Comparative efficacy of sodium-glucose transporter 2 inhibitors on lipid profiles in nonalcoholic fatty liver disease (NAFLD): a comprehensive Bayesian network meta-analysis

Abstract

Background:

Nonalcoholic fatty liver disease (NAFLD) is a global health concern associated with dyslipidemia, obesity, and type 2 diabetes mellitus (T2DM), necessitating effective therapeutic strategies. Sodium-glucose transporter 2 (SGLT-2) inhibitors have shown promise in improving metabolic parameters, but their comparative efficacy in NAFLD remains unclear.

Methods:

We systematically searched PubMed, Cochrane Library, Scopus, and Embase for randomized controlled trials (RCTs) up to 31 December 2024, involving NAFLD patients treated with SGLT-2 inhibitors versus placebo or standard treatments. Outcomes included changes in low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), total cholesterol, triglycerides, BMI, and HbA1c. A Bayesian NMA was performed using a random-effects model, with mean differences (MD) and 95% credible intervals (CrI) reported. Treatments were ranked using Surface Under the Cumulative Ranking Curve (SUCRA) scores.

Results:

We evaluated and compared the efficacy of SGLT-2 inhibitors in improving lipid profiles (LDL-C, HDL-C, total cholesterol, and triglycerides), BMI, and HbA1c in patients with NAFLD. Eleven RCTs involving 805 patients were included. Empagliflozin achieved the greatest LDL-C reduction (MD: −8.07 mg/dL, 95% CrI: −25.92 to 1.41; SUCRA: 95.38%), total cholesterol reduction (MD: −15.08 mg/dL, 95% CrI: −58.87 to 5.38; SUCRA: 94.32%), and HbA1c reduction (MD: −0.69%, 95% CrI: −1.52 to −0.06; SUCRA: 81.84%). Ipragliflozin non-significantly increased HDL-C (MD: 2.28 mg/dL, 95% CrI: −1.21 to 4.83; SUCRA: 85.11%) and significantly reduced triglyceride (MD: −22.39 mg/dL, 95% CrI: −39.25 to −7.27; SUCRA: 82.26%). Dapagliflozin resulted in significant BMI reduction (MD: −1.23 kg/m², 95% CrI: −2.11 to −0.41; SUCRA: 73.28%).

Conclusion:

Empagliflozin and ipragliflozin demonstrated superior efficacy in improving lipid profiles, while dapagliflozin was most effective for BMI reduction in NAFLD patients. These findings align with their cardiovascular and metabolic benefits, offering a multifaceted approach to a complex disease. Further research is needed to confirm long-term effects and optimize treatment strategies for diverse NAFLD populations.

Introduction

Nonalcoholic fatty liver disease (NAFLD) represents a spectrum of liver conditions characterized by excessive hepatic fat accumulation in the absence of significant alcohol consumption. It ranges from simple steatosis to nonalcoholic steatohepatitis (NASH), which can progress to fibrosis, cirrhosis, and hepatocellular carcinoma^[1]. NAFLD is closely associated with metabolic syndrome, obesity, type 2 diabetes mellitus (T2DM), and dyslipidemia, making it a global public health concern^[2]. The prevalence of NAFLD has risen dramatically in parallel with the obesity epidemic, with estimates suggesting it affects 25–30% of the global population, and up to 70–90% of individuals with T2DM or severe obesity^[3,4]. This high prevalence underscores the urgent need for effective therapeutic strategies to address NAFLD and its associated metabolic complications.

HIGHLIGHTS

• Bayesian NMA compared SGLT-2 inhibitors’ efficacy on lipid and metabolic outcomes in NAFLD.

• Treatments were ranked using SUCRA scores.

• Empagliflozin resulted in LDL-C reduction (MD: −8.07 mg/dL; SUCRA: 95.38%).

• Ipragliflozin non-significantly increased HDL-C (MD: 2.28 mg/dL; SUCRA: 85.11%) in NAFLD.

• Dapagliflozin reported the greatest BMI reduction in NAFLD (MD: −1.23 kg/m²; SUCRA: 73.28%).

The pathophysiology of NAFLD is complex and multifactorial, involving insulin resistance, lipotoxicity, oxidative stress, and inflammation, which contribute to hepatic fat accumulation and systemic metabolic dysregulation^[5]. Dyslipidemia is a hallmark of NAFLD, characterized by elevated low-density lipoprotein cholesterol (LDL-C), triglycerides, and total cholesterol, as well as reduced high-density lipoprotein cholesterol (HDL-C)^[6]. These lipid profile alterations not only exacerbate liver injury but also increase cardiovascular risk, which is the leading cause of mortality in NAFLD patients^[7]. Additionally, NAFLD is associated with increased body mass index (BMI) and poor glycemic control, as evidenced by elevated HbA1c levels, particularly in patients with concurrent T2DM^[8]. Addressing these metabolic parameters is critical for improving liver health and reducing the overall burden of NAFLD-related complications.

Current management of NAFLD primarily focuses on lifestyle interventions, such as weight loss and dietary modification, which have shown modest efficacy in improving hepatic steatosis and metabolic parameters^[9]. However, achieving and sustaining significant weight loss remains challenging for many patients, necessitating pharmacological interventions to complement lifestyle changes^[10]. To date, no specific pharmacological therapy has been universally approved for NAFLD, and treatments targeting associated metabolic abnormalities, such as dyslipidemia, obesity, and hyperglycemia, are often employed^[11]. Among these, sodium-glucose transporter 2 (SGLT-2) inhibitors have emerged as a promising therapeutic class due to their multifaceted effects on glucose metabolism, weight reduction, and cardiovascular outcomes^[12].

