Abstract
Oxygen is commonly administered in hospitals, with poor adherence to treatment recommendations.
We conducted a multicentre randomised controlled study in patients admitted to the emergency department requiring O2 ≥3 L·min−1. Patients were randomised to automated closed-loop or manual O2 titration during 3 h. Patients were stratified according to arterial carbon dioxide tension (PaCO2) (hypoxaemic PaCO2≤45 mmHg; or hypercapnic PaCO2>45–≤55 mmHg) and study centre. Arterial oxygen saturation measured by pulse oximetry (SpO2) goals were 92–96% for hypoxaemic, or 88–92% for hypercapnic patients. Primary outcome was % time within SpO2 target. Secondary endpoints were hypoxaemia and hyperoxia prevalence, O2 weaning, O2 duration and hospital length of stay.
187 patients were randomised (93 automated, 94 manual) and baseline characteristics were similar between the groups. Time within the SpO2 target was higher under automated titration (81±21% versus 51±30%, p<0.001). Time with hypoxaemia (3±9% versus 5±12%, p=0.04) and hyperoxia under O2 (4±9% versus 22±30%, p<0.001) were lower with automated titration. O2 could be weaned at the end of the study in 14.1% versus 4.3% patients in the automated and manual titration group, respectively (p<0.001). O2 duration during the hospital stay was significantly reduced (5.6±5.4 versus 7.1±6.3 days, p=0.002).
Automated O2 titration in the emergency department improved oxygenation parameters and adherence to guidelines, with potential clinical benefits.
Abstract
Automated oxygen titration is superior to manual administration in terms of time within oxygenation targets http://ow.ly/pgWC30c2sLv
Introduction
Oxygen therapy is administered daily to a wide number of patients in the emergency care setting, either during pre-hospital transportation or hospital care [1]. Current standards for prescribing oxygen recommend providing adequate flows to correct hypoxaemia and avoid hyperoxia [2, 3]. While deleterious effects of hypoxaemia are well known, the potential harmful effects of hyperoxia are underappreciated. Yet, hyperoxia may increase mortality in severe chronic obstructive pulmonary disease (COPD) patients [4–6] and may cause cardiac and neurological adverse toxicities in certain situations [7–9]. In conditions where access to O2 is limited (pre-hospital care and/or military operations), it is also crucial to minimise its use as much as possible, and to promote rapid O2 weaning [10].
Precise control of O2 flows is difficult to achieve in clinical practice and is time-consuming [11]. Automated O2 administration allows arterial oxygen saturation measured by pulse oximetry (SpO2) to be maintained within a predetermined range using variable O2 flows, as opposed to manual O2 administration where the flow is kept constant with variable SpO2 values. In preterm infants receiving mechanical ventilation, automated O2 control results in more time spent within the intended SpO2 target [12–14]. In a healthy adult model, such a system was more efficient at maintaining SpO2 within the target, while ensuring a significant reduction of hypoxaemia and hyperoxia periods compared with constant O2 flows [15].
We conducted a prospective, multicentre, randomised, controlled trial involving adult patients admitted to the emergency department (ED) for acute hypoxaemic respiratory failure. We hypothesised that automated O2 administration would improve oxygenation and promote better adherence to clinical guidelines during the early emergency care of adult patients than conventional O2 therapy [2].
Methods
Study design
From August 2011, to October 2014, we recruited adult patients that had been evaluated in four French and Canadian University Hospital EDs for an acute respiratory failure. The study protocol was approved by human research ethics committees in France and in Canada.
The trial was overseen by a steering committee and an independent safety monitoring board. All centres were monitored to check adherence to the protocol and the accuracy of recorded data. An investigator at each centre was responsible for enrolling patients and ensuring adherence to the protocol. Research assistants were responsible for patients' follow-up, and for completing the electronic case report form. Although assignment to the study could not be blinded, all clinicians in charge of the patients remained unaware of the study outcomes.
The study was registered at ANSM ID RCB 2010-A00927-32 and ClinicalTrials.gov identifier NCT02027181.
Patients
Patients were eligible if they were adult and admitted to the ED within less than 2 h for acute respiratory disorder still requiring ≥3 L·min−1 of O2 for maintaining a SpO2 ≥92%, whatever the original pathology. Patients were ineligible if they presented life-threatening hypoxaemia, any clinical signs of ventilatory assistance requirement or if they required emergent surgery or coronary angiography. Details of the exclusion criteria are provided in the supplementary material.
All patients or their next of kin provided written informed consent.
Randomisation
A computer-generated block-randomisation sequence with random block sizes was used. All patients were stratified according to arterial carbon dioxide tension (PaCO2) (purely hypoxaemic respiratory failure, PaCO2 ≤45 mmHg; or mildly hypercapnic respiratory failure PaCO2 >45 and ≤55 mmHg) and study centre. Patients were randomly assigned in a 1:1 ratio to either automated oxygen administration (FreeO2) or conventional oxygen administration (manual O2) using a centralised web-based electronic datafile system (Clinsight, Ennov, Paris, France).
