Sodium-glucose co-transporter 2 inhibitors improve insulin resistance and β-cell function in type 2 diabetes: A meta-analysis

Abstract

BACKGROUND

Sodium-glucose cotransporter 2 (SGLT2) inhibitors are widely used for the treatment of type 2 diabetes (T2D).

AIM

To evaluate the influence of SGLT2 inhibitors on homeostasis model assessment of insulin resistance (HOMA-IR) and β-cell function (HOMA-β) in patients with T2D in a meta-analysis.

METHODS

Randomized controlled trials (RCTs) comparing SGLT2 inhibitors to placebo in T2D patients, with a minimum treatment duration of 12 weeks, were searched using the PubMed, EMBASE, and Cochrane Library databases. Risk of bias was assessed using the Cochrane Risk of Bias Tool, and the certainty of evidence was evaluated using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) system. Changes in HOMA-IR and HOMA-β were the outcomes analyzed. Meta-analyses were performed using a random-effects model by incorporating the potential influences of heterogeneity.

RESULTS

Of 1388 articles identified, 24 RCTs met the inclusion criteria. 23 of the included studies were double-blind RCTs with low risk of bias. Pooled results including 2272 patients showed that SGLT2 inhibitors significantly reduced HOMA-IR compared to placebo [mean difference (MD) = -0.81, 95% confidence interval (CI): -1.11 to -0.52, P < 0.001; I² = 82%], indicating reduced insulin resistance. Additionally, meta-analysis with 2845 patients suggested that SGLT2 inhibitors significantly increased HOMA-β (MD = 7.90, 95%CI: 5.44-10.37, P < 0.001; I² = 74%) compared to placebo in patients with T2D, indicating improved β-cell function. Based on GRADE assessment, the certainty of evidence was rated moderate for both outcomes due to heterogeneity. Subgroup analyses showed that HOMA-β increased more substantially in non-Asian studies than in Asian studies (P for subgroup difference < 0.01). Subgroup analyses according to the individual medications of SGLT2 inhibitors all showed significant improvement of HOMA-IR and HOMA-β (P all < 0.05). No significant publication bias was detected (P for Egger’s test all > 0.05).

CONCLUSION

SGLT2 inhibitors are associated with improvements in insulin resistance and β-cell function in patients with T2D, although the certainty of evidence is moderate due to heterogeneity.

Key Words: β-cell function; Homeostasis model assessment; Insulin resistance; Meta-analysis; Sodium-glucose co-transporter 2 inhibitors

Core Tip: This meta-analysis, based on 24 randomized controlled trials, showed that sodium-glucose cotransporter 2 (SGLT2) inhibitors significantly improve insulin resistance and β-cell function in patients with type 2 diabetes (T2D), as evidenced by reductions in homeostasis model assessment of insulin resistance and increases in β-cell function. These findings highlight the potential of SGLT2 inhibitors to alleviate insulin resistance and improve β-cell function beyond glucose lowering. These results provide novel insights into the therapeutic role of SGLT2 inhibitors in managing T2D and underscore their ability to modify disease progression by targeting key pathophysiological mechanisms.

INTRODUCTION

Diabetes is a prevalent metabolic disorder among the global population[1,2]. According to the International Diabetes Federation, there were 537 million adults aged 20-79 years with diabetes in 2021, with projections reaching 643 million by 2030 and 743 million by 2045[3]. Type 2 diabetes (T2D) accounts for over 90% of these cases[4,5]. The hallmark of T2D is chronic hyperglycemia due to insulin resistance, where cells fail to respond adequately to insulin[6]. Initially, insulin production increases to compensate; however, over time, pancreatic β-cell function deteriorates, leading to insufficient insulin secretion[7].

The progressive worsening of insulin resistance and decline in β-cell function are central to the pathogenesis of T2D[8]. Many antidiabetic agents, such as sulfonylureas and metformin, target these key processes[9]. Recently, sodium-glucose cotransporter 2 (SGLT2) inhibitors have emerged as a novel class of antidiabetic drugs[10,11]. These inhibitors work by competitively inhibiting the SGLT2 protein in the renal tubules, reducing glucose reabsorption and increasing glucose excretion, thereby lowering the blood glucose levels[10]. Beyond their glucose-lowering effects, SGLT2 inhibitors offer additional benefits, including improvements in blood pressure, weight management, and lipid profiles[12-14]. Evidence suggests that they may reduce cardiovascular risk by mitigating atherosclerosis processes[15] and decrease hospitalizations for heart failure and progression of renal disease[16-18]. Interestingly, although SGLT2 inhibitors do not directly target insulin resistance or β-cell function, they have been hypothesized to benefit these aspects in T2D patients[19].

Clinically, homeostatic model assessment (HOMA) is a widely used method for evaluating insulin resistance (HOMA-IR) and β-cell function (HOMA-β) from fasting glucose and insulin levels[20]. Since its introduction in 1985[21], HOMA has become a convenient and reliable indicator in clinical practice. SGLT2 inhibitors have garnered significant attention for their multifaceted benefits, including cardiovascular and renal protection, beyond glucose-lowering effects[10]. Although SGLT2 inhibitors are primarily known for their glucose-lowering effects, emerging studies suggest potential benefits in insulin resistance and β-cell function. However, findings from individual randomized controlled trials (RCTs) have been inconsistent, with some reporting significant improvements in HOMA-IR and HOMA-β[22-40], while others found no significant changes[41-45]. Given these conflicting results, a comprehensive meta-analysis is warranted to systematically evaluate the influence of SGLT2 inhibitors on insulin resistance and β-cell function. This study aims to fill this gap by synthesizing evidence from RCTs, providing clarity on the metabolic benefits of SGLT2 inhibitors. By addressing this critical aspect of diabetes management, the findings of this meta-analysis have the potential to inform clinical decision-making and highlight the broader therapeutic value of SGLT2 inhibitors in T2D care.

