Oxygen titration therapy and hypercapnia risk in COPD

Management of COPD

Evidence suggests that high amounts of oxygen administered to patients with COPD can lead to hypercapnia (high levels of arterial carbon dioxide), potentially resulting in respiratory acidosis (Perrin et al, 2011; Wijesinge, 2011). The belief is that this process may be caused by inadequate gaseous exchange, resulting in hypoxia and hypercapnia (Brinkman and Sharma, 2018). In patients without COPD, changes in pH are recognised by medullary sensors, resulting in hypercapnia being the primary drive for respiration.

However, as COPD is a progressive, chronic process, the body adapts to the altered levels of carbon dioxide, causing resistance to develop (Brinkman and Sharma, 2018). Hypoxia becomes the primary drive of respiration, ultimately meaning that administering excessive amounts of supplemental oxygen to hypoxic patients with COPD can reduce respiratory rates to dangerously low levels (Brinkman and Sharma, 2018).

For this reason, NICE (2010) advocates titrating oxygen levels to achieve a percentage of saturated oxygen (SpO₂) of 88–92%. In the UK, gold standard treatment is to titrate oxygen as stated and to gain arterial blood gas (ABG) samples for analysis on all patients who present to accident and emergency hospital units with the presenting complaint of AECOPD (NICE, 2010). Despite the national recognition of these guidelines, adherence is often low (Fitzpatrick et al, 2011).

Edwards et al (2012) suggest this could be due to an ingrained ethos of administering high levels of oxygen to all respiratory patients without recognising the need to make exceptions for those suffering with COPD. Further, ambulance clinicians may only witness the benefits of high-flow oxygen as treatment for COPD because they will not be with patients long enough to observe any dangerous implications (Fitzpatrick et al, 2011).

Methodology

To ensure an efficient literature search, three electronic databases were used (Williamson and Whittaker, 2017): MEDLINE, Academic Search Premier and Cumulative Index to Nursing and Allied Health Literature (CINAHL). Inclusion criteria included full text and English language. All searches specified the date period of 2013–2018 unless this produced insufficient results, in which case the start date was extended to 2011. This was to ensure that no results included articles published before the publication of the NICE (2010) guidelines, which were the most recent at the time of writing.

Of the final eight papers, five are from Australasia and the other three are from Sweden, Brazil and China. On further examination, it was found that European, South American and Asian guidelines recognise the risks of hyper-oxygenation but there is no specific guidance on oxygen titration other than recommendations from GOLD (2017) (Nascimento et al, 2007; Fang et al, 2011; Wedzicha et al, 2017). However, the Australasian guidelines for COPD (Yang et al, 2017) recommend titrated oxygen therapy the same as, and often referenced alongside, NICE (2010) guidelines, which increases the generalisability of the Australasian papers.

Risk of hypercapnia in COPD and benefits of oxygen titration

The randomised controlled trial by Austin et al (2010) explores the evidence behind the NICE (2010) recommendation of oxygen therapy in relation to patients with COPD, with the BTS basing its guidelines on the findings (O'Driscoll et al, 2017).

The study consisted of 62 paramedics divided into two groups between June 2006 and July 2007. One group treated patients with COPD by titrating low-flow oxygen (exact dose not specified) through nasal prongs to maintain SpO₂ in the range of 88–92%, while administering any required bronchodilators through nebulisers driven by compressed air. The other group administered high-flow oxygen (8–10 litres per minute) delivered through non-rebreather face masks, with bronchodilators being delivered by oxygen-driven nebulisers with a flow rate of 6–8 litres per minute. Only those with confirmed COPD diagnoses by later lung function tests were included in the analysis, leaving a final sample size of 117 (92 in the high-flow arm and a significantly lower 43 in the titration arm).

The findings support the NICE (2010) guidelines, with Austin et al (2010) stating that titrating oxygen administration to maintain SpO₂ of 88–92% reduced the mortality rate by respiratory failure by 78% in patients diagnosed with COPD, compared with high-flow oxygen administration. There was also a significant difference in overall mortality rates between the two groups; mortality was 9% when high-flow oxygen was administered compared with 2% in those receiving titrated oxygen (relative risk 0.22, 0.05–0.91; P=0.04). Patients receiving titrated oxygen were also less likely to develop respiratory acidosis (P=0.01) due to acute hypercapnia (P=0.02). However, the results of this study relied on arterial blood gas (ABG) samples to be taken within 30 minutes of arriving at hospital, which was only achieved in 11% of patients.

