European Respiratory Society


It is unknown whether acute response to ambulatory oxygen (O2) predicts longer term improvement in health-related quality of life (HRQL) in chronic obstructive pulmonary disease (COPD) patients.

The aims of this study were 1) to assess the short-term clinical impact, as determined by HRQL, of ambulatory O2 in a 12-week, double-blind, randomised crossover study of O2 (versus cylinder compressed air) of dyspnoeic but not chronically hypoxic COPD patients with exertional desaturation ≤88% (n=41), and 2) to determine whether either baseline characteristics or acute response to O2 predicts short-term (12 weeks) response. Primary outcome measures were Chronic Respiratory Questionnaire (CRQ), Hospital Anxiety and Depression scale and the short form (SF)-36.

Improvements were seen in all domains of the CRQ for cylinder O2 compared with cylinder air. Significant improvements were also noted in anxiety and depression and in certain domains of the SF-36. There were 28 (68%) acute responders to cylinder O2 (defined as increase in 6‐min walk ≥54 m or decrease in post-Borg dyspnoea ≥1) and 23 (56%) short-term responders (defined as clinically significant improvement in CRQ). However, acute and short-term responses were not correlated with no predictors of short-term response identified. At study completion, 14 (41%) of acute or short-term responders did not want to continue therapy, with 11 citing poor acceptability or tolerability.

Short-term ambulatory oxygen is associated with significant improvements in health-related quality of life. These benefits cannot be predicted by baseline characteristics or acute response. Despite acute or short-term response, a substantial proportion of patients declined ambulatory oxygen.

This work was fully funded by the Health Research Council of New Zealand.

There are a considerable number of patients, not currently satisfying criteria for long-term oxygen (O2) therapy (LTOT) 1, 2, who demonstrate significant exertional desaturation. 3, 4. These patients may benefit from the prescription of ambulatory O2. The American Thoracic Society (ATS) guidelines for domicilary O2 state that “ambulatory O2 should be prescribed to patients normoxaemic at rest with evidence of exertional desaturation to 88% or less” 5. More recently, the Royal College of Physicians report on Domicilary Oxygen Services stated that “ambulatory O2 should be prescribed when there is evidence of significant exercise desaturation and improvement of at least 10% in either walking distance or dyspnoea score when walking with cylinder O2 compared with cylinder air” 6.

However, guidelines for ambulatory O2 are limited and not based on randomised controlled studies. The majority of ambulatory O2 studies are acute, concentrating on physiological measures with variable results 712. The longer term benefits of ambulatory O2 remain uncertain. Hence, there is a need for randomised controlled studies of the domicilary use of ambulatory O2, which incorporate not only traditional physiological measures but comprehensive health-related quality of life (HRQL) outcome measures. This should allow a more realistic assessment of the balance between practical benefits which may ensue from the use of ambulatory O2 versus the drawbacks of such an intervention. Only with such an evidence base, can guidelines be developed.

The aims of this study were 1) to assess the short-term clinical impact, as determined by HRQL, of ambulatory O2 in patients with severe chronic obstructive pulmonary disease (COPD) and significant exercise desaturation, who did not fulfil criteria for LTOT, and 2) to determine whether either baseline characteristics or acute response predicts short-term response.


Study design and patient recruitment

Figure 1 outlines the study design and patient recruitment details. A 12-week, randomised, double-blind, crossover study of cylinder air and O2 was performed. Eligible patients were identified consecutively from outpatient clinics and pulmonary rehabilitation referrals to Green Lane Hospital, Auckland, NZ. Inclusion criteria were COPD as defined by ATS criteria 5, exertional dyspnoea impacting on daily activities, not fulfilling criteria for LTOT (resting oxygen tension in arterial blood (Pa,O2) ≥7.3 kPa on air), exertional desaturation (arterial oxygen saturation (Sa,O2) ≤88% on air; ATS criteria for ambulatory O2), exsmoker and clinically stable for over 2 months with standard optimal medical care. All patients were required to have completed a formal 6‐week pulmonary rehabilitation program 13. Patients with important comorbidities (e.g. limiting angina or significant musculoskeletal disability) were excluded. Approval was granted by the North Health Ethics Committee with signed informed consent obtained from all patients.

