Short‐term ambulatory oxygen for chronic obstructive pulmonary disease

Abstract

Background

Ambulatory oxygen is defined as the use of supplemental oxygen during exercise and activities of daily living. Ambulatory oxygen therapy is often used for patients on long term oxygen therapy during exercise, or for non long term oxygen therapy users who achieve some subjective and/or objective benefit from oxygen during exercise. The evidence for the use of ambulatory oxygen therapy is extrapolated from two sources: longer term studies and single assessment studies. Longer term studies assess the impact of ambulatory oxygen therapy used at home during activities of daily living. Single assessment studies compare performance during an exercise test using oxygen with performance during an exercise test using placebo air.

Objectives

To determine the efficacy of ambulatory oxygen in patients with COPD using single assessment studies.

Search methods

The Cochrane Airways Group COPD register was searched with predefined search terms. Searches were current as of March 2005.

Selection criteria

Only randomised controlled trials were included. Studies did not have to be blinded. Studies had to compare oxygen and placebo when administered to people with COPD who were undergoing an exercise test.

Data collection and analysis

Two reviewers (JB, B'ON) extracted and entered data in to RevMan 4.2.

Main results

Thirty one studies (contributing 33 data sets), randomising 534 participants met the inclusion criteria of the review. Oxygen improved all pooled outcomes relating to endurance exercise capacity (distance, time, number of steps) and maximal exercise capacity (exercise time and work rate). Data relating to VO₂ max could not be pooled and results from the original studies were not consistent. For the secondary outcomes of breathlessness, SaO₂ and V_E, comparisons were made at isotime. In all studies except two the isotime is defined as the time at which the placebo test ended. Oxygen improved breathlessness, SaO₂/PaO₂ and V_E at isotime with endurance exercise testing. There was no data on breathlessness at isotime with maximal exercise testing. Oxygen improved SaO₂/PaO₂ and reduced V_E at Isotime.

Authors' conclusions

This review provides some evidence from small, single assessment studies that ambulatory oxygen improves exercise performance in people with moderate to severe COPD. The results of the review may be affected by publication bias, and the small sample sizes in the studies. Although positive, the findings of the review require replication in larger trials with more distinct subgroups of participants. Maximal or endurance tests can be used in ambulatory oxygen assessment. Consideration should be given to the measurement of SaO₂ and breathlessness at isotime as these provide important additional information. We recommend that these outcomes are included in the assessment for ambulatory oxygen. Future research needs to establish the level of benefit of ambulatory oxygen in specific subgroups of people with COPD.

Plain language summary

Short‐term ambulatory oxygen for chronic obstructive pulmonary disease

Short‐term studies indicate that people with chronic obstructive pulmonary disease respond to the administration of oxygen when they do exercise tests. Ambulatory oxygen is the use of supplemental oxygen during exercise and activities of daily living. One way to assess if ambulatory oxygen is beneficial for a patient with COPD is to compare the effects of breathing oxygen and breathing air on exercise capacity. Some people with COPD may benefit more than others, and trials should take account of whether people who do not already meet criteria for domiciliary oxygen also respond. This review shows that there is strong evidence that ambulatory oxygen (short‐term) improves exercise capacity. Further research needs to focus on which COPD patients benefit from ambulatory oxygen, how much oxygen should be provided and the long‐term effect of ambulatory oxygen.

Background

Chronic obstructive pulmonary disease (COPD) is a slowly progressive disorder characterised by airflow obstruction. The degree of airway narrowing is largely fixed but may be partially reversed with bronchodilator therapy (NICE 2004, GOLD 2001; Romain 2001; BTS 1997; ATS 1995). Patients often present with breathlessness, a chronic cough and sputum production. As the disease progresses gas exchange becomes more abnormal and adequate oxygenation may no longer be maintained. Home oxygen may be prescribed as long term oxygen therapy (LTOT), short burst or as ambulatory oxygen therapy. Long term oxygen therapy has been used in hypoxaemic patients with COPD to increase life expectancy, quality of life, exercise tolerance and decrease shortness of breath. The efficacy of LTOT in patients with chronic hypoxaemia has been shown in a Cochrane review and its use and prescription are well established (Crockett 2000).

Ambulatory oxygen is defined as the use of supplemental oxygen during exercise and activities of daily living (RCP 1999). Ambulatory oxygen therapy is often used by patients on LTOT during exercise, or for non LTOT users with or without resting hypoxaemia if they show evidence of exercise desaturation and demonstrate improvement in exercise capacity with supplemental oxygen. There are criteria for assessment and use of ambulatory oxygen therapy (ATS 1995; RCP 1999; Young 1998). The RCP guidelines provide the most objective criteria which can be used to ascertain if ambulatory oxygen is required (RCP 1999). These criteria include (i) a fall in SaO₂ of at least 4% to reach a reading below 90% during a baseline walking test whilst breathing air; (ii) an improvement of at least 10% in walking distance and/or breathlessness score when walking with supplemental oxygen compared with an air cylinder; and (iii) the level of oxygen prescribed should be adequate to maintain SaO₂ above 90%.

Ambulatory oxygen therapy is usually provided in small oxygen cylinders lasting up to 4 hours at 2 L/min or liquid oxygen systems which have a higher oxygen carrying capacity lasting up to 6‐10 hours at 2 L/min. The evidence for the use of ambulatory oxygen therapy is extrapolated from two sources: longer term studies and single assessment studies. Longer term studies assess the impact of ambulatory therapy used at home during activities of daily living. Single assessment studies compare performance during an exercise test using oxygen with performance during an exercise test using placebo air.

A Cochrane review investigating the long term efficacy of ambulatory therapy showed that there was little evidence of its effectiveness (Ram 2003) however this review did not include the single assessment studies which are recommended for assessment of ambulatory oxygen in clinical practice.

Objectives

To determine the efficacy of ambulatory oxygen in patients with COPD using single assessment studies.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials comparing ambulatory oxygen therapy versus placebo air. Only assessment studies comparing performance during a single exercise test using ambulatory oxygen compared to performance during a single exercise test with placebo air were considered. Longer term studies assessing the efficacy of ambulatory oxygen therapy or studies assessing the efficacy of LTOT were not included.

Types of participants

Studies were included only when adult patients with stable COPD were randomised to either ambulatory oxygen or placebo. COPD should have been diagnosed according to criteria included in internationally accepted guidelines e.g. GOLD 2001; Romain 2001; BTS 1997; ATS 1995. Studies published prior to 1995 were included if patients had a diagnosis equivalent to COPD. Patients who did not have a diagnosis of COPD were excluded. Studies with mixed populations where excluded if data from patients with COPD were not analysed separately.

Types of interventions

The intervention in the actively treated group was ambulatory oxygen therapy, provided either via oxygen cylinders or a reservoir system. In the control group, the intervention should have been delivered via air cylinders or a reservoir system.

Types of outcome measures

Primary outcomes

1. Exercise capacity (e.g. distance, time or steps during maximal tests or endurance tests)

Secondary outcomes

Secondary outcomes measured at isotime:

1. Dyspnoea scores (e.g. Borg scores or visual analogue scale)

2. Arterial oxygen saturation (pulse oximetry or arterial blood gases) during or post exercise

3. Physiological measurements (e.g. V_E during exercise)

4. Patient preference

Search methods for identification of studies

Electronic searches

The Cochrane Airways Group Specialised Register of RCTs was searched using the search terms:

(portable* or ambulat* or oxygen* or O2 or hypoxaemia* or hypoxemia*) and (therap*)

The Register contains records downloaded from CENTRAL, MEDLINE, EMBASE and CINAHL, as well as records identified through hand‐searching journals and meeting abstracts, including the American Thoracic Society, British Thoracic Society and European Respiratory Society meetings.

In order to minimise the chance of missing potential studies separate searches were also completed on the Cochrane Central Register of Controlled Trials (CENTRAL). In addition, other electronically available databases and search engines were searched for trials (CINAHL, Science Citation Index, EMBASE, MEDLINE, Scirus, UK National Research Register, PEDro, ClinicalTrials.gov, Google.com). Electronic web sites of the following journals were 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). All databases were searched from their inception up until February 2004.

Searching other resources

Following this the bibliographies of each included RCT as well as any review articles found were searched for additional papers that may contain further RCTs. All authors of identified RCTs were contacted and asked to confirm that the data extracted and the assessment of quality was correct. Where relevant they were also requested to provide further information. Eight authors responded (Bye 1985; Garrod 1999; Dean 1992; O'Donnell 1997; O'Donnell 2001; Light 1989; Swinburn 1984; Fujimoto 2002a) and confirmed that the data extracted and assessment of quality was correct. Four studies (Garrod 1999; Dean 1992; O'Donnell 1997; O'Donnell 2001) provided further information (Table 1 Included Studies).