Study intervention
Automated O2 administration was performed using the FreeO2 system (Oxynov Inc., Québec, QC, Canada) that was set to maintain SpO2 between 92% and 96% for purely hypoxaemic respiratory failure or between 88% and 92% for hypercapnic respiratory failure. FreeO2 is equipped with a SpO2 monitor and an electronically controlled valve that automatically adjusts O2 flows on a per second basis, according to a closed-loop algorithm, in order to reach the predetermined SpO2 goal [15].
Conventional O2 was administered using manual flowmeters, according to standard procedures. All participating units were encouraged to use the same standardised SpO2 goals as in the automated O2 administration group. The intervention period was set to 3 h, considering that such duration may enable stabilisation of the patient with medical treatment, and/or to determine any deterioration with a high level of confidence.
In both groups, we used an O2 mask to administer either low or high O2 flow. Considering the variety of pathological cases requiring O2 administration in an ED, medical treatment was determined by the attending physicians on the basis of clinical needs assessment, except for ventilatory assistance requirements, which were standardised to avoid delaying endotracheal intubation. Following this initial treatment period, patients could be hospitalised to receive further treatment if deemed necessary by the attending physician.
Data collection and study outcomes
At the time of enrolment, physiological characteristics, coexisting medical conditions and oxygen flow rates were recorded. In both study groups, SpO2, respiratory rate (RR) and heart rate (HR) were continuously monitored using dedicated software (FreeOview v2, Oxynov Inc.) and an oximeter (Nonin OEM III, Plymouth, MN, USA) to calculate the time spent within a given SpO2 range.
The primary outcome was the percentage of time spent within the target SpO2 range during the 3-h study period. Secondary outcomes included the percentage of time spent in hypoxaemia as defined by a SpO2 value 2% lower than the minimum SpO2 target (SpO2<90% for purely hypoxaemic respiratory failure and <86% for mildly hypercapnic respiratory failure), or hyperoxia as defined by a SpO2 value 2% higher than the maximum SpO2 target (SpO2≥98% for purely hypoxaemic respiratory failure and ≥94% for mildly hypercapnic respiratory failure), duration of oxygen supplementation duration in the entire hospital stay and hospital length-of-stay. Serious adverse events rate at day 28 and during the entire ED and hospital stay. Serious adverse events were predefined and reported to the safety-monitoring board as they occurred. Data were collected until death, or the first discharge from the hospital.
Statistical analysis
Based on preliminary data, we anticipated a percentage of time within the SpO2 range of 70% using the automated system compared with 60% using conventional O2. We determined that a sample of 190 patients was required to show the superiority of automated oxygenation with a power of 90%. We allowed the switch to conventional oxygen when there were technical problems, but there was no crossover to FreeO2. The primary outcome was short-term treatment efficacy rather than death or morbidity, because of the lack of data on outcome after O2 administration in the ED, and the absence of clinical data on automated systems. All analyses were performed on an intention-to-treat basis, and patients remained in their assigned group for all outcomes.
The primary outcome was analysed using a generalised linear model with normal distribution, adjusted for the study site (random effect) and the subgroup (hypoxaemic or hypercapnic). Homogeneity between subgroups was tested adding an interaction term in the model. Secondary outcomes were analysed using the same statistical method, except that binary data were evaluated using a binomial distribution with identity link and exchangeable correlation structure to account for study centre. Hospital length of stay (LOS) was analysed using the raw data, and subsequently censored for survivors; log-rank tests were used for the univariate analysis and a Cox proportional hazards model was adjusted for study centre and subgroup. All analyses were performed by an independent study statistician using SAS, version 9.3 (SAS Institute, Cary, NC, USA). p-values equal or less than 0.05 were considered statistically significant.
Results
Study patients
From August 2011, to October 2014, 1247 patients were screened for eligibility. A total of 190 patients were included and a total of 187 patients underwent randomisation (93 to the automated O2 group (FreeO2)and 94 to the conventional O2 group (Manual O2)) (figure 1). Hypoxaemic respiratory failure represented the most frequent clinical presentation (n=137; 73.3%). Sixty-five patients had a diagnosis of COPD at enrolment (34.8%), of whom 17 patients were receiving long-term oxygen therapy. Initial mean oxygen flow rates were as follows: in the FreeO2 group, 6.2±3.1 L·min−1; in the Manual O2 group, 5.5±3.1 L·min−1 (table 1).
Primary outcome
FreeO2 was found superior to Manual O2 to maintain SpO2 within the assigned range (adjusted between-group difference of 29.4 percentage points; 95% confidence interval, 25.7–33.2), either for the entire group or for any of the subgroups of respiratory failure (hypoxaemic and hypercapnic) (table 2).