Fig. 1.—

Study design and patient recruitment details.

Physiological measures

Forced expiratory volume in one second and forced vital capacity (R‐Model; Vitalograph Limited, Buckingham, UK) were performed at all visits according to ATS standards and expressed as % of predicted using the European Coal and Steel Community prediction equations. Resting arterial blood gases on air (Quik ABG arterial blood sampler; Marquest Medical Products, Inc., Englewood, CO, USA) were performed at all visits other than at crossover. The 6‐min walk test (6MWT) was performed in triplicate on room air prior to the baseline visit to ensure repeatability 14, 15. At baseline, crossover and study completion, patients performed the 6MWT carrying cylinder gas (the majority with a back pack although some patients chose a shoulder bag). Cylinder air and O2 was delivered in double-blind random order using a standard flow rate (4 L·min−1) with each test separated by 20–30 min rest. Cutaneous oximetry (Nellcor N‐20PA; Puritan Bennett Inc., Pleasanton, CA, USA) was obtained with a finger probe. The following measures were recorded; resting, 2‐min and 6‐min Sa,O2, walk distance, pre- and postwalk modified Borg dyspnoea scores 16. Adequate correction of exertional desaturation was defined as Sa,O2 >90% breathing cylinder O2.

Health-related quality of life measures

Chronic respiratory questionnaire

The chronic respiratory questionnaire (CRQ) 17, 18 is a well-validated, repeatable and responsive disease-specific instrument with four domains; dyspnoea, fatigue, emotional function and mastery.

Hospital anxiety and depression scale

The hospital anxiety and depression scale (HAD) 19 scale is a reliable self-assessment scale for use in the physically ill, incorporating 14 items with separate scores for anxiety and depression: the subscales are also valid measures of the severity of emotional disorder.

Short form-36 health survey questionnaire

The short form (SF)-36 health survey questionnaire 20 is a generic HRQL tool with an eight-dimensional profile which has been demonstrated to be reliable and sensitive to low levels of morbidity.

Domicilary programme

Patients were randomly assigned in a double-blinded manner to cylinder air or O2 (light weight aluminum: standard 2,000–2,200 psi fill: 145 L, weight 2.04 kg (4.5 lbs) fitted with a conserving demand gas delivery system (Oxymatic; Chad Therapeutics, Inc., Chatsworth, CA, USA). All cylinders were painted pink, prefilled with either air or O2 and identifiable only by a unique cylinder number, ensuring blinding of both participants and observers. Instructions were; to use a flow rate of 4 L·min−1 intranasally for any activity during which they would normally experience dyspnoea. After 6 weeks, patients crossed over to the other cylinder gas for a further 6 weeks. Patients received comprehensive practical and written education in the care and usage of their ambulatory gas. Cylinder use was quantified by weighing pre- and postuse with calibrated scales, accepted as an accurate, effective method of quantifying an amount of gas. Self-reported cylinder use was also recorded.


The following definitions were based on published criteria for minimal clinically important differences 14, 16, 18.

Acute response to cylinder O2: 1) an improvement of ≥54 m in the 6MWT on cylinder O2 compared to cylinder air or 2) a decrease in post 6MWT Borg dyspnoea score of ≥1.

Short-term response to cylinder O2: an improvement in dyspnoea (≥3), fatigue (≥2), emotional function (≥4) or mastery (≥2) domains of the CRQ using the validated definition of a clinically important difference.

At study completion patients were asked whether they wished to be considered for the clinical prescription of ambulatory O2. If their response was no, they were asked to state their reason(s).

