American Journal of Respiratory and Critical Care Medicine

Rationale: There are no randomized controlled trials comparing different oxygenation targets for intensive care unit (ICU) patients.

Objectives: To determine whether a conservative oxygenation strategy is a feasible alternative to a liberal oxygenation strategy among ICU patients requiring invasive mechanical ventilation (IMV).

Methods: At four multidisciplinary ICUs, 103 adult patients deemed likely to require IMV for greater than or equal to 24 hours were randomly allocated to either a conservative oxygenation strategy with target oxygen saturation as measured by pulse oximetry (SpO2) of 88–92% (n = 52) or a liberal oxygenation strategy with target SpO2 of greater than or equal to 96% (n = 51).

Measurements and Main Results: The mean area under the curve and 95% confidence interval (CI) for SpO2 (93.4% [92.9–93.9%] vs. 97% [96.5–97.5%]), SaO2 (93.5% [93.1–94%] vs. 96.8% [96.3–97.3%]), PaO2 (70 [68–73] mm Hg vs. 92 [89–96] mm Hg), and FiO2 (0.26 [0.25–0.28] vs. 0.36 [0.34–0.39) in the conservative versus liberal oxygenation arm were significantly different (P < 0.0001 for all). There were no significant between-group differences in any measures of new organ dysfunction, or ICU or 90-day mortality. The percentage time spent with SpO2 less than 88% in conservative versus liberal arm was 1% versus 0.3% (P = 0.03), and percentage time spent with SpO2 greater than 98% in conservative versus liberal arm was 4% versus 22% (P < 0.001). The adjusted hazard ratio for 90-day mortality in the conservative arm was 0.77 (95% CI, 0.40–1.50; P = 0.44) overall and 0.49 (95% CI, 0.20–1.17; P = 0.10) in the prespecified subgroup of patients with a baseline PaO2/FiO2 less than 300.

Conclusions: Our study supports the feasibility of a conservative oxygenation strategy in patients receiving IMV. Larger randomized controlled trials of this intervention appear justified.

Clinical trial registered with Australian New Zealand Clinical Trials Registry (ACTRN 12613000505707)

Scientific Knowledge on the Subject

Recommendations and practices related to oxygenation targets for mechanically ventilated patients are based on weak evidence. Conventional practice follows a liberal approach to oxygen therapy, often resulting in hyperoxia that may adversely affect outcomes. However, evidence from randomized trials is lacking.

What This Study Adds to the Field

A conservative oxygenation strategy is a feasible alternative to the usual liberal oxygenation strategy used in mechanically ventilated patients. No harmful effects were observed with the use of a conservative approach to oxygen therapy. It can significantly reduce exposure to hyperoxia compared with standard care. Larger randomized trials of this intervention seem justified.

Each year 2–3 million intensive care unit (ICU) patients receive invasive mechanical ventilation (MV) (1, 2) at an estimated annual cost of $15–27 billion in high-income nations alone (1, 3) and with a high associated mortality (1, 4) and morbidity (5). Nearly all ICU patients who receive MV also receive supplemental oxygen therapy. Despite the universal use of oxygen therapy, no randomized controlled trials (RCTs) have investigated the effects of different oxygenation targets during MV (6).

In the absence of RCTs, the recommended oxygenation targets for mechanically ventilated ICU patients are largely based on normal physiologic values. For example, in healthy adults at sea level, the usual ranges for SaO2 and PaO2 are approximately 95–97% and 88–100 mm Hg, respectively (7). Moreover, in healthy humans during sleep, the nadir for pulse oximeter measured oxygen saturation (SpO2) is approximately 90% (8). Accordingly, for acutely ill patients, recommendations vary from near-normal SpO2 targets of 94–98% (9) to values greater than or equal to 90% (10). In addition, SpO2 targets of 88–95% are often accepted in patients with acute respiratory distress syndrome (ARDS) (1113).

Despite these recommendations, conventional practice of oxygen therapy is often more liberal and results in hyperoxia (1420) or in the delivering of supplemental oxygen during nonhypoxemic conditions, without any evidence of benefit (21). For example, the lower limits of the 95% confidence intervals (CI) for daily time-weighted mean SpO2 were greater than 96%, with FiO2 ranging from 0.35 to 0.44, on each of the first 7 MV days observed during standard practice at two Australian ICUs (14, 15). This liberal approach may relate to the perception that, outside of very high levels of FiO2, oxygen therapy is safe. This perception of safety, however, is now being challenged by the increasing recognition of the potential harm of excessive FiO2 (16), hyperoxemia, and tissue hyperoxia (6, 2125). Although a liberal use of oxygen may provide a margin of safety against hypoxia (26), a more conservative approach might reduce potentially harmful exposure to excessive FiO2, hyperoxemia, and tissue hyperoxia. However, the relative merits and risks of these two approaches to oxygen therapy in terms of patient-centered endpoints remain undefined, suggesting the need for RCTs. On the other hand, RCTs focusing on patient-centered outcomes can only be ethically justified if pilot RCTs demonstrate a separation in treatment and protocol compliance (feasibility) and a degree of safety associated with a conservative oxygen therapy approach.

Accordingly, we performed a pilot multicenter, multinational RCT to test the hypothesis that conservative oxygen therapy is feasible, and to obtain preliminary data on the safety of such an approach, with the aim of using such pilot data to inform the design of potential subsequent larger clinical trials.