Secondary outcomes during the 3 h of emergency care
More time was spent with hypoxaemia in the Manual O2 group than in the FreeO2 (adjusted difference 2 percentage points; 95% confidence interval 0–4). At 3 h, 24.1% of the patients were experiencing hyperoxia despite the use of oxygen in the Manual O2 group (18.2% and 26.2% for hypercapnic and hypoxaemic respiratory failure patients, respectively) compared with 7.9% in the FreeO2 group (6.7% and 8.2% for hypercapnic and hypoxaemic respiratory failure patients, respectively; p<0.001). The time was spent with hyperoxia was higher in the Manual O2 group than in the FreeO2 (adjusted difference 17 percentage points; 95% confidence interval 15–19), this effect being prominent in the hypercapnic subgroup (adjusted difference 35 percentage points; 95% confidence interval 26–44). No significant worsening of hypercapnia was detected within groups. Oxygen administration longer than 3 h was lower (adjusted difference 15.2 min; 95% confidence interval 18.1–12.4) and weaning was more frequent (adjusted difference 10 percentage points; 95% confidence interval, 8–11) under FreeO2; this effect on oxygen duration was more pronounced in the hypercapnic subgroup (absolute time difference 29.6 min.; 95% confidence interval, 38.5–20.8). Mean oxygen flow during the 3 h was lower under FreeO2 in the hypercapnic subgroup (absolute flow difference 0.85 L·min−1; 95% confidence interval, 1.27–0.44) (table 2), while flow changes were more homogeneously distributed in the hypoxaemic subgroup (supplementary figure S1).
Secondary outcomes during the entire hospital stay and adverse events
Oxygen administration duration during the entire hospital stay was lower in the FreeO2 group (absolute duration difference 1.4 days; 95% confidence interval, 2.3–0.5) (table 3). While a difference was observed for the raw overall length of stay, it was not significant when considering censored data for death. No difference for the invasive and noninvasive mechanical ventilation requirements or any other serious adverse events was observed between groups during the entire hospital stay (Supplementary table S1).
Discussion
The use of automated O2 administration was found superior to manual O2 administration to improve the time spent within oxygenation targets in adult patients attending the ED for the evaluation and treatment of an acute respiratory failure episode, with a between-group absolute difference of 29.4 percentage points. Patients experienced less time with hypoxaemia and hyperoxia in the FreeO2 group, this effect being prominent in the hypercapnic subgroup. When receiving automated oxygen, partial or complete oxygen weaning was more frequent during the 3 h of care than manual O2 administration, and patients were less exposed to oxygen during the overall hospital stay, with a between-group difference of 1.5 days.
The main objective of oxygen therapy should be to maintain stable oxygenation within a target chosen by clinicians. In a retrospective before–after study conducted in an ED, the SpO2 target for COPD patients was achieved in 16% of patients before a quality improvement campaign, and in 32% of them following the campaign [16]. In pre-hospital studies, 56–88% of the patients did not receive the assigned treatment in the titrated oxygen arm [13, 17]. In the present study, time spent within the oxygenation target was achieved more than 80% of the time while using automated titration compared with 50% in the manual group.
Despite extensive literature about the adverse events of inadequate use of oxygen, this therapy is still poorly prescribed and junior doctors poorly understand the effects and dangers of oxygen [18–20]. Changing clinical practice in response to research data (knowledge translation or implementation) is difficult in many areas of medicine, resulting in excess morbidity and mortality and significant additional financial costs [14]. This additional cost is difficult to quantify, but is related to the incidence of non-compliance to treatment guidelines [2, 3] and is proportional to disease prevalence. Even though several studies have shown that compliance to oxygen therapy recommendation may be partially improved through complex multidisciplinary programmes [21], these programmes require frequent training sessions and may only be partially efficient [22]. Automated titration of oxygen may therefore improve compliance to the recommendations.
There is a good level of evidence that hypoxaemia is harmful in various populations, such as adult patients with myocardial ischaemia [23], trauma [24] or in neonates [25]. Such facts are well accepted by the medical community [26, 27], and data also suggest that short periods of hypoxaemia may promote significant negative haemodynamic effects [28]. In an animal model, right ventricular dilation was observed with only 2 h of daily hypoxaemia [29]. In the present study, as also observed in several studies evaluating similar devices, hypoxaemia was less frequent with automated oxygen titration [15, 30, 31].
The physiological risks associated with hyperoxia are also well described, especially in COPD patients, its effect being particularly marked during the acute phase of exacerbations [5]. The first recommendation to adjust oxygen flow rates to reduce the risks of hyperoxia was published in the early 1960s [32], but several recent guidelines have reiterated this recommendation [2, 3, 33]. The recent demonstration of increased mortality because of over-oxygenation during pre-hospital transport in a large randomised controlled trial has revived the debate about the potential harm of excessive oxygen therapy in COPD patients [6]. Other adverse effects of hyperoxia such as increased coronary [7, 8] and carotid artery resistance have been demonstrated [9], all being associated with a potential clinical impact [34, 35].
