Article Text

Download PDFPDF

Closed-loop oxygen control for patients with hypoxaemia during hospitalisation: a living systematic review and meta-analysis protocol
  1. Caroline Gomes Mol1,
  2. Aléxia Gabriela da Silva Vieira1,
  3. Bianca Maria Schneider Pereira Garcia1,
  4. Emanuel dos Santos Pereira1,
  5. Raquel Afonso Caserta Eid1,
  6. Ana Carolina Pereira Nunes Pinto2,3,
  7. Ricardo Kenji Nawa1
  1. 1Hospital Israelita Albert Einstein, São Paulo, SP, Brazil
  2. 2Biological and Health Sciences Department, Universidade Federal do Amapá, Macapá, AP, Brazil
  3. 3Evidence-Based Health Program, Universidade Federal de São Paulo - UNIFESP, São Paulo, SP, Brazil
  1. Correspondence to Ricardo Kenji Nawa; ricardo.nawa{at}


Introduction Oxygen is the most common drug used in critical care patients to correct episodes of hypoxaemia. The adoption of new technologies in clinical practice, such as closed-loop systems for an automatic oxygen titration, may improve outcomes and reduce the healthcare professionals’ workload at the bedside; however, certainty of the evidence regarding the safety and benefits still remains low. We aim to evaluate the effectiveness, efficacy and safety of the closed-loop oxygen control for patients with hypoxaemia during the hospitalisation period by conducting a systematic review and meta-analysis.

Methods and analysis MEDLINE, CENTRAL, EMBASE, LILACS, CINAHL and LOVE evidence databases will be searched. Randomised controlled trials and cross-over studies investigating the PICO (Population, Intervention, Comparator and Outcome) framework will be included. The primary outcomes will be the time in the peripheral oxygen saturation target. Secondary outcomes will include time for oxygen weaning time; length of stay; costs; adverse events; mortality; healthcare professionals’ workload, and percentage of time with hypoxia and hyperoxia. Two reviewers will independently screen and extract data and perform quality assessment of included studies. The Cochrane risk of bias tool will be used to assess risk of bias. The RevMan V.5.4 software will be used for statistical analysis. Heterogeneity will be analysed using I2 statistics. Mean difference or standardised mean difference with 95% CI and p value will be used to calculate treatment effect for outcome variables.

Ethics and dissemination Ethical approval is not required because this systematic review and meta-analysis is based on previously published data. Final results will be published in peer-reviewed journals and presented at relevant conferences and events.

PROSPERO registration number CRD42022306033.

  • Adult intensive & critical care
  • RESPIRATORY MEDICINE (see Thoracic Medicine)

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Strengths and limitations of this study

  • This study will be reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols guidelines.

  • The steps of data screening, extraction and methodological quality assessment will be performed by two reviewers independently.

  • Standardised methodological evaluation tools will be used to assess the risk of bias of included studies in the review.

  • The absence of sufficient high-quality studies, heterogeneity in the interventions, high missing or dropout and small sample size might be the limitations for this systematic review.


Oxygen is vital for cellular metabolism and it is considered to be the most common drug used in critical care patients to correct episodes of hypoxaemia.1 2 Low levels of oxygen in the arterial blood are frequently associated with impairment of adequate gas exchange1 2, and prolonged cellular hypoxia promotes rapid and severe organ injuries triggered by natural adaptive mechanisms.2 3 Thus, supplemental oxygen administration can be considered a life-saving treatment, and may reduce the morbidity and mortality associated with hypoxaemia.1 4

Despite the benefits of oxygen therapy indication, both hypoxaemia and hyperoxia, have potential harmful side effects and complications.5–8 The literature suggests that safe and acceptable targets of peripheral oxygen saturation (SpO2) are between 92% and 98% for patients without lung diseases, and 88% and 92% for patients with previous lung diseases.9 10 However, patients’ need for oxygen varies during hospital stay, and the manual adjustment to promote adequate oxygen delivery titration has been shown to be ineffective.9 A precise delivery oxygen method for maintaining the SpO2 within the target is challenging.10–12 It is even more challenging when we look at patients requiring invasive mechanical ventilation support admitted to the intensive care unit (ICU) who often need supplemental oxygen administration during ICU and hospital stay.1 2 4

