Oxygen therapy during exercise training in chronic obstructive pulmonary disease

Abstract

Background

Exercise training within the context of pulmonary rehabilitation improves outcomes of exercise capacity, dyspnea and health‐related quality of life in individuals with chronic obstructive pulmonary disease (COPD). Supplemental oxygen in comparison to placebo increases exercise capacity in patients performing single‐assessment exercise tests. The addition of supplemental oxygen during exercise training may enable individuals with COPD to tolerate higher levels of activity with less exertional symptoms, ultimately improving quality of life.

Objectives

To determine how supplemental oxygen in comparison to control (compressed air or room air) during the exercise‐training component of a pulmonary rehabilitation program affects exercise capacity, dyspnea and health‐related quality of life in individuals with COPD.

Search methods

All records in the Cochrane Airways Group Specialized Register of trials coded as 'COPD' were searched using the following terms: (oxygen* or O2*) AND (exercis* or train* or rehabilitat* or fitness* or physical* or activ* or endur* or exert* or walk* or cycle*). Searching the Cochrane Central Register of Controlled Trials (CENTRAL, The Cochrane Library), MEDLINE, EMBASE and CINAHL databases identified studies. The last search was carried out in June 2009.

Selection criteria

Only randomized controlled trials (RCTs) comparing oxygen‐supplemented exercise training to non‐supplemented exercise training (control group) were considered for inclusion. Participants were 18 years or older, diagnosed with COPD and did not meet criteria for long‐term oxygen therapy. No studies with mixed populations (pulmonary fibrosis, cystic fibrosis, etc) were included. Exercise training was greater than or equal to three weeks in duration and included a minimum of two sessions a week.

Data collection and analysis

Two review authors independently selected trials for inclusion in the review and extracted data. Weighted mean differences (WMD) with 95% confidence intervals (CI) were calculated using a random‐effects model. Missing data were requested from authors of primary studies.

Main results

Five RCTs met the inclusion criteria. The maximum number of studies compared in the meta‐analysis was three (31 on oxygen versus 32 control participants), because all included studies did not measure the same outcomes. When two studies were pooled, statistically significant improvements of oxygen‐supplemented exercise training were found in constant power exercise time, WMD 2.68 minutes (95% CI 0.07 to 5.28 minutes). Supplemental oxygen increased the average exercise time from 6 to 14 minutes; the control intervention increased average exercise time from 6 to 12 minutes. Constant power exercise end‐of‐test Borg score (on a scale from 1 to 10) also showed statistically significant improvements with oxygen‐supplemented exercise training, WMD ‐1.22 units (95% CI ‐2.39 to ‐0.06). One study showed a significant improvement in the change of Borg score after the shuttle walk test, by ‐1.46 units (95% CI ‐2.72 to ‐0.19). There were no significant differences in maximal exercise outcomes, functional exercise outcomes (six‐minute walk test), shuttle walk distance, health‐related quality of life or oxygenation status. According to the GRADE system most outcomes were rated as low quality because they were limited by study quality.

Authors' conclusions

This review provides little support for oxygen supplementation during exercise training for individuals with COPD, but the evidence is very limited. Studies with larger number of participants and strong design are required to permit strong conclusions, especially for functional outcomes such as symptom alleviation, health‐related quality of life and ambulation.

Plain language summary

Oxygen therapy during exercise training in chronic obstructive pulmonary disease (COPD)

People with COPD often have severe shortness of breath that prevents them from performing their everyday activities. Formal rehabilitation programs that include exercise training improve stamina, decrease breathlessness and enhance quality of life. Use of oxygen during exercise may provide additional benefit. We conducted a review to determine the effectiveness of adding oxygen to exercise training in comparison to exercise training without oxygen supplementation in people with COPD. Five studies addressed the question although, because of measurement of different outcomes, the maximum number available for looking at any individual outcome was three (31 patients receiving oxygen versus 32 not). People with COPD may exercise longer and have less shortness of breath when using oxygen during an exercise‐training program. These studies did not look at the effect of oxygen on shortness of breath in daily life. From the evidence to date, it is not possible to determine whether individuals with COPD should use oxygen during exercise training. Stronger studies with more participants are required in order to understand how oxygen‐supplemented exercise training for people with COPD will affect their shortness of breath, activity and quality of life.

Background

Chronic obstructive pulmonary disease (COPD) is a leading cause of morbidity and mortality worldwide and results in economic and social burden (Celli 2004). The prevalence may be greatly underestimated because COPD is usually not diagnosed until it is clinically apparent and moderately advanced (Anto 2001; Celli 2004). In people aged 25 to 75 years in the USA, the estimated prevalence is about 7% and it is the fourth leading cause of death in North America and Europe (Celli 2004; O'Donnell 2003b).

COPD is a slow progressive disorder that is characterized by a gradual loss in lung function due to airway obstruction, which does not change markedly with time or treatment. The first presentation of COPD usually includes shortness of breath, chronic cough and sputum production (NHLBI 2003).

In some individuals, gas exchange becomes more impaired as COPD progresses and normal oxygenation can no longer be maintained. Oxygen has been administered under a number of conditions. The benefits of long‐term oxygen therapy (LTOT) for individuals with chronic resting hypoxemia are well recognized (MRC 1981; NOTT 1980) and have been reported in a previous Cochrane review (Crockett 2005). Oxygen has also been used to treat those who do not have chronic resting hypoxemia but experience hypoxemia only during exercise or activities of daily living (ambulatory oxygen). Studies of short‐term ambulatory oxygen for COPD include single assessment studies comparing performance during an exercise test using oxygen with performance during an exercise test using placebo air. A recent Cochrane review found that short‐term ambulatory oxygen improves single assessment laboratory exercise performance in people with moderate to severe COPD (Bradley 2005). Long‐term ambulatory oxygen studies include those in which oxygen is used at home during activities of daily living (Bradley 2005; Ram 2002). A Cochrane review found only two studies (a total of 70 patients), one which included patients with severe resting hypoxemia and the other with resting normoxemia (Ram 2002). Finally, oxygen has been used as a temporary measure for those who are moderately hypoxemic or normoxic (during both rest and exertion) in situations such as short‐burst therapy for breathlessness, acute therapy during an exacerbation and for exercise training.

Exercise intolerance is an important problem for individuals with COPD, often leading to a sedentary lifestyle and a downward spiral of further deconditioning and disability (Casaburi 1993). Exercise limitation in individuals with COPD has been attributed to a reduction in ventilatory capacity (expiratory airflow limitation and decreased elastic recoil causing lung hyperinflation) and increased ventilatory requirements (ventilation to perfusion mismatch) (Casaburi 1993; Wasserman 1993). Exercise training for individuals with COPD decreases ventilatory requirements by yielding more efficient exercise breathing patterns and lowering ventilation, allowing for improved exercise tolerance (Casaburi 1997). Exercise training is an integral component of pulmonary rehabilitation (PR), a multidisciplinary program of care that is directed to persons with COPD and their families and is designed to optimize physical and social performance and autonomy (ATS 2004; Lacasse 2006). PR results in reduced dyspnoea, increased exercise ability and improvement in both disease‐specific and generic health status (Chronic Respiratory Disease Questionnaire, St. George's Respiratory Questionnaire) (ATS 2004).

Oxygen may improve exercise tolerance by increasing arterial oxygen content thereby directly inhibiting carotid body stimulation, improving muscle oxygenation, allowing reduction of lactic acid production (thus decreasing carotid body stimulation) and causing pulmonary vasodilation, leading to increased cardiac output and muscle oxygen delivery (Fujimoto 2002; Somfay 2002). The combination of exercise training and supplemental oxygen may interrupt the downward spiral by enabling individuals with COPD to tolerate higher levels of activity with less exertional symptoms, ultimately improving quality of life.

Objectives

The objective of this systematic review was to determine how supplemental oxygen use in comparison to a control intervention (compressed air or room air) during the exercise component of a rehabilitation program affects exercise capacity, dyspnea and health‐related quality of life in individuals with COPD.

Methods

Criteria for considering studies for this review

Types of studies

Only randomized controlled trials comparing oxygen‐supplemented exercise training to non‐supplemented exercise training (control group) were considered for inclusion. Pre and post (crossover) studies were excluded.

Types of participants

Inclusion 
 Studies with adult individuals (18 years or older) diagnosed with stable COPD.

Study participants had COPD defined as: 
 1) a clinical diagnosis of COPD; and 
 2) one of the following, 
 a. best recorded FEV1/FVC ratio of individual study participants < 0.70, 
 b. best recorded FEV1 of individuals study participants < 80% (ATS 2004; Pauwels 1998; Romain 2001).

Exclusion 
 We excluded trials that included any patients who met criteria for long‐term oxygen therapy (LTOT) unless the trial specified the proportion of patients who used LTOT and that proportion was less than 10%. We looked for the following criteria for LTOT 1): resting PaO₂ of less than or equal 55 mm Hg (7.33 kPa) when breathing room air; 2) resting PaO₂ less than or equal to 59 mm Hg (7.87 kPa) with presence of secondary polycythemia, cor pulmonale or peripheral edema.

Types of interventions

Inclusion: 
 The intervention for the actively‐treated group was oxygen therapy, provided through wall oxygen units (in a hospital or clinic), portable oxygen cylinders or liquid oxygen canisters. Intervention for the control group was either placebo (compressed air provided by a wall unit or cylinders) or no intervention (room air). All interventions were given during exercise training which was greater than or equal to three weeks in duration and included a minimum of two sessions a week. Formal (in hospital or a clinic) exercise training as a part of rehabilitation was accepted.

Exclusion: 
 Long‐term (at home for activities of daily living) or short‐term (single assessment) studies assessing the efficacy of ambulatory oxygen therapy and studies assessing the efficacy of LTOT without training or rehabilitation were excluded.

