Background Prehospital high concentration oxygen therapy leads to worse clinical outcomes in patients presenting with acute exacerbations of chronic obstructive pulmonary disease (AECOPD). Less is known about the risks of hypoxaemia despite oxygen treatment. Current respiratory and ambulance guidelines recommend titration of supplemental oxygen to a target oxygen saturation range of 88%–92%.
Aim To explore the association between PaO2 and risk of serious adverse clinical outcomes in AECOPD.
Methods A retrospective review of consecutive patients presenting via ambulance to the Wellington Regional Hospital Emergency Department with AECOPD between June 2005 and January 2008. Patients with an arterial blood gas taken within 4 h of triage were included in the study and were categorised as hypoxaemic (PaO2<60 mm Hg), normoxaemic (PaO2 60–100 mm Hg) or hyperoxaemic (PaO2>100 mm Hg). Serious adverse outcome was defined as a composite of hypercapnic respiratory failure, assisted ventilation or inpatient death. Multivariate logistic regression analysis examined the association between PaO2 category and the composite outcome.
Results Of the 680 patients presenting with AECOPD in the review period, 254 presentations in 180 patients had data suitable for analysis. Hyperoxaemia occurred in 61/254 (24%) presentations and was strongly associated with serious adverse outcome compared with normoxaemia (OR 9.17, 95% CI 4.08 to 20.6). Hypoxaemia was also associated with an increased risk of serious adverse outcome compared with normoxaemia (OR 2.16, 95% CI 1.11 to 4.20). Compared with the recommended target oxygen saturation range of 88%–92%, the risk of a serious adverse outcome was increased in both the <88% group (OR 2.0, 95% CI 1.03 to 3.80) and the >96% group (OR 2.37, 95% CI 1.34 to 4.20).
Conclusions In patients presenting via ambulance to the Emergency Department with AECOPD, serious adverse clinical outcomes are associated with both hypoxaemia and hyperoxaemia. These data provide further support for the principle of titrating supplemental oxygen therapy to target oxygen saturations.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
High concentration oxygen therapy is often administered to patients with acute exacerbations of chronic obstructive pulmonary disease (AECOPD) en route to hospital despite evidence that this approach is associated with worse outcomes.1–9 It has been suggested that this practice may be driven by the belief that hypoxaemia must be avoided at all costs, the desire to relieve breathlessness, the difficulties in differentiating chronic obstructive pulmonary disease (COPD) from other cardiorespiratory conditions and lack of recognition of the risks of hyperoxaemia.9 Current guidelines recommend that oxygen should be titrated to achieve oxygen saturations of 88%–92%.10 The effect of this approach was investigated in a recent randomised controlled trial,8 in which high concentration oxygen therapy caused a twofold increased risk of mortality when compared with a regime in which oxygen was titrated to achieve oxygen saturations of 88%–92%. In this study, high concentration oxygen resulted in a 34 mm Hg increase in PaCO2 and 0.12 reduction in pH compared with titrated oxygen therapy in patients with confirmed COPD who received prehospital treatment as per protocol and had an arterial blood gas (ABG) measured within 30 min of presentation to the Emergency Department (ED). These findings provide strong evidence that the routine use of high concentration oxygen is now contraindicated in AECOPD and that titrated oxygen therapy represents optimal treatment.11 However, they also raise the question of the relative risks of hypoxaemia and hyperoxaemia in this clinical situation.
This issue was partially addressed in our previous review6 which found an association between prehospital high flow oxygen administration and serious adverse outcomes. Although higher oxygen saturations at ambulance pick up were associated with better clinical outcomes, higher PaO2 following presentation to the ED was associated with worse clinical outcomes. To further explore the relationship between PaO2 at ED presentation and clinical outcome, we extended the original review period by a further 12 months and only included patients who had an ABG within 4 h of ED triage. Our hypothesis was that the risk of a serious adverse outcome, as measured by a composite of hypercapnic respiratory failure, assisted ventilation or death, would be associated with both hypoxaemia and hyperoxaemia.
We performed a retrospective review of patients presenting with AECOPD to the Wellington Regional Hospital ED via ambulance. This represents an extension of our previous review by a period of 12 months. Patients were identified by the Decision Support Unit which provided a list of all patients who were admitted to Wellington Regional Hospital consecutively between June 2005 and January 2008, and had a primary discharge diagnosis coded as COPD (ICD Code J440 or J441). In accordance with Capital and Coast District Health Board policy, this clinical review was registered prospectively (MCC2011-01) and patient confidentiality was maintained throughout.