MATERIALS AND METHODS

During the design and implementation of this study, we followed the guidelines set forth by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRIMSA 2000)[46,47] and the Cochrane Handbook[48]. The protocol of the meta-analysis has been registered at the International Prospective Register of Systematic Reviews with ID: CRD42024577384.

Study inclusion and exclusion criteria

This meta-analysis included studies that met the inclusion criteria specified in the PICOS principle, as follows.

Patients: Patients with T2D, either drug naive, or on background therapy with metformin or/and other antidiabetic agents, including insulin.

Intervention: The intervention was SGLT2 inhibitors (including canagliflozin, dapagliflozin, empagliflozin, ertugliflozin, tofogliflozin, bexagliflozin, henagliflozin, ipragliflozin, licogliflozin, luseogliflozin, remogliflozin, sergliflozin, or sotagliflozin) as monotherapy or add-on therapy with approved dosages where both arms have the same background therapy, and the minimal treatment duration was 12 weeks. The 12-week minimum treatment duration was chosen because it represents a clinically relevant timeframe in which meaningful metabolic changes, including improvements in insulin resistance and β-cell function, can be observed in RCTs of T2D.

Control: Placebo. We excluded studies that used active comparators [e.g., metformin, sulfonylureas, dipeptidyl peptidase 4 (DPP4) inhibitors, glucagon-like peptide-1 receptor agonists) to ensure that the observed effects were attributable to SGLT2 inhibitors alone rather than differences between drug classes.

Outcome: Between-group difference of changes of either HOMA-IR or HOMA-β. HOMA-IR was calculated using the formula fasting plasma glucose (FPG) (mmol/L) × fasting insulin (FI) (Mu/mL)/22.5; HOMA-β was calculated using the formula 20 × FI (Mu/mL)/[FPG (mmol/L) - 3.5].

Study design: RCTs published in English in peer-reviewed journals. Excluded from the analysis were reviews, editorials, preclinical studies, studies not designed as RCTs, studies involving patients with type 1 diabetes, studies of patients treated with SGLT2 inhibitors for less than 12 weeks, studies with a control group that was not a placebo, or studies that did not report the outcome of interest. For monotherapy and add-on therapy, studies with active treatment comparisons to a SGLT2 inhibitor were excluded. For add-on therapy, studies without the same background therapies in both arms were also excluded. If studies with overlapping patients were retrieved, the one with the largest sample size was analyzed in the meta-analysis.

Database search

The MEDLINE (PubMed), EMBASE (Ovid), and CENTER (Cochrane Library) databases were searched using a combination of the following terms: (1) (“Sodium glucose transporter 2 inhibitor” OR “sodium glucose transporter ii inhibitor” OR “SGLT 2 inhibitor” OR “SGLT-2 inhibitor” OR “SGLT2” OR “sodium glucose cotransporter 2 inhibitors” OR “canagliflozin” OR “dapagliflozin” OR “empagliflozin” OR “ertugliflozin” OR “tofogliflozin” OR “bexagliflozin” OR “henagliflozin” OR “ipragliflozin” OR “licogliflozin” OR “luseogliflozin” OR “remogliflozin” OR “sergliflozin” OR “sotagliflozin”); (2) “Type 2 diabetes” OR “T2D” OR “T2DM”; (3) “Placebo”; and (4) “Random” OR “randomly” OR “randomized” OR “control” OR “allocated”, limited to clinical studies in humans. Search terms for HOMA-IR or HOMA-β were not incorporated in order to perform an expanded literature search to avoid missing potentially relevant studies. Only studies that included human subjects and were published as full-length articles in peer-reviewed journals in English were considered. The exclusion of non-English studies was intended to minimize potential errors in data interpretation, though this may limit the generalizability of the findings. Additionally, references to related reviews and original articles were screened as part of the final database search. The final database search was conducted on April 6, 2024.

Data collection and quality evaluation

Two authors conducted independent database searches, data extraction, and quality assessment in a blinded manner. Discrepancies were resolved through discussion with the corresponding author, who was independent of the funder. All methodological decisions and statistical analyses were performed by the research team without influence from the funding sponsor, to minimize potential bias. The data collected encompassed various aspects, including overall study information (e.g., first author, publication year, and study country), study design (e.g., double-blind or single blind, parallel-group or crossover), patient characteristics [e.g., number of patients, mean age, sex, baseline hemoglobin A1c (HbA1c), duration of T2D, and concurrent antidiabetic treatments], details of intervention with SGLT2 inhibitors (e.g., individual medication, daily dosage, and treatment duration), control details, and outcomes reported. The quality of the included RCTs was assessed using the Cochrane Risk of Bias Tool[48]. This tool evaluated various aspects such as random-sequence generation, allocation concealment, blinding of participants and outcome assessment, addressing incomplete outcome data, selective reporting, and other sources of bias. In addition, two reviewers evaluated the certainty of evidence using the GRADE system, which includes risk of bias, inconsistency, indirectness, imprecision and publication bias[49]. The certainty of evidence was classified as very low, low, moderate or high. Disagreements were resolved by discussion with the corresponding author.