Automated oxygen titration has already been promoted for infant and neonates [12–14], but only three systems have been evaluated in spontaneously breathing adults [15, 30, 31]. In one study, the authors evaluated a closed-loop system (O2 flow regulator, Dima, Italy), adjusting low-flow oxygen during exercise in COPD patients receiving long-term oxygen therapy [30], and showed better oxygenation and reduced workload with automated oxygen titration. In another study with a similar device (AccuO2, Optisat medical, Minneapolis, MN), the authors demonstrated better maintenance within the target and a reduction in oxygen consumption [31]. The third device (FreeO2, Oxynov, Québec, Canada) was originally validated in a pilot study on healthy volunteers under experimental hypoxaemia [15]. In the present study, we found better oxygenation with more time within the SpO2 target, less hypoxaemia and less hyperoxia. In addition, we also demonstrated automated oxygen weaning, leading to reduced time under oxygen and safety of the procedure (supplementary table S1). The use of automated oxygen administration was well accepted by clinicians, without differences in terms of premature interruption of treatment.
Our trial, which is the largest study to date evaluating an automated oxygen titration device, has several strengths that suggest that it may be generalised to patients with similar conditions. These strengths include the multicentre design and sealed randomisation to the assigned strategy, a well-defined study protocol over a short period of observation, predefined efficacy criteria, complete follow-up of patients by dedicated research assistants and an intention-to-treat analysis. Several potential limitations can however be highlighted. Since O2 was continued using a standard flowmeter when needed after the 3-h study period, the potential benefits of automated O2 flow adjustments over time may have been missed. The choice of a 3-h time period to study the effect of automated O2 administration is also debatable, since a longer period may be appropriate to elicit significant outcome improvement. However, such duration may be more appropriate when considering the acute care setting of the study and the fact that no such trial has yet been performed using such devices. Second, sample size assumptions were a little different from expected (smaller proportion of hypercapnic patients), but the study was still adequately powered. Third, while the intervention was conducted during the first 3 h of treatment, there are no data to extend these benefits for automated oxygen administration use outside the ED. Fourth, the fact that only one FreeO2 device was available in each centre was responsible for a low randomisation rate, which may limit the external validity of our results. While only 57% of eligible patients were enrolled, these results may not be representative of the total eligible group. Fifth, the unblinded characteristic of the study may have biased clinicians’ care.
In conclusion, the use of automated O2 administration was superior to manual O2 administration for patients attending the ED with acute hypoxaemic respiratory failure. The potential benefits of automated O2 titration may exist for either the patients (better control of oxygenation, better monitoring, reduced hospital length-of-stay) or the healthcare system (reduced number of O2 manual adjustments, better adherence with treatment recommendations, reduced oxygen use). Additional data are warranted to demonstrate the safety and cost effectiveness of these systems in different clinical settings.
Supplementary material
Supplementary Material
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Supplementary material ERJ-02552-2016_Supplement
Disclosures
Acknowledgements
We thank the patients, their families, and the ED staff members of our four centres in Canada and France who cared for them, and all research assistants for their help in recruitment and data collection. We would also like to thank François Maltais at the Institut Universitaire de Cardiologie et de Pneumologie de Québec, Québec, Canada for his extensive reviewing of the manuscript, and Julien Coadic from Brest University Hospital (CIC INSERM 1412) for data storage and management.
Author contributions: E. L'Her and F. Lellouche designed the prototype, conceived, coordinated the study in France and Canada respectively, analysed and interpreted the data and wrote the article. P. Dias, A. Riou, L. Souquiere, N. Paleiron, P. Archambault and P-A. Bouchard included the patients, supervised monitoring of the data and corrected the manuscript. M. Gouillou was responsible for data analysis. E. L'Her had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Footnotes
This article has supplementary material available from erj.ersjournals.com
Earn CME accreditation by answering questions about this article. You will find these at erj.ersjournals.com/journal/cme
Support statement: The study was funded by the Programme Hospitalier de Recherche Clinique National 2010 from the French Ministry of Health for the French centres and by the Ministry of Finance in Québec for Canadian centres. The study was sponsored by Brest University Hospital, which was responsible for data management and control. Funding information for this article has been deposited with the Crossref Funder Registry.
Conflict of interest: Disclosures can be found alongside this article at erj.ersjournals.com
The study was registered at ANSM ID RCB 2010-A00927-32 and ClinicalTrials.gov identifier NCT02027181.
- Received December 28, 2016.
- Accepted April 11, 2017.
- Copyright ©ERS 2017