We included studies that tested the effects of metformin on ALI or ARDS in in vivo studies, irrespective of whether there was a clinical trial or not: 1) preclinical experiments or randomized controlled trials; 2) animal models preclinically compatible with ARDS or patients with ARDS; 3) prophylactic or therapeutic administration of metformin as the intervention; 4) routine treatment or placebo as the control; and 5) the primary outcome was indicators of ARDS such as W/D and lung injury scores. The second outcome was mortality.

The exclusion criteria are as follows: 1) Observational studies; 2) duplicated publication; 3) in vitro or ex vivo study; 4) metformin combined with other medicines; 5) compared metformin with other medicines; 6) evidence of lung injury was obtained 7 days after induction in preclinical experiments; 7) evidence of lung injury could not be found in the literature; and 8) data could not be extracted or obtained even after contacting the original authors by email.

From their inception to July 2023, databases including PubMed, Web of Science, Embase, Scopus, and the Cochrane Central Register of Controlled Trials were inquired to retrieve the relevant literature by using search terms as follows: 1) metformin, dimethylbiguanidine, dimethylguanylguanidine, and glucophage; 2) acute lung injury, ALI, ARDS, acute respiratory distress syndrome, and respiratory distress syndrome (Supplementary Table S1). Moreover, ProQuest Dissertations and Theses Global, Google scholar, and Baidu Scholar were also inquired. We also examined reference lists of reviews and potential eligible studies to find relative studies.

The retrieved literature was screened using endnote software (Clarivate Analytics, United States) first to delete duplicates. Then, titles and abstracts of remains were screened by pairs of reviewers. After that, the full texts of potential eligible studies were reviewed and judged. If any conflicts exist, it would be adjudicated by a third reviewer. Reference lists were also examined to find the potential eligible literature.

ALI in animals was defined by using authors’ definitions or evidence of lung injury, such as lung histopathological analysis, protein content or neutrophils in BALF, and W/D, provided within 7 days after ALI induction. ARDS was defined by using Berlin definition (Ranieri et al., 2012).

To evaluate the methodological quality of the eligible literature, the Systematic Review Center for Laboratory Animal Experimentation (SYRCLE) risk of bias tool (Hooijmans et al., 2014) or Cochrane’s “Risk of bias” tool (Higgins et al., 2011) was used for preclinical studies or randomized controlled trials, respectively.

We adopted a standardized form for data extraction. The following items were extracted by two reviewers independently: characteristics of studies including the first author’s names and countries, the publication year, animals, ALI models, and strategies of metformin administration. The primary outcome was lung injury indicators such as lung injury scores, W/D, indicators of pulmonary edema or neutrophil infiltration in the lung, and an index of oxidative stress. The secondary objective was mortality. The primary and secondary objectives were evaluated as pulmonary edema, inflammation, and destruction of lung tissues are the evidence of lung injuries, and mortalities are the clinical concerns.

The standard error of the mean value was converted to the standard deviation value. The most effective result was pooled, if results from multiple time points or various doses intervention were provided in one study.

We performed meta-analysis using STATA SE 14.0 statistical software for Windows (StataCorp LP, TX, United States). Homogeneity between studies was judged using Cochran’s Q statistic and I ² statistic. It denotes a significant or high heterogeneity, if p ≤ 0.10 (Cochran’s Q statistic) or I ² statistic ≥50%, respectively. A random-effects model will be used if significant or high heterogeneity was identified. Otherwise, a fixed-effects model was used. We also performed subgroup analysis to investigate potential heterogeneity.

We present the characteristics of 15 preclinical studies in Table 1. As shown in Table 1, the preclinical studies were published between 2008 and 2023. Healthy mice or rats were used to establish ALI models, which were induced by using lipopolysaccharide (LPS), cecal ligation and puncture, paraquat, or gas explosion. Almost half of the studies used rats including Sprague Dawley rats and Wistar rats to establish animal models. In addition, three types of mice were used in other studies. Either prophylactic or therapeutic administration of metformin was investigated in 14 studies. One study reported both the prophylactic and therapeutic administration of metformin. Results from the therapeutic administration of metformin were reported in seven studies. Metformin was administrated via intraperitoneal injection in most studies. Oral administration of metformin was investigated in three studies. Lung injuries were evaluated from 8 h to 7 days after induction. Nine studies reported lung injury indicators at less than or equal to 24 h. Two studies reported indicators of lung injuries at 7 days.