Types of outcome measures

1) Maximal exercise capacity (peak exercise capacity measured in an exercise laboratory using an incremental exercise test), constant power exercise capacity (e.g. cycle endurance) and functional exercise capacity (e.g. timed walking tests), combined or individually 
 2) Maximum training level achieved during exercise training 
 3) End‐of‐test dyspnoea scores before and after training sessions (e.g. Borg scores or a visual analogue scale) 
 4) Health‐related quality of life (e.g. Chronic Respiratory Questionnaire) 
 5) Arterial oxygen saturation (pulse oximetry or arterial blood gases) post exercise test

Search methods for identification of studies

Electronic searches

Relevant studies were identified with the assistance of the Cochrane Airways Group. All records in the Cochrane Airways Group Specialized Register of trials coded as 'COPD' were searched using the following terms:

(oxygen* or O2*) AND (exercis* or train* or rehabilitat* or fitness* or physical* or activ* or endur* or exert* or walk* or cycle*).

This register contains records identified by systematic searching of the Cochrane Central Register of Controlled Trials (CENTRAL, The Cochrane Library), MEDLINE, EMBASE and CINAHL as well as studies identified through handsearching of respiratory journals and meeting abstracts, including annual meetings of the American Thoracic Society, British Thoracic Society and European Respiratory Society.

In order to minimise the chance of missing potential studies, separate searches were also done on CENTRAL. In addition, other electronically available databases and search engines were searched for trials (MEDLINE, EMBASE, CINAHL, Science Citation Index, Scirus, UK National Research Register, PEDro, Web of Science, ClinicalTrials.gov and Google.com).

Electronic websites of the following journals were also searched: American Journal of Respiratory and Critical Care Medicine, Annals of Internal Medicine, British Medical Journal, Chest, European Respiratory Journal, Lancet, Respiratory Care, Respiratory Medicine, Thorax, American Journal of Physical Medicine & Rehabilitation, Archives of Physical Medicine & Rehabilitation, Clinical Rehabilitation, International Journal of Rehabilitation and Health, International Journal of Rehabilitation Research, Journal of Cardiopulmonary Rehabilitation, Journal of Rehabilitation, Journal of Rehabilitation & Research, Journal of Rehabilitation Medicine).

All databases were searched from their inception up until June 2009.

Searching other resources

Bibliographies of each included RCT as well as any review article or texts found were searched for additional relevant papers that may contain further RCTs.

Authors of included RCTs were contacted for other published and unpublished studies. All authors responded (Emtner 2003; Garrod 2000; Rooyackers 1997; Wadell 2001) except (Fichter 1999).

Language

The search identified relevant studies of all languages. Only one non‐English paper (German) was included and it was translated.

Data collection and analysis

Selection of studies

Two review authors (DB and MLN) independently selected trials for inclusion in the review. They first reviewed titles and abstracts and obtained full articles of studies judged potentially eligible. Both DB and MLN screened the full articles for inclusion. There was total agreement between the review authors for study inclusion at each level (abstract and full report). An attempt was made to contact an author of one abstract (Patessio 1996) to determine if it was fully published or if unpublished data were available. The author could not find the complete data. Please see the 'Characteristics of excluded studies' table for a list of reasons for exclusion.

Data extraction and management

Two review authors (MLN and DB) independently extracted all data from the trials onto a predefined form. The extracted data included: 1) general information about the studies (for example eligibility criteria, study design), 2) quality assessment including participant dropout and withdrawals, 3) description of participants, 4) description of interventions (exercise training programs and gas treatment), and 5) description of outcome measures (exercise testing, health‐related quality of life). Although exercise test outcome measures were performed both on oxygen and on room air (or compressed air), only the latter was considered for the review.

All five authors were contacted to provide missing information in a table or check‐box format (similar to the review authors' data extraction forms) that were individually developed for each study. Four of the five authors responded (Emtner 2003; Garrod 2000; Rooyackers 1997; Wadell 2001), confirming some of the data extracted and providing further information.

Assessment of risk of bias in included studies

Two review authors (DB and MLN) assessed the internal validity of the studies, considering the sources of bias that have major influence on the magnitude of effect size in clinical trials: unconcealed randomization, blinding, dropouts and withdrawals (in this review unblinded trials were accepted).

For the 2008 update of this review, we have included an assessment of the risk of bias for each study included in the review. This is in line with recommendations from the Cochrane Handbook regarding the assessment and presentation of the risk of bias for each eligible study (Handbook 2008).

Each item in the validity assessment was described separately and ratings or grading were provided in addendum. For example, the allocation of concealment was graded according to the following scale.

Grade A: adequate concealment. 
 Grade B: unclear concealment. 
 Grade C: clearly inadequate concealment. 
 Grade D: not used

The quality assessment was rated according to a 0 to 5 scale suggested by Jadad (Jadad 1996). 
 1) Was the study described as randomized (1 = yes, 0 = no). 
 2) Was the study described as double blind (1 = yes, 0 = no). 
 3) Was there a description of withdrawals and dropouts (1 = yes, 0 = no). 
 4) Was the method of randomization well described and appropriate (1 = yes, 0 = no). 
 5) Was the method of double‐blinding well described and appropriate (1 = yes, 0 = no). 
 6) Deduct one point if methods for randomization or blinding were inappropriate.

There was disagreement on one occasion, regarding dropouts and withdrawals. Agreement was reached by consensus. Clarification was sought from authors where needed.

The GRADE system was used to summarize findings and rate the overall quality of evidence for each outcome (Atkins 2004; Guyatt 2006; Schunemann 2006).

Measures of treatment effect

1) Exercise capacity

There are a number of different protocols used to test exercise capacity, for different purposes. Maximal (or incremental or progressive) exercise tests assess the integrated systematic responses to gradually increasing workload. Maximal exercise capacity is measured in terms of peak exercise level (Liesker 2002). Examples of this protocol include cycle and treadmill ergometer tests. Functional exercise tests reflect a person's physiological function but relate closely to their ability to undertake physical demands of daily living (Guyatt 1985). Examples of this protocol include the 2‐minute, 6‐minute and 12‐minute walk tests (Solway 2001). Constant power (or submaximal or endurance) exercise tests assess parameters of cardiopulmonary function specifically under conditions of constant metabolic demand. They are useful for assessing responses to a known workload and to evaluate effectiveness of various therapies on exercise ability (Ruppel 1994). An example of this protocol is the cycle endurance test.

The shuttle walk test, used by one of the selected studies (Garrod 2000), is difficult to categorize. It has been considered a measure of maximal exercise capacity (Singh 1992; Singh 1994), functional exercise capacity (Solway 2001) and even as a 'halfway house' between endurance and maximal exercise capacity (Calverley 2000). It is considered in a separate analysis in this review.

The results from different exercise test protocols were not combined because their respective outcomes represent different constructs (Lacasse 2006). Only results from the same exercise test protocols were combined, that is all maximal exercise tests, all constant power exercise tests, all functional exercise tests and all shuttle walk tests.

The following change in exercise capacity outcome measures (before and after training) were used in the analyses.

• Maximal exercise protocol: power (watts); energy expenditure (oxygen consumption = VO₂ max in L/min); measures at identical power and duration (isotime) for energy expenditure (VO₂ in L/min).

• Functional exercise protocol: distance walked (metres).

• Constant power exercise protocol: exercise time (minutes); energy expenditure (VO₂ in L/min); measures at identical duration (isotime) for energy expenditure (VO₂ in L/min).

• Shuttle walk test exercise protocol: distance walked (metres).

2) Maximum training level, achieved during the training program

Compared between the oxygen‐supplemented group and the control group.

3) Changes in the end‐of‐test dyspnoea (Borg) scores for the tests of exercise capacity

Compared between the oxygen‐supplemented trained group and the control group. A higher Borg score indicates worse dyspnea.

4) Changes in the end‐of‐test arterial oxygen saturation (pulse oximetry or arterial blood gases) for the tests of exercise capacity

Compared between the oxygen‐supplemented trained group and the control group. This measure, at identical power and duration (isotime), was also included.

5) Health‐related quality of life (HRQL)

Instruments that have been shown to be valid and responsive over time were considered in the analysis. Changes in scores were compared between the oxygen‐supplemented trained group and the control group. In this review the Chronic Respiratory Disease Questionnaire (CRQ), the Medical Outcomes Study Short Form 36 (SF‐36), the Hospital Anxiety and Depression Scale (HAD) and the London Chest Activities of Daily Living Scale (LCADL) were used. A higher score in the CRQ and SF‐36 indicates better outcomes, while a lower score in the HAD and LCADL indicates better outcomes.

For certain outcomes, the estimates and confidence limits for the difference between groups were assessed to see if they exceeded the minimal important difference (MID). The MID has been defined as the smallest difference in score that patients perceive as a change and that would mandate, in the absence of troublesome side effects and excessive cost, modification in the patient's management (Jaescheke 1989). For the CRQ domains (dyspnea, fatigue, emotion, mastery and total) the MID (on a 1 to 7 Likert point scale) is 0.5 (Lacasse 1997). For the 6‐minute walk test the MID is 54 metres with a 95% confidence interval of 37 to 71 metres (Redelmeier 1997). For the shuttle walk test the suggested MID is 48 metres with 95% confidence interval of 33.6 to 63.6 metres (Bradley 2005).