Statistical methods

A mixed model approach to crossover trials was employed, which used information from all patients, including those who did not complete both time periods. Treatment and order of treatment (to exclude a carryover effect) were included in the model with the patient as a random effect. Age, baseline Pa,O2 and carbon dioxide tension in arterial blood were also included as was the level of the outcome at the baseline visit. A p‐value <0.05 was considered statistically significant. Logistic regression was used to investigate factors which might predict short-term response to cylinder O2. Included in the model were: age, sex, baseline Borg dyspnoea score pre-6MWT, difference in Borg pre- and post-6MWT on cylinder O2, difference in distance walked on cylinder O2 and cylinder air, change in Sa,O2 pre‐2‐min and post-6MWT on air, correction of exertional desaturation, cylinder O2 use (weights) and treatment order.


Patient characteristics

A total of 50 patients were available for randomisation at baseline. They had severe COPD, did not fulfil for LTOT and demonstrated significant exertional desaturation and dyspnoea (table 1). There were nine withdrawals (comorbidities n=3, personal reasons n=6); results from the remaining 41 patients are reported. Clinical stability over the 2‐month run-in period was established with no significant differences between resting Pa,O2 on air, 6MWT and resting Borg dyspnoea score (run-in and baseline visit).

View this table:
Table 1—

Patient characteristics including acute responses to cylinder gas

Acute and short-term responses in functional indices

Acute functional responses to cylinder O2 and cylinder air (6MWT and post-Borg dyspnoea scores) are shown in table 1 with significant improvement in 6MWT distance and post-6MWT Borg dyspnoea score observed with cylinder O2. Over the study period, there was no difference in Borg dyspnoea score post 6MWT on cylinder O2 depending on whether they had been assigned cylinder air or cylinder O2 in the preceding 6 weeks (p=0.4). Likewise, there was no difference in the distance walked on O2 depending on whether the patient had been using cylinder air or O2 (p=0.9).

Short-term responses in health-related quality of life

Group analysis of mean difference between cylinder O2 and cylinder air demonstrated significant improvements in all domains of the disease-specific CRQ, both domains of the HAD and several domains of the generic SF-36, for cylinder O2 compared with cylinder air (table 2).

View this table:
Table 2—

Health‐related quality of life (HRQL) measures for 41 patients

Cylinder gas use

A treatment order effect was apparent in that patients who were randomised to cylinder O2 first had higher weekly use 12.25±2.06 h compared to cylinder air 6.95±2.12 h (p=0.004). There was no change in the pattern of use over the weeks (p=0.3).

Responders to cylinder oxygen

Using validated definitions of a minimally clinically important difference, the authors found 28 (68%) acute responders to cylinder O2 and 23 (56%) short-term responders; 17 (41%) demonstrating both acute and short-term responses (table 3). Only seven (17%) patients were defined as neither acute nor short-term responders.

View this table:
Table 3—

Acute and short-term responders to cylinder oxygen

Predictors of short-term response to cylinder oxygen

No predictors of short-term response were identified. Baseline characteristics (Sa,O2 post-6MWT, Sa,O2 at 2 min, change in Sa,O2 pre- and post-6MWT, Borg dyspnoea score pre- and post-6MWT) did not predict short-term response. Despite a standard prescription of relatively high flow (4 L·min−1), correction of exercise desaturation was only achieved in 26 (54%) of patients and this factor also did not predict short-term response. Quantity of O2, as measured by cylinder weights (correlation between hours of use and cylinder weights, r=0.7) did not predict short-term response. Neither was an association seen between acute response and short-term improvement in HRQL (p=0.4).

Clinical preference for ambulatory oxygen

At study completion, 14 (41%) of patients, objectively defined as having either an acute or short-term response, did not want to be considered for the clinical provision of ambulatory O2, with 11 citing poor acceptability or tolerability. Of those, 20 patients that elected to continue with ambulatory O2 therapy, six (30%) were defined as acute responders only, two (10%) were short-term responders only and 12 (60%) were both acute and short-term responders.