The risk of bias from preclinical studies was assessed using SYRCLE’s risk of bias tool. Detailed information is shown in appendix Supplementary Table S1.

Fifteen preclinical studies reported the effects of metformin on experimental ARDS. Lung injury was evaluated with the indicators of lung edema, inflammatory cell infiltration, oxidative stress, and lung histopathological analysis. In general, a declining value of indicators suggests an improved result, in addition to the value of anti-oxidative enzymes. W/D is a sensitive indicator for the measurement of lung edema. The W/D was reported in eight studies. Metformin effectively decreased W/D [SMD = −2.67 (−3.53 to −1.81); Figure 2]. Potent publication bias was investigated using the funnel plot (Figure 3). Protein content in BALF could be used to evaluate endothelial permeability. The protein content in BALF was reported in two studies. He et al. (2019) reported the results of protein content in BALF using two different types of animal models, a total of three results were pooled. The pooled result suggests that metformin dampened protein content in BALF [SMD = −3.74 (−6.76 to −0.72); Figure 4]. MDA is an oxidative stress indicator, while superoxide dismutase (SOD) is an important anti-oxidative stress substance. Three studies reported both MDA and SOD. As shown in Figure 4, pooled results showed that metformin reduced MDA [SMD = −1.98 (−3.77 to −0.20)] but did not elevate SOD [SMD = 3.17 (−0.15–6.49)]. MPO is an enzyme that is mainly generated in neutrophils. The activity of MPO is used as an indicator of neutrophil activation and infiltration. The pooled result of MPO including four studies showed reduced MPO activity in metformin-treated animals [SMD = −3.15 (−4.79 to −1.52); Figure 4]. In parallel, counts of neutrophils in BALF were also decreased in metformin-treated animals [SMD = −3.47 (−4.69 to −2.26) and I² = 0%]. The lung injury score is a quantified strategy to investigate the extent of lung injury based on pulmonary tissue sections. Six studies reported lung injury scores in animals that received metformin versus vehicle [SMD = −4.19 (−5.65 to −2.74) and I² = 69.1%; Figure 5].

Mortality is a major concern of ARDS treatment. The effect of metformin on mortality was reported in experimental ARDS animal models. Seven studies reported mortality at 24 h after the induction of experimental ARDS in animals that received metformin versus vehicle [RR = 0.32 (0.15–0.68); Figure 6]. Moreover, improved mortality by metformin at 48 and 72 h was reported in four studies (Figure 6).

It is interesting that a traditional anti-diabetic agent can reduce lung injury. How did metformin do it? Several studies have shed light on it.

Refractory hypoxemia is a feature of ARDS. Increased endothelial permeability due to damage of alveolar capillaries contributes to substantial edema fluid gathering in the alveolar space and pulmonary interstitium that causes impairment of gas exchange (Meyer et al., 2021). Activation of AMP-activated protein kinase (AMPK)-1 isoforms dampened the deterioration of endothelial permeability triggered by LPS (Jian et al., 2013). Metformin can activate AMPK-1 isoforms and reverse the LPS-induced decrease in the pulmonary microvascular endothelial cell barrier resistance and impairment of wound repair in the endothelium that resulted in the improvement of endothelial permeability and pulmonary edema (Jian et al., 2013). Moreover, enhanced transendothelial resistance of endothelial monolayers by metformin was reported in an in vitro study (Uddin et al., 2021). In an ex vivo study, the high-pressure ventilation-induced increase in microvascular permeability and edema formation in rabbit lung was notably improved in the metformin-pretreated group (Tsaknis et al., 2012).