SGLT-2 inhibitors, including dapagliflozin, empagliflozin, ipragliflozin, canagliflozin, sotagliflozin, ertugliflozin, and others, act by inhibiting glucose reabsorption in the proximal renal tubule, promoting glycosuria and reducing hyperglycemia^[13]. Beyond glycemic control, SGLT-2 inhibitors have demonstrated benefits in reducing body weight, improving insulin sensitivity, and modulating lipid profiles, which are particularly relevant for NAFLD patients^[14]. Preclinical and clinical studies suggest that SGLT-2 inhibitors may reduce hepatic fat content by decreasing de novo lipogenesis, enhancing fatty acid oxidation, and mitigating inflammation^[15,16]. For instance, empagliflozin has been shown to reduce liver fat in patients with T2DM and NAFLD, potentially through improvements in insulin sensitivity and lipid metabolism^[17]. Similarly, dapagliflozin and canagliflozin have been associated with reductions in BMI and triglycerides, alongside improvements in HDL-C levels^[18,19]. Ipragliflozin, widely used in East Asian populations, has also shown promise in ameliorating hepatic steatosis and dyslipidemia^[20]. Sotagliflozin, a dual SGLT-1/2 inhibitor, and ertugliflozin have been less extensively studied in NAFLD but may offer similar benefits due to their shared mechanisms^[21,22].

Despite these promising findings, the comparative efficacy of different SGLT-2 inhibitors in improving lipid profiles and other metabolic parameters in NAFLD remains poorly understood. Randomized controlled trials (RCTs) evaluating SGLT-2 inhibitors in NAFLD populations have reported varying effects on LDL-C, HDL-C, total cholesterol, triglycerides, BMI, and HbA1c, likely due to differences in study design, patient populations, and treatment regimens^[23]. For example, a 24-week RCT of empagliflozin in NAFLD patients with T2DM reported significant reductions in total cholesterol and triglycerides compared to placebo^[24], while a study of Ipragliflozin showed more pronounced effects on HDL-C^[25]. These inconsistencies highlight the need for a systematic synthesis of evidence to determine which SGLT-2 inhibitors offer the greatest benefits for specific metabolic outcomes in NAFLD.

Traditional pairwise meta-analyses have provided insights into the efficacy of individual SGLT-2 inhibitors compared to placebo or standard care but are limited in their ability to compare multiple treatments simultaneously^[26]. Network meta-analysis (NMA) overcomes this limitation by integrating direct and indirect evidence from RCTs, allowing for the simultaneous comparison of multiple interventions within a single framework^[27]. Bayesian NMA, in particular, offers advantages over frequentist approaches by incorporating prior knowledge, quantifying uncertainty through credible intervals, and providing probabilistic rankings of treatment efficacy, such as Surface Under the Cumulative Ranking Curve (SUCRA) scores^[28]. This approach is well-suited for evaluating the comparative efficacy of SGLT-2 inhibitors, given the heterogeneity of available RCTs and the need to rank treatments based on their effects on multiple metabolic outcomes.

To date, no comprehensive Bayesian NMA has evaluated the comparative efficacy of SGLT-2 inhibitors, including dapagliflozin, empagliflozin, ipragliflozin, canagliflozin, sotagliflozin, ertugliflozin, and others, in patients with NAFLD. Existing meta-analyses have focused primarily on glycemic control or cardiovascular outcomes in broader populations, with limited attention to lipid profiles and BMI in NAFLD-specific cohorts^[29,30]. Given the critical role of dyslipidemia and obesity in NAFLD progression and cardiovascular risk, a focused analysis of these outcomes is warranted. Furthermore, the inclusion of all available SGLT-2 inhibitors in a single analysis allows for a more complete understanding of their relative benefits, which could inform clinical decision-making and guideline development.

The objective of this systematic review and Bayesian NMA was to evaluate and compare the efficacy of SGLT-2 inhibitors in improving lipid profiles (LDL-C, HDL-C, total cholesterol, and triglycerides), BMI, and HbA1c in patients with NAFLD. By synthesizing data from RCTs, we aimed to provide robust evidence on the relative effectiveness of these agents compared to placebo or standard treatments, rank their efficacy using SUCRA scores, and identify potential sources of heterogeneity through meta-regression. The findings of this study could guide the selection of SGLT-2 inhibitors for NAFLD patients, particularly those with concurrent T2DM, and contribute to the development of personalized treatment strategies to mitigate the metabolic and cardiovascular burden of this disease.

Methods

This Bayesian network meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Network Meta-Analyses (PRISMA-NMA) guidelines to ensure methodological rigor and transparency^[31]. The primary objective was to evaluate and compare the effectiveness of sodium-glucose transporter 2 (SGLT-2) inhibitors, including dapagliflozin, empagliflozin, ipragliflozin, canagliflozin, sotagliflozin, ertugliflozin, and other SGLT-2 inhibitors, in improving metabolic parameters in patients with nonalcoholic fatty liver disease (NAFLD). These parameters included reductions in low-density lipoprotein cholesterol (LDL-C), total cholesterol, triglycerides, body mass index (BMI), and HbA1c, as well as increases in high-density lipoprotein cholesterol (HDL-C). The study was registered with PROSPERO to ensure transparency in the study design and methodology^[32]. The work has been reported in line with AMSTAR (Assessing the methodological quality of systematic reviews) Guidelines^[33].

Study selection criteria

Eligibility criteria

Predefined inclusion and exclusion criteria were applied to select eligible studies. Only randomized controlled trials (RCTs) were included, as they provide the highest level of evidence for intervention efficacy. Studies were required to involve adult participants (aged ≥18 years) diagnosed with NAFLD, with or without concurrent type 2 diabetes mellitus (T2DM). The interventions consisted of any SGLT-2 inhibitor, including but not limited to dapagliflozin, empagliflozin, ipragliflozin, canagliflozin, sotagliflozin, and ertugliflozin, administered as monotherapy or in combination with standard treatments (e.g. metformin and insulin). The comparator group included placebo or standard treatments without SGLT-2 inhibitors. Primary outcomes encompassed changes in LDL-C (mg/dL), HDL-C (mg/dL), total cholesterol (mg/dL), triglycerides (mg/dL), BMI (kg/m²), and HbA1c (%). Studies published in English were eligible, with no restrictions on publication dates up to 31 December 2024, consistent with the search cutoff. Exclusion criteria included non-randomized studies, observational studies, and trials that did not report relevant efficacy outcomes. Studies involving patients with severe liver diseases unrelated to NAFLD (e.g. viral hepatitis and alcoholic liver disease) were excluded to maintain specificity to NAFLD^[34].