This prospective randomized parallel-group trial was conducted at four university-affiliated, multidisciplinary ICUs in Australia, New Zealand, and France. The study was prospectively registered (ACTRN12613000505707). The Human Research Ethics Committee at each site approved the study (approval number 12/07/18/4.03, 12/STH/2/AM01, and 121491A-31). Informed consent was obtained from the patient where possible, or from a legal surrogate. This study was monitored by an independent data and safety monitoring board. Additional details of the methods are provided in the online supplement.

ICU patients aged greater than or equal to 18 years were eligible if they had been receiving invasive MV for less than 24 hours and their treating clinician expected MV to continue for at least the next 24 hours. Exclusion criteria included known pregnancy, imminent risk of death, or if the treating clinician lacked equipoise for the patient to be enrolled in this trial.

Randomization was done in a masked fashion, using opaque sealed envelopes, with a unique computer-generated, permuted block randomization method with random block sizes. After treatment allocation, the bedside nurse titrated the FiO2 within a range of 0.21 to 0.80 to achieve the assigned targets of 88–92% SpO2 for the conservative oxygenation group or greater than or equal to 96% SpO2 for the liberal oxygenation group. The study intervention was continued for the entire duration of MV. Positive end-expiratory pressure levels were determined by the treating clinicians in accordance with usual clinical care. The treating ICU physician could alter oxygenation targets at any time if deemed necessary according to the patient’s clinical status. Data on oxygenation parameters and ventilator settings were recorded every 4 hours from Day 0 to 7.

Primary endpoints were the mean area under the curve (AUC) for SpO2, SaO2, PaO2, and FiO2 on Days 0–7. Secondary endpoints were change from baseline (Δ) Sequential Organ Failure Assessment (SOFA) score, Δ PaO2/FiO2, new-onset ARDS (27), Δ creatinine, incidence of hemodynamic instability (i.e., cardiac arrest or addition of ≥2 new vasopressor/inotrope agents), vasopressor-free days, arrhythmia-free days, and ventilator-free days until Day 28, ICU mortality, and 90-day mortality.


Analysis plan and outline of data presentation were prespecified and reported on the trial registration page ( Analysis was conducted on an intention-to-treat basis. Based on our previous observational studies (14, 15), we estimated that a sample size of 100 would provide more than 350 MV days of exposure to both oxygenation strategies. We deemed this sufficient to assess feasibility in this pilot phase. Categorical variables were compared using chi-square tests or Fisher exact tests, and reported as n (%). Continuous normally distributed variables were compared using Student t test and reported as means (SD), whereas nonnormally distributed data were compared using Wilcoxon rank sum test and reported as medians (interquartile range).

The AUC, a summary index of longitudinal data, was assessed as an integrated expression of mean oxygenation levels achieved over the active treatment period, and was determined using mixed linear modeling fitting main effects for group and time. Preplanned subgroup analysis was performed on patients with baseline PaO2/FiO2 less than 300. Survival analysis was presented as Kaplan-Meier curves. Multivariate time-to-event analysis using Cox regression models, adjusted for baseline variables (SOFA score, Acute Physiology and Chronic Health Evaluation III score, chronic obstructive pulmonary disease, and ARDS), was performed with results reported as hazard ratios (95% CI). A two-sided P value of less than 0.05 was considered statistically significant.


We screened 357 patients and enrolled 104 patients between June 2013 and October 2014. Of these, 53 patients were assigned to the conservative oxygenation group and 51 patients to the liberal oxygenation group (Figure 1; see online supplement). One patient in the conservative arm withdrew consent, and was excluded. The remaining 103 patients were followed up to Day 90. Demographic and clinical characteristics at baseline were similar in the two groups (Table 1). Most patients had a medical diagnosis and the mean duration of MV before randomization was 13 ± 7 hours.

Table 1. Baseline Characteristics of the Patients

CharacteristicsConservative Oxygenation Arm (n = 52)Liberal Oxygenation Arm (n = 51)P Value
Age, mean (SD), yr62.4 (14.9)62.4 (17.4)1.00
Male sex, n (%)32 (62%)33 (65%)0.74
BMI, mean (SD), kg/m227.6 (10.3)27.6 (10.1)0.98
Diagnosis type   
 Trauma, n (%)3 (6%)2 (4%)1.00
 Medical, n (%)39 (75%)41 (80%)0.51
 Surgical, n (%)10 (19%)8 (16%)0.64
APACHE III, median (IQR)79.5 (61–92.5)70 (50–84)0.06
SOFA score, mean (SD)7.9 (2.9)7.4 (3.1)0.44
Active smoker, n (%)10 (19%)14 (27%)0.32
COPD, n (%)11 (21%)5 (10%)0.11
Ischemic heart disease, n (%)6 (12%)5 (10%)0.78
Prerandomization MV period, mean  (SD), h13.6 (7.2)13.2 (7.4)0.78
SpO2, mean (SD), %95 (3)96 (3)0.17
SaO2, mean (SD), %95.5 (3)96 (2.7)0.37
PaO2, median (IQR), mm Hg81 (68–109)82 (75–104)0.54
Hemoglobin, mean (SD), g/L110 (23)115 (23)0.3
PaCO2, mean (SD), mm Hg38 (7)39 (6)0.35
pH, mean (SD)7.36 (0.07)7.37 (0.07)0.6
Lactate, median (IQR), mmol/L1.99 (1.4–2.9)1.65 (1.2–2.6)0.24
FiO2, mean (SD), %0.44 (0.2)0.44 (0.18)0.93
PEEP, mean (SD), cm H2O8.2 (3)7.3 (3)0.14
VT, mean (SD), ml/kg PBW8 (1.8)8 (1.9)0.95
PaO2/FiO2, mean (SD)248 (112)247 (113)0.96
ARDS, n (%)17 (33%)10 (20%)0.13
Peak airway pressure, mean (SD),  cm H2O22 (6)21 (5)0.71
Minute ventilation, mean (SD), L9.2 (2.3)9.1 (2.6)0.8

Definition of abbreviations: APACHE III = Acute Physiology and Chronic Health Evaluation score III; ARDS = acute respiratory distress syndrome; BMI = body mass index; COPD = chronic obstructive pulmonary disease; IQR = interquartile range; MV = mechanical ventilation; PBW = predicted body weight; PEEP = positive end-expiratory pressure; SOFA = Sequential Organ Failure Assessment; SpO2 = oxygen saturation as measured by pulse oximetry.