Data collection and definitions
Patients with an ABG sample taken within 4 h of triage in the ED were included in the analysis. ABG results documented as venous samples were excluded. The paper records reviewed included the ambulance, ED and hospital clinical records. Patients were excluded if the ambulance record was missing or if the patient was transferred from another hospital. If patients presented on more than one occasion during the review period, only the first presentation was analysed when describing the patient population. All hospital presentations were included when describing the outcome variables.
PaO2 values were divided into predetermined categories of hypoxaemia (PaO2<60 mm Hg), normoxaemia (PaO2 60–100 mm Hg) or hyperoxaemia (PaO2>100 mm Hg) for comparative analysis.
For the administration of oxygen during ambulance transfer, high concentration oxygen was defined as documented flow ≥3 l/min via any device or administration of oxygen via mask. If oxygen therapy was documented but no device or flow rate recorded, then oxygen administration was not categorised as low or high concentration. All other scenarios were considered to represent low concentration oxygen administration.
Outcome measures and statistical methods
The primary outcome measure was a composite of hypercapnic respiratory failure (defined as a PaCO2≥45 mm Hg and a pH<7.35), administration of assisted ventilation (either invasive or non-invasive positive pressure ventilation) or inpatient death.6 This composite outcome was used as our primary outcome measure because we anticipated the death rate for the review period would be too low to provide sufficient power to detect associations. In secondary analyses, each component was analysed separately to examine the association between PaO2 and each part of the composite adverse outcome measure.
Logistic regression was used to calculate ORs for the association between PaO2 and the composite dichotomous serious adverse outcome as well as each component of the composite outcome. The reference level for analyses of the PaO2 categories was the normoxaemia group (PaO2 60–100 mm Hg). The following prespecified variables were included in the multivariate analysis as potential confounding markers of acute or chronic disease severity: the first recorded respiratory rate, previous documented hypercapnic respiratory failure, assisted ventilation (invasive or non-invasive) and use of home oxygen. Transit time in the ambulance, defined as the time between patient pick up and ED triage, and the time between triage and ABG sampling were also included in the multivariate analysis. The univariate and multivariate analyses include all presentations and general estimating equations were used to account for correlated measurements on individual patients.
In a post hoc analysis, the OR for the oxygen saturation (in groups defined by saturations <88%, 88%–92%, 93%–96% and >96%) and the composite serious adverse outcome were calculated, using the reference group 88%–92%, representing the target oxygen saturation recommended in the treatment of AECOPD.
A general additive model using a spline smoother was used to fit a curve to the relationship between ABG saturations and PaO2 for those patients who had an ABG within 1 h of triage, in order to predict the oxygen saturation that corresponded to a PaO2 of 60 mm Hg in our sample.
There were 680 presentations with AECOPD in the review period of which 254 presentations were included in the analysis, comprising 180 individual patients. Of the 426 presentations excluded, 223 did not receive an ABG within 4 h of triage, 163 did not arrive by ambulance or were transfers from another hospital and 40 had incomplete medical records (figure 1).
The characteristics of the 180 patients, including acute and chronic severity markers, are shown in table 1. The patients had severe COPD with one in three patients having had a previous episode of hypercapnic respiratory failure, one in five patients on domiciliary oxygen and one in six having received previous assisted ventilation. The mean room air oxygen saturation on ambulance pick up was 84.4%; however, this was only documented in 20% of cases. Glasgow Coma Score on ambulance arrival was 15/15 in 92% of documented cases (229/250).
The mean (SD) ambulance transit time was 51.5 (22.8) min. Prehospital oxygen flow rate was documented for 76% of patients, with a mean (SD) flow rate administered in the ambulance of 6.7 (2.9) l/min. High concentration oxygen was administered to 80% of patients. At least one nebuliser was administered in 60% of ambulance transfers. Compressed air to drive nebulisers was not available in ambulances during the review period. If the use of oxygen to drive nebulisers was included in the definition of high concentration oxygen, the rate of prehospital high concentration oxygen administration was 93%.
The mean (SD) time between triage and the first ABG was 72 (59) min. An ABG was taken within an hour of ED triage in 142/254 (56%) presentations. The overall mean PaO2 was 85 mm Hg with a range from 32 to 300 mm Hg. Overall, 83/254 (33%) presentations were hypoxaemic, 110/254 (43%) normoxaemic and 61/254 (24%) hyperoxaemic. The hypoxaemic threshold (PaO2 60 mm Hg) corresponded to an oxygen saturation (measured on ABG) of 88% in our sample of patients. The mean, median and range of PaO2, PaCO2 and pH values corresponding with the hypoxaemia, normoxaemia and hyperoxaemia groups are shown in table 2. Hypercapnia and acidosis were more pronounced in the hyperoxaemia group.
