Oxygen Versus Air-Driven Nebulisers for Exacerbations of Chronic Obstructive Pulmonary Disease

A Randomised Controlled Trial

George Bardsley; Janine Pilcher; Steven McKinstry; Philippa Shirtcliffe; James Berry; James Fingleton; Mark Weatherall; Richard Beasley


BMC Pulm Med. 2018;18(157) 

In This Article


Trial Design and Patients

This was a parallel-group double-blind randomised controlled trial at Wellington Regional Hospital, New Zealand. The full study protocol is available in the online supplement.

Participants were hospital inpatients, ≥40 years of age, with an admission diagnosis of AECOPD. Exclusion criteria included requirement for ≥4 L/min of oxygen via nasal cannulae to maintain SpO2 between 88 to 92%; current requirement for non-invasive ventilation (NIV); baseline transcutaneous partial pressure of carbon dioxide (PtCO2) > 60 mmHg; inability to provide written informed consent; and any other condition which at the Investigator's discretion, was believed may present a safety risk or impact on the feasibility of the study results. Written informed consent was obtained before any study-specific procedures. The study was undertaken on the ward during the hospital admission. Ethics approval was obtained from the Health and Disability Ethics Committee, New Zealand (Reference 14/NTB/200). The full study protocol (original and updated version) can be found on the OLS (see Additional file 1 and Additional file 2).


After written consent, participants had continuous PtCO2 and heart rate monitoring using the SenTec® (SenTec AG, Switzerland) device and oxygen saturation (SpO2) measured by pulse oximetry (Novametrix 512, Respironics, Carlsbad, USA). Participants were randomised to receive two nebulisations, both driven either by air or oxygen, at 8 L/min, each delivered over 15 min with a five minute break in-between. Randomisation was 1:1 by a block randomised computer generated sequence (block size six), provided in sealed opaque envelopes by the study statistician who was independent of recruitment and assessment of participants.

The participants and blinded investigator, who recorded heart rate and PtCO2 were masked to the randomised treatments. If both oxygen and air ports were available in hospital on the wall behind the participant, these were used for driving nebulisation. If only oxygen ports were available, identical portable oxygen and air cylinders were placed behind the participant's bed prior to randomisation and used instead. Both the participant and blinded investigator faced forward for the full duration of the study. In addition, the blinded investigator sat towards the end of the bed - ahead of the participant, such that they could not see the participant's interventions. Likewise, the blinded investigator and patient could not view the SpO2 on the Sentec device, as this was covered during the interventions, or the pulse oximeter which could only be viewed by the unblinded investigator. Interaction between blinded and unblinded investigators would only occur if a rise in PtCO2 of ≥10 mmHg was demonstrated (a predefined safety criterion to abort intervention).

An initial 15 min wash-in and titration period was administered by the unblinded investigator using nasal cannulae, if required, to ensure that participant's SpO2 were within 88 to 92%. If saturations were ≥ 88% on room air, no supplemental oxygen was required. Randomisation was performed after the 15 min wash-in period, when both patient and blinded investigator were already in a forward-facing position to maintain blinding. The unblinded investigator recorded SpO2 on a separate pulse-oximeter from then onwards.

Immediately before the first nebulisation, denoted by the baseline reading at time-point zero, PtCO2, SpO2 and heart rate were recorded. Participants then received two administrations of 2.5 mg salbutamol by nebulisation, delivered by either air or oxygen - each for 15 min duration at a flow rate of 8 L/min. Hudson RCI Micro Mist Nebuliser Masks (Hudson RCI, Durham, North Carolina, USA) were used. The nebulisations were delivered by the unblinded investigator at time zero and at 20 min, allowing for a five minute interval between nebulisations. Recordings were continued for 45 min after completion of the last nebulisation (80 min after baseline). Measurements of PtCO2, SpO2 and heart rate were recorded at five minute intervals, and at six minutes after the start of each nebulisation, in view of the British Thoracic Society (BTS) guideline's recommendation for limiting oxygen-driven nebulisation to six-minutes in ambulance care, if air-driven nebulisation is unavailable.[6]

Immediately before the first nebulisation and just before completion of the second nebulisation, at 35 min, a capillary blood gas sample was taken from the fingertip for measurement of PcapCO2 and pH.

Oxygen Delivery

During the wash-in and between the nebulisations oxygen was titrated, if required, via nasal prongs to maintain oxygen saturations between 88 to 92%. Participants in the air-driven group who were receiving oxygen at the start of nebulisation continued to receive titrated supplemental oxygen via nasal prongs underneath the nebuliser mask. Those in the oxygen-driven group had the prongs removed at the start, and reapplied after the completion of each nebulisation. At 35 min, oxygen was delivered via nasal prongs to participants at the flow rate they last received during titration (i.e. at 35 min and 20 min in the air-driven and oxygen-driven groups, respectively). From 35 min until 80 min, the oxygen flow rate was only increased (or initiated) if a participant's SpO2 fell below 85%.


The primary outcome was originally planned to be PcapCO2 at 35 min, at completion of the second nebulisation. However, after the first 14 participants had been studied, it was evident that obtaining adequate amounts of blood to fill the capillary tubes from some participants was difficult. At this stage of recruitment 4/14 (29%) of participants had missing data. The primary outcome variable was therefore changed to PtCO2 at 35 min, with PcapCO2 at 35 min reverting to a secondary outcome variable. Other secondary outcomes were the individual PtCO2 measurements at each time point; the proportion of participants who had a rise in PtCO2 or PcapCO2 of ≥4 and ≥ 8 mmHg; capillary pH at 35 min, and heart rate and SpO2 measurements at each time point.

Sample Size Calculation and Statistical Analysis

A rise in PtCO2 from baseline of ≥4 mmHg is considered a physiologically significant change and ≥ 8 mmHg a clinically significant change, based on previous criteria.[7,8] In our study of oxygen versus air-driven nebulisers in stable COPD patients, the standard deviation (SD) of baseline PtCO2 was 5.5 mmHg.[3] With 90% power and alpha of 5%, 82 patients were required to detect a 4 mmHg difference. Assuming a drop-out rate of 10% our target recruitment was 90 patients.

The primary analysis used a mixed linear model with fixed effects of the baseline PtCO2, time, the randomised intervention, and a time by intervention interaction term; to estimate the difference between randomised treatments at 35 min. A power exponential in time correlation structure was used for the repeated measurements. The secondary outcome variables of PtCO2 at the other time points, SpO2 and heart rate used similar mixed linear models. PcapCO2 and pH were compared by Analysis of Covariance with the baseline measurement as a continuous co-variate. As a post-hoc analysis we compared the difference in PtCO2 between the 15 and 6 min, and the 35 and 26 min time points.

Comparison of categorical variables, PtCO2 or PcapCO2 change of ≥4 and 8 mmHg, was by estimation of a risk difference, and Fishers' exact test. As a post-hoc analysis we also compared the difference in paired proportions for those with PtCO2 change of ≥4 mmHg in the oxygen arm only using McNemar's test and an appropriate estimate for the difference in paired proportions. The time for PtCO2 to return to baseline during the observation period (defined as the time until the PtCO2 was first equal to or below the baseline value, between 40 and 80 min), was compared using Kaplan-Meier survival curves and a Cox Proportional Hazards model. A simple t-test was used to compare the lowest value of the SpO2 between 40 and 80 min, compared to baseline. SAS version 9.4 was used.