Statistical analysis

Differences of the changes of HOMA-IR and HOMA-β between patients with T2D allocated to SGLT2 inhibitors vs placebo were summarized as the mean difference (MD) and the corresponding 95% confidence interval (CI)[48]. For studies with multiple intervention groups with different doses of SGLT2 inhibitors, the sample size in the control group was equally split to avoid unit-of-analysis errors, as detailed in the Cochrane Handbook[48]. Specifically, when a study included multiple SGLT2 inhibitor doses but shared a single control group, we equally split the control group sample size to avoid double-counting while maintaining statistical integrity. This approach is commonly used in meta-analyses to ensure balanced comparisons between intervention arms without over representing control data. Heterogeneity was assessed using the Cochrane Q test[48]. The I² statistic was also calculated, with I² > 50% indicating significant statistical heterogeneity[50]. A random-effects model was used to pool the results because this model can incorporate the potential influence of heterogeneity[48]. Sensitivity analysis was performed to evaluate the potential influence of an individual study on the results of the meta-analysis[51]. In addition, subgroup analyses were also conducted to evaluate the study characteristics on the outcomes, such as the study country (Asian vs non-Asian), sample size, mean age of the patients, proportion of men, baseline HbA1c, duration of T2D, whether concurrent antidiabetic treatments were used, individual medications of SGLT2 inhibitors, dosage of SGLT2 inhibitors (categorized as “low” or “high” based on the range of approved daily doses per clinical guidelines and product labeling; e.g., canagliflozin at 100 or 300 mg/day; dapagliflozin at 5 or 10 mg/day; luseogliflozin at 2.5 or 5 mg/day; ipragliflozin at 50 or 100 mg/day; ertugliflozin at 5 or 15 mg; empagliflozin at 10 or 25 mg/day), and treatment durations. The medians of continuous variables were selected as cutoffs for defining the subgroups. The P values for subgroup differences were generated using RevMan’s χ² test. No multiple comparison corrections were applied, as subgroup analyses were predefined based on prior hypotheses. In addition, univariate meta-regression analyses were also performed to evaluate the influence of study characteristics in continuous variables on the outcomes, such as sample size, mean ages of the patients, proportions of men, baseline HbA1c, duration of T2D, and treatment durations[48]. An evaluation of the publication bias was conducted via a visual inspection using funnel plots and by performing Egger’s regression asymmetry test[52]. A value of P < 0.05 was considered statistically significant. Statistical analyses were conducted using RevMan (Version 5.1; Cochrane, Oxford, United Kingdom) and Stata software (version 12.0; Stata Corporation, College Station, TX, United States).

RESULTS

Literature search

Figure 1 depicts the flowchart outlining the process of database searching and study identification, ultimately leading to the selection of studies for inclusion. Initially, a total of 1388 articles were obtained through the database searches, which was subsequently reduced to 870 articles after eliminating duplicate records. Subsequently, 620 articles were excluded based on an evaluation of their titles and abstracts, primarily due to their lack of relevance to the objective of the present meta-analysis. Following this screening process, 226 out of the remaining 250 articles were excluded after full-text reviews for the reasons outlined in Figure 1. Ultimately, 24 RCTs[22-45] were deemed suitable for quantitative analysis.

Figure 1 Flowchart for the literature search and study inclusion. DPP4: Dipeptidyl peptidase 4; SGLT2: Sodium-glucose cotransporter 2; HOMA: Homeostasis model assessment.

Study characteristics and data quality

An overview of the included studies is listed in Table 1. Since five studies included two intervention groups of SGLT2 inhibitors with different dosages[22,24,31,32,42], these datasets were included in the meta-analysis independently, making 29 datasets available for the meta-analysis. These studies were published between 2013 and 2023. The mean ages of the patients ranged between 51.1 and 71.5 years, and the proportions of men were 40.3-78.6%. The baseline HbA1c level was 7.2%-8.4%, and the duration of diabetes was from newly diagnosed T2D to 11.5 years. Patients from 10 studies were not being treated with concurrent antidiabetic medications[22-25,28,30,39,40,42,44]; meanwhile, in the other 14 studies[26,27,29,31-38,41,43,45], concurrent medications such as metformin, sulfonylureas, dipeptidyl-peptidase 4 inhibitors, thiazolidinediones, glucagon-like peptide-1 agonists, or insulin were used. A placebo was used as the control in all of the included studies. The treatment duration was 12-52 weeks. The outcome of HOMA-IR was reported in 17 studies[23-25,29,30,33-42,44,45], and the outcome of HOMA-β was reported in 15 studies[22-29,31,32,34,39,40,42,43]. The details of the study quality evaluation for the RCTs are shown in Table 2. All of the included studies were double-blind RCTs except for one study, which was an open-label study[35]. The details of random-sequence generation were reported in nine studies[30-34,37,38,44,45], while the details of allocation concealment were reported in 11 studies[22-27,33,37,38,42,44].

Table 1 Characteristics of the included studies[22-45].