Information sources

A comprehensive literature search was conducted across major electronic databases, including PubMed, Cochrane Library, Scopus, and Embase, covering studies from inception to 31 December 2024. The search strategy utilized Medical Subject Headings (MeSH) and keywords relevant to the target population (“nonalcoholic fatty liver disease,” “NAFLD,” “steatosis”), interventions (“sodium-glucose transporter 2 inhibitors,” “Dapagliflozin,” “Empagliflozin,” “Ipragliflozin,” “Canagliflozin,” “Sotagliflozin,” “Ertugliflozin”), and study design (“randomized controlled trial,” “RCT”), following established systematic review protocols^[35]. All records were managed using EndNote software, and duplicates were removed prior to screening^[36]. The search aimed to identify all studies meeting the predefined eligibility criteria. Detailed search strategy is provided in Supplementary Digital Content Table 1, available at: http://links.lww.com/MS9/A894.

Study selection process

Two independent reviewers screened the titles and abstracts of all identified records based on the eligibility criteria. Full-text articles of potentially eligible studies were retrieved for further evaluation. Discrepancies during screening or full-text review were resolved through discussion, with a third reviewer consulted if consensus could not be reached, as recommended for minimizing bias in study selection^[37]. A PRISMA flow diagram was generated to summarize the study selection process, as depicted in the manuscript (Fig. 1)^[31].

Figure 1.

PRISMA 2020 flow diagram for systematic reviews which included searches of databases and registers.

Data extraction

Data extraction was performed independently by two reviewers using a standardized, pre-piloted form to ensure consistency and accuracy^[38]. Extracted data included study characteristics (e.g. authors, publication year, country, study design, and funding sources), participant-level characteristics (e.g. sample size, age, sex distribution, baseline NAFLD severity, T2DM status, and concurrent medications), and intervention details (e.g. type of SGLT-2 inhibitor, dosage, and treatment duration). Control group characteristics, including placebo or standard treatment, were also extracted. Outcome data included mean differences (MD) for LDL-C (mg/dL), HDL-C (mg/dL), total cholesterol (mg/dL), triglycerides (mg/dL), BMI (kg/m²), and HbA1c (%) compared to baseline or placebo. Discrepancies in data extraction were resolved through consensus or consultation with a third reviewer.

Data analysis

Network meta-analysis

A Bayesian network meta-analysis (NMA) was conducted using R (version 4.4.3), which facilitates Bayesian NMA for multiple treatment comparisons^[39,40]. A random-effects model was employed to account for between-study heterogeneity, using a Markov chain Monte Carlo (MCMC) method, as it provides robust estimates in the presence of variability across studies^[41]. Non-informative normal priors were applied for treatment effects, and weakly informative priors were used for variance components to minimize bias in posterior estimates^[42]. Model convergence was assessed using convergence diagnostics, including the Gelman-Rubin Diagnostic, Deviance Information Criterion (DIC), and Potential Scale Reduction Factor (PSRF), to ensure reliable results^[43]. The MCMC process involved 10 000 adaptation steps, a burn-in period of 20 000 iterations, and 250 000 iterations to ensure robust estimates, following standard Bayesian NMA protocols^[44]. Mean differences (MDs) compared to placebo were calculated for continuous outcomes (LDL-C, HDL-C, total cholesterol, triglycerides, BMI, and HbA1c). Results were reported as posterior medians with 95% credible intervals (CrIs). Network geometry was visualized using network plots, with node sizes reflecting sample sizes and edge widths indicating the number of direct comparisons. Surface Under the Cumulative Ranking Curve (SUCRA) values were computed to rank treatments, with higher SUCRA values indicating greater effectiveness, as described by Salanti et al^[45].

Model specifications and parameters

A Bayesian multilevel regression model was used to estimate treatment effects across the network, accounting for multiple comparisons and network geometry^[46]. Convergence was evaluated using trace plots, Gelman-Rubin statistics, and DIC to ensure stable estimates^[43]. Sensitivity analyses tested the robustness of findings by varying model assumptions, such as prior distributions, to assess the impact on results^[47].

Statistical significance and uncertainty

Results were expressed as posterior median estimates with 95% CrI, which provide a probabilistic interpretation of treatment effects in Bayesian frameworks^[48]. SUCRA values ranked treatments for each outcome, with values closer to 100% indicating higher efficacy^[45]. Uncertainty was quantified through the width of CrI and posterior distributions, ensuring transparent reporting of variability^[48]. SUCRA values were used to rank treatments, but interpretations considered the width and overlap of 95% CrI to account for uncertainty in treatment effects.

Risk of bias assessment

The risk of bias for each included study was evaluated using the Cochrane Risk of Bias Tool (RoB 2.0), assessing domains including random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, and selective reporting^[49]. Studies were classified as low, unclear, or high risk of bias^[50].

Reporting

The Bayesian NMA and meta-regression results were reported in accordance with PRISMA-NMA guidelines^[31]. A PRISMA flow diagram detailed the study selection process, and findings were presented using forest plots, network diagrams, and SUCRA rankings to illustrate the relative effectiveness of SGLT-2 inhibitors, including dapagliflozin, empagliflozin, ipragliflozin, canagliflozin, sotagliflozin, ertugliflozin, and others, versus placebo across metabolic outcomes^[45].

