The effect of sodium‐glucose cotransporter‐2 inhibitors on inflammatory biomarkers: A meta‐analysis of randomized controlled trials

To conduct a meta‐analysis of randomized controlled trials (RCTs) to assess the effect of sodium‐glucose cotransporter‐2 (SGLT2) inhibitors on inflammatory biomarkers.


| INTRODUCTION
In 2015, sodium-glucose cotransporter-2 (SGLT2) inhibitors were found to significantly reduce cardiovascular events in people with type 2 diabetes mellitus (T2DM) who are at the highest risk of experiencing such events. 1When given to individuals with heart failure with reduced ejection fraction, SGLT2 inhibitors reduced cardiovascular mortality and hospitalizations for acute heart failure by approximately 25% 2,3 and in heart failure with preserved ejection fraction by approximately 20%. 4 Whilst SGLT2 inhibitors were initially designed as a medication for the treatment of T2DM, where they promote renal excretion of glucose, it remains unexplained how SGLT2 inhibitors exert their cardiorenal-protective effects.Multiple explanations for the underlying cardiovascular benefits have been described that extend beyond improved glycaemic control. 5These include early natriuresis, reductions in plasma volume, improved vascular structure and function, renal collecting tubular extension, reduced blood pressure, modifications to tissue sodium handling, favouring of ketone body metabolism, reduced uric acid levels, reduced adipose tissuemediated inflammation, reduced body mass and reduced oxidative stress. 5,6 the mechanisms listed, inflammation is of particular interest as it has a significant role in the pathophysiology of T2DM, [7][8][9] and is increasingly recognized as a key player in the pathogenesis of cardiovascular disease (CVD). 10,11Research from basic science models suggests that SGLT2 inhibitors may be anti-inflammatory.SGLT2 inhibitors may reduce tumour necrosis factor-alpha (TNF-α), interleukin-6 (IL-6) and monocyte chemoattractant protein-1 (MCP-1) 12 in apolipoprotein E knockout mice, IL-6 and tumour necrosis factor receptor-1 (TNFR1) in human proximal tubular cells, 13 and IL-6, TNF-α and MCP-1 in mouse models of diabetic kidney disease. 14Furthermore, SGLT2 inhibitors may upregulate the production of adipokines in obese mice. 15However, it remains uncertain whether these mechanisms apply to humans.Previous reviews have sought to understand whether inflammation plays a role in the cardiorenal-protective effects of SGLT2 inhibitors in humans, but none has been able to provide a quantitative, minimally biased assessment of the effect of SGLT2 inhibitors on biomarkers of inflammation. 16

| METHODS
This review is written in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines (Table S1) and registered with PROSPERO (CRD42022363880). 17

| Search strategy
Medline, Embase and the Cochrane Library were searched from inception to January 2024 for trials investigating the use of SGLT2 inhibitors and measuring biomarkers of inflammation.The full search strategy can be found in Table S2.Medical subject heading (MeSH) terms were used where feasible.Following removal of duplicates, the results of the search were screened independently by three reviewers, before full-text eligibility assessment was performed independently by two reviewers (Figure 1).

| Study selection
Eligibility was restricted to prospective randomized controlled trials (RCTs), of either parallel or crossover design, that used SGLT2 inhibitors as intervention compared to any control other than different SGLT2 inhibitor drugs or doses.Observational studies, case reports and basic science reports without human participants were excluded.
Adults were included if they were eligible for SGLT2 inhibitor prescription, including patients with T2DM, symptomatic chronic heart failure and chronic kidney disease.Trials were excluded if they included individuals with type 1 diabetes mellitus or paediatric participants.Trials of any study duration were included providing they reported the measurement of inflammatory biomarkers, regardless of the primary outcome measured.Trials were also excluded if they did not possess data that could be quantitatively analysed using metaanalysis.
F I G U R E 1 Flow diagram based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines showing the method of identifying trials and reasons for exclusion.PCI, percutaneous coronary intervention; RCT, randomized controlled trial; SGLT2, sodium-glucose cotransporter-2; T1DM, type 1 diabetes mellitus.

| Outcomes of interest and comparisons
The following biomarkers were selected a priori based on published evidence linking these biomarkers to inflammatory pathways.We specify where there were no data available to report a biomarker.
Insulin sensitivity markers-homeostatic model assessment of insulin resistance index (HOMA-IR).
9][20][21][22][23][24][25] Of note, adiponectin is thought to increase IL-6. 26sulin resistance is a key promoter of chronic inflammation, therefore, HOMA-IR (a direct measure of insulin resistance) was included. 27mparisons are made between study arms that were exposed to SGLT2 inhibitors compared with controls.Controls were defined as standard care including other glucose-lowering medications or placebo.Subgroup analysis was performed between comparisons made with placebo and other diabetes medications in an attempt to reduce heterogeneity.

