To the Editor: While Joosten et al highlight the uncommon but serious problem of potential carbon dioxide (CO2) narcosis after emergency management of respiratory illness,1 it is important that their findings are kept in perspective and do not lead to inadequate administration of oxygen to patients with acute dyspnoea.

Their findings are based on a retrospective chart review. The main claim that the administration of oxygen causes increased length of hospital stay and possibly death for those presenting to emergency departments with exacerbation of chronic obstructive pulmonary disease (COPD) can be challenged by selection bias, sample size, assessment of severity of illness, and the definition of clinically significant hypercapnia.

Ninety per cent of their study patients arrived by ambulance, presumably indicating the relatively sudden onset of acute distressing symptoms — a call for urgent help, not the “killing me slowly” drowsiness and confusion of CO2 retention. Of those who received more than 4 litres of oxygen (O2) per minute, 57% (16 of 28) were in triage category 1 and 2, but only 31% (4 of 13) of those who received O2 at a lower flow rate were in triage category 1 and 2. Sixty per cent (12 of 20) of those with a high partial pressure of arterial oxygen (Pao2), when measured after arrival and treatment were in triage category 1 or 2, but only 14% (3 of 21) of those with a lower Pao2 were in triage category 1 or 2 (P = 0.002; Fisher’s exact test). Clearly the first group was a sicker group on arrival, and the increased length of stay of these patients was more likely to be the result of this, rather than of O2 therapy supervised by emergency specialists in an emergency room of a teaching hospital.

The contention that oxygen therapy in emergency departments is “often uncontrolled” is not supported by any data supplied. Critical care staff, including ambulance and emergency personnel, are acutely aware of the challenges posed by patients with chronic respiratory disease. However, they are also aware of the need to achieve adequate oxygenation in patients with acute dyspnoea. Patients are observed closely for signs of clinically significant hypercapnia and respiratory support is adjusted accordingly. Some patients may require a higher fraction of inspired oxygen (Fio2), particularly in the initial phases of care, to achieve this. As the patient’s condition improves, the Fio2 is often reduced. The methods in the study by Joosten et al fail to account for this. Respiratory rate, for example, was not reported.

Treating the patient, not the chart, is of most importance. It would be a pity if the article by Joosten et al resulted in the withholding of oxygen from acutely dyspnoeic patients with a rapid respiratory rate and adequate respiratory drive because of some fear that they could be retaining CO2.

We agree that a better and seamless patient-centred information system with cooperation between sectors of the health system, the patient, the patient’s general practitioner, and ambulance, emergency and in-hospital services, would assist in identifying those at risk of CO2 narcosis and improve patient care.

In reply: We performed a retrospective audit as part of a quality improvement program following a number of serious adverse events in various areas of our hospital. Our article showed that carbon dioxide retention in acute exacerbation of chronic obstructive pulmonary disease (AECOPD) is common (41 of 65 patients admitted with chronic obstructive pulmonary disease [COPD] over 4 months), and that guidelines on blood gas measurement and oxygen use were not being followed.

Dent and colleagues state that more patients in our study who received more than 4 litres of oxygen per minute were in a triage category that indicated a more serious condition. However, the multivariate analysis showed that triage category did not predict length of stay. In contrast, partial pressure of arterial oxygen (Pao2) did, and patients with a Pao2 of less than 74.5 mmHg (range, 36.7–74.0 mmHg) had a shorter length of stay than those with a Pao2 of 74.5 mmHg or higher (range, 74.5–452.0 mmHg). Many patients had a Pao2 much higher than neccessary to achieve a haemoglobin saturation of about 90%.

Dent and colleagues state that our data did not support the claim that oxygen therapy is often uncontrolled in the emergency setting. In fact, only 68% of the patients receiving more than 4 litres of oxygen per minute had arterial blood gas measurements performed.

We agree with Dent et al that the management of AECOPD may not be as simple as following guidelines. However, we hope to raise awareness of the fact that hypercapnia in COPD is common, requires careful assessment, and that oxygen therapy should be titrated to physiological endpoints.