Ref.	Country	Design	Patient number	Mean age (years)	Male (%)	Baseline HbA1c (%)	T2D duration (years)	Concurrent antidiabetic treatment	Intervention	Control	Treatment duration (weeks)	Outcomes reported
Inagaki et al[22]; 100 mg	Japan	R, DB, PC	111	57.7	71.2	8	NR	None	Canagliflozin, 100 mg/day	Placebo	12	HOMA-β
Inagaki et al[22]; 300 mg	Japan	R, DB, PC	113	57.3	72.6	8.1	NR	None	Canagliflozin, 300 mg/day	Placebo	12	HOMA-β
Jabbour et al[41]1	6 countries	R, DB, PC	447	54.9	54.8	8	5.7	Sitagliptin ± metformin	Dapagliflozin, 10 mg/day	Placebo	24	HOMA-IR
Kaku et al[23]	Japan	R, DB, PC	114	56.7	66.7	8.4	6.2	None	Tofogliflozin, 20 mg/day	Placebo	24	HOMA-IR and HOMA-β
Seino et al[42]; 2.5 mg	Japan	R, DB, PC	86	58.1	62.5	8.1	6.5	None	Luseogliflozin, 2.5 mg/day	Placebo	12	HOMA-IR and HOMA-β
Seino et al[42]; 5 mg	Japan	R, DB, PC	87	57	72.7	8.1	6.2	None	Luseogliflozin, 5 mg/day	Placebo	12	HOMA-IR and HOMA-β
Seino et al[24]; 2.5 mg	Japan	R, DB, PC	85	57.3	69.4	8	4.8	None	Luseogliflozin, 2.5 mg/day	Placebo	12	HOMA-IR and HOMA-β
Seino et al[24]; 5 mg	Japan	R, DB, PC	82	57.2	57.2	7.9	4.7	None	Luseogliflozin, 5 mg/day	Placebo	12	HOMA-IR and HOMA-β
Seino et al[25]	Japan	R, DB, PC	155	59.3	73.4	8.2	6.3	None	Luseogliflozin, 2.5 mg/day	Placebo	12	HOMA-IR and HOMA-β
Kashiwagi et al[26]	Japan	R, DB, PC	151	56.2	74.2	8.3	6.8	Pioglitazone	Ipragliflozin, 50 mg/day	Placebo	24	HOMA-β
Kashiwagi et al[27]	Japan	R, DB, PC	240	59.7	65.8	8.4	10.5	Sulfonylurea	Ipragliflozin, 50 mg/day	Placebo	24	HOMA-β
Kashiwagi et al[28]	Japan	R, DB, PC	129	59.4	69.8	8.3	6.7	None	Ipragliflozin, 50 mg/day	Placebo	16	HOMA-β
Kashiwagi et al[29]	Japan	R, DB, PC	168	56.7	58.9	8.3	7.7	Metformin	Ipragliflozin, 50 mg/day	Placebo	24	HOMA-IR and HOMA-β
Liao et al[30]	China	R, DB, PC	162	54	54.3	8	Newly diagnosed	None	Dapagliflozin, 10 mg/day	Placebo	12	HOMA-IR
Wang et al[43]	China	R, DB, PC	28	60.2	46.4	8.3	Newly diagnosed	With or without metformin	Dapagliflozin, 10 mg/day	Placebo	24	HOMA-β
Terra et al[31]2; 5 mg	7 countries	R, DB, PC	233	56.6	55.8	8.1	5	With or without metformin	Ertugliflozin, 5 mg/day	Placebo	26	HOMA-β
Terra et al[31]2; 15 mg	7 countries	R, DB, PC	228	56.2	57.5	8.3	5	With or without metformin	Ertugliflozin, 15 mg/day	Placebo	26	HOMA-β
Dagogo-Jack et al[32]3; 5 mg	12 countries	R, DB, PC	233	58.9	56.2	8.1	9.7	Metformin and sitagliptin	Ertugliflozin, 5 mg/day	Placebo	52	HOMA-β
Dagogo-Jack et al[32]3; 15 mg	12 countries	R, DB, PC	229	59.2	57.6	8	9.3	Metformin and sitagliptin	Ertugliflozin, 15 mg/day	Placebo	52	HOMA-β
Eriksson et al[33]	Sweden	R, DB, PC	40	65.3	78.6	7.4	6.6	Sulfonylurea or metformin	Dapagliflozin, 10 mg/day	Placebo	12	HOMA-IR
Han et al[34]	Korea	R, DB, PC	139	57.5	49.6	7.9	11.5	Metformin and sitagliptin	Ipragliflozin, 50 mg/day	Placebo	24	HOMA-IR and HOMA-β
Hattori[35]	Japan	R, OL, PC	109	NR	64.2	7.2	NR	Sulfonylureas, metformin, or an α-glucosidase inhibitor	Empagliflozin, 10 mg/day	Placebo	12	HOMA-IR
Brown et al[36]	United States	R, DB, PC	48	65.5	57.6	7.7	10	Metformin, sulfonylureas, DPP4 inhibitors, thiazolidinediones, or GLP-1 agonists	Dapagliflozin, 10 mg/day	Placebo	52	HOMA-IR
Chehrehgosha et al[44]	Iran	R, DB, PC	72	51.1	40.3	8	6.2	None	Empagliflozin, 10 mg/day	Placebo	24	HOMA-IR
Phrueksotsai et al[45]	Thailand	R, DB, PC	38	59.2	31.6	8	5	Metformin, sulfonylureas, DPP4 inhibitors, or thiazolidinediones,	Dapagliflozin, 10 mg/day	Placebo	12	HOMA-IR
Cheng et al[37]	China	R, DB, PC	124	71.5	53.4	7.7	NR	Insulin, sulfonylureas, DPP4 inhibitors, or metformin	Empagliflozin, 25 mg/day	Placebo	12	HOMA-IR
Gohari et al[38]	Iran	R, DB, PC	82	62.8	41.1	7.9	NR	Insulin, sulfonylureas, DPP4 inhibitors, or metformin	Empagliflozin, 10 mg/day	Placebo	26	HOMA-IR
Kwak et al[39]	Korea	R, DB, PC	152	59.8	50.3	7.8	6.4	None	Enavogliflozin, 0.3 mg/day	Placebo	24	HOMA-IR and HOMA-β
Yang et al[40]	Korea	R, DB, PC	96	58.3	54.2	8.1	5.9	None	Enavogliflozin, 0.3 mg/day	Placebo	12	HOMA-IR and HOMA-β

¹The study by Jabbour et al[41] was performed in Argentina, Germany, Mexico, Poland, the United Kingdom, and the United States.