| Data extraction and synthesis
Data were independently extracted into a preformatted Excel spreadsheet from eligible RCTs.Continuous outcomes for biomarkers were converted into equivalent, appropriate units.Where data were missing, these were sought via email from authors, and failing this, were considered missing at random.Furthermore, the following participant characteristics were extracted: age, sex, glycated haemoglobin (HbA1c), fasting plasma glucose, weight, body mass index (BMI), and diabetes duration.Data on the mean change in biomarkers of interest, alongside relevant standard deviations (SDs), and numbers of individuals in each relevant arm were collected.

| Associations of biomarker changes with clinically relevant outcomes
Where possible, clinically relevant outcomes were also collected such that analysis could be made for an association between changes in biomarkers with changes in clinically relevant outcomes.Analysis of a potential association with a particular biomarker would not be sought if there was no evidence of a change in this biomarker with SGLT2 inhibitors.

| Quality and risk of bias assessment
The Cochrane risk-of-bias tool for randomized trials (RoB 2) was used to assess risk of bias. 28The Grading of Recommendations Assessment, Development and Evaluation (GRADEpro) tool was used to assess outcome quality for each biomarker of interest. 29

| Statistical analysis
Random-effects meta-analysis was used to assess the change in biomarkers with and without exposure to SGLT2 inhibitors in Stata (17.0, StataCorp LLC, College Station, TX, USA).Mean difference (MD) was used as default, unless different collection or measurement methodologies were used between trials for the same biomarker, in which case standardized mean difference (SMD) using Hedges' G was used. 30Heterogeneity was quantified using the I 2 measure and the p value from the chi-squared test.I 2 > 50% was considered to represent moderate-to-high heterogeneity. 31Small study effects were examined using funnel plots where the number of included trials was greater than 10, accompanied by Egger's regression test. 32If the change in biomarker mean and SD were not available, the SD was calculated from the standard error of the mean, or values were estimated using methodology from the Cochrane handbook. 33In cases where the median and interquartile range were provided in place of mean and SD, the mean and SD were estimated using methodology described by Wan et al. 34 In a minority of cases, if the SD was missing and could not be estimated, data were sought from the authors and failing this, values were imputed using the validated methodology described by Ma et al. 35 Descriptive statistics are reported as means ± SD.Baseline characteristic averages were calculated as the mean for each trial, weighted by the number of participants in the trial.

| C-reactive protein
From the analysis of 16 trials and 1435 participants at follow-up, use of SGLT2 inhibitors was associated with no significant MD in CRP levels compared to control (MD À0.10 mg/L, 95% confidence interval [CI] À0.35, 0.15).There was also no significant difference between groups in subgroup analysis.Overall heterogeneity was high (I 2 = 81.0%,p < 0.1) and remained moderate in the placebo subgroup (I 2 = 61.1%,p < 0.1) and high in the diabetes medications subgroup (I 2 = 93.0%,p < 0.1; Figure 2A).

| Fibroblast growth factor-21
From the analysis of four trials and 157 participants at follow-up, use of SGLT2 inhibitors was associated with no significant SMD in FGF21 levels versus placebo (SMD À0.17 [95% CI À0.47, 0.14]).There were no included trials using diabetes medications as control (Figure 2B).

| Monocyte chemoattractant protein-1
From the analysis of three trials and 291 participants at follow-up, use of SGLT2 inhibitors was associated with no significant SMD in MCP-1 levels compared to control (SMD À0.07 [95% CI À0.29, 0.16]).There was no significant difference between groups in subgroup analysis (Figure 2C).

| Adiponectin
From the analysis of 20 trials and 2789 participants at follow-up, use of SGLT2 inhibitors was associated with no significant SMD in adiponectin levels compared to control (SMD À0.24 [95% CI À1.01, 0.53]).
In subgroup analysis, adiponectin was significantly increased versus T A B L E 1 (Continued)

| Tumour necrosis factor-alpha
From the analysis of five trials and 259 participants at follow-up, use of SGLT2 inhibitors was associated with no significant SMD in TNF-α levels compared to control (SMD À0.30 [95% CI À0.67, 0.08]).There was no significant difference between groups in subgroup analysis.

| Interleukin-6
From the analysis of four trials and 228 participants at follow-up, use of SGLT2 inhibitors was associated with no significant mean difference in IL-6 levels compared to control (MD À0.34 pg/mL

| Plasminogen activator inhibitor-1
From the analysis of three trials and 277 participants at follow-up, use of SGLT2 inhibitors was associated with no significant SMD in PAI-1 levels compared to control (SMD À0.07 [95% CI À0.30, 0.17]).There was no significant difference between groups in subgroup analysis (Figure 3E).

| Tumour necrosis factor receptor-1
From the analysis of two trials and 3561 participants at follow-up, use of SGLT2 inhibitors was associated with a significant standardized mean reduction in TNFR1 levels versus placebo (SMD À0.13 [95% CI À0.20, À0.06]).There were no included trials using diabetes medications as control (Figure 3F).
There were not enough data available in the literature to analyse TNFR2.

| Homeostatic model assessment of insulin resistance
From the analysis of 13 trials and 1066 participants at follow-up, use of

| Clinically relevant outcomes
There were insufficient data available on clinically relevant outcomes.