Assessment of heterogeneity

Heterogeneity was tested with the Cochrane chi‐square (Q) test of homogeneity and with the I² statistic, both available through RevMan 4.2. The chi‐square test measures the deviation of observed effect sizes from an underlying overall effect. A higher Q value or lower P‐value (relative to its degrees of freedom) provides evidence of heterogeneity of treatment effects (variation in effect estimates beyond chance) (Deeks 2005). This test has low power at detecting true heterogeneity, especially with a small number of studies, therefore, a cut‐off P‐value of 0.10 was used (Beyene 2005). The I² statistic quantifies inconsistency across studies and moves the focus away from testing whether heterogeneity is present to assessing its impact on the meta‐analysis (Deeks 2005). It approximates the proportion of total variation in study estimates that is due to heterogeneity rather than by random error or chance. As a rough guide of magnitude: 25% is considered low, 50% moderate and 75% high (Beyene 2005). In this systematic review, these quantitative interpretations were limited because of the few studies available for comparison.

Data synthesis

All trial data were combined and analyzed using RevMan 4.2. The weighted mean difference (WMD) was selected when combining data. For all studies, the before‐training and after‐training means and respective standard deviations were provided. In order to use the change in outcome measures, the following was carried out.

a) The before‐training means were subtracted from the after‐training means:

(meanX_after) ‐ (meanX_before)

b) The standard deviations (SD) were pooled according to this equation (Dunst 2004):

square root((SD²_after ‐ SD²_before)/2)

The random‐effects model was used for all analyses and forest plots were used to compare results across studies.

Subgroup analysis and investigation of heterogeneity

The small number of studies also precluded subgroup and sensitivity analyses and the creation of funnel plots to test for the presence of publication bias. If in future updates there are more studies to compare, the following subgroup analyses will be considered: 
 a) blinding of the study participant; 
 b) individuals who do not have evidence of resting hypoxemia but who demonstrate a fall in SpO₂ to at least 90% during a baseline walking test whilst breathing room air versus those who maintain a SpO₂ of greater than 90%.

Results

Description of studies

Results of the search

There were 1621 publications identified from the initial search of databases. From this list, 169 full articles were retrieved for closer inspection. After examining the full articles, 47 additional articles were identified from the bibliographies and also retrieved for closer inspection (216 in total). From these full articles, five were deemed eligible for inclusion. An update search run in June 2009 did not identify any new studies for inclusion in the review.

Included studies

The 'Characteristics of included studies' table presents detailed descriptions of the methods, participants, interventions and outcomes.

Excluded studies

Most of the studies identified were excluded for several reasons. The most common reason was that they were not randomized controlled trials (n = 109) and/or did not have an exercise‐training program as part of their intervention (n = 91). A full log of reasons for exclusion of studies was tabulated ('Characteristics of excluded studies') and provided in a flow diagram (Figure 1). Four of five abstracts were later published in full and included in the review. An attempt was made to contact the author of the fifth abstract (Patessio 1996) to determine if it was fully published or if unpublished data were available, but the author could not find the complete data. One paper was written in German (Fichter 1999) and was translated.

1.

Flow Diagram Showing Progress Through the Systematic Review

Risk of bias in included studies

Characteristics of included studies includes details of assessment of methodological quality and the Jadad instrument. A plot of our judgements and the information we used to generate them are also given. A graphical overview of our judgements for three domains (allocation ‐ generation, concealment; blinding) are provided in Figure 2.

2.

Methodological quality summary: review authors' judgments about each methodological quality item for each included study.

All five included studies were parallel RCTs. Four of the five had an adequate randomization process: three were computer random‐number generated (Emtner 2003; Garrod 2000; Rooyackers 1997) and one picked allocation from a hat or box (Wadell 2001). The method of the fifth trial was not reported (Fichter 1999). The authors were contacted but did not provide clarification on the randomization process.

Two of the five studies had adequate concealment of allocation (rating A): numbered, opaque sealed envelopes (Garrod 2000) and central randomization (Rooyackers 1997). One had sequentially numbered envelopes but was declared unclear (rating B) because the envelopes were not specified as opaque or sealed (Emtner 2003). The other two studies stated allocation was random but did not provide further details, even after contacting the authors for clarification (rating B) (Fichter 1999; Wadell 2001).

All studies had some blinding of the interventions. In three studies only the study participants were blind (Fichter 1999; Garrod 2000; Wadell 2001). This was not confirmed in two of these studies, even after contacting the authors for clarification (Fichter 1999; Wadell 2001). The fourth study was blind only to the data analysts (Wadell 2001). The final study (Emtner 2003) was blind to the participants, healthcare providers, data collectors, assessors of outcomes and the data analysts.

Three of the five studies reported withdrawals and dropouts (Emtner 2003; Garrod 2000; Rooyackers 1997). Two of these studies specified which group they were withdrawn from (oxygen or control) and that they were excluded from both the numerator and denominator in the analysis (Emtner 2003; Wadell 2001). The third study (Garrod 2000) did not specify which group their patients were withdrawn from, but did specify they were excluded from both the numerator and denominator in the analysis. Fichter 1999 did not report any withdrawals or dropouts. One study reported there was no loss to follow up and all patients randomized were analyzed (Rooyackers 1997). Only one study (Garrod 2000) reported the number of individuals approached to take part in the study and who chose not to participate: 1 out of 26.

Effects of interventions

The 'Table of comparisons' and Table 1 summarize the primary results of the meta‐analysis. Figure 3; Figure 4; Figure 5; Figure 6; Figure 7 summarize the findings and assess the overall quality of evidence for each outcome using the GRADE system (Atkins 2004; Guyatt 2006; Schunemann 2006). Five of the comparisons had three studies pooled for the meta‐analysis, 15 comparisons had two studies pooled and 23 comparisons had only one study included.

1. Table of comparison: HAD and LCADL individual domains.

Outcome measure	Trials	# Participants (O/C)	Mean difference	95% CI
HAD: Anxiety	1	11 / 11	1.72	‐0.59 to 4.05
HAD: depression	1	11 / 11	0.00	‐2.90 to 2.90
LCADL: self care	1	11 / 11	‐1.36	‐3.46 to 0.73
LCADL: domestic	1	11 / 11	0.09	‐3.86 to 4.06
LCADL: physical	1	11 / 11	‐0.73	‐1.96 to 0.50
LCADL: leisure	1	11 / 11	‐0.18	‐1.87 to 1.49

3.

GRADE Evidence Table: Primary Results of Meta‐Analysis, Part A

4.

GRADE Evidence Table: Primary Results of Meta‐Analysis , Part B

5.

GRADE Evidence Table: Single‐Trial Meta‐analysis, Part A

6.

GRADE Evidence Table: Single‐Trial Meta‐analysis, Part B

7.

GRADE Evidence Table: Single‐Trial Meta‐analysis, Part C

Please note: all exercise outcomes (pre and post) reported in this review (for both intervention and control groups) were from tests performed on compressed air (Emtner 2003; Garrod 2000; Wadell 2001) or room air (Fichter 1999; Rooyackers 1997).

Exercise Capacity

Maximal exercise capacity, measured by incremental cycle ergometry, was used in three of the five trials (Emtner 2003; Fichter 1999; Rooyackers 1997): a total of 31 participants in the oxygen‐trained group and 32 in the control group. The common effect (weighted mean difference) in power was 3.59 watts in favour of the control group, but was not statistically significant (95% CI ‐18.03 to 10.86 watts, Analysis 1.5). The energy expenditure (VO2 max) was 0.01 L/min in favour of the control group, but was not statistically significant (95% CI ‐0.17 to 0.16 L/min, Analysis 1.6). The isotime VO2 was measured by two studies (Emtner 2003; Fichter 1999) in which 19 participants in the oxygen‐trained group were compared to 20 participants in the control. The isotime VO2 was 0.03 L/min in favour of the oxygen group, but was not significant (95% CI ‐0.20 to 0.14 L/min, Analysis 1.12).

1.5. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 5 Maximal Test ‐ power (watts).

1.6. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 6 Maximal Test ‐ energy expenditure, VO2 max (L/min).

1.12. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 12 Maximal Test ‐ isotime energy expenditure, VO2 (L/min).

Two trials measured functional exercise capacity using the six‐minute walk test, (Rooyackers 1997; Wadell 2001): a total of 22 participants in the oxygen‐trained group and 22 participants in the control group. The common effect in distance walked was 23.87 metres in favour of the control group, but was not statistically significant (95% CI ‐81.55 to 33.82 metres, Analysis 1.15). The common effect size did not exceed the minimal clinically important difference (MCID) of 54 metres (Redelmeier 1997), however the smallest possible effect went beyond the confidence interval around the estimate of the MCID (95% CI 37 to 71 metres). Therefore, the clinical significance of the result obtained through meta‐analysis remains uncertain.

1.15. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 15 6 Minute Walk Test ‐ distance (metres).

Constant power exercise capacity, measured by cycle ergometry, was used in two trials (Emtner 2003; Rooyackers 1997): a total of 26 participants in the oxygen‐trained group and 27 in the control group. The effect on exercise time was significantly in favour of the oxygen‐trained group: 2.68 minutes (95% CI 0.07 to 5.28 minutes, Analysis 1.1). This effect may have been underestimated because Emtner 2003 terminated 16 tests in the oxygen‐trained group at 30 minutes versus only 7 tests in the control group. The energy expenditure (VO2) and the isotime VO2 were measured only by one study (Emtner 2003) where 14 participants in the oxygen‐trained group were compared with 15 participants in the control. The VO2 was 0.05 L/min in favour of the control group, but was not statistically significant (95% CI ‐0.25 to 0.15 L/min, Analysis 1.2). The isotime VO2 was 0.06 L/min in favour of the oxygen group, but was not significant (95% CI ‐0.26 to 0.14 L/min, Analysis 1.10).

1.1. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 1 Cycle Endurance ‐ exercise time (minutes).

1.2. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 2 Cycle Endurance ‐ energy expenditure, VO2 (L/min).

1.10. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 10 Cycle Endurance ‐ isotime energy expenditure, VO2 (L/min).