²The study by Terra et al[31] was performed in the United States, Canada, Israel, Italy, Mexico, South Africa, and the United Kingdom.

³The study by Dagogo-Jack et al[32] was performed in the United States, Argentina, Colombia, Czech Republic, Hungary, Israel, Romania, Slovakia, Korea, Malaysia, Bulgaria, and Finland.

T2D: Type 2 diabetes; R: Randomized; DB: Double-blind; PC: Placebo-controlled; OL: Open-label; HbA1c: Hemoglobin A1c; NR: Not reported; DPP4: Dipeptidyl-peptidase 4; GLP-1: Glucagon-like peptide-1; HOMA-IR: Homeostasis model assessment of insulin resistance; HOMA-β: Homeostasis model assessment of β-cell function.

Table 2 Study quality evaluation via the Cochrane risk of bias tool[22-45].

Ref.	Random sequence generation	Allocation concealment	Blinding of participants	Blinding of outcome assessment	Incomplete outcome data addressed	Selective reporting	Other sources of bias
Inagaki et al[22]	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk
Jabbour et al[41]	Unclear	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk
Kaku et al[23]	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk
Seino et al[42]	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk
Seino et al[24]	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk
Seino et al[25]	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk
Kashiwagi et al[26]	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk
Kashiwagi et al[27]	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk
Kashiwagi et al[28]	Unclear	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk
Kashiwagi et al[29]	Unclear	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk
Liao et al[30]	Low risk	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk
Wang et al[43]	Unclear	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk
Terra et al[31]	Low risk	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk
Dagogo-Jack et al[32]	Low risk	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk
Eriksson et al[33]	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk
Han et al[34]	Low risk	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk
Hattori[35]	Unclear	Unclear	Low risk	High risk	Low risk	Low risk	Low risk
Brown et al[36]	Unclear	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk
Chehrehgosha et al[44]	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk
Phrueksotsai et al[45]	Low risk	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk
Cheng et al[37]	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk
Gohari et al[38]	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk
Kwak et al[39]	Unclear	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk
Yang et al[40]	Unclear	Unclear	Low risk	Low risk	Low risk	Low risk	Low risk

Influences of SGLT2 inhibitors on HOMA-IR

The meta-analysis of 17 studies[23-25,29,30,33-42,44,45] involving 19 datasets of 2272 patients showed that compared to placebo, SGLT2 inhibitors significantly reduced HOMA-IR (MD = -0.81, 95%CI: -1.11 to -0.52, P < 0.001; I² = 82%; Figure 2A) in patients with T2D, suggesting the attenuation of insulin resistance. Sensitivity analyses by excluding one dataset at a time showed similar results (-0.72 to -0.87, all P < 0.05). Further subgroup analyses did not support that the effect of SGLT2 inhibitors on HOMA-IR could be significantly influenced by other study characteristics, such as the study country, sample size, mean age, proportion of men, baseline HbA1c, duration of T2D, with or without concurrent antidiabetic treatments, individual medications of SGLT2 inhibitors, dosage of SGLT2 inhibitors, or treatment durations (all P for subgroup differences > 0.05; Table 3). Summarized certainty of evidence using the GRADE system is shown in Table 4. We downgraded evidence by one level for the inconsistency results due to significant heterogeneity observed. We judged the evidence to be of moderate certainty. In addition, as shown in Supplementary Table 1, the results of univariate meta-regression analyses did not show any of the following parameters could significantly modify the influence of SLGT2 inhibitors on HOMA-IR, including factors such as sample size, mean ages of the patients, proportions of men, baseline HbA1c, duration of T2D, or treatment durations (P all > 0.05).

Figure 2 Forest plots for the meta-analyses comparing the influence of sodium-glucose cotransporter 2 inhibitors vs placebo on homeostasis model assessment of insulin resistance and β-cell function in patients with type 2 diabetes. A: Forest plot for the outcome of homeostasis model assessment of insulin resistance; B: Forest plot for the outcome of homeostasis model assessment of β-cell function. SGLT2: Sodium-glucose cotransporter 2; CI: Confidence interval.

Table 3 Subgroup analyses for the influence of sodium-glucose co-transporter 2 inhibitors on homeostasis model assessment for insulin resistance in patients with type 2 diabetes.