A serious adverse outcome (comprising hypercapnic respiratory failure, assisted ventilation or death) occurred in 108/254 (42.5%) presentations. Hypercapnic respiratory failure occurred in 83 (32.7%) presentations, assisted ventilation was administered in 75 (29.5%) presentations and 17 (6.7%) presentations were associated with death in hospital. Of the patients who required assisted ventilation, all 75 received non-invasive ventilation (NIV) and none were intubated. Assisted ventilation was administered to 63.9% of patients who met the criteria for respiratory failure and 12.9% of patients who did not meet the criteria for respiratory failure. The proportion of patients who had a serious adverse outcome in each of the three oxygen groups is shown in table 3.
In multivariate analysis, hyperoxaemia was strongly associated with a serious adverse outcome when compared with normoxaemia (OR 9.17, 95% CI 4.08 to 20.6; table 4). Hypoxaemia was also associated with an increased risk of a serious adverse outcome when compared with normoxaemia (OR 2.16, 95% CI 1.11 to 4.20; table 4). In a sensitivity analysis including only first presentations during the review period (n=180), the OR for hyperoxaemia and hypoxaemia were 8.96 (95% CI 3.37 to 23.8) and 3.31 (95% CI 1.40 to 7.82), respectively.
As shown in table 4, there were statistically significant associations between serious adverse outcome and previous hypercapnic respiratory failure, current home oxygen use and ‘triage to ABG time’ on both univariate and multivariate analyses.
The analyses of the primary composite outcome broken down to its three components are shown in table 5. Hyperoxaemia was associated with a statistically significant increased risk for assisted ventilation and hypercapnic respiratory failure. For the other components, although the point estimates for association of hypoxaemia and hyperoxaemia suggested an increased risk, none of the associations were statistically significant with wide CIs.
Multivariate analysis was used for all OR calculations including the following variables: First respiratory rate, previous hypercapnic respiratory failure, previous assisted ventilation, home oxygen use, pick up to triage time and triage to ABG time.
The proportion of subjects who had a serious adverse outcome according to four defined oxygen saturation ranges is shown in table 6. There was no difference in the risk of a serious adverse outcome between the 88%–92% and 93%–96% oxygen saturation groups. Compared with the 88%–92% group, the risk of a serious adverse outcome was increased in both the <88% group (OR 2.0, 95% CI 1.03 to 3.80, p<0.041) and the >96% group (OR 2.37, 95% CI 1.34 to 4.20, p<0.003).
Hyperoxaemia following arrival to the ED occurred in a quarter of patients presenting via ambulance with AECOPD and was associated with a ninefold increased risk of a composite serious adverse outcome when compared with normoxaemia. Hypoxaemia was also significantly associated with a poor clinical outcome, although not as strongly as for hyperoxaemia. These findings support the principle of titrating oxygen therapy to avoid both hypoxaemia and hyperoxaemia in AECOPD.
There are a number of methodological issues relevant to the interpretation of our findings. Our analysis was based on PaO2 rather than oxygen saturations in which a cut-off point to define hyperoxaemia would have been problematic. Standard levels of PaO2 were used to define hypoxaemia (<60 mm Hg) and hyperoxaemia (>100 mm Hg).10 ,12 In our cohort, the PaO2 of 60 mm Hg was equivalent to an oxygen saturation of 88%, as measured from the same ABG test, which is the lower limit of the recommended 88%–92% target range.8 This differs from an expected value (90.6%) calculated via the Severinghaus equation,13 due to the concomitant acidosis (mean pH 7.34) and associated rightwards shift of the oxyhaemoglobin dissociation curve.14 The mean time between triage and the first ABG was 72 min, and as a result the findings reflect the influence of oxygen therapy administered both prehospital and following arrival to the ED.
The AECOPD inclusion criteria were based on the diagnosis at discharge, and as a result are likely to be accurate with all admission information available to the inpatient team. A proportion of patients were readmitted with an AECOPD during the review period and to account for the repeated measurements on the same patients that could lead to underestimation of the CIs, the multivariate analysis used general estimating equations in the logistic regression to adjust for correlated measurements. In a sensitivity analysis which included only first presentations, the ORs for a serious adverse outcome were of a similar magnitude for both hyperoxaemia and hypoxaemia. The inhospital death rate of 6.7% was comparable with the inpatient death rate of 7.7% reported in a recent large UK audit.7
In the analysis of the components of the composite outcome, hypercapnic respiratory failure and assisted ventilation remained significantly associated with hyperoxaemia. As expected, we lacked statistical power to detect an association between death and hypoxaemia or hyperoxaemia, for which the point estimates were consistent with an association, but the CIs were wide. We acknowledge that there is a degree of overlap between respiratory failure and the need for assisted ventilation; however, some subjects received NIV on clinical grounds alone without the availability of an ABG which is an important outcome variable to capture.