Ahmadi et al (2014) conducted a population-based prospective study to identify whether hypo and hypercapnia can predict mortality rates in patients with COPD who are oxygen dependent. This study had a sample size of 2249 patients and was undertaken in Sweden between October 2005 and June 2009. Data were collected from a large health register (Swedevox), which includes approximately 85% of the Swedish population who started long-term oxygen therapy (LTOT) since 1987. Further data were gathered from the National Patient Register for inpatient and outpatient care and the Swedish Prescribed Drug Register (Wettermark et al, 2007).

Cox regression (adjusted for age, sex, arterial partial pressure of oxygen (PaO₂) on air, WHO performance status, body mass index (BMI), comorbidity and medications) was used to estimate the association between arterial carbon dioxide tension while breathing air (PaCO₂ on air) and mortality in patients with COPD (Ahmadi et al, 2014). This study examined only patients with COPD who were beginning LTOT for the first time with PaCO₂ levels assessed before oxygen therapy commenced.

At baseline, 39% of patients were hypercapnic when breathing air (PaCO₂ >6.5 kPa). Long-term results presented in a U-shaped pattern, which recognised the increased risk of mortality to be at PaCO₂ levels less than 5.0 kPa and more than 7.0 kPa. The lowest mortality rate was in patients with a PaCO₂ of approximately 6.5 kPa.

The most significant difference in the findings of Ahmadi et al (2014) is that changes in PaCO₂ caused by the beginning of LTOT did not affect the rate of mortality. Although PaCO₂ levels did increase in some patients after starting LTOT, a significant number of patients had a reduction in PaCO₂ levels, which suggests that increased oxygen administration levels are not the cause of hypercapnia leading to death. Reported oxygen flow levels in LTOT are low at 1–1.5 litres per minute (Ahmadi et al, 2014). It is therefore not possible to compare these results directly to Austin et al (2010), as the exact dose of low-flow oxygen was not documented. Ahmadi et al (2014) do not specify how PaCO₂ levels were achieved; optimal results would be obtained from arterial blood gas samples from each patient (Weaver, 2007).

The findings suggest that the main predictor of mortality associated with hypo- and hypercapnia in patients with COPD may be the level of alveolar ventilation (often compromised by other comorbidities), as opposed to gas exchange inefficiency (Ahmadi et al, 2014).

Yang et al (2015) also conducted a prospective study in China of 275 patients with AECOPD who presented at two medical centres between May 1993 and October 2006. All patients had ABGs taken and lung function tests every 2 years, and took part in 6-monthly follow-ups—either over the phone or through access to medical records—until a final follow-up in October 2011. The aim was to determine whether hypercapnia was a reliable prognostic tool for patients with COPD. Those eligible for inclusion in the study were discharged from hospital with clinical improvements and remained stable at the initial 6-week follow-up. The final sample size of 275 consisted of 98 normocapnic and 177 hypercapnic patients. During the follow-up period, the mortality rate for the normocapnic group was 53% compared with 72% in the hypercapnic group.

While the results of this study imply that hypercapnia is a life-threatening condition secondary to COPD, lack of data on any oxygen treatment received by the subjects results in inconclusive causality. Yang et al (2015) also suggest that hypercapnia may not be a condition developed in more progressed COPD but rather an individual COPD phenotype. Yang et al (2015) suggests that the number of hypercapnic patients is much higher than in similar studies (62% versus approximately 25% (Burge and Wedzicha, 2003)) owing to the high proportion of smokers included.

Yang et al (2015) did not investigate potential prognostic relevance of PaO₂ and this is recognised as a limitation. Although the 2-year follow-up consisted of definitive ABGs and lung function tests, the 6-month follow-ups did not include any clinical tests. Therefore, these follow-ups relied upon medical records being updated sufficiently or phone calls with patients.

Methods used to prevent hypercapnia in patients with COPD

Savi et al (2014) conducted a prospective study, which investigated the response of patients with COPD with known CO₂ retention suffering acute respiratory acidosis being treated with non-invasive ventilation (NIV). Seventeen subjects were included in the study from an intensive care unit in Brazil. Standard procedure during NIV maintains fraction of inspired oxygen (FiO₂) levels at or below 0.5 (Davidson et al, 2016).

This study looked into the response of PaCO₂ when increasing FiO₂ to 1.0. All patients were stabilised with routine NIV at FiO₂ of 0.25 or 0.5 for 40 minutes before being increased to 1.0 for a further 40 minutes. ABGs were taken before FiO₂ was increased and after 40 minutes of the intervention. The increase of FiO₂ significantly increased the mean PaO₂ from 101.4 to 290 mmHg, but there was no significant change to PaCO₂, with a mean of 51.5 mmHg post intervention compared with a pre-intervention measurement of 52.5 mmHg. Therefore, Savi et al (2014) conclude that increasing FiO₂ during NIV does not cause any clinically important changes to PaCO₂.