Characteristic	Datasets (n)	Patients (n)	MD (95%CI)	I2 (%)	P for subgroup difference
Countries
Asian	16	1737	-0.83 (-1.08 to -0.58)	63
Non-Asian	3	535	-0.93 (-2.13 to 0.28)	85	0.88
Sample size
< 100	10	712	-0.61 (-0.89 to -0.32)	36
≥ 100	9	1560	-0.93 (-1.41 to -0.45)	91	0.26
Mean age
< 58 years	9	1342	-0.65 (-1.02 to -0.28)	84
≥ 58 years	9	821	-1.03 (-1.54 to -0.51)	75	0.24
Men
< 55%	9	1312	-0.79 (-1.37 to -0.21)	88
≥ 55%	10	960	-0.78 (-1.03 to -0.52)	54	0.97
Baseline HbA1c
< 8%	8	773	-1.07 (-1.51 to -0.64)	72
≥ 8%	11	1499	-0.62 (-0.97 to -0.27)	81	0.12
T2D duration
< 6.3 years	9	1169	-0.58 (-0.96 to -0.19)	80
≥ 6.3 years	7	788	-1.14 (-1.61 to -0.66)	76	0.08
Concurrent antidiabetic treatments
No	10	1077	-0.86 (-1.22 to -0.50)	75
Yes	9	1195	-0.74 (-1.17 to -0.31)	70	0.67
SGLT2 inhibitor medications
Dapagliflozin	5	735	-0.86 (-1.67 to -0.05)	87
Luseogliflozin	5	491	-0.62 (-0.94 to -0.29)	64
Ipragliflozin	2	307	-0.95 (-1.32 to -0.58)	0
Empagliflozin	4	387	-0.56 (-0.88 to -0.24)	1
Enavogliflozin	2	248	-1.69 (-2.60 to -0.77)	46	0.13
SGLT2 inhibitor dose
Low dose	8	895	-0.75 (-1.01 to -0.50)	34
High dose	8	1025	-0.65 (-1.07 to -0.22)	83	0.67
Treatment duration
12 weeks	11	1060	-0.69 (-0.90 to -0.47)	43
24-52 weeks	8	1212	-1.04 (-1.75 to -0.34)	90	0.34

T2D: Type 2 diabetes; HbA1c: Hemoglobin A1c; SGLT2: Sodium glucose transporter 2; MD: Mean difference; CI: Confidence interval.

Table 4 Summarized certainty of evidence using the Grading of Recommendations, Assessment, Development and Evaluation system.

Outcome	Quality assessment							Absolute effect MD (95%CI)	Quality
	Number of studies	Design	Risk of bias	Inconsistency	Indirectness	Imprecision	Other considerations
MD for HOMA-IR	17	RCTs	No serious risk of bias	Significant heterogeneity observed	No serious indirectness	No serious imprecision	None	-0.81 (-1.11 to -0.52)	Moderate
MD for HOMA-β	15	RCTs	No serious risk of bias	Significant heterogeneity observed	No serious indirectness	No serious imprecision	None	7.90 (5.44 to 10.37)	Moderate

Specific reasons for each Grading of Recommendations, Assessment, Development and Evaluation domain, including: Risk of bias: Downgraded if a significant proportion of studies had unclear or high risk of bias in key domains (e.g., random sequence generation, allocation concealment, or selective reporting); Inconsistency: Downgraded if substantial heterogeneity was observed (I² > 50%) and could not be explained by subgroup analyses or meta-regression; Indirectness: Evaluated but not downgraded, as all included studies directly assessed the population and outcomes of interest; Imprecision: Downgraded if confidence intervals were wide, overlapping no effect, or if the overall sample size was small; Publication bias: Assessed using funnel plots and Egger’s test, downgraded if significant asymmetry suggested potential bias. RCTs: Randomized controlled trials; MD: Mean difference; CI: Confidence interval; HOMA-IR: Homeostasis model assessment of insulin resistance; HOMA-β: Homeostasis model assessment of beta-cell function.

Influences of SGLT2 inhibitors on HOMA-β

The pooled results of 20 datasets from 15 studies[22-29,31,32,34,39,40,42,43] involving 2845 patients showed that SGLT2 inhibitors significantly increased HOMA-β compared to placebo in patients with T2D (MD = 7.90, 95%CI: 5.44-10.37, P < 0.001; I² = 74%; Figure 2B), suggesting the improvement of β-cell function. Sensitivity analyses by omitting one dataset at a time retrieved similar results (MD = 7.14-8.38, all P < 0.05). The subgroup analysis suggested a more remarkable improvement of HOMA-β in non-Asian studies as compared to Asian studies (MD: 20.86 vs 5.54, P for the subgroup difference < 0.001). Moreover, subgroup analyses according to the individual medications of SGLT2 inhibitors all showed significant improvement of HOMA-β (P all < 0.05; Table 5). Further subgroup analyses did not show that other study characteristics could significantly influence the effect of SGLT2 inhibitors on HOMA-β, such as the sample size, mean age, proportion of men, baseline HbA1c, duration of T2D, with or without concurrent antidiabetic treatments, dosage of SGLT2 inhibitors, or treatment duration (all P for subgroup differences > 0.05; Table 4). The certainty of evidence assessed using the GRADE system is presented in Table 4. Due to significant heterogeneity, we downgraded the evidence by one level for inconsistency, resulting in a judgment of moderate certainty. Similarly, the results of univariate meta-regression analyses did not show any of the predefined parameters could significantly modify the influence of SLGT2 inhibitors on HOMA-β (P all > 0.05, Supplementary Table 1).

Table 5 Subgroup analyses for the influence of sodium-glucose co-transporter 2 inhibitors on homeostasis model assessment for β-cell function in patients with type 2 diabetes.