1. Details of Exercise Tests Used.

Study ID	Test	Practice Tests	Time Between Tests	Predosing	Pretest checks
Leggett (1977)	Distance walked in 12 mins on the level and at own pace (including stops if desired): not carrying cylinder	Not documented	30mins (up to 4 tests per day)	Not documented	Not documented
McKeon (1988)	Treadmill test level grade at 1.5km/h and increasing 0.5km/h each min; ?not carrying cylinder	Initial practice walk	30 mins (4 tests on the same day)	Not documented	Bronchodilators withheld during study
Dean (1992)	Endurance cycle ergometer test based on 80% of incremental cycle test; Incremental test 10W incraesing 10‐20 W every 2 mins to symptom limited max; not carrying oxygen	Not documented	At least 30mins and up to 60 mins	No	Light breakfast; usual medications including bronchodilators
Garrod (2000)	Shuttle walk test carrying small portable tank	Practice walk	20 mins	Not documented	Not documented
ODonnell (1997)	Endurance cycle ergometer at 50% work of maximum work rate. Maximum work rate established on cycle test at 50‐70rpm. (Not carrying oxygen)	Thorough familiarisation	60‐90mins	At least 10mins resting breathing air/oxygen	Avoid caffeine alcohol heavy exertion and heavy meal 4 hours pre testing
McDonald (1995)	6 Minute Walk Step Test Test (Not carrying cylinder)	2 practice tests 6 minute walk test and step test	20 mins (for 6 minute walk test ‐repeated at least three times on each gas until within 5 % agreement between 2 tests)	Not documented	Not documented
Wadell (2001)	Treadmill 6 minute walk using non motorised treadmill and speed driven by patients own walking speed (Not carrying cylinder)	Demonstration given and 1‐2 minute practice before test	1 hour	Not documented	Not documented
Raimondi (1970)	Cycle ergometer test start at 100kpm/min and increased by 100kpm/min each minute Single load cycle test at 70% of maximum achieved on air test	Not documented	30 mins (3 tests on two day)	Not documented	Isoprenaline 10 mins pre exercise
Somfay (2001)	Cycle ergometer at a constant work rate equal to 75% of highest work rate achieved during a symptom limited test breathing room air	Not documented but were familiaris with exercise tests	1 hour (2 tests per visit with 3 visits 3‐5 days apart)	Not documented	No caffeine alcohol meals pre testing
ODonnell (2001)	Endurance cycle ergometer at 50% work of maximum work rate. Maximum work rate established on cycle test at 50‐70rpm. (Not carrying oxygen)	Thorough familiarisation	60‐90mins	At least 10mins resting breathing air/oxygen	Avoid caffeine alcohol heavy exertion and heavy meal 4 hours pre testing
Davidson (1988)	Endurance cycle test at 50‐70% max work rate. Maximum work rtae assessed during a proressive cyscle test at 25W increments at 1 minut intervals 6 minute walk test Endurance walk‐ walk as far as possible at a pace as though late for appointment" and stop when unable to go further (Cyclinder carried by patients during 4Lmin 6 minute walk and endurance tests	3 practice walks for 6 minute walk test. No practice for endurance or cycle test	3 hours (no more than 2 cycling and 2 walk tests on each day with testing on 3 consecutive days)	Not documented	Fasted for 2 hours. no bronchodilators pre testing
King (1973)	Treadmill walking at 0.84 (0.18) mph	1‐2 minute practice followed by 5 minute rest	10minutes between tests	Not documented	2 hours after light lunch
Eaton (2002)	6 Minute Walk Test oxygen carried in back pack or shoulder bag	3 practice tests	20‐30 minutes between tests	Not documented	Not documented
Bradley (1978)	Treadmill test‐ initial speed chosen by patient as one comforatable and unlikely to lead to intolerable dysponea after several mins. Treadmill grade increased by 2% per min starting at zero.	Demonstration and practice walk on 3 consecutive days prior to test	10 mins but some needed 1‐2 hours (3 tests per day)	Not documented	Not documented
Stein (1982)	Incremental treadmill test	Demonstration and oportunity to walk on treadmill before test	15 mins	7 mins prior to exercise	Breakfast before tests‐ all tests carried out in the morning
Mannix (1992)	Incremental cycle ergometer 4 min stages (0W and progressed by 12.5W)	Not documented	1 day	Not documented	same time of day
Bye (1985)	Cycle ergometer test at 80% of previously determined max	Not documented	2 days	10mins pre dosing	same time of day
Light (1989)	cycle ergometer with 10 to 20 W increases each min	At least 1 practice test	45 mins	10mins pre dosing	Two inhalations of metaproterenol on arrival unless he had received this within 90 last mins
Leach (1992)	6 Minute Walk Test Endurance test walking as far as possible at a pace as though late for an appointment	3 practices of each test	45min	Not documented	Fasted for at least 90 ins standardised for time of testing bronchodilator treatment
Kurihara (1989)	6 Minute Walk Test Endurance test	No documented	30minutes	Not documented	Not documented
Ishimine (1995)	6 Minute Walk Test Endurance test	1 practice test	30minutes	Not documented	Oral bronchodilator stopped on evening before trial and inhaled bronchodilators stopped on morning of test
Fujimoto (2002)	6 Minute Walk Test Endurance test	1 practice walk	20 minutes	Not ducumented	Not documented
Garrod (1999)	Shuttle walk test carrying small portable tank	1 practice walk	Not documented	Not documented	Not documented
Swinburn (1984)	Maximal cycle ergometer test	Familiar with test	3 hours	4 minutes	salbutamol 1 hour pre test
Vyas (1971)	Maximal cycle ergometry at 60kgm per min work load increased each minute	Familiar with test	1 hour	20 minutes	Not documented
Woodcock 1981	Maximal Treadmill Test 6 Minute Walk Test Endurance test	Practice sessions	30 minutes	Not documented	Not documented
Palange 1995	Submaximal Cycle ergometer	Not documented	1 hour	Not documented	Not documented
Gosselin 2004	Maximal Cycle ergometer	Not documented	1 day	15 mins	Light breakfast 2 hours before
Knebel 2000	Submaximal 6 MWT	3 practice walks	45 minutes	Not documented	2 hours after meals or after waking
Maltais 2001	Maximal cycle ergometer	Not documented	2 hours	Not documented	Not documented
Criner (1987)	Maximal Symptom limited unloaded leg ergometry	Not documented	Not documented	Not documented	Not documented

Data collection and analysis

Selection of studies

Two reviewers independently selected trials for inclusion in the review. Disagreement did not arise on the suitability of a trial for inclusion in the review or in its quality however if this occurs for future updates of this review a consensus will be reached by the two reviewers.

Data extraction and management

Data from the trials was independently extracted by two reviewers (JB, BON) using standard data extraction forms.

Assessment of risk of bias in included studies

The methodological quality of each trial was assessed by each reviewer using two scales: the Jadad scale (Jadad 1996) and the PEDro scale. The PEDro is an 11 item scale based on the previously validated Delphi list (Verhagen 1998). The Pedro scale examines issues of randomisation, allocation, whether patients were similar at baseline, extent of blinding in the study, proportion of patients for which data was available and the method in which this data was reported.

The PEDro scale has been shown to be reliable measure of study quality (Moseley 1999). The PEDro scale is available at http://ptwww.cchs.usyd.edu.au/pedro. The PEDro scale scores studies out of 10.

The quality of included studies was also assessed using the Cochrane allocation concealment scale.

Unit of analysis issues

Due to the crossover design of the studies, we opted to enter data based as generic inverse variance (GIV) data. This takes the mean difference between treatment and control, with a standard error (SEM) for the difference. Where possible, we have taken the published SEM, but where this was not available, we have used the published P value to estimate a standard error. Where data were reported as non‐significant, and no other means of obtaining the variance was possible, we have entered the published means and SDs as a weighted mean difference (WMD ‐ see below) as reported in the published paper and used that as a basis of calculating the SEM for between treatment group differences. This would tend to underestimate the treatment effect as it would overlook the sensitivity of paired data for within patient differences.

Data synthesis

For continuous outcomes (exercise capacity, breathlessness scores), a WMD was used when combining data. Data from the same variable (for example the FEV₁), but expressed differently in different trials (for example as change in litres and change expressed as percentage of baseline) were combined using a standardised mean difference (SMD). If dichotomous outcomes had been identified they would have been analysed using the relative risk or odds ratio. Random and/or fixed effects model were used depending upon the level of statistical heterogeneity observed. Where I² exceeded 0% we applied Random Effects modelling, and compared this with a Fixed Effects model in order to determine whether taking account of within and between study variation impacts upon the overall pooled effect estimate.

Note: For many of the outcomes, the focus is on a favourable outcome. Here the aim of the treatment is to increase the outcome, rather than decrease it. This requires the graph labels to be reversed from the standard format of "favours treatment" on the left of the graph and "favours control" on the right.

All trial data was combined using Review Manager. Funnel plots were carried out to test for the presence of publication bias where possible.

Subgroup analysis and investigation of heterogeneity

If enough studies had been available we would have carried out the following sub group analyses: disease severity; and method of oxygen delivery.

Due to the way in which studies were conducted and reported, we opted to perform a post‐hoc subgroup analysis based upon the level of hypoxaemia and the dose of oxygen delivered. A sensitivity analysis was performed to examine the impact of blinding on the study results (single versus double blind studies).

Sensitivity analysis

If significant heterogeneity was found a sensitivity analyses would have been conducted on study quality (Cochrane scores and or Pedro scale).

Results

Description of studies

A total of 60 studies were retrieved literature search results. Of these, 29 studies failed to meet the inclusion criteria (see Table 'Characteristics of Excluded Studies'). Four new studies have been included in this update of the review (see 'What's New'). Thirty‐one studies (contributing 33 randomised data sets) met the inclusion criteria.

Study design

Thirty one studies (contributing 33 data sets) with a total of 534 participants met the inclusion criteria for this review. All the trials included in this review were cross‐over in design.

Freeman et al (1989) suggest that only on the grounds of biological reasoning can it can be assumed that there is no carry‐over in cross over studies. They argue against using statistical tests to assess if there is a crossover effect. No studies in this review tested for a carry‐over effect although the duration between successive tests is variable. There is no consensus in the literature regarding the minimum time required to eliminate the carry‐over effect of exercise testing.

Participants

Sample size varied from 5 to 41. The mean age reported in the included studies ranged from 47 to 73 years. With the exception of one study which included a group of mild patients (Fujimoto 2002a) all studies included patients with moderate to severe airflow obstruction.

The mean PaO₂ ranged from 6.9 KPa to 11.3 KPa (52 mmHg to 85 mmHg). Judging by the baseline mean PaO₂ , the following loose categorisations can be made: seven studies included patients who likely met the criteria for LTOT on the grounds that baseline mean kPa was lower than 7.3 (see Table characteristics of included studies): (O'Donnell 2001; Garrod 2000; Bye 1985; King 1973; Leach 1992; Leggett 1977; Mannix 1992). Twenty three studies included patients who did not or likely did not meet the criteria for LTOT on the grounds that mean baseline kPa exceeded 7.3 (see Table characteristics of included studies):Eaton 2002; O'Donnell 1997, Dean 1992; Garrod 1999; Wadell 2001; Bradley 1978; Criner 1987; Davidson 1988; Fujimoto 2002a; Fujimoto 2002b; Fujimoto 2002c; Light 1989; McKeon 1988; McDonald 1995; Somfay 2001; Stein 1982; Ishimine 1995; Kurihara 1989; Swinburn 1984; Vyas 1971; Woodcock 1981; Knebel 2000; Maltais 2001; Palange 1995; Gosselin 2004). In one study there was no information relating to baseline oxygen status (Raimondi 1970). Knebel 2000 cited current oxygen usage as an exclusion criterion.

These categorisations form the basis of the subgroup analyses which are based upon baseline oxygen status (mean baseline PaO2 <7.3 kPa/55 mmHg; mean baseline PaO2 >=7.3 kPa/55 mmHg or status unclear). This value (PaO2 <7.3 kPa) is based on the recommendation from NICE for LTOT (NICE 2004).

Intervention and setting

In thirteen studies the oxygen was delivered via nasal specs, in one study the oxygen was delivered via a face mask, in one study the oxygen was delivered via a face mask or nasal specs and in sixteen studies oxygen was delivered via a mouthpiece with a reservoir system (*see Table characteristics of included studies). It was decided that oxygen delivery of less than or equal to 4 L/min or 35% would be termed "low dose oxygen" and oxygen delivery of greater than this would be termed "high dose oxygen". Twenty studies used low dose oxygen (Criner 1987; Eaton 2002; Fujimoto 2002a; Fujimoto 2002b; Fujimoto 2002c; Garrod 1999; Garrod 2000; Ishimine 1995; King 1973; Kurihara 1989; Leach 1992; Leggett 1977; Light 1989; Mannix 1992; McDonald 1995; McKeon 1988; Raimondi 1970; Stein 1982; Woodcock 1981; Gosselin 2004; Palange 1995; Knebel 2000), nine studies used high dose oxygen (Bradley 1978; Bye 1985; Dean 1992; O'Donnell 1997; O'Donnell 2001; Swinburn 1984; Vyas 1971; Wadell 2001; Maltais 2001) and 2 studies used both (Davidson 1988; Somfay 2001).

Twenty‐nine studies compared one oxygen intervention with a placebo intervention. Two studies compared more than one oxygen intervention with a placebo intervention (Davidson 1988; Somfay 2001); in these two studies only one low dose and only one high dose oxygen intervention was selected. The Davidson 1988 study had more than one low dose intervention and the Somfay 2001 study had more than one high dose intervention. The oxygen intervention closest to the critical level (4 L/min or 35%) was chosen.

Exercise Tests

Twelve studies used maximal exercise tests only and 17 used endurance exercise tests and two studies used two tests. For details of each study please see Table characteristics of included studies. Various types of exercise tests were used: treadmill (6 studies); cycle ergometry (12 studies); incremental shuttle walk test (2 studies); 6MW test (10 studies); step test (one study); other walk tests (three studies). A number of studies assessed response to more than one type of exercise test.