Results

Baseline characteristics of included studies

This systematic review and Bayesian network meta-analysis included 11 randomized controlled trials (RCTs) involving a total of 805 patients with nonalcoholic fatty liver disease (NAFLD)24,51–60. The studies evaluated the efficacy of sodium-glucose transporter 2 (SGLT-2) inhibitors – dapagliflozin, empagliflozin, and ipragliflozin – in lipid profiles, body mass index (BMI), and HbA1c, compared to placebo or standard treatments. No eligible RCTs were identified for canagliflozin, sotagliflozin, or ertugliflozin, limiting the analysis to dapagliflozin, empagliflozin, and ipragliflozin. The baseline characteristics of the included studies are summarized below.

The number of participants per study ranged from 38 to 160, with intervention groups ranging from 18 to 80 participants and control groups from 15 to 80 participants. Specifically, Han et al enrolled 44 patients (29 intervention, 15 control), Eriksson et al included 40 patients (20 intervention, 20 control), Elhini et al had 160 patients (80 intervention, 80 control), Shi et al included 78 patients (40 intervention, 38 control), Hussain et al enrolled 138 patients (67 intervention, 71 control), Phrueksotsai et al had 38 patients (18 intervention, 20 control), Chehrehgosha et al included 72 patients (35 intervention, 37 control), Taheri et al enrolled 90 patients (43 intervention, 47 control), Kuchay et al had 42 patients (22 intervention, 20 control), Takahashi et al included 46 patients (21 intervention, 25 control), and Aso et al enrolled 57 patients (33 intervention, 24 control).

The interventions primarily involved SGLT-2 inhibitors: dapagliflozin (5–10 mg) was used by Eriksson et al, Shi et al, Hussain et al, Phrueksotsai et al, and Aso et al; empagliflozin (10–25 mg) was used by Elhini et al, Chehrehgosha et al, Taheri et al, and Kuchay et al; and ipragliflozin (50 mg) was used by Han et al and Takahashi et al. Shi et al combined dapagliflozin with metformin, while other studies used SGLT-2 inhibitors as monotherapy. Control groups varied: most studies (Eriksson et al, Elhini et al, Hussain et al, Phrueksotsai et al, Chehrehgosha et al, Taheri et al) used placebo; Shi et al and Han et al used metformin-based regimens (metformin + antidiabetic treatment or metformin + pioglitazone); Kuchay et al and Aso et al used standard treatments without SGLT-2 inhibitors; and Takahashi et al used diet/exercise therapy with antidiabetic drugs.

The mean age of participants in the intervention groups ranged from 29 to 65 years, and in the control groups, it ranged from 31 to 65 years. The youngest cohort was in the study by Hussain et al (intervention: 29 ± 16 years, control: 31 ± 14 years), while the oldest was in the study by Chehrehgosha et al (intervention: 51 ± 8 years, control: 51 ± 8 years) and Eriksson et al (intervention: 65 ± 6 years, control: 65 ± 6 years). Other studies reported mean ages between 43 and 57 years, with standard deviations ranging from 6 to 14 years, indicating a relatively broad age distribution across the included studies.

Six studies were double-blinded, placebo-controlled (Eriksson et al, Elhini et al, Hussain et al, Phrueksotsai et al, Chehrehgosha et al, Taheri et al), three were placebo-controlled with unclear blinding (Han et al, Shi et al, Aso et al), one used standard treatment without blinding (Kuchay et al), and one was open-label (Takahashi et al.). Treatment durations varied from 12 to 72 weeks, with most studies lasting 12 to 24 weeks: 12 weeks in studies by Eriksson et al, Hussain et al, and Phrueksotsai et al; 20 weeks in a study by Kuchay et al; 24 weeks in studies by Han et al, Elhini et al, (26 weeks) Shi et al, Chehrehgosha et al, Taheri et al, and Aso et al; and 72 weeks in a study by Takahashi et al.

The studies were conducted across diverse geographical regions, reflecting a global perspective: four in Asia (Han et al in Korea, Takahashi et al and Aso et al in Japan, and Shi et al in China), three in the Middle East (Elhini et al in Egypt and Chehrehgosha et al and Taheri et al in Iran), two in South Asia (Hussain et al in Pakistan and Kuchay et al in India), one in Southeast Asia (Phrueksotsai et al in Thailand), and one in Europe (Eriksson et al in Sweden). This geographical diversity highlights the applicability of SGLT-2 inhibitors in NAFLD across different populations, though it may also contribute to heterogeneity in baseline characteristics and outcomes.

In summary, the included studies encompassed a diverse range of participants, interventions, and study designs, with a focus on SGLT-2 inhibitors in NAFLD patients (Table 1). The variation in treatment duration, control groups, and geographical settings provides a robust foundation for evaluating the comparative efficacy of SGLT-2 inhibitors in lipid profiles, BMI, and HbA1c in this population. Risk of bias of included studies is provided in Supplementary Digital Content Fig. 1, available at: http://links.lww.com/MS9/A894.

Table 1.

Population baseline characteristics of included studies with references.