The major finding of this randomised, controlled, double-blind study was improvement in HRQL with ambulatory O2 compared to cylinder air in dyspnoeic COPD subjects with exertional desaturation. This short-term benefit could not be predicted by either baseline characteristics or acute beneficial response to cylinder O2.

The current authors elected to study patients with significant exercise desaturation, consistent with current ATS recommendations for ambulatory O2 5. Studies of ambulatory O2 have generally been very small and concentrated on acute physiological responses 712. The negative impact of COPD on HRQL is substantial 2123, and the use of HRQL as an appropriate primary outcome measure is increasingly acknowledged. The determinants of HRQL include a complex interplay of functional disability and psychosocial variables. This study was specifically designed to examine the effect of cylinder O2 versus cylinder air on HRQL allowing a potentially powerful cylinder placebo effect to be controlled. Significant improvements between cylinder O2 and cylinder air in all domains of the disease-specific CRQ and several domains of the generic SF-36 were demonstrated.

Dyspnoea is arguably the single most distressing symptom for COPD patients. This study found small but significant improvements in the CRQ dyspnoea domain with cylinder O2 compared with cylinder air, consistent with a specific O2 effect and not supportive of alternative mechanisms, such as nonspecific stimulation of nasal receptors 24. The mechanism of any benefit remains uncertain, but is likely to be multifactorial. There is evidence that supplemental O2 reduces minute ventilation and respiratory rate for a given workload, thus, delaying respiratory muscle fatigue 11. Improvements in dyspnoea might relate to reduced ventilatory demand in association with reduced blood lactate 12.

However, supplemental O2 for exercise training in pulmonary rehabilitation has not to date been shown to enhance field testing outcomes 25. Although the prescription of ambulatory O2 is not strictly analogous the current authors were also unable to demonstrate any change in functional capacity, arguing against a physiological benefit. However, this may relate to factors such as amount of use or choice of outcome measure. Each field test has strengths and weaknesses. The 6MWT was chosen in this study as it is the most widely employed field test, repeatable and responsive to change with a validated minimal clinically important difference 14. However, the current authors acknowledge there is increasing support for endurance tests, as they are more sensitive to change 26. Recovery time, as an outcome measure, could also be considered in future trials 27.

A specific measure of anxiety and depression in recognition of their prevalence in COPD was included in this study 21, 23. Psychological factors may affect individual perception of dyspnoea and thus impact on HRQL. A significant decrease in anxiety and to a lesser extent depression, was demonstrated with ambulatory O2, an observation strengthened by significant improvements noted in the emotion domains of both the CRQ and SF-36. Notably emotional function and mastery were the CRQ domains where the larger impact was apparent. Mastery is defined as the degree to which a patient feels control over a disease and therefore might be expected to be a more robust and clinically relevant measure in this setting than traditional physiological indices.

The improvement in HRQL with cylinder O2, albeit significant, was modest in terms of clinical significance. Indeed, in comparison with room air the benefits may become marginal as it seems likely that there is a significant negative impact related to carrying the cylinder. However, this requires a different study design, e.g. a parallel group study incorporating a room air control arm, involving larger patient numbers. These studies are labour intensive which may explain in part the dearth of long-term clinical studies. In this study, only relatively modest group differences were demonstrated compared with improvements attained in some rehabilitation programs 28, 29. However, ambulatory O2 afforded some patients measurable benefit of clinical relevance over a relatively short period of time and this effect was over and beyond any gains from an optimal therapeutic regimen including a formal period of pulmonary rehabilitation. In a disease with limited therapeutic options this is of considerable clinical importance. This study was relatively short and it remains to be determined whether any response is maintained or further incremental response accrues. Long-term studies of ambulatory O2 are warranted, incorporating additional outcome measures more critically examining activities of daily living.

