Effect of liberal or conservative oxygen therapy on the prognosis for mechanically ventilated intensive care unit patients: a meta-analysis

ABSTRACT BACKGROUND: For critically ill patients, physicians tend to administer sufficient or even excessive oxygen to maintain oxygen saturation at a high level. However, the credibility of the evidence for this practice is unclear. OBJECTIVE: To determine the effects of different oxygen therapy strategies on the outcomes of mechanically ventilated intensive care unit (ICU) patients. DESIGN AND SETTING: Systematic review of the literature and meta-analysis conducted at Jiangxi Provincial People’s Hospital, Affiliated to Nanchang University, Nanchang, China. METHODS: We systematically searched electronic databases such as PubMed and Embase for relevant articles and performed meta-analyses on the effects of different oxygen therapy strategies on the outcomes of mechanically ventilated ICU patients. RESULTS: A total of 1802 patients from five studies were included. There were equal numbers of patients in the conservative and liberal groups (n = 910 in each group). There was no significant difference between the conservative and liberal groups with regard to 28-day mortality (risk ratio, RR = 0.88; 95% confidence interval, CI = 0.59-1.32; P = 0.55; I2 = 63%). Ninety-day mortality, infection rates, ICU length of stay, mechanical ventilation-free days up to day 28 and vasopressor-free days up to day 28 were comparable between the two strategies. CONCLUSIONS: It is not necessary to use liberal oxygen therapy strategies to pursue a higher level of peripheral oxygen saturation for mechanically ventilated ICU patients. Conservative oxygen therapy was not associated with any statistically significant reduction in mortality.

Despite this, the guidelines available regarding oxygen therapy standards and targets are contradictory and inconsistent. [20][21][22] Studies on this topic have evaluated the effects of different oxygen therapy strategies on the prognosis for mechanically ventilated patients. However, the conclusions that they reached have not been completely coherent. 17,[23][24][25][26] Therefore, we decided to conduct a secondary analysis.

OBJECTIVE
We performed a systematic review of the literature to determine the effects of different oxygen therapy strategies on the outcomes of mechanically ventilated intensive care unit (ICU) patients.

We followed the Preferred Reporting Items for Systematic
Reviews and Meta-analyses (PRISMA) statement 27 and the Cochrane Handbook 28 for the design, method and presentation of the results of this systematic review and meta-analysis.

Database search
In this systematic review and meta-analysis, we searched the PubMed, Embase, Cochrane Library and Web of Science databases. The following keywords were used for the search: "oxygen inhalation therapy", "liberal*", "conservative*", "conventional*", "respiration", "artificial" and "mechanical ventilation". We set the publication type to clinical trial only, and the publication language was limited to English. We searched for related literature from the time of database inception up to and including July 25, 2021. The search strategy is presented in Appendix 1.

Study selection
Two authors independently assessed all titles and abstracts for inclusion and then assessed the full texts of the studies considered.
The studies included had to satisfy the following criteria. leave-one-out sensitivity analysis was conducted.

Studies retrieved and included
We identified 200 studies from PubMed, Embase, Cochrane Library and Web of Science. After screening the titles and abstracts, 16 studies were included for full-text review. In three of the studies, some patients were not mechanically ventilated. In two studies, the number of deaths and mortality rate were not reported. In three other studies, oxygen therapy strategies could not be classified. In the end, four randomized controlled trials (RCTs) and one cohort study 17,[23][24][25][26] were included in the meta-analysis (Figure 1).

Study characteristics and quality evaluation
The main characteristics of the eligible RCTs and cohort study are shown in Table 1. Five studies and 1806 mechanically ventilated ICU patients were included in the meta-analysis. The quality of the studies included in this meta-analysis was medium. The quality of the studies included, as assessed using the Cochrane risk-of-bias tool is shown in Figure 2. Because the interventions needed the cooperation of doctors, there was a lack of use of blinding methods. As such, there may have been some bias during implementation of the interventions.

Primary outcomes
Short-term mortality is shown in Figure 3. Three studies 17,25,26 provided data regarding 28-day mortality. Since there was high heterogeneity among the studies (P = 0.07; I 2 = 63%), the random-effects model was adopted. The result showed that there was no statistical significance in 28-day mortality between the conservative and liberal groups (RR = 0.88; 95% CI = 0.59-1.32, P = 0.55). Sensitivity analysis was performed to evaluate the effect of a single study on the overall estimate by sequentially excluding each study. The heterogeneity decreased significantly (I 2 = 24%; P = 0.25). After excluding one of the studies 25 and making adjustments, oxygen therapy strategy was found to be significantly associated with 28-day mortality, such that the conservative group performed better than the liberal group (RR = 0.78; 95% CI = 0.63-0.98; P = 0.03) (Figure 4).