APACHE III is the sum of three components at the time of randomization: an acute physiology score (0–252), chronic health evaluation score (0–23), and age score (0–24), with total score ranging from 0 to 299, where higher score indicates more severe disease.

SOFA score includes subscores ranging from 0 to 4 for each of five organ system (circulation, lungs, liver, kidneys, and coagulation), with score ranging from 0 to 20, and higher scores indicating more severe organ failure.

ARDS as defined according to the Berlin definition.

Process of Care

During the study period, ventilator parameters (tidal volume, minute ventilation, positive end-expiratory pressure, and peak airway pressure) and the net fluid balance did not differ between the two groups (see Table E1 in the online supplement). The percentage time points spent on any mandatory mode of ventilation during the first week of MV in the conservative and liberal arm were 34% and 46%, respectively. The odds ratio, adjusted for repeated measures, for the use of mandatory mode of MV within the first week in the conservative arm, as compared with liberal arm, was 0.36 (95% CI, 0.12–1.04; P = 0.06). Arterial blood gases were performed more often in the conservative versus liberal arm during the first week of MV (see Table E1). There were no significant between-group differences with regard to mean hemoglobin level or the number of units of red cells transfusion during the first week of MV (see Table E1).

Feasibility Outcomes

Participants spent the majority of the time within the intended target range in both groups (Figure 2A). The mean AUC and 95% CI for SpO2, SaO2, PaO2, and FiO2 were significantly lower in the conservative group compared with liberal group (Table 2). Overall, participants spent a median of 6% (interquartile range, 0–25%) time off target, but more time was spent off target in the conservative arm than in the liberal arm (14% vs. 3%; P < 0.001). Daily mean SpO2, PaO2, and FiO2 for the groups were well separated on all 7 days of MV (Figures 2B–2D). Participants in the conservative group spent more time at a FiO2 of 0.21 than those in the liberal group (Figure 3).

Table 2. Primary and Secondary Outcomes

CharacteristicsConservative Oxygenation Arm (n = 52)Liberal Oxygenation Arm (n = 51)P Value
SpO2 for Day 0 to 7, mean AUC* (95% CI), %93.4 (92.9 to 93.9)97 (96.5 to 97.5)<0.001
SaO2 for Day 0 to 7, mean AUC* (95% CI), %93.5 (93.1 to 94)96.8 (96.3 to 97.3)<0.001
FiO2 for Day 0 to 7, mean AUC* (95% CI), %0.26 (0.25 to 0.28)0.36 (0.34 to 0.39)<0.001
PaO2 for Day 0 to 7, mean AUC* (95% CI), mm Hg70 (68 to 73)92 (89 to 96)<0.001
Percentage time spent off target, median (IQR), %14 (4 to 36)3 (0 to 10)<0.001
Incidence of new-onset ARDS, n (%)11 (32%)11 (28%)0.65
Δ§ PaO2/FiO2, mean (SD)50 (97)21 (102)0.15
Δ§ worst PaO2/FiO2, mean (SD)−50 (115)−66 (114)0.46
Number of significant episodes of arterial desaturation per patient||, median (IQR)1 (0 to 5)0 (0 to 0)<0.001
MV-free days until Day 28, mean (SD)14.7 (10.3)16.4 (11.3)0.42
Pneumothorax-free days until Day 28, median (IQR)28 (16 to 28)28 (20 to 28)0.96
Incidence of hemodynamic instability,** n (%)9 (17%)12 (24%)0.43
Vasopressor-free days until Day 28, median (IQR)25.3 (6.7 to 27.3)25.8 (14.6 to 27)0.71
Arrhythmia-free days until Day 28, median (IQR)28 (16 to 28)28 (20 to 28)0.78
Vasopressor dose†† during first week, median (IQR), µg/kg/min0.08 (0.04 to 0.16)0.04 (0.02 to 0.09)0.009
Hours on vasopressors, median (IQR)49 (11 to 101)35 (14 to 73)0.52
Δ§ serum creatinine, mean AUC* (95% CI), μmol/L−5 (−34 to 25)3 (−31 to 37)0.74
RRT-free days until Day 28, median (IQR)28 (9 to 28)28 (11 to 28)0.81
Serum lactate, mean AUC* (95% CI), mmol/L1.9 (1.6 to 2.1)1.7 (1.4 to 1.9)0.23
Δ§ SOFA score, mean AUC* (95% CI)−1.4 (−2.2 to −0.6)−1.9 (−2.7 to −1.1)0.41
ICU length of stay, median (IQR)9 (5 to 13)7 (4 to 12)0.19
Hospital length of stay, median (IQR), d20 (10 to 25)16 (7 to 30)0.80
ICU mortality rate, n (%)13 (25%)12 (24%)0.86
90-d mortality rate, n (%)21 (40%)19 (37%)0.74

Definition of abbreviations: ARDS = acute respiratory distress syndrome; AUC = area under the curve; CI = confidence interval; ICU = intensive care unit; IQR = interquartile range; MV = mechanical ventilation; RRT = renal-replacement therapy; SOFA = Sequential Organ Failure Assessment; SpO2 = oxygen saturation as measured by pulse oximetry.