The shuttle walk test was used in one trial (Garrod 2000) where 11 participants in the oxygen‐trained group were compared with 11 participants in the control group. They found a mean difference in change of distance walked between groups of 23.6 metres in favour of the control, but this was not significant (95% CI ‐70.7 to 23.5 metres, Analysis 1.18). The common effect size did not exceed the MCID of 48 metres (Singh 1992), however the smallest possible effect (‐70.7 metres) went beyond the lower confidence interval around the estimate of the MCID (95% CI: 33.6 to 63.6 metres). Therefore, the clinical significance of the mean difference in change of shuttle distance walked obtained through meta‐analysis remains uncertain.

1.18. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 18 Shuttle Walk Test ‐ distance (metres).

Tests of heterogeneity on all measures of exercise capacity were not significant and all I² values were zero per cent.

Maximum Training Level Achieved

Maximal training level achieved during the training program was measured in two of the five trials (Emtner 2003; Rooyackers 1997): a total of 26 participants in the oxygen‐trained group and 27 in the control group. The common effect was 2.30 watts in favour of the oxygen‐trained group, which was not significant (95% CI ‐15.16 to 19.75 watts, Analysis 1.30). Based on the chi‐square test (Q = 1.99, P‐value 0.16) and the I2 statistic (49.7%) these two studies had moderate amounts of heterogeneity. In the Emtner 2003 trial, the type of exercise‐training program was cycle ergometry. In the Rooyackers 1997 trial, the exercise‐training program was interval training that included cycling for 20 minutes where participants cycled for 2 minutes and rested for 2 minutes. This difference in training may have influenced the heterogeneity.

1.30. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 30 Maximum training work level achieved (watts).

End‐of‐Test Borg Dyspnea Score

Maximal exercise end‐of‐test Borg dyspnea score was measured in two of the five trials (Emtner 2003; Rooyackers 1997): a total of 26 participants in the oxygen‐trained group and 27 in the control group. The common effect was 0.25 units in favour of the control group, but this was not significant (95% CI ‐0.82 to 1.32 units, Analysis 1.7).

1.7. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 7 Maximal Test ‐ end‐of‐test borg dyspnea score.

Functional exercise end‐of‐test Borg dyspnea was measured in two of the five trials (Rooyackers 1997; Wadell 2001): a total of 22 participants in the oxygen‐trained group and 22 in the control group. The common effect was 0.06 units in favour of the control group, but this was not significant (95% CI ‐0.89 to 1.00 units, Analysis 1.16).

1.16. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 16 6 Minute Walk Test ‐ end‐of‐test borg dyspnea score.

Constant power exercise end‐of‐test Borg dyspnea was measured in two of the five trials (Emtner 2003; Rooyackers 1997): a total of 26 participants in the oxygen‐trained group and 27 in the control group. The common effect was 1.22 units in favour of the oxygen‐trained group, and was significant (95% CI ‐2.39 to ‐0.60 units, Analysis 1.3). Based on the chi‐Square test (Q = 1.43, P‐value 0.23) and the I² statistic (30%) these two studies had moderate amounts of heterogeneity. The reasons may be similar to those found in the maximum training level achieved (see above).

1.3. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 3 Cycle Endurance ‐ end‐of‐test borg dyspnea score.

The shuttle walk test was used in one trial (Garrod 2000) where 11 participants in the oxygen‐trained group were compared to 11 participants in the control group. The study found a significant change in mean difference of Borg score 1.46 units in favour of the oxygen‐trained group (95% CI ‐2.72 to ‐0.19 units, Analysis 1.19).

1.19. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 19 Shuttle Walk Test ‐ end‐of‐test borg dyspnea score.

Tests of heterogeneity on all measures of end‐of‐test Borg dyspnea scores were not significant. All I² values were zero per cent with the exception of the constant power exercise test (see above).

End‐of‐Test Arterial Oxygen Saturation

Maximal exercise end‐of‐test SpO₂ was measured in two of the five trials (Emtner 2003; Rooyackers 1997): a total of 26 participants in the oxygen‐trained group and 27 in the control group. The common effect was 1.00 % in favour of the control group, but this was not statistically significant (95% CI ‐2.96 to 0.96%, Analysis 1.8). The isotime end‐of‐test SpO₂ was measured by only one group (Emtner 2003) where 14 participants in the oxygen‐trained group were compared to 15 participants in the control group. The isotime end‐of‐test SpO₂ was 1.00 % in favour of the control group, but was not significant (95% CI ‐3.49 to 1.49%, Analysis 1.13).

1.8. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 8 Maximal Test ‐ SpO2 (%).

1.13. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 13 Maximal Test ‐ isotime SpO2 (%).

Maximal exercise end‐of‐test PaO₂ was measured in two of the five trials (Fichter 1999; Rooyackers 1997): a total of 17 participants in the oxygen‐trained group and 17 in the control group. The common effect was 0.37 mm Hg (0.05 kPa) in favour of the oxygen‐trained group, but this was not statistically significant (95% CI ‐3.82 to 4.56 mm Hg; ‐0.51 to 0.61 kPa, Analysis 1.9). The isotime end‐of‐test PaO₂ was measured by only one group (Fichter 1999), where five participants in the oxygen‐trained group were compared to five participants in the control group. The isotime end‐of‐test PaO₂ was 1.40 mm Hg (0.17 kPa) in favour of the control group, which was not significant (95% CI ‐11.11 to 8.31 mm Hg; ‐1.48 to 1.11 kPa, Analysis 1.14).

1.9. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 9 Maximal Test ‐ PaO2 (mmHg).

1.14. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 14 Maximal Test ‐ isotime PaO2 (mmHg).

Functional exercise end‐of‐test SpO₂ was measured in two of the five trials (Rooyackers 1997; Wadell 2001): a total of 22 participants in the oxygen‐trained group and 22 in the control group. The common effect was 0.00% and was not significant (95% CI ‐2.66 to 2.66%, Analysis 1.17).

1.17. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 17 6 Minute Walk Test ‐ SpO2 (%).

Constant power exercise end‐of‐test SpO₂ was measured in two of the five trials (Emtner 2003; Rooyackers 1997): a total of 26 participants in the oxygen‐trained group and 27 in the control group. The common effect was 1.47% in favour of the control group, but was not significant (95% CI ‐3.45 to 0.51%, Analysis 1.4). The isotime end‐of‐test SpO₂ was measured by only one group (Emtner 2003), where 14 participants in the oxygen‐trained group were compared to 15 participants in the control group. The isotime end‐of‐test SpO₂ was 1.00 % in favour of the control group, which was not significant (95% CI ‐1.26 to 3.26 %, Analysis 1.11).

1.4. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 4 Cycle Endurance ‐ SpO2 (%).

1.11. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 11 Cycle Endurance ‐ isotime SpO2 (%).

Tests of heterogeneity on all measures of end‐of‐test arterial oxygen saturation were not significant and all I² values were zero percent.

Health‐Related Quality of Life

Only the CRQ measure had enough studies (three) to pool in a meta‐analysis. The SF‐36 (Emtner 2003), HAD and LCADL (Garrod 2000) were measured by only one study.

In all three studies, the CRQ was measured at baseline and immediately following treatment. The only exception was Rooyackers 1997 who administered the CRQ six weeks after training. For each of the four domains of the CRQ, two of the five trials were pooled (Emtner 2003; Rooyackers 1997): a total of 26 participants in the oxygen‐trained group and 27 in the control group. The weighted mean differences represent differences in the CRQ score per item (that is on a seven‐point scale). The common effect in CRQ dyspnea, fatigue, emotion and mastery were all in favour of the oxygen‐trained group, but were not significant. For the total CRQ score (of all domains) three of the five trials were pooled (Emtner 2003; Garrod 2000; Rooyackers 1997): 37 participants in the oxygen‐trained group and 38 in the control group. The common effect was in favour of the oxygen‐trained group, but this was not significant. Tests of heterogeneity on all measures of the CRQ were not significant and all I2 values were zero per cent. Please see 'Table of comparisons' (plots 31 to 35) for more details.

The SF‐36 was used in one trial (Emtner 2003), where 14 participants in the oxygen‐trained group were compared to 15 participants in the control group. Most of the common effects favoured the oxygen‐trained group, with general health reaching statistical significance (15.00 points: 95% CI 1.51 to 28.49). Please see 'Table of comparisons' (plots 20 to 27) for more details.

The HAD and the LCADL were used in one trial (Garrod 2000), where 11 participants in the oxygen‐trained group were compared to 11 participants in the control group. With respect to the HAD, the mean difference in change of all scores between groups favoured the control group but none were significant. With respect to the LCADL, the mean difference in the change of most scores between groups favoured the oxygen‐trained group, but none were significant. Please see 'Table of comparisons' (plots 28 and 29) and Table 1 for more details.

Discussion

Pulmonary rehabilitation that includes exercise training improves outcomes of exercise capacity, dyspnea and health‐related quality of life in individuals with COPD (Casaburi 1997; Lacasse 2006). Oxygen improves outcomes relating to endurance exercise capacity (distance, time, number of steps) and maximal exercise capacity (exercise time and work rate) during single‐assessment constant power and maximal exercise testing in individuals with COPD (Bradley 2005). In this review, five studies comparing oxygen‐supplemented and control exercise training in individuals with COPD were included. The overall effect of oxygen‐supplemented exercise training did not reveal consistent benefits on the outcomes of interest. There were a few significant improvements of oxygen‐supplemented exercise training in the meta‐analysis; these results were associated with methodological weakness because of low to moderate quality of evidence. There were no significant differences in maximal exercise outcomes, functional exercise outcomes (six‐minute walk test), shuttle walk distance, health‐related quality of life or oxygenation status.