Inflammasome activation contributes to inflammation. Metformin inhibited LPS-triggered activation of the NLRC4 inflammasome by upregulating AMPK phosphorylation (He et al., 2019). Activation of AMPK by metformin may also trigger mitophagy-mediated inhibition of NLRP3 inflammasome activation, as metformin is capable of induction of mitophagy (Boyle et al., 2018). Moreover, the LPS-induced increase in cleaved caspase-1, IL-1β production, and the N-terminal fragment of gasdermin D was reduced in metformin-treated animals (Zhang Y. et al., 2022). The effect of metformin on pyroptosis was associated with elevating sirtuin-1 expression and inhibiting NLRP3 inflammasome activation (Zhang Y. et al., 2022). In LPS-challenged macrophages, the inhibiting effects of metformin on NLRP3 inflammasome activation and pyroptosis were also detected (Xian et al., 2021). However, the inhibiting effect of metformin on NLRP3 inflammasome activation in LPS-challenged macrophages was not AMPK-dependent but ATP-dependent mitochondrial DNA synthesis (Xian et al., 2021). Mitochondrial DNA is the ligand of NLRP3. The conjunction of mitochondrial DNA and NLRP3 resulted in NLRP3 inflammasome activation (Zhong et al., 2018). As metformin inhibits ATP production (Wheaton et al., 2014), a decrease in the ATP-dependent mitochondrial DNA production was detected in metformin-treated and LPS-primed macrophages (Xian et al., 2021).

NF-κB triggers a key role in inflammation such as the regulation of cytokine production. LPS-stimulated NF-κB p65 subunit phosphorylation, degradation of IκB-α, and accumulation of NF-κB in the nucleus of neutrophils or alveolar macrophages could be inhibited by metformin via inhibiting mitochondrial respiratory complex I activation (Zmijewski et al., 2008).

Moreover, in the present meta-analysis, although pooled results showed that metformin did not increase SOD levels; the SOD level was elevated in two of the three trials included. As an anti-oxidative stress substance, SOD plays an important role in protecting the lung. In addition to having anti-oxidative effect within cells, extracellular superoxide dismutase in alveolar fluid is protective for methicillin-resistant Staphylococcus aureus inoculation-triggered inflammation and acute lung injury (Sul et al., 2023).

Collectively, anti-inflammatory and anti-edematogenic effects of metformin favor the amelioration of lung injury.

Improved mortality at 24–72 h was detected in pooled preclinical studies. Further subgroup analysis showed that only pre-treatment of metformin has benefits on 24-h mortality in indirect ALI. However, post-treatment has more clinical practice meanings. To our surprise, a high dose of metformin (≥100 mg/kg) did not improve 24 h mortality in indirect ALI; on the contrary, a low dose of metformin (<100 mg/kg) significantly reduced 24-h mortality. Its potential mechanisms are still needed to be elucidated. These preclinical studies only reported mortality during a short period. Two retrospective clinical studies reported that all-cause mortality at 30 or 90 days was not improved in preadmission metformin users compared with metformin non-users in diabetic patients with ARDS (Jo et al., 2017; Oh and Song, 2020). Nevertheless, there are several limitations in the two clinical studies. First, all the two clinical studies were observational and conducted in the same country, Korea. Importantly, it is unclear that whether the preadmission metformin users continued using metformin during ARDS. Commonly, oral anti-diabetic agents will be substituted with insulin when diabetic patients suffer a critical illness. Accordingly, both the previous metformin users and non-users will use insulin, giving the result no meaning.

In addition to its anti-diabetic use, metformin has shown its effects in treating a variety of disorders such as cancer, polycystic ovary syndrome, and COVID-19 (Triggle et al., 2022). For critical illness, metformin has been suggested as a potent therapeutic intervention for patients with severe traumatic brain injury (Taheri et al., 2019). Improved mortality was reported in septic patients with diabetes mellitus that were treated with metformin (Li et al., 2021; Gómez et al., 2022). This evidence suggests the potential use of metformin for critically ill patients. Nevertheless, well-designed randomized controlled trials are still needed to assess the effects of metformin on ARDS in diabetic or non-diabetic patients.