acute severe asthma. Ram 2003 included only six studies, which did not involve the same breathing exercise techniques that were examined by the present review, such as the Papworth method.
Implications for practice
This review indicates that breathing exercises are a safe and well-tolerated intervention for people with asthma. Also, meta-analysis of two studies showed that breathing exercises may have positive effects on asthma symptoms and quality of life (more specifically, on AQLQ score). Even though outcomes that were reported from individual trials show that breathing exercises may have a role in the treatment and management of asthma, no conclusive evidence is provided in this review to support or refute the benefits of these techniques in terms of quality of life, asthma symptoms, number of exacerbations and lung function. This is a result of the small number of participants in most of the included studies, the small number of sessions, the methodological differences among included studies, trials with poor methodology and the statistical heterogeneity noted among the studies for three of the four outcomes assessed by meta-analysis. No data are available regarding the effects of breathing exercises on inpatient hospitalisation episodes, reduction in GP and hospital outpatient appointments, days off work and participants' subjective evaluation of the intervention.
Implications for research
Well-conducted randomised controlled trials are needed to assess the clinical benefit of breathing exercises in the management of asthma, including those that were not assessed by the studies included in this review such as inpatient hospitalisation episodes, reduction in GP and hospital outpatient appointments, days off work and participants' subjective evaluation of the intervention. Furthermore, in the future, much more attention needs to be paid to good reporting and high-quality study design, including items such as adequate random sequence generation and allocation concealment, blinding of outcome assessor and determination of the trial sample size before the study is begun.
The authors would like to thank Emma Welsh (the Managing Editor of the Cochrane Airways Group) for providing assistance throughout the review process and Elizabeth Stovold (the Trials Search Co-ordinator/Information Specialist of the Cochrane Airways Group) for performing the search.
We would also like to thank all the authors who responded to our enquiries.
Anne Holland was the Editor for this review. Anne commented critically on the review and assisted the Co-ordinating Editor with signing off on the review for publication.
Data and analyses
- Top of page
- Summary of findings [Explanations]
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Index terms
Appendix 1. Sources and search methods for the Cochrane Airways Group Specialised Register (CAGR)
Electronic searches: core databases
Handsearches: core respiratory conference abstracts
MEDLINE search strategy used to identify trials for the CAGR
1. exp Asthma/
3. (antiasthma$ or anti-asthma$).mp.
4. Respiratory Sounds/
6. Bronchial Spasm/
8. (bronch$ adj3 spasm$).mp.
10. exp Bronchoconstriction/
11. (bronch$ adj3 constrict$).mp.
12. Bronchial Hyperreactivity/
13. Respiratory Hypersensitivity/
14. ((bronchial$ or respiratory or airway$ or lung$) adj3 (hypersensitiv$ or hyperreactiv$ or allerg$ or insufficiency)).mp.
15. ((dust or mite$) adj3 (allerg$ or hypersensitiv$)).mp.
Filter to identify RCTs
1. exp "clinical trial [publication type]"/
2. (randomised or randomised).ab,ti.
11. 9 not (9 and 10)
12. 8 not 11
The MEDLINE strategy and RCT filter are adapted to identify trials in other electronic databases.
Last assessed as up-to-date: 30 January 2013.
Protocol first published: Issue 4, 1998
Review first published: Issue 3, 2000
Contributions of authors
Diana Freitas: selected the studies, extracted data, entered data into RevMan, carried out the analysis, interpreted data and drafted the final review.
Elizabeth Holloway: drafted the original review and contributed her clinical expertise.
Selma Bruno: contributed with clinical expertise, carried out the analysis, interpreted data and drafted the final review.
Gabriela Chaves: selected the studies, extracted data, entered data into RevMan, carried out the analysis, interpreted data and drafted the final review.
Guilherme Fregonezi: contributed with clinical expertise and drafted the final review.
Karla Mendonça: coordinated the review, made an intellectual contribution, interpreted data and drafted the final review.
Declarations of interest
Medical Subject Headings (MeSH)
MeSH check words
* Indicates the major publication for the study