Exercise Capacity and Maximum Training Level Achieved

The effects of oxygen in individuals with COPD may be more evident in terms of constant power exercise capacity (compared to maximal exercise) because this test is more responsive to changes in treatment (ATS 2003; Calverley 2000; Oga 2000). Individuals are less affected by reduced ventilatory capacity and are able to reach their true peak exercise level. Considering that the activities of daily living of individuals with COPD are submaximal, measuring at maximal exercise may have little bearing on functional status (Oga 2000). In this review, constant power was the only exercise capacity outcome measure that showed differences. After pooling two studies (Emtner 2003; Rooyackers 1997) the constant power exercise time was significantly greater after oxygen training. This effect may have been underestimated because Emtner 2003 terminated 16 tests in the oxygen‐trained group at 30 minutes versus only 7 tests in the control group. The study by Garrod 2000 included participants with greater hypoxemia. Their inclusion criteria required participants to desaturate with exercise and 11 of them were receiving long‐term oxygen therapy (five trained with compressed air, six trained with supplemental oxygen). This may have resulted in a larger response to training with oxygen. A subgroup analysis was not possible to elucidate this issue.

Exercise workload performance, specifically in maximal exercise testing, may be expressed quantitatively in several ways: work, power and energy (Wasserman 2005). Energy expenditure (VO₂) during maximal exercise testing was analyzed in this review (two studies contributed to the meta‐analysis). The VO₂ was not significantly different between the oxygen‐trained group and the control group. This is consistent with the lack of change in the maximal test power results between the two groups.

Constant power exercise time increased in favour of the oxygen‐supplemented group, however this difference in effect may have been overestimated. Two studies limited participants' exercise‐training levels on the basis of SpO₂. Rooyackers 1997 limited the work rate to a level that would not allow the SpO₂ to fall below 90%. Wadell 2001 stopped participants' exercise if the SpO₂ fell below 90%. It has been suggested that muscle hypoxia may act as a stimulus to training in individuals (Sundberg 1993). This would likely occur more frequently in the control‐trained group, allowing them to reach a higher level of exercise capacity and further decreasing the difference of effect between the two groups. It is not clear if this muscle hypoxia stimulus would occur in the participants of the trials in this systematic review. Garrod 2000 did not have saturation limitations imposed on their control group despite their participants having the lowest mean PaO₂ out of all the studies. This did not enhance exercise tolerance in their study sample, nor was it associated with adverse effects.

In other measures of exercise capacity (maximal exercise power, maximal exercise VO₂, six‐minute walk distance), trends toward improvement favoured the control group. This may indicate exercise training without supplemental oxygen may provide benefit in these outcomes.

Dyspnea

Dyspnea is one of the most important symptoms individuals with COPD experience and often limits exercise tolerance. Many clinicians regard it as a primary outcome measure in the rehabilitation of patients with COPD. A one unit of change has been suggested to represent the minimum important difference (MID) for dyspnea ratings using the modified Borg scale (Solway 2002). In this review oxygen training significantly reduced the Borg dyspnea rating after performing the constant power exercise test. However, one of the studies in the pooled analysis (Rooyackers 1997) did not have a placebo arm (the control group had no intervention). Spence 1993 found the administration of compressed air had a cooling effect on the face and diminished the sensation of dyspnea. This effect may not have been distinguishable from the oxygen‐training effect seen in the Rooyackers 1997 study. Garrod 2000 found the mean change in Borg dyspnea after the shuttle walk test was significantly reduced by oxygen and the effect was greater than the MID. Since Garrod 2000 did not find any significant reductions in exercise capacity or activities of daily living, these results in Borg dyspnea may be due to other unknown factors or low statistical power. In all other outcomes the dyspnea rating favoured the control group and the upper (indicating worse dyspnea) 95% confidence level went beyond the suggested MID. These results support the fact that oxygen influences dyspnea to an unpredictable degree and the mechanisms involved have not been clearly established (O'Donnell 2002; Sweer 1990).

Arterial Oxygen Saturation

In this review oxygen‐supplemented exercise training showed no significant effect on the indices of oxygenation (SpO₂, PaO₂). The evidence regarding the effects of exercise training or oxygen supplementation on oxygen saturation or tension is limited. It is likely the participants in this review did not change in their oxygenation status because they were stable (no acute exacerbations) over the relatively short duration of the exercise‐training program. Even in trials of long‐term oxygen therapy, individuals with COPD and chronic hypoxemia who were stable and survived over time did not change in their oxygenation status, regardless of if they were using supplemental oxygen, or not (MRC 1981; NOTT 1980). Oxygen therapy has been shown to increase arterial oxygen tension in individuals with COPD receiving LTOT over the course of a year(PaO₂ = 55 mm Hg; 7.33 kPa), however this study was based on retrospective data analysis and is limited in its conclusions (O'Donohue 1991).

Health Related Quality of Life

For the most part, oxygen‐supplemented exercise training did not significantly change HRQL measures compared to the control group. The overall trends favoured the oxygen‐supplemented trained group. For instance, the CRQ upper bound 95% confidence interval around the common effect surpassed the MID and Emtner 2003 found a significant improvement in the SF‐36 general health domain.

The effects of oxygen‐supplemented exercise training on health‐related quality of life are not well documented in the literature. The separate effects of exercise training (in the context of pulmonary rehabilitation) and supplemental oxygen on health‐related quality of life have been addressed.

It is well established that pulmonary rehabilitation brings important improvements in health‐related quality of life (Lacasse 2006). These programs include an exercise therapy component and an education or psychological support aspect, or both, though the evidence regarding non‐exercise interventions alone (such as in self‐management programs) are inconclusive (Monninkhof 2003). The evidence regarding the effect of oxygen therapy on health‐related quality of life in individuals with COPD is not well established. One of the reasons is that these types of studies are difficult to carry out. Initial studies of the use of long‐term oxygen showed that indicators such as general sense of well‐being and neuropsychological function improved with oxygen (MRC 1981; NOTT 1980). Since then, studies of long‐term domiciliary and ambulatory oxygen in COPD have shown no significant benefit in HRQL (Lacasse 2005; Okubadejo 1996; Ram 2002). In contrast, Eaton 2002b found improvements in all domains of the CRQ, but the mean differences between the oxygen and compressed air were small and did not reach MID.

In this review the composition of the overall exercise training programs in each study differed. All trials included an exercise‐training component, with three of the five trials included additional education or psychological support (Emtner 2003; Garrod 2000; Rooyackers 1997). This review did not find supplemental oxygen provided additional benefit to the exercise training program, despite the differences in program composition.

Methodological Issues

The main issue with this review was the lack of homogeneity (preventing the pooling of data) and small sample sizes. These studies differed in many ways: clinical diversity was evident in the participants and interventions; methodological diversity was also apparent in the different designs and quality of the studies.

In three studies (Garrod 2000; Rooyackers 1997; Wadell 2001) participants were required to desaturate with exercise. In one trial (Emtner 2003) the SpO₂ was not allowed fall below 88% during exercise and in the final trial (Fichter 1999) the oxygenation status requirement was not specified. Because there were not enough studies to do a subgroup analysis and the participants who were desaturators were not separated from non‐desaturators within the studies, it is uncertain how oxygen‐supplemented training would affect these two populations.

The types and intensities of the exercise‐training programs varied along with the different range of outcome measures. The optimal type, frequency and duration of an exercise‐training program for individuals with COPD remain uncertain. Training in general improves exercise performance; however different outcomes can be expected depending on whether the person undertakes aerobic endurance versus strength training, whether high or low‐intensity training and whether upper‐limb or respiratory muscle training is pursued in addition to lower‐limb training (Casaburi 1993; Maltais 2002; Rochester 2003). There were not enough studies to do post‐hoc subgroup analysis comparing different exercise‐training programs.

The amount and method of oxygen delivery varied but in general was low in concentration or flow rate. Providing a higher concentration of oxygen or titrating oxygen to a specific SpO₂ may have improved outcomes for the treatment groups. Two studies limited exercise‐training intensity based on participant's SpO₂ (Rooyackers 1997; Wadell 2001). As mentioned in the 'Exercise capacity and maximum training level achieved' discussion above, it is not clear if abolishing a minimum saturation limit may induce benefit.

Performing practice tests for the six‐minute walk test and the shuttle walk tests improves the quality of data obtained from these outcome measurements (ATS 2002; Singh 1992). Rooyackers 1997 and Wadell 2001 performed the six‐minute walk test but only the former had at least two practice tests. Wadell 2001 had participants perform the test on a non‐motorized treadmill but also tested at the same time. Garrod 2000 performed the shuttle walk test and did not specify that a practice walk was done prior to testing.

In all studies included, the sample sizes were small and may have underpowered the overall effect. The maximum number of studies compared in the meta‐analysis was three, with 32 participants in each comparison arm. The other concern regarding design was blinding: one trial had participants, research and healthcare staff and analysts blinded (Emtner 2003); three had participants only blinded (Fichter 1999; Garrod 2000; Wadell 2001) and the fifth study only had data analysts blinded because the control arm did not receive any intervention (Rooyackers 1997). Including a placebo arm when assessing effects of oxygen therapy eliminates the possibility of this type of bias; not being placebo controlled may have led to an overestimation of the effect. In the absence of more placebo‐controlled studies, the degree to which this study overestimates treatment efficacy is difficult to ascertain reliably. There were not enough trials to perform a subgroup analysis and to explain differences that may have occurred due to blinding or not blinding, or with absence of a placebo arm.