While Savi et al (2014) suggest there is no added risk to patients with COPD when FiO₂ is increased, it is unclear if the increase has any clinical benefit. Another limitation is the small sample size, which included COPD patients with combined presentations of pneumonia or AECOPD.

Pilcher et al (2017) conducted a randomised controlled cross-over trial to investigate the effect of nasal high-flow cannulae (NHF) on PaCO₂ in AECOPD as an alternative to NIV treatment. The study involved adults presenting to one New Zealand hospital with an admission diagnosis of AECOPD who were receiving oxygen therapy via standard nasal prongs (SNP). Between May 2013 and December 2014, 24 patients who met the initial selection criteria remained on the same oxygen flow rate as when they arrived at hospital for the first 15 minutes of the study, before being randomly assigned to two intervention groups. The machine used for NHF had a manually controlled flow of oxygen and could be set to allow for 0.21–1.0 FiO₂, administered with a maximum of 60 litres/minute (Nishimura, 2015). The first intervention was 30 minutes of NHF at 35 litres/minute at 37°C with additional oxygen if required, titrated to maintain the SpO₂ on study entry. The second intervention was 30 minutes of SNP with titrated supplemental oxygen to maintain initial SpO₂ of study entry. Both interventions were followed by a 15-minute observation period where oxygen therapy was administered via SNP at the baseline flow rate. The primary outcome was difference in transcutaneous carbon dioxide tension (PtCO₂) after 30 minutes of either intervention. Results showed that NHF only produced a reduction in PtCO₂ of 1.4 mmHg, and such a small difference between the control groups is unlikely to have any clinical significance (Pilcher et al, 2017).

In the NHF group, FiO₂ measurements were taken directly from the device, whereas FiO₂ measurements from the SNP group were estimated from the flow rate of oxygen administered. Another limitation is that CO₂ levels were taken transcutaneously rather than arterially, making the results less definitive (Weaver, 2007).

While the study did not prove that NHF decreases PtCO₂, it recognises that the small changes in PtCO₂ could be owing to physiological effects of the NHF device, as opposed to changes to oxygen administration. It is suggested that NHF increases alveolar ventilation and that this could result from an increase in tidal volume and a reduction in dead space, which is also recognised by Savi et al (2014) as a subject of interest for future studies.

When considering titrating oxygen therapy in the prehospital setting, further complications arise by the recommendation to administer bronchodilators and corticosteroids through oxygen-driven nebulisers as a vital component of the treatment for AECOPD (Brown et al, 2016).

Heys et al (2018) carried out a clinical audit of paramedics in a New Zealand ambulance service carrying oxygen and air-driven nebulisers in order to note any differences in the SpO₂ on arrival to hospital in patients being treated for AECOPD. The inclusion criteria of being over 50 years old with a presenting complaint of AECOPD or shortness of breath, wheeze, cough or other respiratory symptom in combination with a history of COPD were met by 474 patients. All information was taken from the patient report forms, which were completed by paramedics. After a further 74 patients were omitted due to exclusion criteria, 400 subjects were split into two intervention groups. The first group received medication through an oxygen-driven nebuliser for 5 minutes followed by a resting period of 5 minutes breathing room air, while the second group received it through an air nebuliser, with supplemental oxygen administered through nasal prongs if SpO₂ dropped below 88%. The primary outcome was SpO₂ on arrival to the hospital, with measurements documented throughout the treatment divided into three groups: hypoxic (<88%), target range (88–92%) and high saturation (>92%).

The study showed significantly lower SpO₂ measurements in patients receiving air-nebuliser treatment than the oxygen arm, with a higher percentage of patients arriving at hospital within the target range group (24% versus 16.5% respectively) but this did not reach statistical significance (p=0.82). Significant numbers of hypercapnic and acidotic COPD patients was the main impetus behind the rationale for this study; Heys et al (2018) concluded that the use of air-driven nebulisers can prevent patients developing these conditions and help to maintain SpO₂ within the target range.

While this is an encouraging outcome, the number of hypoxic patients also slightly increased (8.5% versus 7.0%). Patients being treated in the air nebuliser arm were less likely to have any oxygen administered both at initial assessment and throughout consultation, even if initial assessment indicated hypoxia (p<0.001), although all patients who were hypoxic on arrival at hospital had been treated with oxygen in some form. Heys et al (2018) argue that those who arrived at hospital in the high saturation group were not necessarily overtreated with oxygen as the majority presented with SpO₂ readings >92% from initial assessment so the true validity of the results remains unknown.

Limitations to this study include the reliance on pulse oximetry readings and manual documentation. Unlike Savi et al (2014) and Pilcher et al (2017), no effort was made to take more accurate tests such as ABGs to gain a greater understanding of exactly how the different treatment arms affected PaO₂ or PaCO₂, or further tests for FiO₂. However, this study has a bigger sample size of 400 and is targeted specifically at prehospital management of AECOPD, without the risk of data being compromised by inclusion of hospital treatment.

