Page 9 - GTM-2-3

Global Translational Medicine Hydrogen for acute lung injury

(v) Other: Two evaluators (LYK and MY) independently extracted
(a) Experimental studies. the data and evaluated the quality scores of each study. Any
(b) Published between 2010 and 2021, disagreements were resolved by the chief reviewer (GSZ).
(c) Published in any publication language,
(d) The cited literature sources should be retrievable. 2.5. Data synthesis and statistics analysis
The meta-analysis was performed using Stata software
2.2.2. Exclusion criteria (version 14.0) and GraphPad Prism (version 9.4). The
The exclusion criteria are as follows: outcomes of W/D and PaO were evaluated using the
2
(i) Types: weighted mean difference (WMD) model to assess the
(a) Studies of non-animal experiments about ALI, improvement in lung function. For MDA and SOD,
(b) Publications that did not comply with or the same model was used to evaluate the antioxidant
adequately describe standard laboratory animal effect of hydrogen, while TNF-α was assessed using the
rules. standardized mean difference (SMD) model to analyze
(ii) Intervention: Did not use 2% H or hydrogen saline hydrogen’s anti-inflammatory effects. In cases where
2
but used other therapy to treat ALI animals. studies reported standard error (SE) instead of standard
(iii) Control: Group(s) given placebo or no treatment. difference (SD), Equation I was used to convert the values:
(iv) Other: SD = SE × √N (I)
(a) The same author published the same articles at
different times, 2.6. Statistical heterogeneity test and exploration
(b) Unclear literature sources, Study heterogeneity was assessed using the Q statistic and
(c) Data are only displayed in figures and unable to I² statistic . Significant heterogeneity was considered
[17]
contact the author to obtain original data. present if the I² value exceeded 50% or if the P-value
After defining our strategy, each study was randomly was less than the predefined significance level (α = 0.05).
given to two independent authors (LYK and MY) and was Conversely, non-significant heterogeneity was indicated
selected by studies of title and abstract. Disagreements when the I² value was 50% or below, and P ≥ α = 0.05.
between the two reviewers were solved by discussing with Considering the possible high heterogeneity due to the
the third author (GSZ). differences between interventions, we used subgroup
[18]
analysis to explore the heterogeneity . Moreover, the
2.3. Study selection and data extraction diversity of species may affect the heterogeneity of the
The information of the included studies, including meta-analysis results. Therefore, meta-regression was
authors, year of publication, species of animal subjects, used to explore the heterogeneity attributable to species.
and language, were recorded and considered in the study Meta-regression required a minimum of ten studies to
[19]
selection based on the inclusion and exclusion criteria. examine one characteristic . Consequently, the present
In addition, we extracted data, including the number of paper included studies using W/D and species as the
animals in the experimental and control groups, types of covariate to explore the sources of heterogeneity (n = 12
ALI animal models, interventions, and outcomes. Data > 10). Publication bias was assessed using Egger’s test,
extraction followed predefined rules, with one author and the stability of the results from the included studies
(LYK) responsible for study data collection. In case of was analyzed through the trim and fill funnel plot and
[18]
uncertainty, the chief author (GSZ) was consulted, and the sensitivity analysis .
final decision was reached collaboratively by all reviewers.
2.7. Grade evidence assessment
2.4. Assessment of Methodological Quality We followed the PRISMA checklists and used the GRADE
In the present paper, we evaluated the research quality of profiler (version 3.6) to evaluate the grade evidence of
included studies using the CAMARADES checklist, which included indicators. All the included studies were animal
covered 10 items related to a series of animal experiment experiments, and their grades were downgraded by 5 impact
designs . Due to differences in the research, three factors. The grade evidence for these included indicators
[16]
evaluation indicators were modified as follows: was divided into 4 levels: (i) High quality, indicating high
(i) Item no. 4: Blinded induction of model, confidence in results; (ii) moderate quality, indicating
(ii) Item no. 6: Use of anesthetic agents without significant general confidence in results; (iii) low quality, indicating
intrinsic neuroprotective activity, limited confidence in certain areas; and (iv) very low
(iii) Item no. 7: Approval of animal ethics for creating the quality, indicating the confidence could not be promised or
ALI model. predicted. The work was completed by GSZ, LYK, and MY.

Volume 2 Issue 3 (2023) 3 https://doi.org/10.36922/gtm.0379