Association of preadmission metformin use and mortality in patients with sepsis and diabetes mellitus: a systematic review and meta-analysis of cohort studies

Sepsis is a life-threatening organ dysfunction caused by dysregulated host responses to infection [1]. Because the high mortality due to sepsis remains a major medical problem [2, 3], exploring the mechanism of its development is particularly important. To date, the exact mechanism remains unclear, but it is widely postulated that the release of inflammatory factors by innate immune cells plays an important role in sepsis disease pathogenesis [4, 5]. Interestingly, recent studies indicated that the active molecular responses in inflammation require intensive metabolic support, and modulation of the metabolic pathways could become a novel strategy to restrict inflammatory diseases [6].

Metformin is a reagent with extensive and strong metabolic regulatory activities, and it is often used as a first-line antidiabetic drug for the treatment of type 2 diabetes [7, 8]. In addition to its well-known hypoglycaemic activities, increasing evidence has suggested that metformin inhibits the expression of pro-inflammatory factors in vitro and ameliorates inflammatory injuries in vivo [9‐14]. The mechanisms underlying the pharmacological effects of metformin remain unknown. It has been suggested that metformin inhibits the activity of the mitochondrial respiratory complex, which decreases the generation of ATP and activates adenosine 5′-monophosphate (AMP)-activated protein kinase (AMPK) [15]. Furthermore, several studies have suggested that AMPK activation by drugs or small molecular compounds protects against experimental sepsis in animals [16‐18], demonstrating that AMPK plays an irreplaceable role in the pathogenesis of sepsis. Moreover, AMPK has been regarded as a major target mediating the effects of metformin.

Given the evidence of AMPK activation by metformin, metformin is a potential treatment for sepsis. In recent years, some studies [19, 20] have suggested that in patients with sepsis and diabetes mellitus, the preadmission metformin users had higher lactate levels but lower mortality than the non-metformin users. Geen et al. [21] reported similar lactate levels between preadmission metformin users and non-metformin users, but metformin users still had significantly lower mortality than nonusers. However, some studies [22, 23] showed that preadmission metformin use had no significant effect on mortality compared with non-metformin use in septic patients with diabetes mellitus. Therefore, it is necessary to consolidate the available information to assess whether metformin is beneficial for improving the outcomes of sepsis in patients with diabetes mellitus.

This meta-analysis included 1282 patients and demonstrated that preadmission metformin users had lower mortality than nonusers in patients with sepsis. This is the first systematic review and meta-analysis to describe the association between preadmission metformin use and mortality in septic adult patients with diabetes mellitus. Our finding indicates that metformin may have therapeutic potential for patients with sepsis and diabetes mellitus.

A beneficial association between metformin and mortality has already been described in selected ICU patients with chronic heart failure and liver disease [28‐31], and a retrospective study [31] of 17 Danish ICUs found that 30-day mortality was lower in metformin users than in non-metformin users, which was the same result as that found in this meta-analysis. In addition, many experimental studies [32‐37] have reported that metformin can ameliorate sepsis or endotoxaemia-induced injuries in various organs and many inflammatory diseases. For example, Wu et al. [38] reported that metformin can alleviate acute lung injury, and Liu et al. [13] showed that metformin can attenuate acute myocarditis. However, some studies [39, 40] have shown that metformin cannot decrease but instead increases the mortality rate of septic animals. And unfortunately, to date, the association between preadmission metformin use and mortality in septic patients with diabetes mellitus remains controversial. Some studies [20, 22, 23] have shown that in patients with sepsis, preadmission metformin users were not significantly different from nonusers, whereas the studies by Green et al. [21] and Doenyas-Bara et al. [19] suggested that the mortality was significantly lower in preadmission metformin users than in nonusers among patients with sepsis. Therefore, this meta-analysis provides evidence to support that preadmission metformin use may decrease mortality in septic patients with diabetes mellitus.