A variety of exercise test protocols were used and details of these are provided in Table 1. In 10 studies there was no familiarization or practice test; six studies used familiarization only and 14 studies used practice tests +/‐ familiarisation (Table 1) and one study used practice tests +/‐ familiarisation in two out of three exercise tests (Davidson 1988).

Up to four submaximal tests were performed on the same day and the rest period between tests ranged from 10 mins to one day. In one study 16 tests where performed over three days (Woodcock 1981). 
 
 17/31 of the studies controlled for factors which can affect performance on exercise test such as time of day, food intake and medication such as bronchodilators.

Outcome

There was a wide range of outcome measures reported in the studies. Exercise capacity was measured using five methods: distance (14 studies); time (12 studies); VO₂ (7 studies); number steps (1 study); watts (3 studies). The following measurements were made at isotime: breathlessness, 4 studies;. SaO₂/PaO₂, 6 studies (1 mmHg = 0.133 kPa); V_E, 9 studies: (Table 2 Outcome measures used).

2. Outcome Measures Used (E=Endurance test; M=maximal test; *=Isotime).

Study	Exercise Distance	Exercise Time	VO2	Number Steps	Isotime breathless		Isotime SaO2/PaO2	Isotime VE	Watts
Bradley 1978	‐	‐	M	‐	‐	‐	‐	‐	‐
Bye 1985	‐	E	‐	‐	‐	‐	E*	E*	‐
Davidson 1988	E6MW , Ewalk	Ecycle, Ewalk	‐	‐	‐	‐	‐	‐	‐
Dean 1992	‐	E	‐	‐	E*	‐	E*	E*	‐
Eaton 2002	E	‐	‐	‐	‐	‐	‐	‐	‐
Garrod 2000	M	‐	‐	‐	‐‐	‐	‐	‐	‐
Ishimine 1995	E	‐	‐	‐	‐	‐	‐	‐	‐
King 1973	‐	‐	‐	‐	‐	‐	‐	‐	‐
Kurihara 1989	E	‐	‐	‐	‐	‐	‐	‐	‐
Leach 1992	E	‐	‐	‐	‐	‐	‐	‐	‐
Legget 1977	E	‐	‐	‐	‐	‐	‐	‐	‐
Light 1989	‐	‐	‐	‐	‐	‐	M*	M*	M
Mannix 1992	‐	M	M	‐	‐	‐	‐	M*	‐
McDonald 1995	E6MW	‐	‐	EStep	‐	‐	‐	‐	‐
McKeon 1988	M	‐	‐	‐	‐	‐	‐	‐	‐
ODonnell 1997	‐	E	[E*]	‐	E*	‐	E*	E*	‐
ODonell 2001	‐	E	‐	‐	E*	‐	‐	E*	‐
Raimondi 1970	‐	M E	‐	‐	‐	‐	‐	‐	‐
Somfay 2001	‐	E	‐	‐	E*	‐	E*	E*	‐
Stein 1982	‐	M	‐	‐	‐	‐	‐	‐	‐
Wadell	E	‐	‐	‐	‐	‐	‐	‐	‐
Fujim oto 2002	E	‐	‐	‐	‐	‐	‐	‐	‐
Garrod 1999	M	‐	‐	‐	‐	‐	‐	‐	‐
Swinburn 1984	‐	M	‐	‐	‐	‐	‐	‐	‐
Vyas 1971	‐	M	M [M*]	‐	‐	‐	‐	M*	‐
Woodcock 1981	M E	‐	‐	‐	‐	‐	‐	‐	‐
Palange 1995	‐	‐	E	‐	‐	‐	‐	‐	‐
Knebel 2000	E	‐	‐	‐	‐		‐	‐	‐
Gosselin 2004	‐	‐	M	‐	‐		‐	M	M
Maltais 2001	‐	‐	M	‐	‐		‐	‐	‐
Criner 1987	‐	M	M	‐	‐	‐	‐	‐	‐

Risk of bias in included studies

Overall, the methodological quality of the included studies as rated by the Jadad score was low (16 studies had score of 1; nine studies had a score of 2; 4 studies had a score of 3; and two studies had a score of 5). These low scores can be attributed to lack of double blinding (14 studies single blind; 15 double blind; two studies no blinding) or lack of detail in the reporting of the blinding or randomisation (Table characteristics of included studies).

The Pedro scale includes a wider variety of quality criteria relating to both internal and external validity. Overall, the methodological quality of the included studies as rated by the Pedro score was good (2 studies has score of 6; 14 studies had a score of 7; 2 studies had a score of 8; 12 studies had a score of 9; 1 study had a score of 10) (Table characteristics of included studies).

One study reported dropouts (Knebel 2000).

One study reported the use of a concealed randomisation procedure (Garrod 1999).

In general the sample size of many individual studies was small (range 5 to 41 participants) and no studies reported that the sample size was based on a power calculation.

29/31 studies included in the review reported on at least 1 of the primary outcome measures. King 1973 and Light 1989 did not report on a primary outcome measure.

Effects of interventions

ENDURANCE TEST STUDIES

Primary outcome measure: Exercise capacity

Distance

(Eaton 2002; Fujimoto 2002a; Fujimoto 2002b; Fujimoto 2002c; Ishimine 1995; Kurihara 1989; McDonald 1995; Davidson 1988;Woodcock 1981; Knebel 2000)

All studies assessed the effects of low dose oxygen. Davidson 1988 reported results from two exercise tests (6MWT and endurance walk test). In order to avoid double‐counting participants, we calculated two pooled estimates. There was significant heterogeneity between the studies with either data‐set used. With fixed effects modelling oxygen significantly improved exercise distance by 18.86 metres (95% CI 13.11 to 24.61, N=238 Davidson 1988 6 MWT; Figure 1), and 18.61 metres [95% CI 12.83 to 24.39], N = 238 Davidson 1988 endurance walk test; Figure 2). Random Effects modelling did not alter the significance of these effects.

1.

Forest plot of comparison: 1 Oxygen versus placebo (crossover studies), outcome: 1.1 Endurance test ‐ exercise distance (Davidson 1988 6MWT).

2.

Forest plot of comparison: 1 Oxygen versus placebo (crossover studies), outcome: 1.2 Endurance test ‐ exercise distance (Davidson 1988 endurance walk).

Three studies did not report data in a usable format for meta‐analysis. All assessed the effects of low dose oxygen. Leach 1992 reported no significant increase in distance walked (mean % increase 28.6 [95% CI ‐ 21.5 to 111, N = 20]. Leggett 1977 reported censored data (N=8) from a study of 26 participants, which showed a significant increase in distance walked 12MWT (mean SEM increase 53 m +/‐ 12.7. Wadell 2001reported that oxygen significantly improved the median distance walked on a 6MWT [median (min‐max) improvement 30 (‐30 to 60) metres, N = 20].

Exercise time

(Davidson 1988; Raimondi 1970; Somfay 2001; Bye 1985; Dean 1992; O'Donnell 1997; O'Donnell 2001).

One study assessed the effects of treatment with low dose oxygen (Raimondi 1970). Two studies reported data on the effects of treatment with both low dose and high dose oxygen (Somfay 2001; Davidson 1988) and four studies assessed the effects of treatment with high dose oxygen (Bye 1985; Dean 1992; O'Donnell 1997; O'Donnell 2001). In one study the effect of low dose oxygen was assessed using two different endurance tests (cycle test and an endurance walk test, Davidson 1988). To avoid over‐estimating the effects of treatment in the placebo arms of the studies, three pooled estimates have been calculated. There was no significant heterogeneity between the subgroups in any instance. With low dose data from Somfay 2001 and Davidson 1988, oxygen significantly improved exercise time irrespective of whether data related to the cycle test (Low dose: WMD 2.70 minutes [95% CI 1.95 to 3.44], N = 77) or endurance walk test (Low dose: WMD 2.63 minutes [95% CI 1.91 to 3.44], N = 77). With high dose data from both studies entered oxygen significantly increased exercise time (High dose: WMD 2.71 minutes [95% CI 1.96 to 3.46], N = 77).

Step

(McDonald 1995)

Data from the original study showed that low dose oxygen significantly improved the mean (SD) number of steps climbed oxygen 35 (21) versus placebo 30 (18), N = 26.

VO₂

(Palange 1995)

Data from the original study showed that low dose oxygen significantly improved VO2 oxygen 2280 (426) versus placebo 1120 (204), N = 9.

Patient Preference

No studies reported on patient preference.

MAXIMAL TEST STUDIES

Primary outcome: Exercise capacity

Distance

(Garrod 1999; Garrod 2000; McKeon 1988; Woodcock 1981).

All studies assessed the effects of low dose oxygen. There was a significant improvement in distance walked during oxygen versus placebo of 32 metres [95% 20.61 to 43.38], N = 70. There was a moderate level of heterogeneity between the studies. Random Effects modelling widened the confidence interval but the result remained significant (39.57 metres [95% CI 17.03 to 62.11]).

Time

(Criner 1987; Mannix 1992; Raimondi 1970; Stein 1982; Swinburn 1984; Vyas 1971).

Fixed effect modelling gave a significant difference of 1.06 minutes in favour of oxygen [95% CI: 0.67, 1.46], N = 50. Although there was a moderate level of heterogeneity (I² 51.7%), Random Effects modelling did not alter the significance of this finding (1.19 minutes [0.53, 1.86]).

VO₂max

(Bradley 1978; Criner 1987; Mannix 1992; Vyas 1971; Gosselin 2004).

Three studies assessed the effects of low dose oxygen (Criner 1987; Gosselin 2004; Mannix 1992) and two studies assessed the effects of high dose oxygen (Bradley 1978; Vyas 1971). As each study used different units of measurement these studies could not be pooled. Data from the original studies showed that oxygen significantly (P<0.05) improved VO₂ max in three studies [Criner 1987; N=6, oxygen 13.4 (0.9) versus placebo 9.6 (1.4)ml kg^{‐1 min‐1}; Gosselin 2004; N=9, oxygen 22.03 (4.2) versus placebo 20.2 (5.7) ml kg‐1 min‐1; Vyas 1971; N=12, oxygen 4.14 (1.21) versus placebo 2.74 (0.86) L 0.5 min^‐1]. Oxygen did not improve VO₂ max in the other two studies [Mannix 1992; N=10, oxygen 740 (335.20) versus placebo 609 (220.39)ml min^‐1; Bradley 1978; N=26, oxygen 8.7 (3.0) versus placebo 8.20 (2.60)ml min^‐1.

Work rate

(Light 1989; Gosselin 2004; Maltais 2001)

Two studies assessed the effects of low dose oxygen (Light 1989; Gosselin 2004) and one study assessed the effects of high dose oxygen (Maltais 2001). There was no significant heterogeneity between the studies and oxygen significantly increased work rate compared to placebo by 8.88 watts [95% CI 5.71 to 12.06] N=40.

Patient Preference

No studies reported on patient preference.

Secondary outcome measure at Isotime

Secondary outcomes e.g. breathlessness, SaO₂, V_E at end exercise are not directly comparable as they are dependant on exercise performance (time/distance/VO2/work rate). Therefore it was decided to compare the effect of oxygen treatment on these secondary outcomes at isotime. In all studies except two the isotime is defined as the time at which the placebo test ended. One study used the point of comparison as the time at which the lesser test ended (Vyas 1971). One study used the point of comparison as the time at which stage 1 (no resistance or zero watts) ended (Mannix 1992). One study reported data on isotimes with both high and low dose oxygen treatment so to avoid double‐counting the effects of treatment two pooled estimates have been calculated (Somfay 2001).