SOFA score includes subscores ranging from 0 to 4 for each of the five organ system (circulation, lungs, liver, kidneys, and coagulation), with score ranging from 0 to 20, and higher scores indicating more severe organ failure.

*AUC, a summary measure of longitudinal data, was determined using mixed linear modeling fitting main effects for group and time.

Derived as percentage of total time points when SpO2 was not within the alarm limits assigned to each arm and there was further room for FiO2 titration (i.e., 0.21 < FiO2<0.80).

New-onset ARDS was defined as subsequent occurrence of ARDS in those patients who did not have ARDS on Day 0, and where ARDS was defined according to the Berlin definition.

§Δ refers to the change in variable value during Day 0–7 as compared with its baseline value.

||Significant hypoxemic episodes were those episodes recorded by the bedside nurses when SpO2 <86% lasted >5 min.

Event-free days were defined as those days when a patient was alive and free of that event.

**Hemodynamic instability was defined as “cardiac arrest” or “addition of two or more new vasopressor/inotrope agents in a day.”

††Sum total of noradrenaline and adrenaline dose.

Safety Outcomes

There were no significant differences between the groups in regard to any of the measures of organ dysfunction (Δ SOFA score, Δ PaO2/FiO2, new-onset ARDS, Δ creatinine, hemodynamic instability, vasopressor-free days, arrhythmia-free days, or ventilator-free days), or ICU or 90-day mortality (Table 2). Vasopressor dose requirement was lower in the liberal arm, but the vasopressor duration and hospital length of stay were similar in both groups (Table 2). The data on percentage time points per patient spent at different SpO2 or PaO2 thresholds are comprehensively presented in Table 3 and Figures E1–E8. The median number of time points that were spent at different SpO2, SaO2, PaO2, or FiO2 thresholds, and the lowest and the highest values for oxygenation parameters in each group during the study period, are described in Table E2. One percent of SpO2 time points in the conservative arm versus 0.3% of SpO2 time points in the liberal arm were spent at SpO2 less than 88% (P = 0.03). Using the hyperoxia threshold of SpO2 greater than 98% (9, 28) while FiO2 greater than 0.21, 4% of SpO2 values in conservative arm versus 22% of SpO2 readings in liberal arm were in hyperoxic range (P < 0.001). Survival analysis curves for the treatment groups were similar (Figure 4A). The adjusted hazard ratio for death by Day 90 in the conservative arm, as compared with liberal arm, was 0.77 (95% CI, 0.40–1.50; P = 0.44).

Table 3. A Post Hoc Analysis of the Percentage of Time Points* Spent at Specified SpO2 or PaO2 Levels

SpO2 or PaO2 LevelsConservative GroupLiberal GroupP ValueSurvivorsNonsurvivorsP Value
SpO2 <88%, while at FiO2 <1, % (n/N)1% (16/1,526)0.3% (3/1,184)0.0260.9% (14/1,599)0.5% (5/1,111)0.22
PaO2 <55 mm Hg, while at FiO2 <1, % (n/N)7% (72/1,006)1% (7/764)<0.0015% (56/1,074)3% (23/696)0.27
SpO2 >98%, while at FiO2 >0.21, % (n/N)4% (41/933)22% (246/1,138)<0.00114% (191/1,334)13% (96/737)0.61
PaO2 >120 mm Hg, while at FiO2 >0.21, % (n/N)3% (22/641)13% (92/734)<0.0019% (79/889)7% (35/486)0.88

Definition of abbreviation: SpO2 = oxygen saturation as measured by pulse oximetry.

*These were punctual prospective predecided 4-hourly time points.

Prespecified Subgroup Analyses

The subgroup analysis for patients with a baseline PaO2/FiO2 less than 300 is presented in Table E3. The separation in mean FiO2 exposure between the two arms was wider in this subgroup. However, outcomes including survival (Figure 4B) were similar. In this subgroup, the adjusted hazard ratio for death by Day 90 in the conservative arm was 0.49 (95% CI, 0.20–1.17; P = 0.10).

One patient in the conservative arm was treated according to liberal oxygenation protocol in error. However, results were unchanged in per-protocol analysis (see Table E4). To probe further for any signal of major harm associated with oxygenation parameters, survivors and nonsurvivors were compared in a post hoc analysis (Table 3; see Tables E2 and E5), but no significant differences were evident.

Key Findings

In this pilot multicenter randomized clinical trial, we assessed the feasibility of a conservative oxygenation strategy (target SpO2 88–92%) compared with a liberal oxygenation strategy (target SpO2 ≥96%) during invasive MV for adult ICU patients. The study protocol was implemented well. We identified clear separation in the mean SpO2, SaO2, PaO2, and FiO2 values between the two groups, confirming treatment feasibility. The conservative oxygenation arm had a significantly lower incidence of hyperoxemia but a higher incidence of hypoxemia. There were no significant between-group differences in the secondary endpoints of new organ dysfunction or mortality, and the use of a conservative SpO2 target was not associated with harm. In the prespecified subgroup of patients with impaired gas exchange, the between-group separation in mean FiO2 exposure was wider, but outcomes were similar.