Clinician compliance with study protocols and COPD guidelines

The study by Austin et al (2010) demonstrated low clinician protocol compliance. Despite pre-trial education to paramedics and medical hospital staff, 56% of the patients in the oxygen-titration arm received high-flow oxygen against protocol and ABGs were only drawn from 11% of patients within 30 minutes of hospital arrival. Austin et al (2010) recognised the benefit of having a representative in the hospital ensuring ABGs were taken straight away, but stated that, practically, this was not possible.

A retrospective review was undertaken by Cameron et al (2012) investigating the association between PaO₂ and risk of adverse clinical outcomes in patients with COPD. The study was based on patients presenting to one hospital by ambulance in New Zealand with AECOPD. There were 680 presentations between June 2005 and January 2008 and all subject information was gathered from a list of patients from the hospital's Decision Support Unit with a discharge diagnosis of COPD.

The resulting sample of 254 patients was split into four groups defined by SpO₂ readings: <88%, 88–92%, 93–96% and >96%. Results concluded that those in groups <88% and >96% were at a significantly increased risk of serious adverse outcomes; however, there was no difference in risk in groups 88–92% and 93–96%.

Analysis demonstrated a strong association between hyperoxaemia and increased risk of mortality. In a sensitivity analysis, the OR for hyperoxaemia was 8.96 (95% CI 3.37–23.8) and 3.31 (95% CI 1.40–7.82) for hypoxaemia. Results showed hyperoxaemic patients to be at a significantly increased risk of hypercapnic respiratory failure and the need for assisted ventilation. This is particularly significant as nearly a quarter of patients brought in by ambulance were hyperoxaemic. It is presumed that this is likely to be caused by high-flow oxygen being administered by ambulance clinicians (Cameron et al, 2012), but as the oxygen flow rate administered by paramedics was only documented in 76% of cases, this cannot be confirmed.

As a retrospective review, reliance on the existing documentation was high and it was not possible to control certain variables, such as ABG sample collection time and documentation of exact oxygen flows administered. The small sample size is also acknowledged as a limitation.

It is an important observation that had national protocol been followed and all patients had ABGs taken within 4 hours, the sample would have been nearly twice the size, significantly increasing the power of the study. The increased length of time between hospital triage and taking of ABGs also means that non-documented hospital oxygen therapy will have influenced the result, making analysis less reliable. While a possible strength is that PaO₂ levels were analysed directly from ABG results rather than relying on transcutaneous devices (Weaver, 2007), Cameron et al (2012) recognise that the guidelines to paramedics refer to SpO₂ measurements, so were working to equivalent units; PaO₂ of 60 mmHg in this study was equivalent to SpO₂ of 88%.

Non-compliance is also particularly evident in the retrospective audit carried out by Susanto and Thomas (2015). The aim of the study was to assess the use of oxygen therapy and FiO₂ in patients with COPD who were admitted to hospital as an emergency. The audit included 150 patients in one Australian hospital between January 2011 and June 2013. Patients assigned were brought in by ambulance with a pre-existing diagnosis of COPD, with ambulance paperwork and digital medical records audited.

Blood gases were taken on arrival to hospital, with a 93% compliance rate. However, 59% were venous blood gas (VBG) samples, which Susanto and Thomas (2015) recognised as a limitation, as VBGs provide less accurate results than ABGs (Weaver, 2007). Time from hospital triage to the taking of blood gases is not documented. Therefore, similarly to Cameron et al (2012), it is impossible to determine if results are solely representative of prehospital treatment. Oxygen flow rates were only documented in 80% of patients.

Conclusion

Studies investigating management of COPD have identified the association between COPD and hypercapnia. However, the cause of hypercapnia remains unclear. While no evidence has been found to suggest that oxygen titration is an incorrect intervention for this patient group, the studies reviewed have questioned whether high-flow levels of oxygen are more likely to cause hypercapnia over rate of alveolar ventilation or specific COPD phenotypes.

In review of studies on oxygen-titration methods, it is suggested again that alveolar ventilation has more influence on hypercapnia than oxygen administration levels. However, the studies analysed were weakened by small sample sizes and data collection methods, and lacked statistical significance. A recurring theme throughout most of the research analysed was clinician non-compliance, not only in experimental protocols but also in longstanding guideline recommendations.

Recommendations

The need for UK-based studies is particularly high as current guidelines are based on studies from other health systems. Further research would benefit from having a large sample size with a long study period in order for results to be clinically accurate and statistically significant. Robust data collection methods must also be controlled, with specific tests performed uniformly throughout the study population.

Key points

CPD Reflection Questions