Characteristic	Datasets (n)	Patients (n)	MD (95%CI)	I2 (%)	P for subgroup difference
Countries
Asian	15	1770	5.54 (3.99 to 7.10)	26
Non-Asian	2	461	20.86 (16.76 to 24.96)	0	< 0.001
Sample size
< 150	11	1056	6.64 (4.87 to 8.40)	0
≥ 150	9	1789	9.81 (5.02 to 14.60)	88	0.22
Mean age
< 58 years	11	1497	8.12 (4.57 to 11.67)	81
≥ 58 years	9	1348	7.61 (4.19 to 11.03)	61	0.84
Men
< 60%	10	1585	10.11 (5.14 to 15.08)	82
≥ 60%	10	1260	5.86 (3.93 to 7.78)	38	0.12
Baseline HbA1c
< 8.2%	12	1642	8.88 (6.08 to 11.69)	61
≥ 8.2%	8	1203	6.44 (2.15 to 10.73)	83	0.35
T2D duration
< 6.5 years	10	1246	8.50 (3.54 to 13.47)	84
≥ 6.5 years	8	1375	7.25 (4.59 to 9.91)	54	0.66
Concurrent antidiabetic treatments
No	11	1196	5.95 (4.19 to 7.70)	12
Yes	6	1160	7.11 (3.64 to 10.58)	64	0.56
SGLT2 inhibitor medications
Canagliflozin	2	224	6.49 (2.40 to 10.57)	29
Luseogliflozin	5	491	5.32 (2.59 to 8.05)	41
Ipragliflozin	5	827	5.72 (2.90 to 8.54)	46
Empagliflozin	4	923	17.13 (12.57 to 21.69)	53
Enavogliflozin	2	248	7.84 (0.48 to 15.19)	0	< 0.001
SGLT2 inhibitor dose
Low dose	11	1729	7.64 (4.52 to 10.77)	78
High dose	6	764	8.70 (3.21 to 14.19)	80	0.74
Treatment duration
12-16 weeks	9	940	6.13 (4.10 to 8.17)	29
24-52 weeks	11	1905	9.04 (4.70 to 13.37)	83	0.23

T2D: Type 2 diabetes; HbA1c: Hemoglobin A1c; SGLT2: Sodium glucose transporter 2; MD: Mean difference; CI: Confidence interval.

Publication bias

The funnel plots for the meta-analyses of the effects of SGLT2 inhibitors on HOMA-IR and HOMA-β in patients with T2D are shown in Figure 3, respectively. These plots are symmetrical on visual inspection, suggesting a low risk of publication bias. Egger’s regression tests also suggested a low risk of publication bias (P = 0.34 for HOMA-IR and P = 0.37 for HOMA-β).

Figure 3 Funnel plots evaluating the publication bias underlying the meta-analyses comparing the influence of sodium-glucose cotransporter 2 inhibitors vs placebo on homeostasis model assessment of insulin resistance and β-cell function in patients with type 2 diabetes. A: Funnel plot for the outcome of homeostasis model assessment of insulin resistance; B: Funnel plot for the outcome of homeostasis model assessment of β-cell function. MD: Mean difference.

DISCUSSION

The findings from our systematic review and meta-analysis of 24 RCTs (23 double-blind RCTs with low risk of bias) indicate that SGLT2 inhibitors significantly improve both insulin resistance and β-cell function in patients with T2D. Our analysis of 17 studies (2272 patients) revealed a marked reduction in HOMA-IR, indicating that SGLT2 inhibitors effectively attenuate insulin resistance. Additionally, the improvement in HOMA-β observed in 15 studies (2845 patients) suggests enhanced β-cell function. Based on GRADE assessment, the certainty of evidence was rated moderate for both outcomes due to heterogeneity. To the best of our knowledge, this study is the first meta-analysis comparing the influence of SGLT2 inhibitors with placebo on HOMA-IR and HOMA-β in patients with T2D. These results support the hypothesis that SGLT2 inhibitors are associated with beneficial changes in insulin resistance and β-cell function beyond their primary glucose-lowering mechanism. However, these effects may be partly mediated by indirect pathways, such as reduced glucotoxicity or weight loss, and do not imply direct causality.

The pharmacological and molecular mechanisms by which SGLT2 inhibitors may improve insulin resistance and β-cell function are multifaceted. One proposed mechanism is the reduction of glucotoxicity. Chronic hyperglycemia can impair insulin signaling pathways, contributing to insulin resistance[53] and reduced β-cell function[54]. By lowering blood glucose levels, SGLT2 inhibitors may reduce glucotoxicity, thereby improving cellular insulin sensitivity and β-cell function. A previous study in a mouse model of human adenosine triphosphate (ATP)-sensitive potassium channel-induced neonatal diabetes mellitus showed that SGLT2 inhibitor therapy protects glucotoxicity-induced β-cell failure through mitigation of oxidative and endoplasmic reticulum stress[55]. Additionally, the reduction in body weight often associated with SGLT2 inhibitor therapy can enhance insulin sensitivity[56], as obesity is a known contributor to insulin resistance. Another potential mechanism involves the improvement of β-cell function through the reduction of lipotoxicity. Elevated free fatty acids in the bloodstream can impair β-cell function and promote apoptosis[54,57]. SGLT2 inhibitors may reduce lipotoxicity by lowering triglyceride levels[58], thus preserving β-cell viability and function. Consistently, empagliflozin was shown to alleviate lipotoxicity and improve islet β-cell function by restoring key transcription factors and promoting β-cell proliferation in diabetic mice[59]. Moreover, SGLT2 inhibitors have been shown to reduce oxidative stress and inflammation[60], both of which are implicated in the deterioration of β-cell function and insulin resistance.