Studies	Study design	Intervention	Age (mean ± SD) Intervention	Age (mean ± SD) Control	Treatment Duration (weeks)	No. of participants Total	No. of participants (Intervention/ Control)
Kuchay et al[24]	Placebo- controlled	Empagliflozin	51 ± 9	49 ± 10	20	42	(22/20)
Phruksotsai et al[51]	Double-blinded, placebo- controlled	Dapagliflozin	57 ± 6	62 ± 7	12	38	(18/20)
Chehrehgosha et al [52]	Double-blinded, placebo- controlled	Empagliflozin	51 ± 8	51 ± 8	24	72	(35/37)
Taheri et al[53]	Double-blinded, placebo- controlled	Empagliflozin	43 ± 8	44 ± 9	24	90	(43/47)
Takahashi et al[54]	Open-label RCT	Ipragliflozin	55 ± 15	57 ± 16	72	46	(21/25)
Aso et al[55]	Placebo-controlled	Dapagliflozin	56 ± 11	51 ± 14	24	57	(33/24)
Han et al[56]	Placebo-controlled	Ipragliflozin	56 ± 8	57 ± 9	24	44	(29/15)
Eriksson et al[57]	Double-blinded, placebo- controlled	Dapagliflozin	65 ± 6	65 ± 6	12	40	(20/20)
Elhini et al[58]	Double-blinded, placebo- controlled	Empagliflozin	47 ± 7	46 ± 8	26	160	(80/80)
Shi et al[59]	Placebo-controlled	Dapagliflozin	49 ± 9	47 ± 10	24	78	(40/38)
Hussain et al[60]	Double-blinded, placebo- controlled	Dapagliflozin	29 ± 16	31 ± 14	12	138	(67/71)

LDL-C change

For LDL reduction in patients with nonalcoholic fatty liver disease (NAFLD), we analyzed data from nine RCTs involving 576 patients. Our Bayesian network meta-analysis demonstrated (Figs. 2,3) that empagliflozin achieved the greatest efficacy compared to placebo, with a mean difference (MD) of −8.07 mg/dL (95% credible interval [CrI]: −25.92 to 1.41; SUCRA: 95.38%). Dapagliflozin followed, showing an MD of −0.10 mg/dL (95% CrI: −11.87 to 7.14; SUCRA: 53.81%) versus placebo. Ipragliflozin exhibited an MD of 11.40 mg/dL (95% CrI: −2.36 to 25.01; SUCRA: 2.80%) compared to placebo, which served as the reference (SUCRA: 48.00%).

Figure 2.

Comparative analysis of LDL-C changes across treatments. (A) Network plot illustrating treatment comparisons with rankings. (B) Forest plot depicting mean differences in LDL-C relative to placebo (posterior median with 95% CrI). (C) Rank plot showing the probability of each treatment achieving specific rankings (higher rankings associated with smaller outcome values). (D) SUCRA plot displaying the probability of treatments being better ranked.

Figure 3.

League table with heatmap for head-to-head comparison of LDL-C outcomes across interventions. Values represent mean differences in LDL-C (mg/dL) with 95% credible intervals. Color intensity reflects the magnitude of change, with blue indicating LDL-C reduction and orange indicating LDL-C increase.

HDL-C change

For HDL cholesterol increases in patients with nonalcoholic fatty liver disease (NAFLD), we analyzed data from nine RCTs involving 576 patients. Our Bayesian network meta-analysis showed that Ipragliflozin had the highest efficacy compared to Placebo (Figs. 4,5), with a mean difference (MD) of 2.28 mg/dL (95% credible interval [CrI]: −1.21 to 4.83; SUCRA: 85.11%). Empagliflozin followed, with an MD of 0.96 mg/dL (95% CrI: −2.11 to 4.28; SUCRA: 53.98%) versus placebo. Dapagliflozin demonstrated an MD of 0.15 mg/dL (95% CrI: −0.70 to 3.40; SUCRA: 42.34%) compared to placebo, which served as the reference (SUCRA: 18.57%).

Figure 4.

Comparative analysis of HDL-C changes across treatments. (A) Network plot illustrating treatment comparisons with rankings. (B) Forest plot depicting mean differences in HDL-C relative to placebo (posterior median with 95% CrI). (C) Rank plot showing the probability of each treatment achieving specific rankings (higher rankings associated with higher outcome values). (D) SUCRA plot displaying the probability of treatments being better ranked.

Figure 5.

League table with heatmap for head-to-head comparison of HDL-C outcomes across interventions. Values represent mean differences in HDL-C (mg/dL) with 95% credible intervals. Color intensity reflects the magnitude of change, with blue indicating HDL-C reduction and orange indicating HDL-C increase.

Total cholesterol change

For total cholesterol reduction in patients with nonalcoholic fatty liver disease (NAFLD), we analyzed data from six RCTs involving 215 patients. Our Bayesian network meta-analysis indicated that empagliflozin exhibited the greatest efficacy compared to placebo (Figs. 6,7), with a mean difference (MD) of −15.08 mg/dL (95% credible interval [CrI]: −58.87 to 5.38; SUCRA: 94.32%). Dapagliflozin showed an MD of 0.06 mg/dL (95% CrI: −28.57 to 28.72; SUCRA: 46.73%) versus placebo. Ipragliflozin had an MD of 9.46 mg/dL (95% CrI: −19.62 to 38.88; SUCRA: 10.47%) compared to placebo, which served as the reference (SUCRA: 48.47%).

Figure 6.

Comparative Analysis of Total Cholesterol (TC) Changes Across Treatments. (A) Network plot illustrating treatment comparisons with rankings. (B) Forest plot depicting mean differences in TC relative to placebo (posterior median with 95% CrI). (C) Rank plot showing the probability of each treatment achieving specific rankings (higher rankings associated with lower outcome values). (D) SUCRA plot displaying the probability of treatments being better ranked.

Figure 7.

League table with heatmap for head-to-head comparison of total cholesterol (TC) outcomes across interventions. Values represent mean differences in TC (mg/dL) with 95% credible intervals. Color intensity reflects the magnitude of change, with blue indicating TC reduction and orange indicating TC increase.

Triglyceride change

For triglyceride reduction in patients with nonalcoholic fatty liver disease (NAFLD), we analyzed data from nine RCTs involving 576 patients. Our Bayesian network meta-analysis demonstrated that ipragliflozin achieved the greatest efficacy compared to placebo (Figs. 8,9), with a mean difference (MD) of −22.39 mg/dL (95% credible interval [CrI]: −39.25 to −7.27; SUCRA: 82.26%). Empagliflozin followed, with an MD of −22.35 mg/dL (95% CrI: −48.38 to 3.86; SUCRA: 79.21%) versus placebo. Dapagliflozin showed an MD of −0.63 mg/dL (95% CrI: −15.41 to 5.10; SUCRA: 26.31%) compared to placebo, which served as the reference (SUCRA: 12.22%).