ENDURANCE TEST ISOTIMES STUDIES

Breathlessness

(Dean 1992; O'Donnell 1997; O'Donnell 2001; Somfay 2001).

Three studies assessed the effects of treatment with high dose oxygen (Dean 1992; O'Donnell 1997; O'Donnell 2001) and one study assessed the effects of both high and low dose oxygen (Somfay 2001). There was a moderate level of heterogeneity between the studies (I² 40‐44%). With low and high dose oxygen data from Somfay 2001 entered separately oxygen significantly decreased breathlessness (Low dose data from Somfay 2001: WMD ‐1.15 [95% CI ‐1.65 to ‐0.66] and high dose data from Somfay 2001: WMD ‐1.15 [95% CI ‐1.66 to ‐0.65]). Random Effects modelling gave a significant result in favour of oxygen (Somfay 2001 low dose: ‐1.46 [95% CI ‐2.30 to ‐0.62]; Somfay 2001 high dose: ‐1.51 [95% CI ‐2.41 to ‐0.61] N=44).

SaO₂/PaO₂

(Bye 1985; O'Donnell 1997; Somfay 2001)

Two studies assessed the effects of treatment with high dose oxygen (Bye 1985; O'Donnell 1997) and one study assessed the effects of both high and low dose oxygen (Somfay 2001). There was no significant heterogeneity between the studies. Oxygen significantly improved SaO₂ with low and high dose data from Somfay 2001 (Somfay 2001 low dose: 8.36% [95% CI 5.08 to 11.64]; Somfay 2001 high dose: 8.80% [95% CI 5.36 to 12.24] N=29).

Data for PaO₂ was reported in two high dose studies (Dean 1992; O'Donnell 2001). PaO₂ was significantly greater with oxygen versus placebo by 15.15 kPa [95% CI 6.42 to 23.89] N=23.

V_E (Dean 1992; Bye 1985; O'Donnell 1997; O'Donnell 2001; Somfay 2001) 
 V_E is the volume of air inhaled per minute. Four studies assessed the effects of treatment with high dose oxygen (Dean 1992; Bye 1985; O'Donnell 1997; O'Donnell 2001) and one study assessed the effects of both high and low dose oxygen (Somfay 2001). Pooled estimates with both high and low dose data from Somfay 2001 indicated that oxygen significantly decreased V_E (Somfay 2001 high dose: ‐3.58 [95% CI ‐4.85 to ‐2.31]; Somfay 2001 low dose: ‐3.60 L/min [95% CI ‐4.88 to ‐2.33] N=52).

MAXIMAL TEST ISOTIMES

Breathlessness

Data from 14 patients showed high dose oxygen significantly reduced breathlessness compared to placebo but no data was provided (Maltais 2001).

SaO₂ and PaO₂

(Light 1989; Maltais 2001)

One study assessed the effects of low dose oxygen (Light 1989) and one study assessed effects of high dose oxygen (Maltais 2001). There was no significant heterogeneity between the studies and oxygen significantly increased SaO2 versus placebo by 7.82% [95% CI 4.89 to 10.74] N=31.

V_E (Light 1989; Mannix 1992; Vyas 1971; Gosselin 2004)

Three studies assessed the effects of treatment with low dose oxygen (Light 1989; Mannix 1992; Gosselin 2004) and one study assessed the effects of treatment high oxygen (Vyas 1971). There was a significant difference in V_E in favour of oxygen, of ‐3.26 L/min [95% CI ‐4.33 to ‐2.19] N=48.

Sensitivity & Subgroup Analyses

It was not possible to conduct a sub group analysis according to disease severity or method of oxygen delivery.

Subgroup analyses according to baseline oxygen status did not provide evidence of different response to treatment between those studies which included patients that did/likely did meet the criteria for LTOT and those studies that did not/ likely did not. This may be explained in part by the nature of the distinctions made. These have been made at the 'trial level', and as such may have been too arbitrary to differentiate between different patient populations. However, such distinctions are based upon internationally approved guidelines (NICE 2004). 
 
 Two trials evaluated the effects of low dose and high dose oxygen (Somfay 2001; Davidson 1988). Both trials reported data on endurance exercise time. Head to head comparison showed no significant improvement in endurance time WMD 1.85 [95% CI ‐0.16 to 3.86] during high dose versus low dose oxygen. A number of factors undermine the validity of this estimate: i) the original studies compared a number of categories (Somfay 2001, 30;50;75;100%: Davidson 1988, 2;4;6 L) of oxygen and did report benefit from higher dose compared to lower dose. In this review data relating to the oxygen intervention closest to the critical level was extracted; ii) the SEM for the Davidson 1988 was imputed, based upon the variance from a P value given for a comparison with placebo; iii) the data for the two studies extracted from different exercise tests (Davidson 1988: endurance cycle; Somfay 2001: endurance walk test). Further studies comparing high with low dose oxygen are required before the dose effect can be more fully explored.

Somfay 2001 reported on endurance isotime. No significant difference was reported in breathlessness or V_E at isotime between high and low dose oxygen. SaO2 was significantly greater at isotime during high dose versus low dose oxygen (high dose: 99.7% (SEM 0.2), low dose: 98% (SEM 0.8), (P < 0.05).

Other Results

There is no documented criteria on the minimal clinically important difference for ambulatory oxygen, however the RCP guidelines state that ambulatory oxygen therapy should be prescribed when there is an improvement of at least 10% in walking distance and/or breathlessness score when walking with supplemental oxygen compared to air. 28/31 studies provided data on mean exercise performance (distance, time. work rate) and/or mean breathlessness score. Of these 23/28 studies (in Fujimoto 2002c patients with severe disease only) demonstrated an improvement of at least 10% in mean exercise performance and/or mean breathlessness score when walking with supplemental oxygen compared with an air cylinder.

Lack of adequate blinding procedures may exert some bias on the effect estimates, if the study investigators are aware as to which intervention participants are receiving during an exercise test. The majority of the studies assembled in the primary outcomes were double‐blind and a post‐hoc sensitivity analysis suggested that the removal of single‐blind studies did not reduce the significance of the summary estimate (Exercise distance: double‐blind studies: 20.68 metres (14.11, 27.25; seven studies); exercise time: 3.22 (2.15, 4.29; four studies).

Discussion

Thirty‐one studies (contributing 33 data sets), comparing performance during a single exercise test using ambulatory oxygen with performance during a single exercise test with placebo have been included in this review. The overall effect of oxygen was estimated (low dose plus high dose). Oxygen improved all outcomes relating to endurance exercise capacity (distance, time, number of steps) when the studies were pooled. Oxygen improved exercise distance, time, and work rate during maximal exercise testing in a mixed population of patients. Data for VO₂ max during maximal exercise testing could not be pooled and data from original studies was not consistent.

The secondary outcomes of breathlessness, SaO₂ and V_E are not directly comparable at end exercise, so comparisons were made at isotime. Oxygen improved breathlessness, SaO₂/PaO₂ and V_E at isotime with endurance exercise testing. Data from studies comparing high dose versus low dose oxygen were analysed but they may have been inadequately powered to explore the dose response in terms of breathlessness or V_E at isotime. No data from maximal exercise test data were reported on breathlessness at isotime; oxygen improved V_E and SaO₂/PaO₂ at isotime.

Methodological issues

The results of this review are noteworthy because of the lack of statistical heterogeneity in many of the meta‐analyses. The studies differed in several ways, notably in the type of exercise test deployed (Table 1), the method and amount of oxygen, whether participants were known to be able to exercise up to a certain capacity, whether studies were single or double‐blind, and the severity of the participants recruited to the studies. Due to the way in which participant characteristics were described we could not answer important questions surrounding the fulfilment of LTOT criteria as a predictor of short‐term response. We could not determine whether the assessments derived from the different exercise tests used differed sufficiently to avoid combining them. Comparisons of the different exercise tests could elucidate this issue further. The studies assembled did not provide evidence that high and low doses of oxygen are too dissimilar to keep apart in a meta‐analysis. However, the entry criteria did not allow us to explore the effects of oxygen in this setting in different patient populations. Rather than interpret this review as indicative of a consistent effect across several subgroups of patients with COPD, we feel that further studies are required in more distinct patient populations before this can be stated with authority.

Publication bias was suggested by funnel plots of primary outcomes (Figure 3; Figure 4). In spite of extensive literature searches, contact with trialists and the incorporation of data from two studies published in Japanese, there remains the possibility that there are several small negative studies that have not been published, or identified by the search strategy employed in this review. The recent identification of Knebel 2000 provides an opportunity to revisit the question of publication bias on the primary outcome. This is the only non‐significant study in this subset of trials, and takes up the third largest weighting for this outcome (16%). The level of heterogeneity has increased to only a modest level (33%) and the fixed effect summary estimate has moved a small amount towards the null compared with the previous version of this review. The addition of this study goes part way to correcting for some of the publication bias suggested by the initial funnel plot published in this review in issue 2, 2005, but additional large studies in future versions of this review will help further to provide a more reliable estimate.

3.

Funnel plot of exercise distance (Comparison 01; outcome 01). The blue dots represent the mean differences of individual trial estimates. The distribution of these dots to the right of the dotted line suggests that there may be the equivalent number of 'negative' trials that have not been included in this analysis.

4.

Funnel plot of exercise distance (Comparison 01; outcome 03). The blue dots represent the mean differences of individual trials. The distribution of these dots to the right of the dotted line suggests that there may be the equivalent number of 'negative' trials that have not been included in this analysis.

External validity

Three characteristics of the studies affect the generalisability of the review: inclusion and exclusion criteria of individual studies, method of delivery of oxygen, and outcome assessment. 
 
 Few of the trials categorised patients at baseline according to level of hypoxaemia. Due to the lack of variation between effect estimates, subgroup analyses did not provide useful insights in to whether baseline oxygen status affected response to ambulatory oxygen in the studies. This may be because the studies recruited mixed populations, rendering trial‐level distinctions on this basis somewhat arbitrary. A baseline 'mean' can be a crude measurement of population health, especially in this review where threshold values indicate treatment with LTOT oxygen on an individual patient level. The subgroup analysis from available data relating to level of hypoxaemia provides some evidence that there is benefit of ambulatory oxygen in patients who meet/likely meet as well as those who do not meet/likely do not meet the criteria for LTOT. The absence of a significant difference between treatment responses could well indicate that these studies have recruited mixed populations, with some variation between them in the percentage of participants who would qualify for LTOT.

Current UK guidelines state that non‐LTOT patients who desaturate by 4% to a point below 90% on a baseline walk should be assessed for ambulatory oxygen (RCP 1999). The majority of studies in this review do not provide information on the level homogeneity of the patients with regard to exercise induced desaturation and do not provide sufficient information to enable us to ascertain whether trials (and/or all the patients in each trial) meet the current UK criteria for ambulatory oxygen assessment. Therefore it was not possible to conduct a sub‐group analysis on the following categories: (a) patients who did not meet criteria for LTOT but have evidence of resting hypoxaemia (desaturation <90%); (b) patients who did not meet criteria for LTOT and did not have evidence of resting hypoxaemia, but who demonstrated a fall in SaO2 of at least to reach a reading below 90% during a baseline walking test whilst breathing air (c) patients who do not meet any of these criteria. Therefore we are unable to determine how accurately the results of this review can be inferred to the population which the UK guidelines recommend that ambulatory oxygen should be considered.