The GRADE system involves an initial assessment of the quality of evidence followed by judgement about the direction and strength of treatment recommendations. The grading scheme classifies quality of evidence as high, moderate, low or very low. Randomized trials are considered the highest quality evidence unless limited by methodological issues, inconsistency, indirectness (generalizability) or imprecision (Atkins 2004; Guyatt 2006; Schunemann 2006). In this systematic review all outcomes were limited by weaknesses related to study quality, such as lack of blinding, small sample size and imprecise estimates of treatment effects. This resulted in a low quality for most outcomes. Outcomes which yielded a moderate quality rating were based only on one trial (Emtner 2003). The little or lack of effect on outcomes makes it unlikely this treatment would have importance. The burdens associated with this treatment for example travel to the facility, cost of membership and cost of oxygen would make any desirable effects closely balanced with undesirable effects. Please see Figure 3; Figure 4; Figure 5; Figure 6; Figure 7 for details regarding GRADE for each outcome.

Authors' conclusions

Implications for practice.

This review does not provide solid conclusions for oxygen‐supplemented exercise training. The evidence at this time indicates that oxygen does little to accentuate the benefit of exercise training. It is well established that pulmonary rehabilitation brings important improvements in health‐related quality of life (Lacasse 2006), thus any general exercise program with or without oxygen is recommended for individuals with COPD.

Implications for research.

Studies with larger number of participants and strong methodological design are required to provide clear, consistent and precise conclusions regarding oxygen‐supplemented exercise training. These studies should ensure that both participants and the providers of the intervention are blinded and should include a placebo arm. Further studies are required to establish which types of training programs and what concentration of oxygen are efficacious. Evaluation is also required to determine the effect of oxygen supplementation to exercise training on specific subgroups of individuals, particularly those who do and do not desaturate.

What's new

Date	Event	Description
12 June 2009	New search has been performed	Literature search re‐run; no new studies found

History

Protocol first published: Issue 3, 2005
 Review first published: Issue 2, 2007

Date	Event	Description
5 June 2008	Amended	Converted to new review format.
5 June 2008	New search has been performed	New search run; no new trials found
5 January 2007	New citation required and conclusions have changed	Substantive amendment

Data and analyses

Comparison 1. Oxygen Training versus Control Training (Random).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Cycle Endurance ‐ exercise time (minutes)	2	53	Mean Difference (IV, Random, 95% CI)	2.68 [0.07, 5.28]
2 Cycle Endurance ‐ energy expenditure, VO2 (L/min)	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
3 Cycle Endurance ‐ end‐of‐test borg dyspnea score	2	53	Mean Difference (IV, Random, 95% CI)	‐1.22 [‐2.39, ‐0.06]
4 Cycle Endurance ‐ SpO2 (%)	2	53	Mean Difference (IV, Random, 95% CI)	‐1.47 [‐3.45, 0.51]
5 Maximal Test ‐ power (watts)	3	63	Mean Difference (IV, Random, 95% CI)	‐3.59 [‐18.03, 10.86]
6 Maximal Test ‐ energy expenditure, VO2 max (L/min)	3	63	Mean Difference (IV, Random, 95% CI)	‐0.01 [‐0.17, 0.16]
7 Maximal Test ‐ end‐of‐test borg dyspnea score	2	53	Mean Difference (IV, Random, 95% CI)	0.25 [‐0.82, 1.32]
8 Maximal Test ‐ SpO2 (%)	2	53	Mean Difference (IV, Random, 95% CI)	‐1.0 [‐2.96, 0.96]
9 Maximal Test ‐ PaO2 (mmHg)	2	34	Mean Difference (IV, Random, 95% CI)	0.37 [‐3.82, 4.56]
10 Cycle Endurance ‐ isotime energy expenditure, VO2 (L/min)	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
11 Cycle Endurance ‐ isotime SpO2 (%)	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
12 Maximal Test ‐ isotime energy expenditure, VO2 (L/min)	2	39	Mean Difference (IV, Random, 95% CI)	‐0.03 [‐0.20, 0.14]
13 Maximal Test ‐ isotime SpO2 (%)	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
14 Maximal Test ‐ isotime PaO2 (mmHg)	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
15 6 Minute Walk Test ‐ distance (metres)	2	44	Mean Difference (IV, Random, 95% CI)	‐23.87 [‐81.55, 33.82]
16 6 Minute Walk Test ‐ end‐of‐test borg dyspnea score	2	44	Mean Difference (IV, Random, 95% CI)	0.06 [‐0.89, 1.00]
17 6 Minute Walk Test ‐ SpO2 (%)	2	44	Mean Difference (IV, Random, 95% CI)	0.0 [‐2.66, 2.66]
18 Shuttle Walk Test ‐ distance (metres)	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
19 Shuttle Walk Test ‐ end‐of‐test borg dyspnea score	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
20 SF36 ‐ physical function	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
21 SF36 ‐ physical role	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
22 SF36 ‐ bodily pain	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
23 SF36 ‐ general health	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
24 SF36 ‐ vitality	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
25 SF36 ‐ social function	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
26 SF36 ‐ emotional role	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
27 SF36 ‐ mental health	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
28 HAD ‐ total	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
29 LCADL ‐ total	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
30 Maximum training work level achieved (watts)	2	53	Mean Difference (IV, Random, 95% CI)	2.30 [‐15.16, 19.75]
31 Chronic Respiratory Questionnaire ‐ dyspnea (1‐7)	2	53	Mean Difference (IV, Random, 95% CI)	0.03 [‐0.54, 0.59]
32 Chronic Respiratory Questionnaire ‐ fatigue (1‐7)	2	53	Mean Difference (IV, Random, 95% CI)	0.22 [‐0.35, 0.79]
33 Chronic Respiratory Questionnaire ‐ emotion (1‐7)	2	53	Mean Difference (IV, Random, 95% CI)	0.33 [‐0.23, 0.88]
34 Chronic Respiratory Questionnaire ‐ mastery (1‐7)	2	53	Mean Difference (IV, Random, 95% CI)	0.20 [‐0.38, 0.79]
35 Chronic Respiratory Questionnaire ‐ total (1‐7)	3	75	Mean Difference (IV, Random, 95% CI)	0.19 [‐0.21, 0.59]

1.20. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 20 SF36 ‐ physical function.

1.21. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 21 SF36 ‐ physical role.

1.22. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 22 SF36 ‐ bodily pain.

1.23. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 23 SF36 ‐ general health.

1.24. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 24 SF36 ‐ vitality.

1.25. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 25 SF36 ‐ social function.

1.26. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 26 SF36 ‐ emotional role.

1.27. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 27 SF36 ‐ mental health.

1.28. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 28 HAD ‐ total.

1.29. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 29 LCADL ‐ total.

1.31. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 31 Chronic Respiratory Questionnaire ‐ dyspnea (1‐7).

1.32. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 32 Chronic Respiratory Questionnaire ‐ fatigue (1‐7).

1.33. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 33 Chronic Respiratory Questionnaire ‐ emotion (1‐7).

1.34. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 34 Chronic Respiratory Questionnaire ‐ mastery (1‐7).

1.35. Analysis.

Comparison 1 Oxygen Training versus Control Training (Random), Outcome 35 Chronic Respiratory Questionnaire ‐ total (1‐7).

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Emtner 2003.

Methods	STUDY DESIGN: parallel RCT INCLUSION CRITERIA: a) COPD; FEV1 < 50% pred b) FEV1/FVC > 65% c) 45‐80 years old d) clinically stable e) physically inactive f) RA rest PO2 > 55mmHg / 7.33kPa g) RA constant work rate test SpO2 > 88% EXCLUSION CRITERIA: a) symptomatic cardiovascular co‐morbidities or other disease that might contribute to exercise limitation b) regular participation in formal exercise program c) participation in formal rehab program w/n past 2 years STATISTICAL ANALYSIS: a) within groups ‐ paired t‐test, Wilcoxon test b) between groups ‐ unpaired t‐test c) relation b/w physiological variable ‐ Pearson's correlation d) difference between proportions ‐ chi‐squared e) p < 0.05 significance.
Participants	SAMPLE SIZE: 14 O2 group, 15 compressed air group AGE MEAN(SD): 66(7) O2 group, 67(10) compressed air group GENDER: 6 female O2 group, 5 female compressed air group REST RA PaO2 MEAN (SD): 71.6(8.4)mmHg / 9.54(1.112)kPa O2 group, 73.5(7.8) / 9.80(1.04)kPa compressed air group FEV1 MEAN (SD): 1.01(0.33)L O2 group, 1.01(0.27)L compressed air group FVC MEAN (SD): 2.70(0.8)L O2 group, 2.58(0.7)L compressed air group FEV1/FVC MEAN (SD): 37(8)% O2 group, 41(10)% compressed air group
Interventions	GAS TREATMENT: 3lpm O2 nasal cannula; 3lpm CA nasal cannula TYPE OF EXERCISE TRAINING: cycle ergometry INTENSITY OF TRAINING: 1st week low level; 2nd week 75% peak work rate (room air baseline test); >2 weeks increased based on dyspnea & fatigue # SESSION PER WEEK: 3 DURATION OF EACH SESSION: 45 minutes ‐ 5 min low resistance, 35 min high intensity, 5 min no resistance TOTAL DURATION OF TRAINING: 7 weeks LEVEL OF SUPERVISION: rehab & study staff (groups of 4‐6)
Outcomes	a) Maximal cycle ergometry ‐ 30% O2 & compressed air; 5 or 10 watts/min b) Constant power cycle ergometry ‐ 30% O2 & compressed air; 75% peak work rate in compressed air max test, stopped at 30min if intol not reached c) CRQ d) SGRQ e) SF‐36 f) PFT ‐ spirometry, lung volumes g) ABG Assessed at baseline (pre) and immediately following treatment phase (post)
Notes	WITHDRAWALS: 1 from O2 group due to illness during postintervention testing period. Exclusion from both the numerator and denominator. # APPROACHED TO TAKE PART & CHOSE NOT TO PARTICIPATE: not reported COUNTRY: USA JADAD SCORE: 5/5 Author contacted: full response
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Low risk	computer random‐number generator
Allocation concealment?	Unclear risk	sequentially numbered envelopes (not clear whether envelopes where sealed)
Blinding? All outcomes	Low risk	participants, healthcare providers, data collectors, assessors of outcomes, data analysts

Fichter 1999.

