This systematic review will be conducted according to the Cochrane Handbook for Systematic Review of Interventions, the Preferred Reporting Items for Systematic Reviews, Meta-Analyses (PRISMA) statement.19,20 The Grading of Recommendations Assessment, Development, and Evaluation (GRADE) system will be used to assess the evidence in cumulative evidence.21 This protocol paper was prepared according to PRISMA-P checklist which was uploaded as a supplementary material.22
Type of study
Randomized controlled trials (RCTs) regardless of cluster or individual randomization between January 1, 1980 and May 31, 2020 will be included restricted to those published in English. In addition, full-text studies, those published as abstracts only, and unpublished data are eligible for inclusion. We will only include RCTs with supplemental oxygen during exercise training for COPD and exclude RCTs related to ambulatory oxygen or long-term oxygen therapy.
Type of participants
We will include participants 18 years of age or older, with a diagnosis of COPD diagnosed according to the GOLD criteria.1 We will exclude participants with other respiratory diseases.
Type of intervention and comparators
The active treatment group will receive supplemental oxygen during exercise training using wall oxygen units in a hospital or clinic, portable oxygen cylinders, liquid oxygen canisters, and oxygen concentrators.
The control group will be provided sham gas, such as compressed air, through a wall unit or cylinders or will receive no intervention under room air.
The exercise training performed in both groups is defined as a supervised or unsupervised inpatient, outpatient, home-based intervention that includes some form of exercise training applied to COPD patients.
Types of outcome
Primary outcome
- Exercise capacity: 6-minute walking distance (6MWD: m) or shuttle walking distance (SWT: m) or other measures, e.g., peak oxygen consumption. (Peak VO2: mL/kg/min)
- Maximum exercise load of exercise training, e.g., workload (Watts).
- Dyspnea scores at the end of the exercise test, e.g. Borg score.
- Health-related QOL assessed by questionnaires, e.g., 36-item Short Form Health Survey (SF-36).
Search methods for identification of studies
Electronic searches
This meta-analysis will be carried out according to the PRISMA Statement.19 We will conduct searches for studies with inception dates up to May 31, 2020 in the Cochrane Central Register of Controlled Trials (CEN-TRAL), MEDLINE, and EMBASE electronic databases. We will also search abstracts and meeting presentations utilizing the same search terms from Medical Subject Headings (MeSH). The search strategy will be based on discussions with an information specialist, and will be modified appropriately for each database. The search strategy includes a combination of free text words, words in titles/abstracts, and medical subject headings, such as “pulmonary disease, chronic obstructive,” “supplemental oxygen,” “oxygen inhalation therapy,” and “exercise.’’ Searches are limited to peer-reviewed research involving human participants. The detailed strategy and details of the formula that will be used to search the databases are shown in table 1.
Searching other resources
We will check the reference lists of all primary studies and review articles for additional references. Further, we will contact experts in the field to ask if they know of any ongoing or unpublished trials.
Selection of studies
Two review authors (SK and SY) will independently screen the titles and abstracts of all potentially relevant studies identified by the search, and code them as “retrieve” (eligible or potentially eligible/unclear) or “do not retrieve.” In the case of disagreements, other review authors will be asked to arbitrate (KN and TY). SK and SY will retrieve the full text reports/publications, and independently screen the full texts and identified studies for inclusion. SK and SY will also identify and record reasons for the exclusion of ineligible studies. Any disagreements will be resolved by discussion or by consulting other review authors (KN and TY) when necessary. The review authors will identify and exclude duplicates, and collate multiple reports of the same study so that the unit of interest in the review will be each study, rather than each report. We will record the selection process in sufficient detail to complete a PRISMA flow diagram and “Characteristics of excluded studies” table. A data collection form piloted in at least one study in the review will be used to record study characteristics and outcome data. Two review authors (SK and SY) will extract characteristics from the included studies as follows:
- Methods: study design, total duration of study, duration of follow-up, number of subjects, study setting, and date of study.
- Participants: sample size, mean age, sex, history of smoking, history of oxygen therapy, forced expiratory volume in 1 s (FEV1) after bronchodilator administration, and arterial partial pressure of oxygen (PaO2) at baseline.
- Interventions: type of supplemental oxygen device, oxygen supplemental settings, such as continuous flow or demand delivery systems, and their flow rates, type of intervention as control, such as compressed air (and the flow rate), or room air, type of exercise training as frequency, duration, and contents.