UK guidelines state that assessment should include titrating oxygen to a level in which SaO2 is kept above 90% (RCP 1999). It is unclear whether the flow rate/percentage of oxygen used in these studies was adequate to achieve this. If the oxygen level was inadequate it may underestimate the effect of ambulatory oxygen.

The studies in this review used a variety of methods to deliver oxygen which could affect the outcome and nasal speculae were the most frequently used method of delivery of ambulatory oxygen.

Subgroup analysis of high versus low dose oxygen did not indicate that this distinction explains any heterogeneity between the studies on any outcome. In studies where these two treatment regimens have been compared directly, there is insufficient evidence to conclude that a dose response effect exists. More studies are required in this area.

The results of our review demonstrate that maximal or endurance tests can be used in ambulatory oxygen assessments. The choice of exercise test should be determined by a number of factors: e.g. the psychometric properties of the test is important and it has been suggested that endurance exercise testing may be more sensitive and more related to functional exercise capacity and activities of daily living than maximal exercise testing (Revill 1999; ATS 2002). Other important factors include the resources available; and the patient activity levels. For example a maximal test may be appropriate for patients who intend to use ambulatory oxygen during intense activities and endurance tests may be more appropriate during functional activities of daily living. 
 
 It has been shown that the use of familiarisation and/or a practice periods prior to exercise testing will improve the quality of the data obtained (ATS 2002; Singh 1992). Some exercise tests have standardised protocols which include information on how many practice tests are required (e.g. 6MWT 2 practice tests, incremental shuttle walk test ‐ 1 practice test) (ATS 2002; Singh 1992). The majority of studies in this review included a familiarisation period and approximately half the studies included a practice period.

There is no specific guidance on length of time between repeated exercise tests and how many tests can be performed on a single day. For maximal exercise testing it has been suggested that 24‐48 hours should be left between tests, and it is unclear how long should be left between endurance tests although there is some evidence to suggest that at least 30 minutes should be left between repeat endurance tests (Marques 1998; ATS 2002). Repeated exercise testing could result in an order effect with patients performing less well on repeated testing (Marques 1998). It is unclear how much impact the order and frequency of testing could have influenced the results.

While the effects of short episodes of hypoxaemia are still unclear (especially in patients adapted to chronic hypoxaemia) single assessment studies are useful in determining the efficacy of ambulatory oxygen in patients with COPD, although it is not known how variable the results of single assessments are in any given individual (Senn 2004), nor how far the response to short term testing predicts uptake or compliance with therapy in the longer term.

Estimates for the Minimal Clinically Important Difference (MCID) are available for the 6MWT and the shuttle walk test. Redelmeier 1997 estimates the MCID for the 6MWT to be 50 metres [95% CI 37 to 71], and Singh 2002 estimates the MCID for the SWT to be 48 metres [95% CI 33.6 to 63.6]. Although effect sizes from the outcomes measuring exercise capacity in this review were statistically significant, the clinical significance of this result remains open to interpretation, especially as the lower confidence interval for our summary estimates are lower than those estimated by the above studies. No estimates on MCID for the other outcome measures have been identified. A mean difference less than the MCID does not exclude the fact that some patients in the study may actually have achieved the MCID and it would be useful for future studies to state how many individual patients achieve the MCID for primary outcome measure.

Internal validity

The methodological quality of the trials in this review were assessed using two different scales ‐the Jadad and PEDro scales. Both scales were used because different quality scales have been shown to generate discrepant results. Both of these scales mostly assess different aspects of internal validity‐ Jadad: randomisation (2 questions, 40%), blinding (2 questions, 40%) and withdrawal (1 question, 20%); PEDro: randomisation (2 questions, 20%), blinding (3 questions, 30%), withdrawal (2 questions, 20%). In addition PEDro includes aspects relating to baseline characteristics (1 question, 10%) and statistics (2 questions, 20%). Jadad gives more weighting to quality of reporting than actual methodological quality, for example Jadad focuses on reporting of randomisation and the reporting of the randomisation procedure where as PEDro focuses on whether there is randomised and concealed allocation. In Jadad a statement on withdrawals will earn a point independently of how many patients were excluded or whether "intention to treat" was used whereas the PEDro scale examines the impact of patient withdrawals in detail. These differences explain any discrepancies in the rank order of quality rating between the Jadad and Pedro scales. A score of 3 on the Jadad scale represents "high quality" (Jadad 1996). Six studies scored 3 or above on the Jadad scale. All studies which scored 3 on the Jadad scale scored 9 or 10 on the PEDro. There is no reference criteria regarding high quality on the PEDro scale‐ all studies in this review scored 6 or more. We would propose that the PEDro scale may be more sensitive than the Jadad scale to aspects of internal validity relevant to the ambulatory oxygen studies included in this review and that a score of 6 or more demonstrates that all the trials in this review have at least reasonable internal validity.

Authors' conclusions

Implications for practice.

This review provides evidence from single assessment studies that ambulatory oxygen improves exercise performance. It is our opinion that it is no longer necessary to compare performance during an exercise test using oxygen versus placebo air but rather an ambulatory assessment should determine whether patients demonstrate objective benefit during an exercise test using oxygen versus room air. The system used in the ambulatory oxygen assessment should be the system that would be potentially made available to the patient and should be used in the same way that the patient would be advised to use it (for example, if the patient will be required to carry/wheel the ambulatory system during activities of daily living then they should carry/wheel it during the ambulatory oxygen assessment).

Maximal or endurance tests could be used in ambulatory oxygen assessment. Consideration should be given to the measurement of signs and symptoms such as SaO₂/SpO₂ and breathlessness at isotime as these provide important information. We recommend that these outcomes are included in the assessment for ambulatory oxygen. Consideration should also be given to the amount of oxygen required to maintain SaO₂/SpO₂ at an acceptable level. The technology for the delivery of ambulatory oxygen is developing rapidly and in the future such systems should be accessible to those who demonstrate objective benefit during an ambulatory oxygen assessment (Law 2004; DOHBTS 2004). Detailed and standardised assessment procedures for ambulatory oxygen therapy should be developed and these should include assessment of potential utilisation by the patient as well as objective benefit.

Implications for research.

This review has established the short‐term efficacy of ambulatory oxygen, although additional studies may help to reduce suspected bias in reporting of the primary outcome. In our opinion research should focus on establishing the long‐term efficacy, and on developing a method for determining the optimal dose and delivery system for ambulatory oxygen. Further trials are also required to establish whether improvements in primary and secondary outcomes which reach statistical significance are clinically important. We remain uncertain of the level of benefit of ambulatory oxygen in specific subgroups of COPD: patients who already meet the criteria for LTOT; (b) patients who do not meet criteria for LTOT but have evidence of resting hypoxaemia (desaturation <90%); (c) patients who do not meet criteria for LTOT and do not have evidence of resting hypoxaemia but who demonstrate a fall in SpO₂ of at least 4% to reach a reading below 90% during a baseline walking test whilst breathing air.

What's new

Date	Event	Description
24 April 2008	Amended	Converted to new review format.

History

Protocol first published: Issue 1, 2003
 Review first published: Issue 2, 2005

Date	Event	Description
8 June 2005	New citation required and conclusions have changed	Substantive amendment

Data and analyses

Comparison 1. Oxygen versus placebo (crossover studies).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Endurance test ‐ exercise distance (Davidson 1988 6MWT)	10		Metres (Fixed, 95% CI)	18.86 [13.11, 24.61]
1.1 Low dose	10		Metres (Fixed, 95% CI)	18.86 [13.11, 24.61]
2 Endurance test ‐ exercise distance (Davidson 1988 endurance walk)	10		Metres (Fixed, 95% CI)	18.61 [12.83, 24.39]
2.1 Low dose	10		Metres (Fixed, 95% CI)	18.61 [12.83, 24.39]
3 Endurance test ‐ exercise time (Davidson 1988 low dose cycle data/Somfay 2001 low dose)	7		Minutes (Fixed, 95% CI)	2.70 [1.95, 3.44]
3.1 Low dose	3		Minutes (Fixed, 95% CI)	2.21 [1.18, 3.25]
3.2 High dose	4		Minutes (Fixed, 95% CI)	3.22 [2.15, 4.29]
4 Endurance test ‐ exercise time (Davidson 1988 high dose cycle/Somfay 2001 high dose)	7		Minutes (Fixed, 95% CI)	2.71 [1.96, 3.46]
4.1 Low dose	1		Minutes (Fixed, 95% CI)	1.91 [0.79, 3.03]
4.2 High dose	6		Minutes (Fixed, 95% CI)	3.37 [2.35, 4.38]
5 Endurance test ‐ exercise time (Davidson 1988 low dose end'rnce walk/Somfay 2001 low dose)	7		Minutes (Fixed, 95% CI)	2.63 [1.91, 3.34]
5.1 Low dose	3		Minutes (Fixed, 95% CI)	2.15 [1.18, 3.12]
5.2 High dose	4		Minutes (Fixed, 95% CI)	3.20 [2.14, 4.25]
6 Endurance test ‐ exercise time (change from baseline)	2		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
6.1 Low dose	2		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
7 Endurance test ‐ exercise steps	1		Number (Fixed, 95% CI)	Totals not selected
7.1 Low dose	1		Number (Fixed, 95% CI)	0.0 [0.0, 0.0]
7.2 High dose	0		Number (Fixed, 95% CI)	0.0 [0.0, 0.0]
8 Maximal test ‐ exercise distance	4		Metres (Fixed, 95% CI)	32.00 [20.61, 43.38]
8.1 Low dose	4		Metres (Fixed, 95% CI)	32.00 [20.61, 43.38]
9 Maximal test ‐ exercise time	6		Minutes (Fixed, 95% CI)	1.06 [0.67, 1.46]
9.1 Low dose	4		Minutes (Fixed, 95% CI)	1.30 [0.56, 2.04]
9.2 High dose	2		Minutes (Fixed, 95% CI)	0.97 [0.50, 1.43]
10 Maximal test ‐ exercise VO2max (SMD)	5		Std. Mean Difference (IV, Fixed, 95% CI)	Totals not selected
10.1 Low dose	3		Std. Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
10.2 High dose	2		Std. Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
11 Maximal test ‐ wattage output	3		Watts (Fixed, 95% CI)	8.88 [5.71, 12.06]
11.1 Low dose	2		Watts (Fixed, 95% CI)	7.77 [4.19, 11.35]
11.2 High dose	1		Watts (Fixed, 95% CI)	13.0 [6.12, 19.88]
12 Endurance test ‐ isotime breathlessness (Somfay 2001 low dose)	4		Borg (Fixed, 95% CI)	‐1.15 [‐1.65, ‐0.66]
12.1 Low dose	1		Borg (Fixed, 95% CI)	‐2.3 [‐4.46, ‐0.14]
12.2 High dose	3		Borg (Fixed, 95% CI)	‐1.09 [‐1.60, ‐0.58]
13 Endurance test ‐ isotime breathlessness (Somfay 2001 high dose)	4		Borg (Fixed, 95% CI)	‐1.15 [‐1.66, ‐0.65]
13.1 Low dose	0		Borg (Fixed, 95% CI)	0.0 [0.0, 0.0]
13.2 High dose	4		Borg (Fixed, 95% CI)	‐1.15 [‐1.66, ‐0.65]
14 Endurance test ‐ isotime SaO2 (Somfay 2001 low dose)	3		% (Fixed, 95% CI)	8.36 [5.08, 11.64]
14.1 Low dose	1		% (Fixed, 95% CI)	7.0 [0.47, 13.53]
14.2 High dose	2		% (Fixed, 95% CI)	8.82 [5.03, 12.62]
15 Endurance test ‐ isotime SaO2 (Somfay 2001 high dose)	3		% (Fixed, 95% CI)	8.80 [5.36, 12.24]
15.1 Low dose	0		% (Fixed, 95% CI)	0.0 [0.0, 0.0]
15.2 High dose	3		% (Fixed, 95% CI)	8.80 [5.36, 12.24]
16 Endurance test ‐ isotime PaO2	2		mmHg (Fixed, 95% CI)	15.15 [6.42, 23.89]
16.1 Low dose	0		mmHg (Fixed, 95% CI)	0.0 [0.0, 0.0]
16.2 High dose	2		mmHg (Fixed, 95% CI)	15.15 [6.42, 23.89]
17 Endurance test ‐ isotime ventilation (Somfay low dose)	5		L/min (Fixed, 95% CI)	‐3.58 [‐4.85, ‐2.31]
17.1 Low dose	1		L/min (Fixed, 95% CI)	‐4.2 [‐8.12, ‐0.28]
17.2 High dose	4		L/min (Fixed, 95% CI)	‐3.51 [‐4.85, ‐2.17]
18 Endurance test ‐ isotime ventilation (Somfay high dose)	5		L/min (Fixed, 95% CI)	‐3.60 [‐4.88, ‐2.33]
18.1 Low dose	0		L/min (Fixed, 95% CI)	0.0 [0.0, 0.0]
18.2 High dose	5		L/min (Fixed, 95% CI)	‐3.60 [‐4.88, ‐2.33]
19 Maximal test ‐ isotime SaO2	2		% (Fixed, 95% CI)	7.82 [4.89, 10.74]
19.1 Low dose	1		% (Fixed, 95% CI)	7.1 [3.40, 10.80]
19.2 High dose	1		% (Fixed, 95% CI)	9.0 [4.24, 13.76]
20 Maximal test ‐ isotime PaO2	2		kPa (Fixed, 95% CI)	7.69 [4.32, 11.06]
20.1 Low dose	0		kPa (Fixed, 95% CI)	0.0 [0.0, 0.0]
20.2 High dose	2		kPa (Fixed, 95% CI)	7.69 [4.32, 11.06]
21 Maximal test ‐ isotime ventilation	4		L/min (Fixed, 95% CI)	‐3.26 [‐4.33, ‐2.19]
21.1 Low dose	3		L/min (Fixed, 95% CI)	‐3.55 [‐5.11, ‐1.99]
21.2 High dose	1		L/min (Fixed, 95% CI)	‐3.0 [‐4.47, ‐1.53]
22 Endurance test VO2	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
22.1 Low dose	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]