Methods	STUDY DESIGN: parallel RCT INCLUSION CRITERIA: a) moderate to severe COPD (based on anamnesis & several PFT measurements) b) clinically stable (no symptoms of acute exacerbation) EXCLUSION CRITERIA: a) known coronary heart disease b) other internal diseases that would limit physical capacity c) previous participation in training program. STATISTICAL ANALYSIS: a) within groups ‐ paired t‐test b) between groups ‐ unpaired t‐test
Participants	N: 5 O2 group, 5 compressed air group AGE MEAN (SD): 58(11) years O2 group; 59(7) years compressed air group GENDER: 0 female O2 group, 0 female compressed air group REST ROOM AIR PaO2 MEAN (SD): 73.5(8.25)mmHg / 9.80(1.10)kPa O2 group; 75.8(9.75)mmHg / 10.11(1.30)kPa compressed air group FEV1 MEAN (SD): 1.36(0.41)L O2 group; 1.54(0.83)L compressed air group FVC MEAN (SD): not reported FEV1/FVC MEAN (SD): not reported
Interventions	GAS TREATMENT: humidified 35% O2; humidified compressed air (21%) TYPE OF EXERCISE TRAINING: cycle ergometry INTENSITY OF TRAINING: 80% highest work rate of pre‐train exercise test # SESSIONS PER WEEK: 5 DURATION OF EACH SESSION: 45 minutes TOTAL DURATION OF TRAINING: 4 weeks LEVEL OF SUPERVSIONS: not reported
Outcomes	a) Maximal cycle ergometry on room air Assessed at baseline (pre) and immediately following treatment phase (post)
Notes	BLINDING: RANDOMIZATION: CONCEALMENT OF ALLOCATION: WITHDRAWALS: not reported # APPROACHED TO TAKE PART & CHOSE NOT TO PARTICIPATE: not reported COUNTRY: Germany JADAD SCORE: 1/5 Author contacted: no response
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Unclear risk	stated random but unable to obtain further details
Allocation concealment?	Unclear risk	no information available
Blinding? All outcomes	Unclear risk	probably participants

Garrod 2000.

Methods	STUDY DESIGN: parallel RCT INCLUSION CRITERIA: a) COPD b) FEV1 < 40% predicted, < 15% reversibility with salbutamol c) stable d) exercise limitation due to dyspnea e) exercise room air SaO2 decrease => 4% from baseline or SaO2 <= 90% EXCLUSION CRITERIA: a) exacerbation in previous 6 weeks b) unstable angina c) intermittent claudification d) mobility limiting conditions STATISTICAL ANALYSIS: a) within groups ‐ paired t‐test b) between groups ‐ unpaired t‐test c) relation between con't variables ‐ Pearson's correlation coefficient d) 95% confidence interval shown
Participants	N: 11 O2 group, 11 compressed air group AGE MEAN (RANGE): 64.3(54‐77) year O2 group, 71.6(52‐81) year compressed air group GENDER: not reported REST ROOM AIR PaO2 MEAN (SD): 63.8(10.5)mmHg / 8.51(1.40)kPa O2 group, 63.6(9.2)mmHg / 8.48(1.23)kPa compressed air group FEV1 MEAN (SD): 0.77(0.26)L O2 group, 0.84(0.26) compressed air group FVC MEAN (SD): 2.30(0.81)L O2 group, 2.14(0.68)L compressed air group FEV1/FVC MEAN(SD): 33(5)% O2 group, 40(8)% compressed air group
Interventions	GAS TREATMENT: 4lpm O2 nasal cannula + conserving device; 4lpm compressed air TYPE OF EXERCISE TRAINING: interval ‐ upper limb at 1kg; lower limb at no resistance; fast walk over 10m at 80% VO2 max (baseline shuttle walk test); cycle ergometer unloaded until intolerance (BORG dyspnea). INTENSITY OF TRAINING: see "Type of exercise training" # SESSION PER WEEK: 3 DURATION OF EACH SESSION: 1 hour TOTAL DURATION OF TRAINING: 6 weeks LEVEL OF SUPERVISION: physical therapist (groups of 6‐8)
Outcomes	a) Shuttle walk test on room air, 4lpm O2, 4lpm compressed air b) CRQ c) HAD d) LCADL Assessed at baseline (pre) and immediately following treatment phase (post)
Notes	WITHDRAWALS: 1 admitted to hospital after initial assessment (group not reported); 2 admitted to hospital & unable to attend follow‐up (group not reported). All 3 were excluded from both the numerator and denominator, except for baseline measures where they were included. # APPROACHED TO TAKE PART AND CHOSE NOT TO PARTICIPATE: 1 COUNTRY: UK JADAD SCORE: 2/5 11 pts receiving LTOT Recruited from London Chest Hospital Author contacted: partial response
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Low risk	computer random generator
Allocation concealment?	Low risk	numbered, opaque, sealed envelopes
Blinding? All outcomes	High risk	participants

Rooyackers 1997.

Methods	STUDY DESIGN: parallel RCT INCLUSION CRITERIA: a) COPD (ATS) b) SaO2 < 90% at maximal exercise c) => 15mmHg / 2.00kPa increase in alveolar‐arterial difference in PO2 from rest to maximal exercise EXCLUSION CRITERIA: a) PaO2 < 64mmHg / 8.53kPa b) mean nocturnal SaO2 < 90% c) mean pulmonary artery pressure > 25mmHg d) neuromuscular or cardiovascular disease STATISTICAL ANALYSIS: a) within groups ‐ Wilcoxon test b) between groups ‐ Kruskal Wallis test c) correlation between various parameters ‐ Spearman correlation coefficient d) p < 0.01 significance
Participants	N: 12 O2 group, 12 room air group AGE MEAN (SD): 63(5) years O2 group, 59(13) years room air group GENDER: 2 female O2 group, 2 female room air group REST ROOM AIR PaO2 MEAN(SD): 76.5(12)mmHg / 10.20(1.60)kPa O2 group, 78.8(8.3)mmHg / 10.51(1.11)kPa room air group FEV1 MEAN(SD): 0.9(0.3)L O2 group, 1.2(0.5)L room air group FVC MEAN(SD): 3.6(1.0)L O2 group, 3.7(0.9) room air group FEV1/FVC MEAN(SD): 26(5)% O2 group, 32(9)% room air group
Interventions	GAS TREATMENT: 4lpm O2 nasal cannula; room air TYPE OF EXERCISE TRAINING: Interval ‐ 20 minutes cycling (2 minutes exercise then 2 minutes rest), 5 minutes rowing, 5 minutes pulley (arm, shoulder), 5 minutes back & abdominals, 3 minutes isometric strength training of arms & shoulders & legs, 3 minutes stair climb, 5 minutes chair sit/stand & slalom walk, 5 minutes arm weights INTENSITY OF TRAINING: start at low workload; > 1 week intensity increased as tolerated; work rate not exceeded to a level where SaO2 < 90% # SESSIONS PER WEEK: 5 DURATION OF EACH SESSION: mean of 80 minutes including rests (51 minutes exercise) TOTAL DURATION OF TRAINING: 10 weeks LEVEL OF SUPERVISION: physical therapist
Outcomes	a) Maximal cycle ergometry ‐ 4lpm O2 and room air; increase 10% of estimated max work rate b) Constant power cycle ergometry ‐ 4lpm O2 and room air; 65% actual max work rate until intolerance or 15 minutes (whichever reached first) c) 6MWT ‐ 4lpm O2 and room air d) Stair climb ‐ up 4, plateau, down 3 e) Weight lift ‐ lift between racks f) CRQ Assessed at baseline (pre) and immediately following treatment phase (post). CRQ also assessed 6 weeks after training.
Notes	WITHDRAWALS: no loss to follow‐up, all pts randomized in analysis # APPROACHED TO TAKE PART AND CHOSE NOT TO PARTICIPATE: not reported COUNTRY: Netherlands JADAD SCORE: 2/5 Patients recruited were all referred for pulmonary rehabilitation Author contacted: full response.
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Low risk	computer random‐number generator
Allocation concealment?	Low risk	central randomization
Blinding? All outcomes	High risk	data analysts

Wadell 2001.