Secondary outcomes
Medium-term mortality is shown in Figure 5. Four studies [23][24][25][26] provided data regarding 90-day mortality. As there was high heterogeneity among the studies (P = 0.1; I 2 = 53%), the randomeffects model was adopted. The result showed that there was no New infections are shown in Figure 6. Three studies 17,25,26 provided data regarding the rate of new infections. Since there was no significant heterogeneity among the studies (P = 0.28; I 2 = 22%), the fixed-effect model was adopted. The result showed that there was no statistically significant difference in the rate of new infections between the conservative and liberal groups (RR = 0.91; 95% CI = 073-1.13; P = 0.73).
ICU length of stay is shown in Figure 7. Two studies 23,26 provided data regarding ICU length of stay. As there was no significant heterogeneity among the studies (P = 0.18; I 2 = 45%), the fixed-effect model was adopted. The result showed that there was no statistically significant difference in the ICU length of stay between the conservative and liberal groups (mean difference, MD = 0.15; 95% CI = -1.52-1.81; P = 0.86).
The mechanical ventilation-free time within 28 days is shown in Figure 8. Three studies 23,24,26 provided data regarding the mechanical ventilation-free time within 28 days. Since there was no significant heterogeneity among the studies (P = 0.18; I 2 = 42%), the fixed-effect model was adopted. The result showed that there was no statistically significant difference in mechanical ventilation-free     The vasopressor-free time within 28 days is shown in Figure 9.
Three studies 23,24,26 provided data regarding the vasopressor-free time within 28 days. Since there was no significant heterogeneity among the studies (P = 0.15; I 2 = 48%), the fixed-effect model was adopted. The result showed that there was no statistically significant difference in vasopressor-free time within 28 days between the conservative and liberal groups (MD = 0.79; 95% CI = -0.71-2.30; P = 0.3).
The risk of bias in the studies included is shown in Figure 10.
The funnel plot of the result showed that the primary outcome was symmetrical. Hence, there was no evidence of significant small-sample effects or publication bias.