Figure 8.

Comparative analysis of triglyceride (TG) changes across treatments. (A) Network plot illustrating treatment comparisons with rankings. (B) Forest plot depicting mean differences in TG relative to placebo (posterior median with 95% CrI). (C) Rank plot showing the probability of each treatment achieving specific rankings (higher rankings associated with lower outcome values). (D) SUCRA plot displaying the probability of treatments being better ranked.

Figure 9.

League table with heatmap for head-to-head comparison of triglyceride (TG) outcomes across interventions. Values represent mean differences in TG (mg/dL) with 95% credible intervals. Color intensity reflects the magnitude of change, with blue indicating TG reduction and orange indicating TG increase.

Change in body mass index (BMI)

For BMI reduction in patients with nonalcoholic fatty liver disease (NAFLD), we analyzed data from 10 RCTs involving 765 patients. Our Bayesian network meta-analysis showed that dapagliflozin had the greatest efficacy compared to placebo (Figs. 10,11), with a mean difference (MD) of −1.23 kg/m² (95% credible interval [CrI]: −2.11 to −0.41; SUCRA: 73.28%). Empagliflozin followed, with an MD of −1.10 kg/m² (95% CrI: −2.00 to −0.27; SUCRA: 63.20%) versus placebo. Ipragliflozin demonstrated an MD of −1.08 kg/m² (95% CrI: −2.23 to 0.04; SUCRA: 62.15%) compared to placebo, which served as the reference (SUCRA: 1.38%).

Figure 10.

Comparative analysis of BMI changes across treatments. (A) Network plot illustrating treatment comparisons with rankings. (B) Forest plot depicting mean differences in BMI relative to placebo (posterior median with 95% CrI). (C) Rank plot showing the probability of each treatment achieving specific rankings (higher rankings associated with lower outcome values). (D) SUCRA plot displaying the probability of treatments being better ranked.

Figure 11.

League table with heatmap for head-to-head comparison of BMI outcomes across interventions. Values represent mean differences in BMI (kg/m²) with 95% credible intervals. Color intensity reflects the magnitude of change, with blue indicating reduction and orange indicating increase.

Change in HbA1c

For HbA1c reduction in patients with nonalcoholic fatty liver disease (NAFLD), we analyzed data from 10 RCTs involving 715 patients. Our Bayesian network meta-analysis revealed that empagliflozin demonstrated the greatest efficacy compared to placebo (Figs. 12,13), with a mean difference (MD) of −0.69% (95% credible interval [CrI]: −1.52 to −0.06; SUCRA: 81.84%). Dapagliflozin followed, with an MD of −0.50% (95% CrI: −1.16 to 0.13; SUCRA: 63.31%) versus placebo. Ipragliflozin showed an MD of −0.37% (95% CrI: −1.24 to 0.47; SUCRA: 48.60%) compared to placebo, which served as the reference (SUCRA: 6.25%).

Figure 12.

Comparative analysis of HbA1c changes across treatments. (A) Network plot illustrating treatment comparisons with rankings. (B) Forest plot depicting mean differences in HbA1c relative to placebo (posterior median with 95% CrI). (C) Rank plot showing the probability of each treatment achieving specific rankings (higher rankings associated with lower outcome values). (D) SUCRA plot displaying the probability of treatments being better ranked.

Figure 13.

League table with heatmap for head-to-head comparison of HbA1c outcomes across interventions. Values represent mean differences in HbA1c (%) with 95% credible intervals. Color intensity reflects the magnitude of change, with blue indicating reduction and orange indicating increase.

Discussion

Nonalcoholic fatty liver disease (NAFLD) represents a global health challenge, affecting 25–30% of adults and up to 70–90% of individuals with type 2 diabetes mellitus (T2DM) or obesity, with significant implications for dyslipidemia, cardiovascular risk, and liver-related morbidity^[1,5]. This Bayesian network meta-analysis (NMA) of 11 randomized controlled trials (RCTs) involving 805 patients evaluates the comparative effectiveness of sodium-glucose transporter 2 (SGLT-2) inhibitors – dapagliflozin, empagliflozin, and ipragliflozin – versus placebo in improving lipid profiles (low-density lipoprotein cholesterol [LDL-C], high-density lipoprotein cholesterol [HDL-C], total cholesterol, and triglycerides), body mass index (BMI), and HbA1c in NAFLD patients. Our findings indicate that empagliflozin excels in reducing LDL-C, total cholesterol, and HbA1c, Ipragliflozin leads in increasing HDL-C and reducing triglycerides, and dapagliflozin is most effective in lowering BMI. These results provide valuable insights for NAFLD management, particularly in patients with concurrent T2DM, and require comparison with existing literature to elucidate mechanisms, contextualize findings, and address limitations.

Lipid profile improvements

Our analysis demonstrated that empagliflozin achieved the greatest reductions in LDL-C (MD: −8.07 mg/dL, 95% CrI: −25.92 to 1.41; SUCRA: 95.38%) and total cholesterol (MD: −15.08 mg/dL, 95% CrI: −58.87 to 5.38; SUCRA: 94.32%) compared to placebo. Ipragliflozin was most effective in increasing HDL-C (MD: 2.28 mg/dL, 95% CrI: −1.21 to 4.83; SUCRA: 85.11%) and reducing triglycerides (MD: −22.39 mg/dL, 95% CrI: −39.25 to −7.27; SUCRA: 82.26%). Dapagliflozin showed modest effects across lipid outcomes, with less pronounced reductions compared to empagliflozin and ipragliflozin. While SUCRA rankings suggest empagliflozin’s superiority for LDL-C and total cholesterol, wide CrIs (e.g. −25.92 to 1.41 for LDL-C) indicate uncertainty, warranting cautious interpretation.