1.1. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 1 Endurance test ‐ exercise distance (Davidson 1988 6MWT).

1.2. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 2 Endurance test ‐ exercise distance (Davidson 1988 endurance walk).

1.3. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 3 Endurance test ‐ exercise time (Davidson 1988 low dose cycle data/Somfay 2001 low dose).

1.4. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 4 Endurance test ‐ exercise time (Davidson 1988 high dose cycle/Somfay 2001 high dose).

1.5. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 5 Endurance test ‐ exercise time (Davidson 1988 low dose end'rnce walk/Somfay 2001 low dose).

1.6. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 6 Endurance test ‐ exercise time (change from baseline).

1.7. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 7 Endurance test ‐ exercise steps.

1.8. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 8 Maximal test ‐ exercise distance.

1.9. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 9 Maximal test ‐ exercise time.

1.10. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 10 Maximal test ‐ exercise VO2max (SMD).

1.11. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 11 Maximal test ‐ wattage output.

1.12. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 12 Endurance test ‐ isotime breathlessness (Somfay 2001 low dose).

1.13. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 13 Endurance test ‐ isotime breathlessness (Somfay 2001 high dose).

1.14. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 14 Endurance test ‐ isotime SaO2 (Somfay 2001 low dose).

1.15. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 15 Endurance test ‐ isotime SaO2 (Somfay 2001 high dose).

1.16. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 16 Endurance test ‐ isotime PaO2.

1.17. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 17 Endurance test ‐ isotime ventilation (Somfay low dose).

1.18. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 18 Endurance test ‐ isotime ventilation (Somfay high dose).

1.19. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 19 Maximal test ‐ isotime SaO2.

1.20. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 20 Maximal test ‐ isotime PaO2.

1.21. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 21 Maximal test ‐ isotime ventilation.

1.22. Analysis.

Comparison 1 Oxygen versus placebo (crossover studies), Outcome 22 Endurance test VO2.

Comparison 2. SUBGROUP ANALYSES: mean baseline kPa/PaO2 & high dose versus low dose studies).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 MEAN BASELINE PAO2 </> 7.3kPa/55mmHG: Endurance test ‐ exercise distance	10		Metres (Fixed, 95% CI)	18.86 [13.11, 24.61]
1.1 Mean baseline PaO2 </=7.3 kPa/55 mmHg	0		Metres (Fixed, 95% CI)	0.0 [0.0, 0.0]
1.2 Mean baseline PaO2 >7.3 kPa/55 mmHg	10		Metres (Fixed, 95% CI)	18.86 [13.11, 24.61]
2 MEAN BASELINE PAO2 </> 7.3kPa/55mmHG: Endurance test ‐ exercise time	7		Minutes (Fixed, 95% CI)	2.99 [2.12, 3.86]
2.1 Mean baseline PaO2 </=7.3 kPa/55 mmHg	1		Minutes (Fixed, 95% CI)	4.7 [1.66, 7.74]
2.2 Mean baseline PaO2 >7.3 kPa/55 mmHg	5		Minutes (Fixed, 95% CI)	3.16 [2.11, 4.22]
2.3 Unclear O2 status	1		Minutes (Fixed, 95% CI)	1.91 [0.15, 3.67]
3 MEAN BASELINE PAO2 </> 7.3kPa/55mmHG: Maximal test ‐ exercise distance	4		Metres (Fixed, 95% CI)	32.00 [20.61, 43.38]
3.1 Mean baseline PaO2 </=7.3 kPa/55 mmHg	1		Metres (Fixed, 95% CI)	27.3 [14.76, 39.84]
3.2 Mean baseline PaO2 >7.3 kPa/55 mmHg	3		Metres (Fixed, 95% CI)	54.01 [26.85, 81.16]
4 MEAN BASELINE PAO2 </> 7.3kPa/55mmHG: Maximal test ‐ exercise time	6		Minutes (Fixed, 95% CI)	1.06 [0.67, 1.46]
4.1 Mean baseline PaO2 </=7.3 kPa/55 mmHg	1		Minutes (Fixed, 95% CI)	2.0 [0.28, 3.72]
4.2 Mean baseline PaO2 >7.3 kPa/55 mmHg	4		Minutes (Fixed, 95% CI)	1.12 [0.69, 1.55]
4.3 Unclear O2 status	1		Minutes (Fixed, 95% CI)	0.05 [‐1.22, 1.32]
5 MEAN BASELINE PAO2 </> 7.3kPa/55mmHG: Maximal test ‐ exercise VO2max (SMD)	5		Std. Mean Difference (IV, Fixed, 95% CI)	Totals not selected
5.1 Mean baseline PaO2 </=7.3 kPa/55 mmHg	1		Std. Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
5.2 Mean baseline PaO2 >7.3 kPa/55 mmHg	4		Std. Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
6 MEAN BASELINE PAO2 </> 7.3kPa/55mmHG: Endurance test ‐ isotime breathlessness	4		Borg (Fixed, 95% CI)	‐1.15 [‐1.65, ‐0.66]
6.1 Mean baseline PaO2 </= 7.3kPa/55mmHg	1		Borg (Fixed, 95% CI)	‐2.0 [‐4.14, 0.14]
6.2 Mean baseline PaO2 >7.3 kPa/55 mmHg	3		Borg (Fixed, 95% CI)	‐1.11 [‐1.62, ‐0.59]
7 MEAN BASELINE PAO2 </> 7.3kPa/55mmHG: Maximal test ‐ isotime ventilation	4		L/min (Fixed, 95% CI)	‐3.26 [‐4.33, ‐2.19]
7.1 Mean baseline PaO2 </= 7.3kPa/55mmHg	1		L/min (Fixed, 95% CI)	‐6.0 [‐11.21, ‐0.79]
7.2 Mean baseline PaO2 >7.3 kPa/55 mmHg	3		L/min (Fixed, 95% CI)	‐3.14 [‐4.23, ‐2.04]
8 HIGH DOSE VERSUS LOW DOSE: Endurance test ‐ exercise time (Davidson imputed SEM)	2		Minutes (Fixed, 95% CI)	1.85 [‐0.16, 3.86]
9 HIGH DOSE VERSUS LOW DOSE: Endurance test ‐ isotime breathlessness	1		Borg (Fixed, 95% CI)	Totals not selected
10 HIGH DOSE VERSUS LOW DOSE: Endurance test ‐ isotime SaO2	1		% (Fixed, 95% CI)	Totals not selected
11 HIGH DOSE VERSUS LOW DOSE: Endurance test ‐ isotime ventilation	1		L/min (Fixed, 95% CI)	Totals not selected
12 HIGH DOSE VERSUS LOW DOSE: Maximal test ‐ isotimeSaO2	0		SaO2 (Fixed, 95% CI)	0.0 [0.0, 0.0]
13 MEAN BASELINE PA02</> 7.3kPa/55mmHg: Maximal test‐ wattage	3		Wattage (Fixed, 95% CI)	8.88 [5.71, 12.06]
13.1 Mean baseline PaO2 </=7.3 kPa/55 mmHg	0		Wattage (Fixed, 95% CI)	0.0 [0.0, 0.0]
13.2 Mean baseline PaO2 >7.3 kPa/55 mmHg	3		Wattage (Fixed, 95% CI)	8.88 [5.71, 12.06]

2.1. Analysis.

Comparison 2 SUBGROUP ANALYSES: mean baseline kPa/PaO2 & high dose versus low dose studies), Outcome 1 MEAN BASELINE PAO2 </> 7.3kPa/55mmHG: Endurance test ‐ exercise distance.

2.2. Analysis.

Comparison 2 SUBGROUP ANALYSES: mean baseline kPa/PaO2 & high dose versus low dose studies), Outcome 2 MEAN BASELINE PAO2 </> 7.3kPa/55mmHG: Endurance test ‐ exercise time.

2.3. Analysis.

Comparison 2 SUBGROUP ANALYSES: mean baseline kPa/PaO2 & high dose versus low dose studies), Outcome 3 MEAN BASELINE PAO2 </> 7.3kPa/55mmHG: Maximal test ‐ exercise distance.

2.4. Analysis.

Comparison 2 SUBGROUP ANALYSES: mean baseline kPa/PaO2 & high dose versus low dose studies), Outcome 4 MEAN BASELINE PAO2 </> 7.3kPa/55mmHG: Maximal test ‐ exercise time.

2.5. Analysis.

Comparison 2 SUBGROUP ANALYSES: mean baseline kPa/PaO2 & high dose versus low dose studies), Outcome 5 MEAN BASELINE PAO2 </> 7.3kPa/55mmHG: Maximal test ‐ exercise VO2max (SMD).

