Protocol for a systematic review and individual participant data meta-analysis of optimizing oxygen therapy in critically ill patients

Background Oxygen therapy is a cornerstone treatment of critically ill patients in the intensive care unit (ICU). Whether lower oxygenation therapy brings superior survival outcomes to higher oxygenation therapy is unknown. Methods We will search electronic databases: PubMed, Embase, Web of Science, the Cochrane Central Register of Controlled Trials (CENTRAL), International Clinical Trials Registry Platform (ICTRP), and ClinicalTrials.gov from inception to 1 January 2024. Two authors will independently screen for all eligible clinical studies. Emails will be sent for individual participant data. The statistical analyses will be conducted using STATA 15.0 software. Results We will evaluate the efficacy of lower oxygenation therapy compared with higher oxygenation therapy based on individual participant data. Conclusion This study will offer clinical evidence for oxygen therapy in ICU patients.


Introduction
Oxygen is commonly used in medical settings, especially for critically ill patients who may have an increased need for oxygen (1).However, it is important to note that too much oxygen, or hyperoxia, can actually be harmful to some patients (2).This is particularly true for those who have had a myocardial infarction (MI) or have been resuscitated from cardiac arrest.Recent studies have shown that hyperoxia can cause further damage to the heart in patients with ST-elevation MI who are not experiencing hypoxia (3).In addition, arterial hyperoxia after cardiac arrest may lead to higher rates of in-hospital mortality (4,5).In a preplanned secondary analysis of targeted hypothermia vs. targeted normothermia after out-of-hospital cardiac arrest, Robba et al. (6) found that the time exposure of hyperoxemia was significantly associated with mortality.A systematic review and meta-analysis examining the association of hyperoxemia with survival and neurological outcomes included 10 observational studies of patients with refractory cardiogenic shock or refractory cardiac arrest treated with venoarterial extracorporeal membrane oxygenation.Tigano et al. (7) found that severe hyperoxemia may be associated with worse survival and neurological outcomes in these patients.As a result, medical professionals now recommend a peripheral oxygen saturation (SpO 2 ) level of 94-98% for these patients (8).Two systematic reviews and meta-analyses on trials of oxygen therapy have been performed (9,10).However, recent studies, such as the ICU-ROX study, the PILOT study, the LOCO 2 study, and the HOT-ICU study, were not included (11)(12)(13)(14).To our knowledge, there has been no meta-analysis of individual participant data from trials of oxygen therapy in critically ill patients.In this study, we will evaluate the efficacy of low oxygen therapy in critically ill patients to provide evidence for oxygen therapy in the ICU.

Study registration
This meta-analysis was registered with PROSPERO on 30 September 2023 (registration number CRD42023464558).The Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-IPD) will be followed.

The inclusion criteria 2.2.1 Types of studies
Only randomized controlled trials (RCTs) are considered for inclusion.Animal experiments, case reports, non-RCTs, secondary analysis of RCTs, and reviews will be excluded.

Types of participants
Adults admitted into ICUs, including cardiovascular ICUs, neurological ICUs, surgical ICUs, medical ICUs, and general ICUs, are eligible for our study.

Types of interventions
The interventions are lower oxygenation therapy and higher oxygenation therapy.

Types of outcomes
The main outcome is 28-day mortality.The secondary outcomes are 60-day mortality and 90-day mortality.

Collection and analysis of data 2.3.1 Search strategy
Author XY will carry out a thorough search in electronic databases: PubMed, Embase, Web of Science, the CENTRAL, ICTRP, and ClinicalTrials.govfrom inception to 1 January 2024.