Relationship to Previous Studies

In recent years, several observational studies from varied critical care settings have reported liberal use of supplemental oxygen in standard practice (1418). In this regard, the oxygenation levels achieved in the liberal arm of our study were similar to those previously reported in conventional practice by other observational studies (14, 15, 18). However, the percentage time spent with hyperoxia in the liberal arm was lower than previously reported (15, 17). Only a single-center prospective before-and-after feasibility study has compared a conservative oxygenation target with conventional practice (28). Our results are consistent with this study in demonstrating protocol compliance, feasibility, and lack of any major adverse events with a conservative oxygenation strategy. In the aggregate, our study and the previous before-and-after study have now exposed 106 patients to a total of more than 800 MV days of conservative oxygenation strategy. In the before-and-after study, a conservative oxygenation strategy was associated with lower incidence of new organ dysfunction (28). In contrast, we did not find any significant between-group differences in any of the measures of new organ dysfunction. In our study, vasopressor dose requirement was lower in the liberal oxygenation arm, although there was no between-group difference in duration of vasopressor therapy or vasopressor-free days. One explanation of this finding might be related to the vasoconstrictor effect of higher oxygenation levels as previously reported for different vascular beds (2933).

In the conservative oxygenation group, we noticed a trend to lower use of mandatory MV mode, which might indicate earlier attempts to wean patients in response to lower FiO2 requirement. However, it did not result in any difference in the duration of MV or ventilator-free days. In our study, the percentage time spent with hypoxemia was higher in the conservative arm, and the percentage time spent with hyperoxia was higher in the liberal arm. These findings are not unexpected, because the likelihood of finding SpO2 values less than 88% is high with a target SpO2 range of 88–92% when compared with a target SpO2 range of greater than or equal to 96%. Likewise, the lack of an upper limit alarm for SpO2, as is often the case in conventional practice, could have led to more exposure to hyperoxia in the liberal arm. Future studies might consider a closed loop feedback system (34, 35) to titrate FiO2 more closely to the intended SpO2 target range. Although there is no defined threshold for permissive hypoxemia (36), the SpO2 range of 88–92% in the conservative oxygenation arm of our study might be considered an approximate approach of permissive hypoxemia. Indirect evidence suggests that permissive hypoxemia might improve outcomes in some patient groups by reducing the potential dose-dependent adverse effects of the traditional liberal oxygen therapy (22, 23). In our study, the point estimate for 90-day mortality was lower with conservative oxygenation strategy. This is consistent with recent metaanalyses that reported an association between hyperoxia and mortality in some patient subgroups (6, 25).

Implication of the Study Findings

Our study findings support the feasibility of delivering conservative oxygen therapy in patients on invasive MV. Assigned SpO2 targets in this study were achieved by titrating FiO2. The lack of a significant difference in positive end-expiratory pressure levels observed between the treatment groups provides reassurance that this approach is feasible and does not result in a major imbalance of a cointervention. Exposure to hyperoxia was significantly reduced with the conservative approach to oxygen therapy. However, exposure to hypoxemia was also marginally higher. These data, and the data from a previous before-and-after study (28), justify continued and prudent investigation of conservative oxygen therapy. Given the unexpected harm evident from a strategy of lower oxygen targets (SpO2 85–89%) in recent RCTs among preterm infants (3739), safety considerations are paramount. Our preliminary data provide low-level evidence in support of the safety of a conservative oxygen approach (SpO2 88–92%) in adult ICU patients requiring MV. However, because of our small sample size, these data should be regarded as exploratory.

Existing data support the hypothesis that conservative oxygen therapy could potentially reduce the risk of pulmonary oxygen toxicity compared with a liberal approach (4042). Our pilot data may also inform the design of the potential subsequent larger clinical trials (43). In our study the observed SD of ventilator-free days (44) is approximately 0.7 of the mean in both the conservative and liberal oxygen groups. Thus, using the ventilator-free days observed in the liberal (standard care) arm as baseline (i.e., 16.4 days [SD 11.3]) and assuming the SD is the same proportion of mean in the experimental group, a sample size of 800 participants provides 90% power to detect a minimum clinically important difference of 2.6 ventilator-free days (45), using a two-tailed hypothesis at an alpha of 0.05.

Strengths and Limitations

The major strength of this study is its multicenter multinational randomized controlled design. Although two other studies evaluating different oxygenation strategies in the ICU have been completed recently (NCT01319643 [OXYGEN-ICU], NCT01722422 [HYPER-2S]), our study is the first multicenter trial to be reported. Our study endpoints were objective, and a priori specified criteria were used to assess secondary outcomes. Protocol adherence was good, separation clear, and detailed longitudinal data available. Despite SpO2 targets of 88–92%, the percentage time spent with SpO2 less than 88% in the conservative arm was low. Additionally, despite SpO2 targets of greater than or equal to 96% in the liberal arm, the percentage time spent with SpO2 greater than 98% was less than in previous observational studies. Our study design was pragmatic and allowed clinicians the freedom to choose a particular oxygenation target if deemed clinically indicated. We studied a broad group of ICU patients to which supplemental oxygen is administered in routine practice.