The use of artificial intelligence (AI) and machine learning (ML) is increasing in health science to make predictions, improve the interpretation of monitored data and support decision-making.10 Closed-loop systems are part of these advances, using a feedback principle to maintain a given variable around a desired set point.10–12 Delivery oxygen devices based on closed-loop technology have been developed and used in patient care in order to provide a real-time adjustment of oxygen titration, based on patients’ SpO2 preventing episodes of hypoxaemia or hyperoxia.11 12 The adoption of new technologies of AI and ML in clinical practice, may reduce the healthcare professionals’ workload at the bedside; however, there is low certainty evidence for their safety and benefits.11 12 It is still unclear whether closed-loop oxygen control devices could improve clinical outcomes, and with the technological advances, new randomised clinical trials (RCTs)13–21 have been published since the last two systematics reviews were conducted.11 12 Thus, the aim of this systematic review is to investigate the effectiveness, efficacy and safety of the closed-loop oxygen control for patients during hospitalisation.

Methods and analysis

This study is reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA-P) statement22 and followed the recommendations of the Cochrane Collaboration Handbook. The protocol is registered at the International Prospective Register of Systematic Reviews (PROSPERO).23

Patient and public involvement

No patient or public involved. Only data already existent in the literature and the aforementioned sources will be used in this study. Patients and/or the public will not be involved in the design, conduct, reporting or dissemination plans of this research.

Search strategy

A search strategy was initially designed for the Medical Literature Analysis and the Retrieval System Online—MEDLINE, via PubMed by an information specialist, responsible for assisting the authors, searching potential studies for inclusion in their reviews, and for keeping up to date with Cochrane methodological developments in information retrieval. The search strategy was independently peer-reviewed by the information specialist, and afterwards will be adapted for use into five databases, as follows: (1) Cochrane Central Register of Controlled Trials (CENTRAL)—via Wiley; (2) Excerpta Medica dataBASE (EMBASE)—via Elsevier; (3) the Latin American the Caribbean Literature in Health Sciences (LILACS)—via Virtual Health Library; (4) Cumulative Index to Nursing and Allied Health Literature (CINAHL)—via Elton Bryson Stephens Company (EBSCO); and (5) LOVE evidence databases. A handsearching will be performed to check preprints, editorials about the included studies, errata of published articles and references lists from the included studies and any relevant systematic review identified. We will track the randomised controlled trials in progress on a specific website ( and on the WHO website. There will be no restrictions to any specific language, date or type of publication. The detailed search strategy for MEDLINE—via PubMed is shown in table 1. The study selection process will be conducted by two reviewers independently, and any disagreement between the reviewers will be resolved by consensus or by consulting a third reviewer.

Table 1

Search strategy for MEDLINE via PubMed

Inclusion criteria

The eligibility criteria were determined using the PICO (Population, Intervention, Comparator and Outcome) framework.24 The studies will be considered eligible based on the following inclusion criteria, as follows: (1) population: hospitalised adult patients requiring supplemental oxygen—either for patients with hypoxaemia (SpO2<92%) or with acute chronic hypoxaemia (SpO2<88%); (2) type of interventions: any devices that allow an automatic oxygen delivery, including invasive and non-invasive devices; low and high flow oxygen devices; (3) type of comparison: manual adjustments of oxygen; (4) type of outcome: time within the SpO2 target, oxygen weaning time, length of stay, costs, adverse events, mortality, healthcare professionals workload – process optimisation, and percentage of time with hypoxia and hyperoxia. Two reviewers (CGM and AGV) will independently assess the titles, abstracts and full-text published RCTs without language restriction.

Study selection

The reviewers will identify and exclude duplicates and collate multiple reports of the same study so that each study, rather than each report, is the unit of interest in the review. For the selection process we will use Rayyan—a web and mobile application for systematic reviews software.25 The selection process will be recorded in sufficient detail to complete the PRISMA flow diagram (figure 1). Two review authors (CGM and AGV) will independently screen the titles and abstracts of all the potential studies we identify as a result of the search. All the potential full-texts of the articles that fulfilled eligibility criteria will be included. If a consensus is not reached, a third reviewer (ACP) will be consulted to solve potential disagreements regarding the included articles.

Figure 1

Preferred Reporting Items for Systematic Reviews and Meta-Analysis flow diagram describing the search strategy.

Outcome measures

The primary outcome of interest will be the time spent within the SpO2 target range. Secondary outcomes of interest will be the time for oxygen weaning time, length of stay, costs, adverse events, mortality, healthcare professionals' workload, percentage of time with hypoxia and hyperoxia.