Selection of studies
All authors will study PRISMA-IPD.Authors YO and JX will independently review the titles and abstracts of all retrieved studies for their eligibility, and the references for other possible eligible studies.All repetitions and studies that do not meet the enrollment criteria will be excluded.The included studies will then be cross-checked, and any uncertainties will be resolved by discussion with authors XY and YS.Emails will be sent by authors XY and YS to the corresponding authors to request data on gender, age, race, country, intervention group, the presence of mechanical ventilation at enrollment, the presence of shock and the type of shock at enrollment, the presence of MI at enrollment, SpO 2 , PaO 2 , time to death since enrollment, and the living status at 28th day, 60th day, and 90th day of each participant from all included studies.For those who do not respond, another five emails will be sent 1 week or 2 weeks after the previous email.Any ambiguous information will be cleared by discussion with the corresponding authors.

Assessment of risk of bias and quality of evidence
The risk of bias and quality of evidence will be assessed using Cochrane's "Risk of bias 2" tool.The following domains will be involved: sequence generation, allocation concealment, blinding, and completeness of outcomes and measures.The diagram of this study is shown in Figure 1.

Statistical analysis 2.4.1 Synthesis of data
If there are two groups of higher oxygenation therapy, they will be combined as one higher oxygenation therapy group.Individual Frontiers in Medicine frontiersin.orgparticipant data will be combined according to the intention-totreat group.

Measures of effect
Author XY will perform the statistical analyses with Stata 15.0 and its command called IPDMETAN, which was designed for two-stage individual participant data meta-analyses of any measures of effect.The hazard ratio (HR) and a 95% confidence interval (CI) of mortality will be calculated.

Assessment of heterogeneity
The heterogeneity will be assessed by Cochrane's Q-test and I 2 will be presented.

Sensitivity analysis
Sensitivity analysis will be conducted in two ways.First, aggregated data from studies without individual participant data will be included to perform a sensitivity analysis.Second, per-protocol individual participant data will be combined to conduct a metaanalysis.Per-protocol is defined as the oxygenation target within the predefined interval of each group from each included study.

Assessment of reporting bias
A funnel plot and the Egger test will be used to assess reporting bias.

Subgroup analysis
Subgroup analyses will be conducted as per primary analysis if sufficient data are available: age ≥65 years vs. age <65 years, invasive mechanically ventilated (IMV) vs. non-IMV, shock vs. non-shock, acute myocardial infarction (AMI) vs. non-AMI, stroke vs. non-stroke, and cardiac arrest vs. non-cardiac arrest at enrollment.

Discussion
Oxygen therapy has been used, and the toxicity of supranormal oxygen has been recognized for more than a century (15).Providing the appropriate amount of oxygen is a balance of potential benefits and risks.For some critically ill patients, oxygen is a life-or-death therapy, and both hypoxia and hyperoxia are associated with an increased risk of death (16).The benefits and risks of appropriate oxygen therapy are most likely to be established in ICU patients because of the convenience of targeted oxygen therapy and the relatively higher mortality compared with patients in general wards.In recent years, several large randomized trials comparing low and high oxygen therapy in critically ill patients have been reported, including the ICU-ROX trial, the PILOT trial, the LOCO 2 trial, and the HOT-ICU trial (11)(12)(13)(14).A metaanalysis including these trials is needed.
Compared with meta-analysis of aggregated data, metaanalysis of individual participant data is more powerful (17).
Another strength of this meta-analysis will be the two ways of sensitivity analysis.It is not easy to obtain individual participant data.Meta-analysis of combinations of individual participant data from some trials and aggregated data from other trials is an important supplement to meta-analysis of individual participant data.In the higher oxygenation group, the oxygenation target is easily achieved.However, in the lower oxygenation group, the actual oxygenation index is usually higher than the predefined upper limit of the target.For example, the predefined upper limit of the SpO 2 target was 92% in Panwar's et al. ( 18) study and 90% in the PILOT study (12); the actual SpO 2 was 93.4 and 94%, respectively, in the two studies.
The different kinds of patients are a challenge for oxygen therapy and this meta-analysis.We aim to include all types of patients admitted to different types of ICUs.The overall effect of oxygen therapy on mortality will be examined.Subgroup analyses will be performed, and the results may serve as hints for further randomized trials.

FIGURE 1 Flow
FIGURE 1Flow diagram of studies selection and data synthesis.