DISCUSSION
This systematic review and meta-analysis enrolled 1806 mechanically ventilated ICU patients. All the studies included were considered to be of high quality. Despite the high heterogeneity,     In clinical practice, oxygen therapy has been widely used to prevent or correct arterial hypoxemia for mechanically ventilated ICU patients. Due to concerns over the possible adverse outcomes of hypoxia exposure among critically ill patients, liberal oxygen therapy and hyperoxia are widely used for mechanically ventilated ICU patients. One study reported that 59% of patients have oxygen saturation greater than 98% most of the time. 32 However, according to the formula of oxygen delivery (DO 2 = cardiac output × arterial oxygen content; arterial oxygen content = (Hb × 1.34 × SaO 2 ) + (0.0031 × PaO 2 )), oxygen delivery is governed by three key factors: arterial saturation (SO 2 ), cardiac output (CO) and hemoglobin (Hb). It is unreasonable to only use SaO 2 as the indicator for evaluating gas exchange in hypoxemic patients. Moreover, the oxygen dissociation curve of hemoglobin is "S-shaped": the upper part of the curve is very gradual, which means that it is very difficult to further increase SaO 2 by increasing blood oxygen content and PaO 2 in the upper part. For example, even when the patient's PaO 2 is increased, at the risk of hyperoxia exposure, from 100 mmHg to 150 mmHg, only an incremental increase (200 ml/l to 201.5 ml/l) in the blood oxygen content results from this. 9 It has also been reported that hyperoxia results in decreased heart rate, reduced CO and increased vascular resistance. 33 Therefore, liberal oxygen therapy that only focuses on arterial oxygen saturation when increasing the oxygen delivery is unhelpful. Hyperoxia caused by liberal oxygen therapy may even be harmful. It can promote production of reactive oxygen species and expression of inflammatory cytokines, thus increasing the risk and severity of pneumonia, 11 epithelial and endothelial damage 13 and pulmonary interstitial edema. 12 The results from the meta-analysis confirm that in acutely ill patients, liberal oxygen therapy is unhelpful and does not improve patient outcomes, but may increase mortality. When the range of SpO 2 is more than 94-96%, patients may be affected adversely. 4 Recent studies have shown that conservative oxygen therapy has no significant adverse effect on ICU patients with respiratory failure and hypoxic ischemic encephalopathy. 34,35 Conservative oxygen therapy is relatively safe for critically ill ICU patients.
Thus, oxygen therapy should be restricted. The goal of oxygen therapy should be to ensure adequate oxygen delivery while minimizing any unnecessary hyperoxia exposure. However, the question is how conservative it should be. The ideal situation is that supplemental oxygen administration should be guided through assessment of tissue oxygen delivery and consumption. However, these two parameters are difficult to obtain in clinical practice.
In clinical trials, conservative oxygen therapy is usually carried out by keeping SpO 2 at the lower limit of normality. 24,25 To define conservative oxygen therapy solely on the basis of SpO 2 seems to ignore assessment of oxygen consumption.
As surrogate parameters for oxygen consumption, blood lactate concentration, central venous-to-arterial CO 2 difference and central venous or mixed venous oxygen saturation can also help in implementation of conservative oxygen therapy. Over recent years, there has been a conservative trend in oxygen therapy practice in some hospitals. 36 However, for mechanically ventilated critically ill ICU patients, there is a lack of consensus and explicit guiding criteria regarding the use of conservative oxygen therapy. Clinicians who worry about hypoxemia will still increase the patient's oxygen saturation as much as possible, even at levels exceeding what they think is reasonable, 8 even though these clinicians are aware of the potential harm of liberal oxygen therapy.
The results from our study on mechanically ventilated ICU patients showed that there was no significant difference in clinical prognosis between use of liberal and use of conservative oxygen therapies. Conservative oxygen therapy did not result in additional risk; therefore, it is feasible and safe. It is worth mentioning that there was great heterogeneity regarding 28-day mortality among the studies reviewed here. By excluding each study one by one, we found that the heterogeneity arose from the study by Barrot et al. 25 Excluding Barrot's study decreased the heterogeneity (I 2 = 24%).
After adjustments, the results from the meta-analysis revealed that conservative oxygen therapy reduced short-term mortality (RR = 0.78; 95% CI = 0.63-0.98; P = 0.03) (Figure 4). The reasons for this may have been related to the fact that the study population comprised acute respiratory distress syndrome (ARDS) patients.
Such patients are characterized by difficult-to-correct hypoxemia.
Hypoxemia arises from a diverse range of factors. 37 There may be little difference in clinical prognosis between liberal and conservative oxygen therapy use until the pathological basis of ARDS has been effectively improved. Moreover, the target for conservative oxygen therapy in the studies reviewed here was set at 88%-92% blood oxygen saturation, which is close to the lower limit recommended in ARDS guidelines. 38,39 In practice, there was some deviation between the actual and target oxygen saturation. This would undoubtedly have increased the risk of hypoxia exposure in the conservative group. The adverse events of mesenteric ischemia seen in the conservative group may indicate that conservative oxygen therapy close to the lower limit recommended may have been inappropriate. Furthermore, apart from three studies 17,26 that only partially included patients with ARDS and the sample population in the study of Barrot et al., 25 all the patients included were classified as presenting ARDS. The results suggest that there may have been a discrepancy between ARDS patients and non-ARDS patients regarding the prognosis from conservative oxygen therapy. In other words, this could imply that conservative oxygen therapy is beneficial for reducing short-term mortality among mechanically ventilated patients who do not present ARDS.

Limitations
The findings reported in this study must be interpreted with caution because of several limitations. Firstly, the definitions of conservative oxygen therapy and liberal oxygen therapy were not quite concordant in the studies that we enrolled, and this may have led to inaccuracies in the relative mortality rates between the conservative and liberal groups. Secondly, the number of studies included was relatively small and, therefore, subgroup analysis according to ARDS status was not possible. Thirdly, we assumed that the respiratory function of mechanically ventilated patients in ICUs would be severely impaired. However, some patients received MV for extrapulmonary reasons, and it was not possible to exclude these patients.

CONCLUSIONS
Liberal oxygen therapy and higher SpO 2 for mechanically ventilated ICU patients are not necessary. For partial MV patients, conservative oxygen therapy was not associated with a statistically significant reduction in mortality.