Data extraction

Two reviewers (CGM and AGV) will independently extract the data on a standard worksheet. Data will be extracted from each included study using a standardised spreadsheet developed at Microsoft Excel, as follows: authors, year, protocol number, Digital Object Identifier (DOI), study type, country of publication; the participants demographics (ie, age, gender, inclusion and exclusion criteria, number of participants, diseases, severity scores, severity of condition, comorbidity, phase of hypoxaemia), Interventions—type of device and form of delivery (ie, mechanical ventilation or conventional oxygen therapy), duration of intervention, follow-up, Comparators and Outcomes—defined in this review. We will also extract variables, as follows: time spent within the target SpO2 and other relevant variables to answer the review question (ie, oxygen weaning time, length of stay, costs, adverse events, mortality, health professionals’ workload—process optimisation). Additionally, we will extract data from the funding, sponsorship of the included studies and notable conflicts of interest of trial authors. For missing data, we will contact the corresponding authors of the studies through the email provided. In case of crossover studies inclusions, we will consider a paired analysis or if carry-over is thought to be a problem, the first period will be used to perform the analysis. If the data are homogeneous for conducting meta-analyses, one review author will transfer data into the Review Manager (RevMan) V.5.4 software. We will double-check if the data are correctly entered by comparing the data presented in the systematic review with those in the study report.

Methodological quality assessment

The risk of bias of the included trials will be assessed using the Cochrane Risk of Bias 2 (RoB2) tool for randomised trials.

Risk of bias

Assessment of the risk of bias of individual studies will be performed as recommended by the Cochrane Collaboration Handbook.26 The RoB227 will be used to evaluate the risk of bias according to five domains: (1) bias arising from the randomisation process; (2) bias due to deviations from intended interventions; (3) bias due to missing outcome data; (4) bias in measurement of the outcome; and (5) bias in selection of the reported result. Each domain will be considered within one of the three levels, as follows: “low risk of bias”, “some concerns” or “high risk of bias”.27 We will involve a third reviewer (ACP) if a consensus cannot be reached. With the concurrence of the reviewers on the final judgement of all the included trials, the result will be displayed in a table or graph.

Assessment of bias in conducting the systematic review

The review will be conducted according to this published protocol and any deviation will be reported in the ‘differences between protocol and review’ section of the systematic review.

Assessment of certainty of evidence

The Grading of Recommendations Assessment, Development and Evaluation (GRADE) system will be used to measure and summarise the overall current evidence of each outcome.28 The GRADE system consists of five items: (1) study limitations—risk of bias; (2) inconsistency of results (heterogeneity); (3) indirectness of evidence; (4) imprecision in effect estimates and (5) reporting bias. The quality of evidence will be classified into four categories, as follows: “high”, “moderate”, “low” and “very low”; and it will be related to the studies that contributed data to the main prespecified outcomes. All analyses will be performed using GRADEpro Guideline Development Tool (GRADEpro GDT) software.29 Two authors will rate it independently and a third author will address any discrepancy found in the study.

Data synthesis

RevMan V.5.4 (Cochrane Collaboration) software will be used to conduct the meta-analysis if appropriate—that is: statistically and clinically homogeneous. A random-effects model will be used. The mean difference or standardised mean difference will be used to analyse continuous variables with 95% CI. Dichotomous outcomes will be presented as risk ratios with 95% CI. Heterogeneity among included trials will be measured using the I2 statistic. If it is identified as substantial heterogeneity, we will report and explore it through a prespecified subgroup analysis. In addition, sensitivity analysis will be performed through separate analyses of studies judged to have a high risk of bias or a methodological weakness considered important. In cases where the combination of data does not make it possible to do the meta-analysis, we will carry out only a qualitative synthesis of each included study, of the ongoing studies identified in our search, and of the publication bias analysis.

Subgroup analysis

We plan to perform analysis of subgroups, as follows: underlying disease; hypoxaemia stage (acute vs chronic); SpO2 target (threshold <92% versus >92%), and devices (mechanical ventilation vs non-invasive devices).

Ethics statements

Patient consent for publication



  • Contributors Conceptualisation—AGV, CGM and RKN. Protocol writing—AGV, CGM and RKN. Methodology—AGV, ACP, BMG, CGM and RKN. Project administration—AGV, ACP, CGM and RKN. Supervision—AGV, CGM and RKN. Validation—AGV, ACP, CGM and RKN. Writing—original draft—AGV, ACP, BMG, CGM and RKN. Writing—review and editing—ACP, AGV, BMG, CGM, EDSP, RAC and RKN.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.