2.6. Analysis.

Comparison 2 SUBGROUP ANALYSES: mean baseline kPa/PaO2 & high dose versus low dose studies), Outcome 6 MEAN BASELINE PAO2 </> 7.3kPa/55mmHG: Endurance test ‐ isotime breathlessness.

2.7. Analysis.

Comparison 2 SUBGROUP ANALYSES: mean baseline kPa/PaO2 & high dose versus low dose studies), Outcome 7 MEAN BASELINE PAO2 </> 7.3kPa/55mmHG: Maximal test ‐ isotime ventilation.

2.8. Analysis.

Comparison 2 SUBGROUP ANALYSES: mean baseline kPa/PaO2 & high dose versus low dose studies), Outcome 8 HIGH DOSE VERSUS LOW DOSE: Endurance test ‐ exercise time (Davidson imputed SEM).

2.9. Analysis.

Comparison 2 SUBGROUP ANALYSES: mean baseline kPa/PaO2 & high dose versus low dose studies), Outcome 9 HIGH DOSE VERSUS LOW DOSE: Endurance test ‐ isotime breathlessness.

2.10. Analysis.

Comparison 2 SUBGROUP ANALYSES: mean baseline kPa/PaO2 & high dose versus low dose studies), Outcome 10 HIGH DOSE VERSUS LOW DOSE: Endurance test ‐ isotime SaO2.

2.11. Analysis.

Comparison 2 SUBGROUP ANALYSES: mean baseline kPa/PaO2 & high dose versus low dose studies), Outcome 11 HIGH DOSE VERSUS LOW DOSE: Endurance test ‐ isotime ventilation.

2.13. Analysis.

Comparison 2 SUBGROUP ANALYSES: mean baseline kPa/PaO2 & high dose versus low dose studies), Outcome 13 MEAN BASELINE PA02</> 7.3kPa/55mmHg: Maximal test‐ wattage.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Bradley 1978.

Methods	Blinding: single; Randomisation: no description; Withdrawals: none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad=1/5 Pedro=7/10 Statistical analysis: Student's paired t test
Participants	Country USA; N=26; 11M; 64.7 (8.7) years; FEV1 22.2 (7.8)%; FVC 47.9 (14)%; SaO2 92 (4.1)%; **PaO2 69 (14.3)mmHg; PaCO2 41.8 (8.1) mmHg;
Interventions	5 L/min O2 or air via cylinder and nasal specs
Outcomes	see table‐ outcome measures used
Notes	No contact address
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Bye 1985.

Methods	Blinding: double; Randomisation: no description; Withdrawals: None; Baseline characteristics: Comparable; Point/ variability estimates for between group analysis: Yes Power calculation: None Jadad=2/5 Pedro=9/10 Statistical analysis: Paired t test
Participants	Country Canada; N=8; 8M; 61 SEM 3 years ; FEV1 32 SEM 4 %, FVC 59 SEM 5 % *PaO2 63 SEM6 mmHg, PaCO2 43 SEM 3mmHg;
Interventions	40% O2 or humidified air via reservoir and mouth piece
Outcomes
Notes	Author contacted: data confirmed
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Criner 1987.

Methods	Blinding: single; Randomisation: no description; Withdrawals: none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad=1/5 Pedro=7/10 Statistical analysis: Paired t test
Participants	Country USA; N=6; 5M; 64 (7) years; FEV1 0.66 (0.13)L; PaO2 67 (6.2) mmHg**, PaCO2 41 (3) mmHg
Interventions	30% O2 or air via reservoir and mouth piece
Outcomes
Notes	Author contacted: no response
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Davidson 1988.

Methods	Blinding: single; Randomisation: no description; Withdrawals:none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad=1/5 Pedro=7/10 Statistical test: Wilcoxon's signed rank test
Participants	Country UK; N=17; 64.4 SEM 2.1 years; FEV1 0.79 SEM 0.03 L, FVC 2.14 SEM 0.11L; **PaO2 8.8 SEM 0.3 kPa, PaCO2 6 SEM 0.4 kPa;
Interventions	6L/min O2 or air via mouthpiece 4L/min O2 or air via nasal specs or mask for walking tests
Outcomes
Notes	Author contacted: no response; Data not extracted relating to 2LO2 Quality scores relate to cycle test. Walking tests were double blinded
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Dean 1992.

Methods	Blinding: double; Randomisation: flip coin; Withdrawals:none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad= 5/5 Pedro = 9/10 Statistical test: Paired t test
Participants	Country USA; N=12; 12M; >50years; FEV1 0.89 SEM 0.09 L; FVC 2.37 SEM 0.2 L; **PaO2 71 SEM 2.6 mmHg
Interventions	40% oxygen or air via reservoir and mouthpiece
Outcomes
Notes	Author contacted: data confirmed further details provided on details of exercise test
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Eaton 2002.

Methods	Blinding: double; Randomisation: no details; Withdrawals: N = 11; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Statistical test: Mixed model approach which used information from all participants. 41 participants reported, but 39 participants completed both arms. Jadad= 3/5 Pedro = 9/10
Participants	Country New Zealand; 29M; N=41; 67.1 (9.3) years; FEV1 25.9% SaO2 94 (1.9)%; **PaO2 9.2kPa; PaCO2 5.8 (0.7)kPa;
Interventions	4 L/min O2 or air via cylinder and nasal specs. Study duration: 12 weeks. Participants prescribed O2 for 2 x 6 week periods. This was used as ambulatory O2.
Outcomes
Notes	Author contacted: no response; This was a 12 week study of ambulatory O2. Data was extracted for the acute responses at baseline of 41/50 patients who completed 12 week study
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Fujimoto 2002a.

Methods	Blinding: double; Randomisation: no description; Withdrawals: none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad=2/5 Pedro=9/10 Statistical analysis: Paired t test
Participants	Country Japan; N=16; 16M; 71 (2) years; FEV1 62.7SEM 2.9%; PaO2 10.07** SEM 0.25, PaCO2 5.27
Interventions	2 L/min O2 or air via cylinder and nasal specs
Outcomes
Notes	Author contacted: data confirmed This study stratifies patients according to disease severity and reports results for each group separately Data not extracted for haemodynamic study as measurements taken 3 mins into exercise only
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Fujimoto 2002b.

Methods	Blinding: double; Randomisation: no description; Withdrawals:None; Baseline characteristics: Comparable; Point/ variability estimates for between group analysis: Yes Power calculation: None Jadad=2/5 Pedro=9/10
Participants	Country Japan; N=25; 25M; 69 (1) years; FEV1 40.9 SEM 1%; PaO2 9.1** SEM 0.1, PaCO2 5.5 SEM 0.1
Interventions	2 L/min O2 or air via cylinder and nasal specs
Outcomes
Notes	Author contacted: ?5/5/04 This study stratifies patients according to disease severity and reports results for each group separately Data not extracted for haemodynamic study as measurements taken 3 mins into exercise only
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Fujimoto 2002c.

Methods	Blinding: double; Randomisation: no description; Withdrawals:none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad=2/5 Pedro=9/10 Statistical analysis: Paired t test
Participants	Country Japan; N=34; 34M; 66 (1) years; FEV1 25 SEM 1.1%; PaO2 9.4** SEM 0.1, PaCO2 5.7 SEM 0.1
Interventions	2 L/min O2 or air via cylinder and nasal specs
Outcomes
Notes	Author contacted: data confirmed This study stratifies patients according to disease severity and reports results for each group separately Data not extracted for haemodynamic study as measurements taken 3 mins into exercise only
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Garrod 1999.

Methods	Blinding: single; Randomisation: no description; Withdrawals:none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad=1/5 Pedro=8/10 Statistical test: paired t test
Participants	Country UK; N=15; mean age 66 range (50‐75) years; FEV1 32 (9.4)%; PaO2 8.38 (1.24) kPa,** PaCO2 5.95 (0.86) kPa
Interventions	2 L/min O2 or air via cylinder and nasal specs
Outcomes
Notes	Author contacted: data confirmed This study also evaluated pulsed dose oxygen delivery
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	Inestigators unaware as to order of treatment group assignment (Cochrane Grade A)

Garrod 2000.

Methods	Blinding: single; Randomisation: no description; Withdrawals: none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad=1/5 Pedro=7/10 Statistical test: paired t test
Participants	Country UK; N=25; 19M; median age 70 range 52‐84years; FEV1 30 (9.89)%; fall in SaO2 by 4% from baseline to 90% or below on exercise testing; *11/25 on LTOT
Interventions	4 L/min O2 or air via cylinder and nasal specs
Outcomes
Notes	Author contacted: data confirmed further details provided on point/ variability estimates for between group analysis This was a 6 week study of supplemental oxygen in pulmonary rehabilitation. Data was extracted for the acute responses at baseline of 25 patients recruited
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Gosselin 2004.

Methods	Blinding: single Randomisation: no description; Withdrawals: none Baseline characteristics: comparable Point/ variability estimates for between group analysis: yes Power calculation: none Jadad= 1/5 Pedro = 7/10 Statistical test: two‐way analysis of variance
Participants	Country France; N=9; 7M; 62 SEM 2.2 years; FEV1 1.5 SEM 0.1 L; FVC 2.9 SEM 0.1 L; PaO2 **9.1 (0.4) kPa; PaCO2 4.2 SEM 0.08 kPa
Interventions	30% O2 or air via mouthpiece
Outcomes
Notes	Authors not contacted
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Ishimine 1995.

Methods	Blinding: double; Randomisation: no description; Withdrawals: none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad= 2/5 Pedro = 9/10 Statistical test: unclear
Participants	Country Japan: N=22; 22M; 69 (7) years; FEV1 44.9 (22.7)%; FVC 2.26 (0.57)L; PaO2 75.9 (8.6)mmHg; PaCO2 43.6 (4.8)mmHg ** no LTOT
Interventions	3 L/min O2 or air via cylinder and nasal specs
Outcomes
Notes	Authors not contacted Japanese article study translated
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

King 1973.

Methods	Blinding: single; Randomisation: no description; Withdrawals: none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad= 1/5 Pedro = 7/10 Statistical test: paired t test
Participants	Country UK: N=10; 7M; 60.2 (7.1) years; FEV1 0.76 (0.32)L; *PO2 48.8 (5.8)mmHg; PaCO2 52.1 (8.8)mmHg;
Interventions	30% O2 or air via reservoir and mouthpiece
Outcomes
Notes	Author contacted: no response; No data available for primary or secondary outcomes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Knebel 2000.

Methods	Blinding: double Randomisation: yes, random numbers table Withdrawals: none Baseline characteristics: comparable Point/ variability estimates for between group analysis: yes Power calculation: none Jadad= 5/5 Pedro = 10/10 Statistical test: Paired t tests
Participants	Country USA: N=33; 22M; 47 (7) years; FEV1 48 (13) %; **SpO2 97.1 (1.7) %. Participants excluded for current use of O2.
Interventions	4 L/min O2 or air via cylinder and nasal specs
Outcomes
Notes	Author not contacted
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Kurihara 1989.

Methods	Blinding: single; Randomisation: no description; Withdrawals: none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad= 1/5 Pedro = 7/10 Statistical test: Wilcoxon t test
Participants	Country Japan: N=14; 11M; 68.8 (8.9) years; FEV1% 30.8 (6.4)%; FVC 58.3 (6.2)%; PaO2 68.8 (8.9)mmHg ** no LTOT
Interventions	3 L/min O2 or air via cylinder and nasal specs
Outcomes
Notes	Authors not contacted Japanese article study translated
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Leach 1992.