| Sensitivity analyses
Sensitivity analyses were conducted investigating the effect of using MD and SMD for each biomarker (Table S5) as well as the effect of stratifying papers by their risk of bias (Table S6).Visual assessment of funnel plots and Egger's regression test showed there was no evidence of small study effects in any of the outcomes (Figure S1-S4).

| DISCUSSION
This is the largest review to date encompassing randomized data that provides Cochrane-standard mitigation of bias, showing that SGLT2 inhibitors likely improve adipokine profiles and insulin sensitivity.
However, in this analysis, SGLT2 inhibitors did not appear to improve other biomarkers of inflammation when compared to placebo and other glucose-lowering medications.Our results demonstrate that SGLT2 inhibitors significantly improved adiponectin, IL-6 and TNFR1 versus placebo, as well as leptin and HOMA-IR versus control.The reduction in HOMA-IR may be secondary to improved glucose handling as SGLT2 inhibitors are known to increase renal glucose excretion and reduce insulin secretion. 78TNFR1 was found to be reduced by SGLT2 inhibitors, but this result should be viewed with caution as it was obtained from the analysis of only two trials.
Obesity is a risk factor for CVD; adipocytes produce immunomodulatory factors that are thought to mediate this link. 79This review shows that SGLT2 inhibitors improve adiponectin and IL-6 versus placebo, and leptin versus control.These results support the hypothesis that SGLT2 inhibitors improve adipokine biomarkers.It is plausible that this could be a contributory mechanism by which SGLT2 inhibitors exert their cardiovascular-protective effects.Nevertheless, contrary to our initial hypothesis, this meta-analysis shows that there is little evidence to support the hypothesis that SGLT2 inhibitors improve inflammatory biomarkers, other than adipokines.This adds weight to the following assertion, but does not prove, that the cardioprotective mechanisms of SGLT2 inhibitors may not be due to an anti-inflammatory mechanism.This is in contrast to our previous publication showing that glucagon-like peptide-1 receptor agonists, which also have cardiorenal-protective effects, improve biomarkers of inflammation including CRP and TNF-α. 80[83] As mentioned in the introduction, there is evidence from animal studies that suggests SGLT2 inhibitors may be anti-inflammatory.This highlights a need for further research, to better understand the difference in the effect of SGLT2 inhibitors in animal models compared to humans.
Our finding that SGLT2 inhibitors significantly affect adipocyte sensitivity profiles is supported by Wang et al., 84 who also reported that, when compared to placebo, adiponectin is significantly raised, and both leptin and PAI-1 levels are significantly reduced.However, the authors conclude that SGLT2 inhibitors are anti-inflammatory, particularly reporting a significant reduction in CRP when compared In terms of limitations, it was necessary to include many trials as all trials on this topic are small.Most trials investigating SGLT2 inhibitors include inflammatory biomarkers as secondary outcomes, often in supplements, occasionally with errors in units.In four cases, biomarker data were only found in post hoc analyses.Despite the evidence base being heterogenous and carrying some risk of bias, this analysis was an effective way to answer our study question using currently published data and, in order to address heterogeneity, subgroup and sensitivity analyses were performed.This analysis would be surpassed by
to placebo.We included additional trials comparing CRP to placebo and it is suspected that Wang et al. may have used median change, as opposed to mean change, or inappropriately converted units when reporting CRP outcomes.Additionally, they reported data from Seino 2018 65 as CRP when the paper investigated C-peptide immunoreactivity (CPR) instead, as well as reporting data from Hao 2022 85 which was not a randomized trial.This may further explain the finding of a significant reduction in CRP compared to placebo reported by Wang F I G U R E 3 (Continued) et al. 84 Despite this difference, the data from Wang et al. support the hypothesis that SGLT2 inhibitors do not have an anti-inflammatory action, but instead alter adipokine profiles as they report SGLT2 inhibitors do not significantly reduce any inflammatory biomarker when compared to other glucose-lowering medications other than the adipokine leptin.

a
dedicated clinical trial, although the number of participants required may prohibit such a study design in this context.Where possible, missing data were estimated or imputed (using validated Cochraneendorsed methods), but in a minority of cases, trials had to be excluded.Follow-up was also short in many trials.Extensive exclusion criteria were often employed in the included RCTs, limiting the generalizability of the results to a wider population.The scope of this review does not include oxidative stress, but this remains a useful future area of investigation.In conclusion, this review has found evidence suggesting that SGLT2 inhibitors improve adipokine profiles and insulin sensitivity, but the analysis shows little evidence of improvement in other inflammatory biomarkers including CRP.Adipokines are important aetiological factors in CVD and thus may be a contributing factor to the cardiovascular-protective effects of SGLT2 inhibitors.F I G U R E 4 Forest plot showing the outcome for sodium-glucose cotransporter-2 (SGLT2) inhibitor versus control groups for the insulin sensitivity marker, homeostatic model assessment for insulin resistance (HOMA-IR) as mean difference.