Clinical responders were identified using a priori validated definitions of clinically important differences. Notably, despite acute improvement in functional indices such as walk distance or in short-term HRQL, a significant proportion elected to decline ambulatory O2. This suggests that they did not feel the clinical benefit afforded was of significant magnitude to overcome the perceived negative aspects of cylinder gas use. This occurred despite careful patient selection, absence of significant comorbidities, comprehensive patient education and individual attention to the carrier device. Intuitively, tolerability would be expected to have been captured by a measure of HRQL, this did not necessarily appear to be the case. A further point was the apparent treatment order effect. Patients randomised to cylinder air for the first 6 weeks used less cylinder O2 in the second 6‐week period. Conceivably, patients who perceived no benefit with cylinder air weighted the tolerability factor higher, which then influenced their approach to the second 6‐week period. The advent of smaller lighter cylinders may result in greater improvements in HRQL.

There is only one other randomised, short-term study of cylinder O2. This reported no significant difference in HRQL between cylinder air or O2 30. Ambulatory O2 remains widely advocated 5, 6 despite this negative result. The difference between the two studies may lie partly in study numbers with only 26 patients completing the study by McDonald et al. 30. This study had larger numbers, employed a minimum run-in period of 2 months to ensure maximal clinical stability and to minimise potential confounders and drop-outs. All patients had completed formal pulmonary rehabilitation, another potential source of confounding. McDonald et al. 30 employed very heavy cylinders (5 kg) compared to the considerably lighter cylinders (2.04 kg) in this study. Whilst this should have been controlled for in the study design, it is possible that the negative effect of such a weight overwhelmed any small benefit from O2. It is also possible that the patient group in the study by McDonald et al. 30 differed in some way which made them less likely to derive benefit from ambulatory O2. The obvious difference in this respect was in degree of exercise desaturation. In this study the ATS criteria (exertional desaturation Sa,O2 ≤ 88% on air) were employed 5. In the other study 30, exercise desaturation was not one of the selection criteria and was generally only mild. Conversely, exercise desaturation per se has not been found previously to predict acute response to cylinder O2.

No predictors of short-term response to cylinder O2 were identified. A number of baseline characteristics including dyspnoea and degree of exertional desaturation were examined; none were able to predict short-term response. Neither was an association seen between acute response and short-term improvement in HRQL. This is not necessarily surprising, bearing in mind the acute response has been traditionally defined by improvement in functional capacity and dyspnoea whereas short-term benefit was captured by measures of HRQL, particularly mastery and emotional function. Arguably such improvements in HRQL are of greater clinical relevance to the patients. To simplify study design and maximise the proportion of patients with adequate correction of exertional desaturation a maximal tolerable flow rate of 4 L·min−1 was chosen here. Despite this, exertional Sa,O2 was adequately corrected in barely one-half of the study population. As discussed, correction of Sa,O2 was not predictive of short-term response. Previous work has generally only examined acute response but a similar lack of predictors has been reported 712.

In conclusion the results in this study have demonstrated that the short-term use of ambulatory oxygen is associated with significant improvements in health-related quality of life in chronic obstructive pulmonary disease patients who do not fulfil criteria for long-term oxygen therapy but who demonstrate significant exertional desaturation. How might these results influence clinical practice? Ambulatory oxygen for this group of patients appears generally acknowledged internationally, particularly in the USA, although this has not previously been evidence-based. This study provides the first evidence supporting these clinical recommendations. However, there are lingering problems. The improvements were modest, and the authors remain disadvantaged in that they are unable to predict which patients may benefit from ambulatory oxygen. Finally, poor tolerability negatively influenced decisions to proceed with ambulatory oxygen in a substantial proportion of patients. The Royal College of Physician guidelines go some way to articulating the issues, however their criteria are based on acute response 6. The results of this study would challenge this. Formal assessment of short-term HRQL response should be mandatory. It is to be hoped that with the advent of further studies, more comprehensive evidence-based guidelines may be developed for the provision of ambulatory oxygen.


The authors would like to thank E. Robinson for assistance with statistical analyses.

  • Received December 4, 2001.
  • Accepted February 27, 2002.


View Abstract