Our study, however, has some limitations. First, the intervention was not blinded. We tried to minimize ascertainment bias by collating data on oxygenation levels measured with two independent methods (pulse oximetry measured SpO2 and CO-oximeter measured SaO2 and PaO2). Furthermore, data analysis and data presentation were performed according to a prespecified analysis plan. Second, our study was not adequately powered to test superiority of different oxygenation strategies or to demonstrate safety of the conservative oxygenation strategy. Therefore, the lack of any significant between-group difference in the safety endpoints may represent a type II error. We regard the observed point estimates of effect for all secondary outcomes as hypothesis-generating and our findings do not provide definitive data in relation to safety or the efficacy of either treatment strategy. Furthermore, the differences in secondary outcomes observed in this feasibility study may be attributable to imbalances in baseline variables that were either measured or unmeasured. Third, we did not assess some of the other potentially important endpoints, such as neurocognitive outcomes (46, 47) and incidence of delirium.

Fourth, because treating clinicians were free to alter oxygenation targets, it could have led to some instances where the decision to alter oxygenation targets might have been influenced more by inherent bias rather than scientific evidence. However, the percentage of time points spent off target in the study was modest. Fifth, 69 out of 357 screened patients were excluded because of lack of equipoise. Although we did not collect specific reasons for lack of equipoise, this may reflect clinical conditions where the most appropriate approach to oxygen management is well established, such as exacerbation of COPD. Sixth, we did not measure any biomarker in this study and this could be a subject of further investigations. Seventh, we did not measure plateau pressure and therefore cannot comment on driving pressure, which has recently been suggested as a predictor of outcome in patients with ARDS (48). Lastly, the mean SpO2 levels that were achieved in the conservative arm were higher than the intended target range. This was primarily caused by the limit of FiO2 titration, because it was not possible to titrate FiO2 below 0.21.


In conclusion, our study demonstrates that a conservative oxygenation strategy is a feasible alternative to the usual liberal oxygenation strategy, while being effective in reducing exposure to hyperoxia. These data justify continued and prudent investigation of conservative oxygen therapy.

The authors thank the following members of the data safety and monitoring committee: Professor Richard Beasley, Medical Research Institute of New Zealand; Dr. Bob Ure, ICU, Wellington Regional Hospital, New Zealand; and Professor Mark Weatherall, Wellington School of Medicine and Health Sciences, New Zealand. They thank Ismail Mohamed, Software Engineering Consultant for SETServices, Australia, for database creation. They thank the following for enrollment and data collection: Anna Hunt, ICU, Wellington Regional Hospital, New Zealand; Emma Pollock, ICU, John Hunter Hospital, New Lambton, Australia; Helen Young, ICU, Austin Hospital, Melbourne, Australia; Leah Peck, ICU, Austin Hospital, Melbourne, Australia; Emily Paton, ICU, Austin Hospital, Melbourne, Australia; Lucie Vettoretti, ICU, University Hospital, Besançon, France; Lynn Andrews, ICU, Wellington Regional Hospital, New Zealand; Rachel Whyte, ICU, John Hunter Hospital, New Lambton, Australia; and Sally Hurford, ICU, Wellington Regional Hospital, New Zealand. They thank Andrew Davies, ICU, The Alfred, Prahran, Victoria, Australia, and Carol Hodgson, Senior Research Fellow, ANZIC-RC, Melbourne, Australia, for contribution to study design and grant application.