These findings align with several studies in the provided literature. Mantovani et al reported that SGLT-2 inhibitors significantly reduce liver fat and improve lipid profiles in NAFLD, with empagliflozin showing reductions in total cholesterol and triglycerides^[23]. This is consistent with our results, suggesting empagliflozin’s robust effect on lipid metabolism. Kuchay et al’s RCT found that empagliflozin (10 mg/day) reduced liver fat and LDL-C by approximately 7.5 mg/dL in NAFLD patients with T2DM over 20 weeks, closely mirroring our MD of −8.07 mg/dL^[24]. The mechanism behind empagliflozin’s efficacy may involve enhanced fatty acid oxidation and reduced de novo lipogenesis, as supported by studies showing SGLT-2 inhibitors’ modulation of hepatic lipid metabolism^[13,17]. Additionally, empagliflozin’s improvement in insulin sensitivity likely reduces circulating lipids by decreasing lipotoxicity, a key driver of NAFLD progression^[5,8].

Ipragliflozin’s superiority in HDL-C and triglyceride outcomes is supported by studies in Asian populations. Ito et al’s RCT reported that ipragliflozin (50 mg/day) increased HDL-C by 2.5 mg/dL and reduced triglycerides by 20 mg/dL in NAFLD patients with T2DM, aligning with our findings^[25]. Similarly, Ohki et al found ipragliflozin effective in improving metabolic parameters in NAFLD, likely due to its modulation of lipoprotein lipase activity, which enhances triglyceride clearance and HDL particle formation^[20]. The ethnic specificity of ipragliflozin’s use in East Asia may contribute to its efficacy, as Asian cohorts may have genetic predispositions (e.g. CETP polymorphisms) that enhance responsiveness to SGLT-2 inhibitors’ lipid effects^[4]. These genetic factors could explain the stronger HDL-C and triglyceride improvements observed in our analysis compared to studies in Western populations^[23].

Dapagliflozin’s modest lipid effects (e.g. LDL-C MD: −0.10 mg/dL, triglycerides MD: −0.63 mg/dL) are consistent with Eriksson et al’s RCT, which found minimal lipid profile changes despite significant liver fat reduction with dapagliflozin^[57]. This may reflect dapagliflozin’s weaker impact on hepatic lipid metabolism compared to empagliflozin, potentially due to differences in SGLT-2 selectivity or downstream signaling pathways^[14]. Unlike empagliflozin, which may upregulate PPAR-α to enhance lipid oxidation, dapagliflozin’s primary mechanism appears to be glycosuria-driven caloric loss, with less direct influence on lipid synthesis^[13,16].

Compared to other NAFLD therapies, SGLT-2 inhibitors offer moderate lipid-lowering effects. Statins, commonly used in NAFLD, reduce LDL-C by 30–50 mg/dL, surpassing empagliflozin’s effect, but lack the glycemic and weight-loss benefits of SGLT-2 inhibitors^[6]. Pioglitazone improves HDL-C and triglycerides but increases BMI, contrasting with SGLT-2 inhibitors’ weight-reducing profile^[11]. These comparisons highlight SGLT-2 inhibitors’ unique role in addressing multiple metabolic abnormalities in NAFLD, complementing their cardiovascular benefits^[7,12].

BMI and HbA1c reductions

Dapagliflozin resulted in BMI reduction (MD: −1.23 kg/m², 95% CrI: −2.11 to −0.41; SUCRA: 73.28%), followed by empagliflozin (MD: −1.10 kg/m²; SUCRA: 63.20%) and ipragliflozin (MD: −1.08 kg/m²; SUCRA: 62.15%). For HbA1c, empagliflozin was most effective (MD: −0.69%, 95% CrI: −1.52 to −0.06; SUCRA: 81.84%), followed by dapagliflozin (MD: −0.50%; SUCRA: 63.31%).

These results are supported by the literature. Eriksson et al’s RCT reported that dapagliflozin (10 mg/day) reduced BMI in NAFLD patients with T2DM, consistent with our MD of −1.23 kg/m^2[57]. This effect likely stems from caloric loss via glycosuria, reducing visceral and hepatic fat, as described by Heerspink et al^[13]. Empagliflozin’s HbA1c reduction aligns with Kahl et al’s findings, which reported a 0.7% decrease in HbA1c, attributed to improved insulin sensitivity and reduced glucotoxicity^[17]. Ipragliflozin’s BMI effect is corroborated by Ohki et al, who noted a significant reduction in NAFLD patients, likely driven by similar glycosuric mechanisms^[20].

Compared to other NAFLD therapies, SGLT-2 inhibitors show superior BMI reduction. Sumida et al noted that GLP-1 receptor agonists reduce BMI by 0.5–1.0 kg/m², less than dapagliflozin’s effect^[11]. For HbA1c, SGLT-2 inhibitors perform comparably to metformin but outperform pioglitazone, which has limited glycemic impact in NAFLD^[23]. The weight loss induced by SGLT-2 inhibitors likely amplifies their metabolic benefits, as reduced visceral fat correlates with decreased steatosis, supporting lifestyle interventions^[9].

Comparison with current literature

Our findings diverge from some studies due to methodological and population differences. Mantovani et al meta-analysis reported smaller LDL-C reductions (approximately 5 mg/dL) with SGLT-2 inhibitors, possibly because it included non-NAFLD T2DM patients with milder dyslipidemia^[23]. Our NAFLD-specific cohort, with higher baseline lipid levels, likely amplified treatment effects. Similarly, Monami et al’s meta-analysis found no significant HDL-C increase with SGLT-2 inhibitors, somewhat contrasting with our Ipragliflozin results, potentially due to fewer Asian studies or shorter trial durations in their analysis^[29].