Although this retrospective study found a strong association between hyperoxaemia and poor clinical outcome, it was not able to demonstrate causality. It was also limited by reliance on adequate documentation. For example, there was a low rate of documenting initial oxygen saturation on room air by ambulance staff, a measure which would have been a useful acute severity marker for inclusion in the multivariate analysis. Ambulance transit time was included in the multivariate analysis with the hypothesis that a longer journey would correspond to longer exposure to hyperoxaemia and subsequent adverse outcome. This was not demonstrated in our data. ABG measurements were not made in about a half of patients presenting by ambulance with AECOPD, which limited the power of the study.
Our findings of an association between serious adverse outcomes and both hypoxaemia and hyperoxaemia provide support for the British Thoracic Society guideline recommendations to titrate oxygen therapy in AECOPD to avoid both hypoxaemia and hyperoxaemia.10 Increased risks were observed in the groups defined by an oxygen saturation <88% and >96%, compared with the recommended 88%–92% target range. The risk was similar for the 88%–92% and 93%–96% groups. The analysis of outcomes based on ABG measurements suggests that the increased risk observed in the >96% group was due to many of these patients having a supranormal PaO2. However, these findings were based on a post hoc analysis and our observational study was retrospective and involved small numbers in each of these two groups. Randomised controlled trials would be required to determine whether there is evidence to broaden the current oxygen saturation target range recommended by the British Thoracic Society guideline. In the interim the current target oxygen saturation range of 88%–92% remains the gold standard of management, supported by level 1 evidence.8 Our finding that around a third of patients with AECOPD had a PaO2>100 mm Hg suggests that there remains a propensity to administer high concentration oxygen rather than titrated oxygen therapy according to British Thoracic Society guidelines.
Titrating supplemental oxygen therapy to maintain target oxygen saturations is also recommended in other acute respiratory conditions.10 In support of this approach, oxygen therapy titrated to achieve oxygen saturations of 93%–96% have been shown to result in lower PaCO2 levels than high concentration oxygen in both asthma15 and community-acquired pneumonia.16
In conclusion, we have demonstrated an association between both hyperoxaemia and hypoxaemia and serious adverse clinical outcomes in patients presenting with AECOPD. These findings illustrate the balance that is required in clinical practice with the administration of oxygen therapy in AECOPD. This concept of titrating oxygen therapy to relieve hypoxaemia without causing hyperoxaemia is promoted in respiratory guidelines that recommend a target oxygen saturation range of 88%–92% in patients at risk of hypercapnic respiratory failure.10 It is also supported by the only randomised controlled trial of oxygen therapy in AECOPD in which oxygen titrated to achieve saturations of 88%–92% resulted in a twofold lower death rate than a high concentration oxygen regime.8
Both hypoxaemia and hyperoxaemia are associated with serious adverse clinical outcomes in acute exacerbations of chronic obstructive pulmonary disease (AECOPD).
These findings demonstrate the critical importance of accurate titration of oxygen in AECOPD to avoid both hypoxaemia and hyperoxaemia.
Current research questions
What is the optimal target oxygen saturation in acute exacerbations of chronic obstructive pulmonary disease (eg, 88%–92%, 85%–90% or 90%–95%)?
What is the optimal oxygen saturation target in other respiratory disorders in which hypercapnia is known to occur?
Wijesinghe M, Perrin K, Healy B, et al. Pre-hospital oxygen therapy in acute exacerbations of chronic obstructive pulmonary disease. Intern Med J 2011;41:618–22.
Austin MA, Wills KE, Blizzard L, et al. Effect of high flow oxygen on mortality in chronic obstructive pulmonary disease patients in prehospital setting: randomised controlled trial. BMJ 2010;341:c5462
O'Driscoll BR, Howard LS, Davison AG. BTS guideline for emergency oxygen use in adult patients. Thorax 2008;63 Suppl 6:vi1–68
Contributors LC, RB and KP: conception, design, analysis and interpretation of data, manuscript write up; MW: design, analysis and interpretation of data, manuscript write up; JP: interpretation of data, manuscript write up.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.