Methods	Blinding: double; Randomisation: no description; Withdrawals:none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad= 2/5 Pedro = 9/10 Statistical test: paired t test Stats test: paired t test
Participants	Country UK: N=20; 12M; 63.4 (7.2) years; FEV1 0.74 (0.25)L, FVC 1.94 (0.51)L; *PaO2 8.74 (2.38)kPa; PaCO2 5.55 (1.26)kPa
Interventions	4 L/min O2 or air via facemask
Outcomes
Notes	Author contacted: no response; Mixed population, data extracted for COPD only Data not extracted relating to 2 or 6LO2 as no comparable placebo
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Leggett 1977.

Methods	Blinding: single; Randomisation: no description; Withdrawals:none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad= 1/5 Pedro = 7/10 Statistical test: paired t test
Participants	Country UK: N=8; assessed for LTOT*
Interventions	4 L/min O2 or air via cylinder and nasal specs
Outcomes	Distance walked in 12 minutes
Notes	Data not extracted relating to 2 L/min O2 or 4L/min O2 when carrying walker. No data available for primary or secondary outcomes on the treadmill/bicycle study as no comparable placebo
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Light 1989.

Methods	Blinding: single; Randomisation: no description; Withdrawals: none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none; Jadad= 1/5 Pedro = 7/10 Statistical test: Student's paired t test
Participants	Country USA: N=17; 16M; 62 (5.3) years; FEV1 0.99 (0.45)L; FVC 2.65 (1)L; **PaO2 68.7 (12.1)mmHg, PaCO2 40.3 (6.6)mmHg SaO2 91.2 (4)%
Interventions	30% O2 or air via mouth piece
Outcomes
Notes	Author contacted: data confirmed
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Maltais 2001.

Methods	Blinding: double Randomisation: no description Withdrawals: none Baseline characteristics:comparable Point/ variability estimates for between group analysis: yes Power calculation: none Jadad=2 Pedro=9 Statistical test: ANOVA
Participants	Country Canada: N=14; FEV1 1.04 SEM (0.07), FVC 2.64 SEM 0.15; **PaO2 85 SEM 4 mmHg; PaCO2 37 SEM 2 mmHg
Interventions	75% O2 or air via a mouthpiece
Outcomes
Notes	Author nor contacted
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Mannix 1992.

Methods	Blinding: none; Randomisation: no description; Withdrawals: none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad=1/5 Pedro=6/10 Statistical test: ANOVA; significant tests followed by paired t test
Participants	Country USA: N=10; 10M; FEV1 0.96 SEM 0.25, FVC 2.27 SEM 0.35; *PaO2 54 SEM 3 mmHg; PaCO2 47 SEM 3 mmHg
Interventions	30% O2 or air via mouth piece
Outcomes
Notes	Author contacted: no response
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

McDonald 1995.

Methods	Blinding: double; Randomisation: no description; Withdrawals: none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad= 3/5 Pedro = 9/10 Statistical tests: paired t test
Participants	Country Australia: N=26; 24M; 73 (6) years; FEV1 0.9 (0.4)L; **PaO2.69 (8.5)mmHg;
Interventions	4 L/min O2 or air via cylinder and nasal specs
Outcomes
Notes	Author contacted: no response; This was a 12 week study of supplemental oxygen. Data was extracted for the acute responses at baseline of 26/36 patients recruited
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

McKeon 1988.

Methods	Blinding: double; Randomisation: no description; Withdrawals: none; Baseline characteristics: Comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad= 3/5 Pedro=9/10 Statistical test: ANOVA
Participants	Country Australia: N=21; 10M; 62 (9) years; FEV1, 29(13)%; FVC 58(20)%; **PaO266.4 (11)mmHg; PaCO2 43.9 (8.8)mmHg; 6 on home O2 15 hours per day
Interventions	4 L/min O2 or air via light weight cylinder and nasal specs
Outcomes
Notes	Author contacted: no response;
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

O'Donnell 1997.

Methods	Blinding: double; Randomisation: no description; Withdrawals: none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad= 2/5 Pedro=9/10 Statistical test: Paired t test
Participants	Country Canada: N=11; 7M; 68 (2) years; FEV1 0.97 (0.13)L, FVC 2.27 (0.25)L; **non LTOT and mildly hypoxemic and did not meet criteria for ambulatory O2
Interventions	60% O2 or air via mouthpiece
Outcomes
Notes	Author contacted: data confirmed and further details provided on exercise test and on point/ variability estimates for between group analysis
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

O'Donnell 2001.

Methods	Blinding: double; Randomisation: no description; Withdrawals: none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad= 2/5 Pedro=9/10 Statistical analysis: Paired t test
Participants	Country Canada: N=11; 4M; 68 (2) years; FEV1 0.65 (0.06)L, FVC 1.59 (0.11)L; *PaO2 52.4 (2.2)mmHg; PaCO2 48.5 (2.1)mmHg; met the criteria for ambulatory O2 in Canada
Interventions	60% O2 or air via mouthpiece
Outcomes
Notes	Author contacted: data confirmed and further details provided on exercise test
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Palange 1995.

Methods	Blinding: none Randomisation: no description Withdrawals: none Baseline characteristics :comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad= 1/5 Pedro=6/10
Participants	Country Italy: N =9; 67 SEM 1 years; FEV11.0 SEM 0.1L; FVC 2.6 SEM 0.2L; **PaO2 64 SEM 2 mmHg; PaCO2 43 SEM 1 mmHg
Interventions	30% O2 in via mouthpiece
Outcomes
Notes	Author nor contacted
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Raimondi 1970.

Methods	Blinding: single; Randomisation: no description; Withdrawals: none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad= 1/5 Pedro=7/10
Participants	Country UK: N =8; 51‐70 years; FEV1 0.74 range 0.55‐0.95L; ***no information re O2 status
Interventions	35% O2 in nitrogen and 21% O2 in helium via mouthpiece
Outcomes
Notes	No contact address
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Somfay 2001.

Methods	Blinding: single; Randomisation: no description; Withdrawals: none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad= 1/5 Pedro=7/10 Statistical test: ANOVA
Participants	Country USA: N=10; 6M; 65 (7) years; FEV1 31 (10)%, FVC 76 (15)%; **mild hypoxaemia SaO2>92% and during exercise >88%, none qualified for ambulatory O2
Interventions	30% and 50% O2 or air via mouth piece
Outcomes
Notes	Author contacted: no response No data extracted for 75% or100% O2
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Stein 1982.

Methods	Blinding: single; Randomisation: no details; Withdrawals: none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad=1 /5 Pedro=7/10 Statistical test: not reported
Participants	Country USA: N=9; 8M; FEV1 29 (3)% **PaO2 63 (10)PaCO2 39 (6)
Interventions	30% O2 in 70% nitrogen or air via mouthpiece
Outcomes
Notes	Author contacted: no response Data for exercise time estimated from graph
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Swinburn 1984.

Methods	Blinding: double; Randomisation: no description; Withdrawals: none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad=2/5 Pedro=9/10 Statistical analysis: Student's paired t test
Participants	Country UK: N=5; 2M; 65 (53‐72) years; FEV1 0.8 (0.2) L; FVC 1.8 (0.4)L; SaO2 93.2 (82‐94)%**
Interventions	60% O2 or air via reservoir and mouth piece
Outcomes
Notes	Author contacted: data confirmed
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Vyas 1971.

Methods	Blinding: double; Randomisation: no description; Withdrawals: yes n=2?; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: none Power calculation: none Jadad=2/5 Pedro=8/10 Statistical test: t test
Participants	N=14; 12M; FEV1 29.5 (10.2)% **PaO2 71.4 (8.78) PaCO2 38.4 (4.6) Country Canada:
Interventions	40% O2 or air via reservoir mouth piece
Outcomes
Notes	Author contacted: Primary author deceased ?sent to second author Demographic data documented in previous article Point/variability estimates for between group analysis calculated from raw data
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Wadell 2001.

Methods	Blinding: single; Randomisation: yes blocks of men and women; Withdrawals: none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad= 2/5 Pedro=7/10
Participants	Country Sweden: N=20; <75 years; FEV1<70%; **PaO2>= 8kPa; SaO2,90% during corridor 6 minute walk test;
Interventions	5 L/min O2 or air via cylinder and nasal specs
Outcomes
Notes	Author contacted: no response This was a 8 week study of physical training with and without oxygen. Data for acute responses at baseline is presented as medians (range) for oxygen training group and air training group extracted separately
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

Woodcock 1981.

Methods	Blinding: double; Randomisation: no details; Withdrawals: none; Baseline characteristics: comparable; Point/ variability estimates for between group analysis: yes Power calculation: none Jadad= 3/5 Pedro=9/10
Participants	Country England: N=10; 9M; FEV1 0.71 (0.29)Lmin **PaO2 9.65 (1.51) PaCO2 4.55 (0.6)
Interventions	4 L/min O2 or air via cylinder and nasal specs
Outcomes
Notes	Author not contacted 16/12/04
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Information not available (Cochrane Grade B)

*Studies in which some patients likely met the criteria for LTOT (mean PaO2 < 7.3kPa or 55mmHg) 
 **Studies in which patients did not/likely did not meet the criteria for LTOT 
 *** no information given re baseline oxygen status

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Arlati 1988	Room air not true placebo
Barach 1966	Not a RCT
Bower 1988	No placebo group
Brambilla 1985	Not a RCT
Braun 1992	No placebo group
Corriveau 1989	Order of gas not randomised
Cotes 1956	Room air not true placebo
Cotes 1963	Mixed population
Cuvelier 2002	Order of gas not randomised Room air not true placebo
Eaton 2001	Not a RCT
Emtner 2002	Order of gas not randomised (in assessment acute responses to exercise)
Guyatt 2001	Not a RCT
Hargarty 1997	No placebo group
Jolly 2001	No specific oxygen flow rate
Lacasse 2003	Not a RCT
Lane 1987	Not a RCT
Lilker 1975	Not short term ambulatory assessment study
Lock 1991	Mixed population
Lock 1992	Mixed population No placebo group
Patessio 1996	Exercise training study
Pierce 1965	Order of gas not randomised Room air not true placebo
Revill 2000	Room air not true placebo
Roberts 1996	Room air is not true placebo
Rooyackers 1997	Order of gas not randomised (in assessment of acute responses to exercise) Room air not true placebo
Scano 1982	Not a RCT
Tiep 2002	No true placebo
Vergeret 1989
Waterhouse 1983	No equivalent control group
Wedzicha 1996	Not a RCT

Contributions of authors

The title for the protocol was conceived by J. Bradley and B. O'Neill in collaboration with the Cochrane Airways Group.

J. Bradley, B. O'Neill and Felix Ram designed and assisted in writing the protocol.

J. Bradley and B. O'Neill designed and wrote the review, and updated it in June 2005

T Lasserson acted as supervisor for the review.

J. Bradley acts as guarantor of the review.

Sources of support

Internal sources

• No sources of support supplied

External sources

• This review was supported by the Northern Ireland Reserach & Development Office Fellowship Scheme, Ireland.

Declarations of interest

There are no known conflicts of interest.

Edited (no change to conclusions)