Empagliflozin’s superior performance in LDL-C and HbA1c may reflect its high SGLT-2 selectivity, enhancing hepatic and systemic metabolic effects^[14]. Ipragliflozin’s efficacy in HDL-C and triglycerides may be linked to its longer half-life, sustaining lipid-lowering effects in Asian populations^[16]. Dapagliflozin’s focus on BMI reduction aligns with its broader glycosuric effects, which may dilute its hepatic lipid specificity^[13]. These variations underscore the importance of tailoring SGLT-2 inhibitor therapy to patient-specific metabolic profiles.

Mechanistic explanations

The differential efficacy of SGLT-2 inhibitors likely arises from their effects on hepatic lipid metabolism and insulin sensitivity. Empagliflozin’s reduction in LDL-C and total cholesterol may result from increased LDL receptor expression and cholesterol clearance, as suggested by preclinical models of SGLT-2 inhibition^[13,17]. Ipragliflozin’s HDL-C and triglyceride improvements could involve enhanced lipoprotein lipase activity and reduced very-low-density lipoprotein (VLDL) secretion, particularly in Asian cohorts^[16,20]. Dapagliflozin’s BMI reduction is likely driven by glycosuria-induced caloric loss, reducing visceral fat and indirectly improving lipid profiles^[13,18].

Heterogeneity in RCT populations also influences outcomes. Studies with higher T2DM prevalence showed larger HbA1c reductions, as glycemic control is more responsive in diabetic cohorts^[5,8]. Asian populations in ipragliflozin trials may exhibit genetic predispositions to HDL-C responsiveness, such as CETP variants, unlike Western cohorts in empagliflozin studies^[4]. Baseline NAFLD severity further modulates effects, with milder disease enhancing lipid and BMI improvements^[3].

Study limitations

Our study has several limitations. First, the inclusion of only 11 RCTs with 805 patients limits statistical power, particularly for total cholesterol and HDL-C outcomes. Second, the absence of data on canagliflozin, sotagliflozin, and ertugliflozin restricts generalizability to all SGLT-2 inhibitors, as these agents were included in our protocol but lacked sufficient RCT evidence^[19,21,22]. Third, heterogeneity in baseline NAFLD severity and T2DM status may confound results, despite our use of a random-effects model to account for variability^[28]. Fourth, the short trial durations (mostly 12–24 weeks) may underestimate long-term effects on lipid profiles and BMI^[31]. Fifth, reliance on indirect comparisons due to the absence of head-to-head SGLT-2 inhibitor trials introduces potential inconsistency, though our network consistency was verified^[27,28]. Finally, the lack of liver-specific outcomes (e.g. steatosis grade and fibrosis scores) limits insights into hepatic benefits beyond metabolic parameters^[2].

Implications and future directions

Our findings suggest that empagliflozin is optimal for reducing LDL-C, total cholesterol, and HbA1c, ipragliflozin for improving HDL-C and triglycerides, and dapagliflozin for lowering BMI in NAFLD patients. Clinically, empagliflozin may be preferred for NAFLD patients with elevated LDL-C and HbA1c, particularly those with T2DM, while ipragliflozin suits Asian patients with low HDL-C. Dapagliflozin is ideal for obese patients prioritizing weight loss. Clinicians should consider side effects (e.g. urinary tract infections and volume depletion) and costs, as SGLT-2 inhibitors vary in affordability. Patient selection should account for T2DM status, baseline lipid profiles, and ethnicity^[29,30]. These SGLT-2 inhibitors complement lifestyle interventions, which significantly reduce NAFLD features, and may mitigate cardiovascular risk^[7,9].

Future RCTs should include canagliflozin, sotagliflozin, and ertugliflozin to broaden the evidence base^[19,21,22]. Longer trial durations (≥52 weeks) and inclusion of histological outcomes (e.g. steatosis and fibrosis) would clarify SGLT-2 inhibitors’ hepatic benefits^[2]. Head-to-head trials comparing SGLT-2 inhibitors would provide direct evidence, reducing reliance on indirect comparisons^[27]. Meta-regression exploring ethnicity, T2DM status, and NAFLD severity could address heterogeneity sources^[28]. Additionally, integrating imaging (e.g. MRI-PDFF) and cardiovascular outcomes would strengthen the case for SGLT-2 inhibitors in NAFLD management^[7].

Conclusion

This Bayesian NMA provides robust evidence that SGLT-2 inhibitors – empagliflozin, ipragliflozin, and dapagliflozin – significantly improve lipid profiles, BMI, and HbA1c in NAFLD patients. Empagliflozin and ipragliflozin stand out for lipid outcomes, while dapagliflozin excels in BMI reduction, supporting their role in personalized NAFLD management. These findings align with their cardiovascular and metabolic benefits, offering a multifaceted approach to a complex disease. Further research is needed to confirm long-term effects and optimize treatment strategies for diverse NAFLD populations.

Ethical approval

This meta-analysis study did not require ethical approval as it involves the secondary analysis of previously published, anonymized data available in the public domain. According to the guidelines of the Institution Ethics Committee, meta-analyses that do not involve direct human participation, collection of new data, or access to identifiable patient information are exempt from formal ethical review. Therefore, no specific ethics committee approval or reference number is applicable for this study. The research was conducted in accordance with the principles outlined in the Declaration of Helsinki.

Conflicts of interest disclosure

The authors declare that they have no competing interests.

Sources of funding

No sources of funding were received for this study.

Author contributions

I.K.: study concept or design, data collection, data analysis or interpretation, and writing of the paper. M.I.H., M.A., S.G.A.-B.: drafting, reviewing, and editing.

Research registration unique identifying number (UIN)

Registration ID: CRD420251032868.

Guarantor

Ibrahim Khalil.

Provenance and peer review

Not applicable.

Data availability statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.