Methods	STUDY DESIGN: parallel RCT INCLUSION CRITERIA: a) COPD (ERS) b) FEV1 < 70% predicted c) stable (no infection in last 3weeks, no change in medications in the last month) d) < 75 years old e) non‐smoker for => 6 months f) SaO2<= 92% during room air 6MWT g) PaO2 => 60mmHg / 7.80kPa room air, rest EXCLUSION CRITERIA: a) past or present major illness interfering with exercise performance STATISTICAL ANALYSIS: a) within groups ‐ Wilcoxon test b) between groups ‐ Mann Whitney U test c) p < 0.05 significance
Participants	N: 10 O2 group, 10 compressed air group AGE MEAN (SD): 64.4(6.1) year O2 group; 68.3(4.1) years compressed air group GENDER: 5 female O2 group; 5 female compressed air group REST ROOM AIR PaO2 MEAN(SD): 73.5(6.9)mmHg / 9.80(0.92)kPa O2 group; 70.2(7.6)mmHg / 9.36(1.01)kPa compressed air group FEV1 MEAN(SD): 1.05(0.36)L O2 group; 1.25(0.36)L compressed air group VC MEAN(SD): 3.26(1.05)L O2 group; 3.30(0.61)L compressed air group FEV1/FVC MEAN(SD): 33(7.8)% O2 group; 38(9.6)% compressed air group
Interventions	GAS TREATMENT: 5lpm O2 nasal cannula; 5lpm compressed air nasal cannula TYPE OF EXERCISE TRAINING: treadmill walk. INTENSITY OF TRAINING: speed adjusted until 6.0 km/hr then incline increased; stop if dyspnea 7/10 or exertion 17/20 and/or SaO2 <90%; restart when ratings lowered and SaO2 > 90%. # SESSIONS PER WEEK: 3 DURATION OF EACH SESSION: 30 minutes TOTAL DURATION OF TRAINING: 8 weeks LEVEL OF SUPERVISION: physical therapist
Outcomes	a) 6 MWT on 5lpm O2 and 5lpm compressed air (single blind and random order) Assessed at baseline (pre) and immediately following treatment phase (post)
Notes	WITHDRAWALS: 1 from O2 group, 1 from compressed air group (both due to airway infection). Exclusion from both the numerator and denominator. # APPROACHED TO TAKE PART AND CHOSE NOT TO PARTICIPATE: not reported COUNTRY: Sweden JADAD SCORE: 3/5 Recruited out‐patients from respiratory medicine & allergy department at University hospital Author contacted: partial response.
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Low risk	picking allocation from hat/box
Allocation concealment?	Unclear risk	stated random but unable to obtain further details
Blinding? All outcomes	Unclear risk	probably participants

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Abdulla 2000	Not an RCT
Ambrosino 2004	Not an RCT
Anthonisen 1983	Not an RCT
Aubier 1980	Not an RCT
Baldwin 1995	Not an RCT
Banner 1986	Not an RCT
Barach 1966	Not an RCT
Barach 1970	Not an RCT
Barach 1971	Not an RCT
Bartels 2004	Not exercise training
Bell 1988	Not an RCT
Bellone 1996	Not an RCT
Berg 1996	Not an RCT
Block 1974	Not an RCT
Bower 1988	Not exercise training
Bradley 1978	Not exercise training
Bradley 2005e	Not an RCT
Braun 1992	Not an RCT
Brezinova 1979	Not an RCT
Bye 1985	Not exercise training
Casaburi 1980	Not exercise training
Chaney 2002	Not an RCT
Chatila 2004	Not an RCT
Christensen 2004	Not an RCT
Christensen 2004b	Not an RCT
Cockcroft 1986	Not exercise training
Corriveau 1988	Not an RCT
Cotes 1956	Not an RCT
Cotes 1963	Not an RCT
Cotes 1963b	Not exercise training
Cottrell 1995	Not exercise training
Criner 1987	Not exercise training
Criner 2000	Not an RCT
Crockett 1999	Not an RCT
Crockett 2000e	Not an RCT
Cuvelier 2002	Not exercise training
Dardes 1990	Not an RCT
Davidson 1988	Not exercise training
Dean 1992	Not exercise training
Degre 1974	No treatment (oxygen) and control (air)
Dewan 1994	Not exercise training
Diaz 1992	Not an RCT
Domingo 2004	Not exercise training
Donner 1992	Not an RCT
Drummond 2001	Not an RCT
Earnest 2002	Not an RCT
Eaton 2001	Not exercise training
Eaton 2002	Not exercise training
Fletcher 1989	Not an RCT
Frey 1998	Not an RCT
Fuhrman 2004	Not exercise training
Fujimoto 2002	Not exercise training
Fussell 2003	Not an RCT
Garrod 1997	Not exercise training
Garrod 1999	Not exercise training
Gautier 1998	No control group
Gautier 2002	Not exercise training
Gibbons 2002	Not an RCT
Gimenez 1977	Not an RCT
Gimenez 1979	Not an RCT
Gislason 2000	Not an RCT
Gorbachenkov 1994	Language (Russian)
Gosselin 2004	Not exercise training
Gosselink 1997	Not an RCT
Guthrie 1970	Not an RCT
Guyatt 2000	Not an RCT
Haas 1966	Not an RCT
Hagarty 1997	Not exercise training
Haidl 2002	Not exercise training
Haidl 2004	Not exercise training
Heimlich 1982	Not an RCT
Hoffman 1988	Not an RCT
Hoffman 1992	Not an RCT
Hoo 2003	Not an RCT
Horsfield 1968	Not an RCT
Ioli 1994	Not an RCT
Ishimine 1995	Not exercise training
Janssens 1997	Not an RCT
Jolly 2001	Not exercise training
Kampelmacher 2001	Not an RCT
Kanazawa 2003	Not an RCT
Kawakami 1982	Not exercise training
Kawakami 1983	Not an RCT
Keller 1985	Not an RCT
Keller 1998	Not an RCT
Killen 2000	Not exercise training
King 1973	Not exercise training
Kitchin 1961	Not an RCT
Kozlowski 1971	Not COPD
Kurihara 1987	Not an RCT
Kurihara 1989	Not exercise training
Lacasse 2005	Not exercise training
Leach 1992	Not exercise training
Leggett 1977	Not exercise training
Lehnigk 1998	Not exercise training
Levin 1980	Not exercise training
Lewczuk 1998	Not exercise training
Lewis 2003	Not exercise training
Light 1989	Not exercise training
Liker 1975	Not exercise training
Limberg 2001	Not an RCT
Little 1999	Not an RCT
Lock 1991	Not exercise training
Longo 1971	Not exercise training
Machida 1992	Not exercise training
MacIntyre 2000	Not an RCT
MacIntyre 2000b	Not an RCT
MacNee 1985	Not exercise training
MacNee 1986	Not exercise training
Maldonado 2004	Not exercise training
Maltais 2001	Not exercise training
Mannix 1992	Not exercise training
Mannix 1992b	Not exercise training
Marchand 1998	Not exercise training
Marques 1998	Not exercise training
Matsuzawa 2000	Language (Japanese)
McDonald 1995	Not exercise training
McDonald 2005	Not exercise training
McKeon 1988	Not exercise training
McKeon 1988b	Not exercise training
Miller 1962	Not an RCT
Minh 1979	Not an RCT
Mitlehner 1992	Not an RCT
Miyagi 1989	Not an RCT
Morrison 1992e	Not an RCT
Muir 1995	Not an RCT
Nandi 2000	Not exercise training
Nandi 2003	Not exercise training
NOTT 1980e	Not exercise training
O'Donnell 1997	Not exercise training
O'Donnell 2000	Not exercise training
O'Donnell 2003	Not an RCT
O'Donohue 1997	Not an RCT
Palange 1995	Not exercise training
Patessio 1996	Abstract only
Payen 1993	Not exercise training
Peckham 1998	Not an RCT
Pepin 1996	Not an RCT
Petty 1968	Not an RCT
Petty 1974	Not an RCT
Petty 1974b	Not an RCT
Petty 1981	Not an RCT
Petty 1987	Not an RCT
Petty 1998	Not an RCT
Petty 2000	Not exercise training
Pflug 1972	Not an RCT
Pierce 1964	Not an RCT
Pierce 1965	Not an RCT
Piszko 1997	Not exercise training
Piszko 2002	Not an RCT
Piszko 2002b	Not exercise training
Puhan 2004	Not an RCT
Raimondi 1970	Not exercise training
Ram 2002e	Not an RCT
Revill 2000	Not exercise training
Roberts 1996	Not exercise training
Roberts 1998	Not an RCT
Roberts 2004	Not an RCT
Rooyackers 1996	Long term outcomes
Sant'Anna 2003	Not an RCT
Scano 1982	Not an RCT
Segard 1992	Not exercise training
Selinger 1987	Not an RCT
Senn 1989	Not exercise training
Skorodin 1989	Not exercise training
Sliwinski 1994	Not an RCT
Smith 2001	Not exercise training
Snider 2002	Not an RCT
Soffer 1985	Not exercise training
Somfay 2001	Not exercise training
Somfay 2001b	Not exercise training
Somfay 2002	Not exercise training
Stein 1982	Not exercise training
Stevenson 2002	Not exercise training
Stevenson 2004	Not exercise training
Stewart 1992	Not an RCT
Swinburn 1984	Not exercise training
Tarpy 1994	Not an RCT
Terrados 1990	Not COPD
Tiep 1987	Not exercise training
Tiep 1990	Not an RCT
Timms 1985	Not exercise training
Twohig 2000	Not exercise training
Veale 1998	Not an RCT
Vergeret 1989	Not exercise training
Vyas 1971	Not exercise training
Waldegger 2003	Not exercise training
Walsh 1990	Not an RCT
Waterhouse 1983	Not exercise training
Wedzicha 1996	Not an RCT
Weitzenblum 1991e	Not exercise training
Weitzenblum 1992	Not an RCT
Weitzenblum 2002	Not an RCT
Welch 1977	Not COPD
Welch 1987	Not an RCT
Wesmiller 1990	Not exercise training
Wilkinson 1996	Not an RCT
Wilson 1975	Not COPD
Wright 1983	Not an RCT
Wurtemberger 2001	Wrong comparison (compares different modes of exercise training)
Yaeger 1994	Not exercise training
Young 1998	Not an RCT
Young 1999	Not an RCT
Zack 1985	Not an RCT
Zhu 2005	Not an RCT
Zielinski 1998	Not an RCT
Zimmerman 1992	Not exercise training

Contributions of authors

The idea for the protocol was conceived by Y. Lacasse

M. Nonoyama, D. Brooks, G.H. Guyatt and R.S. Goldstein designed and assisted in writing the protocol and the review.

M. Nonoyama acts as guarantor of the review.

Declarations of interest

None known.

New search for studies and content updated (no change to conclusions)