- Outcomes: exercise capacity, maximum exercise load of exercise training, dyspnea scores at the end of exercise test, health-related QOL assessed by questionnaire survey.
- Notes: funding for trial and conflicts of interest of the trial authors.
Data collection of studies
Two review authors (SK and SY) will independently extract outcome data from the included studies and check each other’s data extraction. One review author (SK) will transfer data into the Review Manager 5.3 file (RevMan 2019) and the EZR statistical program.23 SK and SY will double-check that data have been entered correctly by comparing the data presented in the systematic review with the study reports. A second review author (KN) will spot-check the study characteristics for accuracy against the trial report.
Assessment of risk of bias
Two review authors (SK and SY) will independently assess risk of bias for each study using the criteria outlined in the (RoB version 2) in the Cochrane Handbook for Systematic Reviews of Interventions.19 Any disagreements will be resolved by discussion, by involving other review authors (KN and YT), or by contacting the authors of the included studies. We will assess risk of bias according to the following domains:
- Bias arising from the randomization process
- Bias due to deviations from intended interventions
- Bias due to missing outcome data
- Bias in measurement of the outcome
- Bias in selection of the reported result
We will grade each potential source of bias as low risk, some concerns, or high risk and provide a quote from the study report together with a justification for our judgment in the “Risk of bias” table. We will summarize the risk of bias judgments across different studies for each of the domains listed.
Measures of treatment effect
We will analyze continuous data as the mean difference (MD) or standardized mean difference (SMD) with 95% confidence interval (95% CI). The MD is the absolute difference between the mean change before and after the intervention in a trial. The SMD is used as a summary statistic in the meta-analysis when the studies all assessed the same outcome but measured it in a variety of ways. We will enter data presented as a value with a consistent direction of effect. When the standard deviation (SD) for change from the baseline is not available, we will calculate the SD using Review Manager (RevMan) version 5.3. All CIs have two-sided probability coverage of 95%. A p-value < 0.05 will be considered significant. We will carry out sensitivity analysis when there is a high risk of bias that could affect the results.
Assessment of heterogeneity
We will assess statistical heterogeneity in each meta-analysis with Tau2, I2, and Chi2 statistics following the Cochrane Handbook for Systematic Reviews.19 We regard heterogeneity as substantial with I2 > 0% and either the Tau2 is > zero or p < 0.10 in the c2 test for heterogeneity. Data from each study will be pooled with fixed-effects modeling when there is no heterogeneity (p > 0.1, I2 ≤ 40%). We will perform meta-analyses with random-effects models when there is heterogeneity (p < 0.1, I2 > 40%).
Subgroup analysis
We plan to carry out the following subgroup analyses for primary outcomes if we obtain an I2 score > 50%.
- Severity of hypoxemia of rest at baseline (PaO2 ≤ 60Torr or PaO2 > 60Torr)
- Severity of exercise-induced desaturation at baseline (exercise induced desaturation: either decrease of SpO2 ≥ 4% and nadir SpO2 during exercise ≤ 88% or not-exercise induced desaturation)
- Severity of airflow limitation (%FEV1 ≥ 50% or %FEV1 < 50%)
We plan to explore differences in outcomes in these subgroups if the number of collected studies is sufficient.
Sensitivity analysis
We plan to carry out sensitivity analysis for primary outcomes if the high risk of bias of some of the included studies affected the results. We will define “high risk” in each study as at least one domain evaluated as “high risk” or some domains are allocated as “some concern” that can decrease the reliability of the results.18 We plan to carry out the following sensitivity analyses.
Meta-regression
If there is any statistically heterogeneity or methodological heterogeneity, we will perform meta-regression using PaO2 at baseline (Torr) and decreases of SpO2 in exercise tests as covariates to investigate whether the severity of desaturation at baseline affects the effectiveness of exercise tolerance.
Assessment of reporting biases
When 10 or more studies are included in a meta-analysis, we will create a funnel plot and examine its asymmetry visually to explore any publication bias.
Assessment of evidence in individual study
We will use the GRADE criteria to assess the quality of evidence for each outcome according to four levels (high, moderate, low, or very low).21 The quality of evidence will be reduced by any one of the following limitations: risk of bias, inconsistency, indirectness, imprecision and publication bias. Two investigators (SK, SY) will independently conduct those assessment. Investigators will resolve disagreements between the two investigators through discussion, with a third reviewer available for adjudication (KN).