The mechanism underlying the ability of metformin to decrease the mortality of septic patients remains unclear. Metformin may supply higher amounts of lactate, serving as an energetic carbon source, thus making energy available to ischaemic tissue; this hypothesis is consistent with the results of a study [41] that indicated that lactic acid is also a key energy source, like glucose, amino acids, and ketones. Furthermore, Friesecke et al. [42] and Protti et al. [43] suggested that lactic acidosis due to metformin use is associated with lower mortality compared to other forms of lactic acidosis, which may explain why preadmission metformin users had lower mortality than nonusers among patients with sepsis and diabetes mellitus. Moreover, the mechanism may also be related to the potential protective effects of metformin, including its anti-endotoxaemic, anti-inflammatory, and vasoactive properties [44‐46]. Additionally, metformin has been shown to decrease the expression of nitric oxide synthase in animal models, which could ameliorate vasodilatation [47]. Most importantly, metformin induces the activation of AMPK, an enzyme that is a key cellular energy sensor that maintains energy homeostasis at the cellular and organismal levels [48]. Once activated, AMPK regulates cellular energy status, switching on the ATP-generating pathways and switching off the ATP-consuming pathways [49]. Furthermore, some studies [50, 51] have shown that metformin may possess antimicrobial effects, which may be the primary mechanism involved in the beneficial effects on sepsis. As a result, cellular function is improved under stressful conditions, leading to improved cardiovascular function and hypoxia. These potential protective effects of metformin use may significantly ameliorate the prognosis of septic patients, but this association requires further study.

Meta-analysis is a method used for the comprehensive statistical analysis of multiple studies on a topic. When the difference between the outcomes of each study is greater than expected, statistical heterogeneity is present in the pooled results of the meta-analysis. The four included cohort studies had a high degree of heterogeneity, reflecting differences in these studies. In this meta-analysis, the risk of bias assessment of the included studies showed a low risk bias. Therefore, the statistical heterogeneity was not considered methodological heterogeneity. The heterogeneity may be derived from many causes, such as the small sample sizes, the different initial lactate levels and the use of other antidiabetic medications.

This meta-analysis has several strengths. In the study by Doenyas-Bara et al. [19], the metformin users had more severe diabetes mellitus at admission than the non-metformin users, thus providing additional strong evidence to support the outcome that metformin users may have lower mortality among patients with sepsis and diabetes mellitus. Second, the risk of bias assessment used the NOS, which showed a low risk of bias among the included studies. Third, we used the random effects model with generic inverse variance, and we extracted the adjusted ORs and 95% CIs so that the pooled estimate of the OR for the effect of metformin on mortality could be computed. Finally, the outcome of the sensitivity analysis showed that this result was robust and reliable.

However, this study has important limitations. Although we did our best to conduct a comprehensive search of the relevant literature, only four studies met our inclusion criteria, so the conclusion may need more original studies for confirmation. Second, the sample size of the septic patients with diabetes mellitus was small, suggesting that the result may not be reliable. Nevertheless, the robustness of our conclusions was supported by the result of the risk of bias assessment. Third, the outcome of this meta-analysis may be confounded by the fact that the inclusion criteria of the four studies varied considerably. Fourth, we studied the relationship between metformin use prior to admission and mortality in septic patients with diabetes mellitus, so it is not clear whether metformin is beneficial during the hospital stay for septic patients with diabetes mellitus previously untreated with metformin; therefore, further studies are needed. Fifth, although septic patients with diabetes may benefit from preadmission metformin use, the effects and safety of metformin treatment initiation and continuation in patients who are critically ill remain to be further clarified. Finally, there were no randomised controlled trials included in this meta-analysis, which is a limitation even though all the included retrospective cohort studies had a low risk of bias.

This is the first systematic review and meta-analysis to describe the association of preadmission metformin use and mortality in septic adult patients with diabetes mellitus. The results suggest that in septic adult patients with diabetes mellitus, preadmission metformin use is associated with lower mortality. However, the current conclusions need to be further supported by more high-quality original studies.