1. Wunsch H, Linde-Zwirble WT, Angus DC, Hartman ME, Milbrandt EB, Kahn JM. The epidemiology of mechanical ventilation use in the United States. Crit Care Med 2010;38:19471953.
2. Adhikari NK, Fowler RA, Bhagwanjee S, Rubenfeld GD. Critical care and the global burden of critical illness in adults. Lancet 2010;376:13391346.
3. Dasta JF, McLaughlin TP, Mody SH, Piech CT. Daily cost of an intensive care unit day: the contribution of mechanical ventilation. Crit Care Med 2005;33:12661271.
4. Metnitz PG, Metnitz B, Moreno RP, Bauer P, Del Sorbo L, Hoermann C, de Carvalho SA, Ranieri VM; SAPS 3 Investigators. Epidemiology of mechanical ventilation: analysis of the SAPS 3 database. Intensive Care Med 2009;35:816825.
5. Kahn JM, Benson NM, Appleby D, Carson SS, Iwashyna TJ. Long-term acute care hospital utilization after critical illness. JAMA 2010;303:22532259.
6. Damiani E, Adrario E, Girardis M, Romano R, Pelaia P, Singer M, Donati A. Arterial hyperoxia and mortality in critically ill patients: a systematic review and meta-analysis. Crit Care 2014;18:711.
7. Crapo RO, Jensen RL, Hegewald M, Tashkin DP. Arterial blood gas reference values for sea level and an altitude of 1,400 meters. Am J Respir Crit Care Med 1999;160:15251531.
8. Gries RE, Brooks LJ. Normal oxyhemoglobin saturation during sleep. How low does it go? Chest 1996;110:14891492.
9. O’Driscoll BR, Howard LS, Davison AG; British Thoracic Society. BTS guideline for emergency oxygen use in adult patients. Thorax 2008;63:vi1vi68.
10. Slutsky AS. Consensus conference on mechanical ventilation--January 28-30, 1993 at Northbrook, Illinois, USA. Part I. European Society of Intensive Care Medicine, the ACCP and the SCCM. Intensive Care Med 1994;20:6479.
11. The Acute Respiratory Distress Syndrome Network. Ventilation with lower tidal volumes as compared with traditional tidal volumes for acute lung injury and the acute respiratory distress syndrome. N Engl J Med 2000;342:13011308.
12. Meade MO, Cook DJ, Guyatt GH, Slutsky AS, Arabi YM, Cooper DJ, Davies AR, Hand LE, Zhou Q, Thabane L, et al.; Lung Open Ventilation Study Investigators. Ventilation strategy using low tidal volumes, recruitment maneuvers, and high positive end-expiratory pressure for acute lung injury and acute respiratory distress syndrome: a randomized controlled trial. JAMA 2008;299:637645.
13. Mercat A, Richard JC, Vielle B, Jaber S, Osman D, Diehl JL, Lefrant JY, Prat G, Richecoeur J, Nieszkowska A, et al.; Expiratory Pressure (Express) Study Group. Positive end-expiratory pressure setting in adults with acute lung injury and acute respiratory distress syndrome: a randomized controlled trial. JAMA 2008;299:646655.
14. Panwar R, Capellier G, Schmutz N, Davies A, Cooper DJ, Bailey M, Baguley D, Pilcher V, Bellomo R. Current oxygenation practice in ventilated patients: an observational cohort study. Anaesth Intensive Care 2013;41:505514.
15. Suzuki S, Eastwood GM, Peck L, Glassford NJ, Bellomo R. Current oxygen management in mechanically ventilated patients: a prospective observational cohort study. J Crit Care 2013;28:647654.
16. Rachmale S, Li G, Wilson G, Malinchoc M, Gajic O. Practice of excessive F(IO(2)) and effect on pulmonary outcomes in mechanically ventilated patients with acute lung injury. Respir Care 2012;57:18871893.
17. de Graaff AE, Dongelmans DA, Binnekade JM, de Jonge E. Clinicians’ response to hyperoxia in ventilated patients in a Dutch ICU depends on the level of FiO2. Intensive Care Med 2011;37:4651.
18. de Jonge E, Peelen L, Keijzers PJ, Joore H, de Lange D, van der Voort PH, Bosman RJ, de Waal RA, Wesselink R, de Keizer NF. Association between administered oxygen, arterial partial oxygen pressure and mortality in mechanically ventilated intensive care unit patients. Crit Care 2008;12:R156.
19. Parke RL, Eastwood GM, McGuinness SP; George Institute for Global Health; Australian and New Zealand Intensive Care Society Clinical Trials Group. Oxygen therapy in non-intubated adult intensive care patients: a point prevalence study. Crit Care Resusc 2013;15:287293.
20. Ihle JF, Bernard S, Bailey MJ, Pilcher DV, Smith K, Scheinkestel CD. Hyperoxia in the intensive care unit and outcome after out-of-hospital ventricular fibrillation cardiac arrest. Crit Care Resusc 2013;15:186190.
21. Iscoe S, Beasley R, Fisher JA. Supplementary oxygen for nonhypoxemic patients: O2 much of a good thing? Crit Care 2011;15:305.
22. Capellier G, Panwar R. Is it time for permissive hypoxaemia in the intensive care unit? Crit Care Resusc 2011;13:139141.
23. Martin DS, Grocott MP. Oxygen therapy in critical illness: precise control of arterial oxygenation and permissive hypoxemia. Crit Care Med 2013;41:423432.
24. Budinger GR, Mutlu GM. Balancing the risks and benefits of oxygen therapy in critically III adults. Chest 2013;143:11511162.
25. Helmerhorst HJ, Roos-Blom MJ, van Westerloo DJ, de Jonge E. Association between arterial hyperoxia and outcome in subsets of critical illness: a systematic review, meta-analysis, and meta-regression of cohort studies. Crit Care Med 2015;43:15081519.
26. Panwar R, Young P, Capellier G. Conservative oxygen therapy in mechanically ventilated patients. Crit Care Med 2014;42:e630e631.
27. Ranieri VM, Rubenfeld GD, Thompson BT, Ferguson ND, Caldwell E, Fan E, Camporota L, Slutsky AS; ARDS Definition Task Force. Acute respiratory distress syndrome: the Berlin Definition. JAMA 2012;307:25262533.
28. Suzuki S, Eastwood GM, Glassford NJ, Peck L, Young H, Garcia-Alvarez M, Schneider AG, Bellomo R. Conservative oxygen therapy in mechanically ventilated patients: a pilot before-and-after trial. Crit Care Med 2014;42:14141422.
29. Gao Z, Spilk S, Momen A, Muller MD, Leuenberger UA, Sinoway LI. Vitamin C prevents hyperoxia-mediated coronary vasoconstriction and impairment of myocardial function in healthy subjects. Eur J Appl Physiol 2012;112:483492.
30. Dallinger S, Dorner GT, Wenzel R, Graselli U, Findl O, Eichler HG, Wolzt M, Schmetterer L. Endothelin-1 contributes to hyperoxia-induced vasoconstriction in the human retina. Invest Ophthalmol Vis Sci 2000;41:864869.
31. Kawamura J, Meyer JS, Terayama Y, Weathers S. Cerebral hyperemia during spontaneous cluster headaches with excessive cerebral vasoconstriction to hyperoxia. Headache 1991;31:222227.
32. Moulder PV, Lancaster JR, Harrison RW, Michel SL, Snyder M, Thompson RG. Pulmonary arterial hyperoxia producing increased pulmonary vascular resistance. J Thorac Cardiovasc Surg 1960;40:588601.
33. McNulty PH, Robertson BJ, Tulli MA, Hess J, Harach LA, Scott S, Sinoway LI. Effect of hyperoxia and vitamin C on coronary blood flow in patients with ischemic heart disease. J Appl Physiol (1985) 2007;102:20402045.
34. Tehrani FT, Bazar AR. A feedback controller for supplemental oxygen treatment of newborn infants: a simulation study. Med Eng Phys 1994;16:329333.
35. Iobbi MG, Simonds AK, Dickinson RJ. Oximetry feedback flow control simulation for oxygen therapy. J Clin Monit Comput 2007;21:115123.
36. Gilbert-Kawai ET, Mitchell K, Martin D, Carlisle J, Grocott MP. Permissive hypoxaemia versus normoxaemia for mechanically ventilated critically ill patients. Cochrane Database Syst Rev 2014;5:CD009931.
37. Schmidt B, Whyte RK, Asztalos EV, Moddemann D, Poets C, Rabi Y, Solimano A, Roberts RS; Canadian Oxygen Trial (COT) Group. Effects of targeting higher vs lower arterial oxygen saturations on death or disability in extremely preterm infants: a randomized clinical trial. JAMA 2013;309:21112120.
38. Stenson BJ, Tarnow-Mordi WO, Darlow BA, Simes J, Juszczak E, Askie L, Battin M, Bowler U, Broadbent R, Cairns P, et al.; BOOST II United Kingdom Collaborative Group; BOOST II Australia Collaborative Group; BOOST II New Zealand Collaborative Group. Oxygen saturation and outcomes in preterm infants. N Engl J Med 2013;368:20942104.
39. Carlo WA, Finer NN, Walsh MC, Rich W, Gantz MG, Laptook AR, Yoder BA, Faix RG, Das A, Poole WK, et al.; SUPPORT Study Group of the Eunice Kennedy Shriver NICHD Neonatal Research Network. Target ranges of oxygen saturation in extremely preterm infants. N Engl J Med 2010;362:19591969.
40. Freeman BA, Crapo JD. Hyperoxia increases oxygen radical production in rat lungs and lung mitochondria. J Biol Chem 1981;256:1098610992.
41. Phillips M, Cataneo RN, Greenberg J, Grodman R, Gunawardena R, Naidu A. Effect of oxygen on breath markers of oxidative stress. Eur Respir J 2003;21:4851.
42. Carpagnano GE, Kharitonov SA, Foschino-Barbaro MP, Resta O, Gramiccioni E, Barnes PJ. Supplementary oxygen in healthy subjects and those with COPD increases oxidative stress and airway inflammation. Thorax 2004;59:10161019.
43. Chow SSJ, Wang H, editor. Sample size calculations in clinical research. New York: Chapman and Hall/CRC; 2007.
44. Schoenfeld DA, Bernard GR; ARDS Network. Statistical evaluation of ventilator-free days as an efficacy measure in clinical trials of treatments for acute respiratory distress syndrome. Crit Care Med 2002;30:17721777.
45. McAuley DF, Laffey JG, O’Kane CM, Perkins GD, Mullan B, Trinder TJ, Johnston P, Hopkins PA, Johnston AJ, McDowell C, et al.; HARP-2 Investigators; Irish Critical Care Trials Group. Simvastatin in the acute respiratory distress syndrome. N Engl J Med 2014;371:16951703.
46. Mikkelsen ME, Christie JD, Lanken PN, Biester RC, Thompson BT, Bellamy SL, Localio AR, Demissie E, Hopkins RO, Angus DC. The adult respiratory distress syndrome cognitive outcomes study: long-term neuropsychological function in survivors of acute lung injury. Am J Respir Crit Care Med 2012;185:13071315.
47. Hopkins RO, Weaver LK, Collingridge D, Parkinson RB, Chan KJ, Orme JF Jr. Two-year cognitive, emotional, and quality-of-life outcomes in acute respiratory distress syndrome. Am J Respir Crit Care Med 2005;171:340347.
48. Amato MB, Meade MO, Slutsky AS, Brochard L, Costa EL, Schoenfeld DA, Stewart TE, Briel M, Talmor D, Mercat A, et al. Driving pressure and survival in the acute respiratory distress syndrome. N Engl J Med 2015;372:747755.
Correspondence and requests for reprints should be addressed to Rakshit Panwar, M.D., D.N.B., M.B. B.S., Intensive Care Unit, John Hunter Hospital, Lookout Road, New Lambton, Newcastle, New South Wales 2305, Australia. E-mail:

Supported by the Intensive Care Foundation (ANZ), University Hospital Besançon (France), and Don Du Souffle (France).

Author Contributions: R.P. and M.H., full access to all the data in the study and responsibility for the integrity of the data. R.P., M.H., G.M.E., R.B., and M.B., access to database, data analysis, and interpretation. R.P., R.B., G.C., G.M.E., and P.J.Y., study concept and design. R.P., M.H., L.B., G.M.E., and P.J.Y., responsibility for data acquisition at respective sites. R.P., manuscript preparation and drafting. R.P., R.B., P.J.Y., and P.W.J.H., critical revision of the manuscript for important intellectual content. R.P. and M.B., statistical analysis. R.P., R.B., G.C., P.J.Y., and P.W.J.H., obtainment of funding. R.P., M.H., G.M.E., R.B., G.C., P.J.Y., and P.W.J.H., administrative, technical, or material support.

This article has an online supplement, which is accessible from this issue's table of contents at

Originally Published in Press as DOI: 10.1164/rccm.201505-1019OC on September 3, 2015

Author disclosures are available with the text of this article at

Comments Post a Comment

New User Registration

Not Yet Registered?
Benefits of Registration Include:
 •  A Unique User Profile that will allow you to manage your current subscriptions (including online access)
 •  The ability to create favorites lists down to the article level
 •  The ability to customize email alerts to receive specific notifications about the topics you care most about and special offers
American Journal of Respiratory and Critical Care Medicine

Click to see any corrections or updates and to confirm this is the authentic version of record