Our subgroup analyses provided further insights into the variability of the response to SGLT2 inhibitors. We observed more pronounced improvements in HOMA-β in non-Asian studies compared to Asian studies, indicating potential ethnic or regional differences in response to therapy. In this analysis, 83.7% of the patients in non-Asian countries were Caucasian, while 100% of the patients in Asian countries were Asian. Prior research has suggested that Asians generally have a lower β-cell mass and reduced insulin secretory capacity compared to Caucasians[61], potentially making them less responsive to interventions that enhance β-cell function. Additionally, genetic polymorphisms in key regulators of β-cell development and insulin secretion, such as TCF7L2, KCNJ11, and CDKAL1 may contribute to interethnic differences in β-cell adaptability to metabolic stress[62]. Moreover, lifestyle-related factors, including differences in dietary composition, physical activity, and obesity prevalence, may also influence insulin secretory response and treatment effects. Caucasian patients with T2D often exhibit a greater degree of obesity and insulin resistance, making β-cell compensation more critical and potentially amplifying the observed response in non-Asian populations[63]. It is important to emphasize that these explanations are hypothesis-driven and derived from population-level observations rather than patient-level data. Given that only two non-Asian studies (n = 2 datasets, 461 patients) were included, caution is warranted in interpreting these findings, and further research, particularly individual patient-level meta-analyses, is needed to validate these hypotheses.

Despite conducting multiple sensitivity, subgroup and meta-regression analyses, no single factor fully accounted for the heterogeneity. These findings may indicate that this heterogeneity could not be solely attributed to individual study characteristics, such as study country (Asian vs non-Asian), sample size, mean age of patients, proportion of men, baseline HbA1c, duration of T2D, concurrent antidiabetic treatments, dosage of SGLT2 inhibitors, or treatment duration. The heterogeneity may be mediated by multiple interrelated factors, including variations in patient populations, baseline disease characteristics, and study methodologies. In addition, although we followed Cochrane recommendations by splitting the control group in studies with multiple SGLT2 inhibitor doses, a formal dose-response analysis was not feasible due to heterogeneity across drug types and dosing regimens. Future meta-analyses based on individual patient data are needed to explore the possible influence of patient and study characteristics-such as obesity status, concurrent medications, and metabolic profiles-on the results, as well as the potential dose-dependent effects of SGLT2 inhibitors, which could not be fully assessed in the current analysis due to the use of aggregated study-level data.

The strengths of our study include the comprehensive and expanded nature of the literature search to avoid missing potentially relevant studies. A substantial number of RCTs were included, thus enhancing the robustness and generalizability of our findings. Additionally, the implementation of sensitivity and subgroup analyses strengthened the reliability of our results. However, our study also had some limitations. First, HOMA-IR and HOMA-β are indirect measures of insulin resistance and β-cell function, as they are derived from fasting glucose and insulin levels rather than dynamic metabolic responses. The gold-standard techniques for assessing these parameters include the hyperinsulinemic-euglycemic clamp for insulin resistance[64] and the glucose-stimulated insulin secretion test for β-cell function[65], which provide a more precise evaluation of insulin action and β-cell response. SGLT2 inhibitors lower fasting glucose primarily through urinary glucose excretion, which may artificially lower HOMA-IR and elevate HOMA-β, potentially overestimating improvements in insulin resistance and β-cell function. This suggests that the observed changes may reflect altered glucose availability rather than intrinsic metabolic improvements. Gold-standard methods such as the hyperinsulinemic-euglycemic clamp and glucose-stimulated insulin secretion tests should be considered in future studies for more accurate assessments. However, a previous study showed that HOMA-IR was well correlated with the results of the hyperinsulinemic-euglycemic clamp during SGLT2 inhibitor treatment[66], suggesting that HOMA-IR may reasonably reflect insulin sensitivity changes in this context. Nonetheless, HOMA metrics remain limited by their reliance on fasting parameters and do not capture dynamic insulin responsiveness. Future studies should incorporate clamp-based or glucose-stimulated insulin secretion testing alongside HOMA to validate and contextualize effect sizes in clinical practice. Second, another limitation of our meta-analysis is the relatively short treatment duration in the included RCTs. While our findings suggest that SGLT2 inhibitors improve insulin resistance and β-cell function in the short term (12-52 weeks), the long-term durability of these effects is unknown. This represents a key research gap. Future large-scale RCTs with extended follow-up are needed to determine whether these improvements are sustained and whether they translate into long-term clinical benefits such as delayed disease progression or complication reduction. Third, while we conducted extensive subgroup and meta-regression analyses, significant heterogeneity remained to exist. Fourth, we included only English-language publications, which may introduce language bias and limit the generalizability of our findings to some extent. In addition, studies of patient groups using agents that increase insulin release or sensitivity such as sulfonylureas, metformin, and DPP4 inhibitors were also included in the meta-analysis, which may confound the results of the meta-analysis. However, the subgroup analysis limited to studies of patient group without concurrent antidiabetic treatments showed similar results to the overall meta-analysis. Moreover, it is important to compare the effects of SGLT2 inhibitors on HOMA-IR and HOMA-β with other antidiabetic treatments. However, our study excluded trials with active comparators (e.g., metformin, sulfonylureas) to isolate the effect of SGLT2 inhibitors, which may limit the comparability with other therapies. Future head-to-head studies are needed to evaluate the relative metabolic benefits across different drug classes, which may provide more clinically relevant guidance for treatment decisions. Lastly, as for the publication bias, visual assessment of funnel plots is subjective, and small-study effects cannot be entirely ruled out. Therefore, despite formal tests indicating low risk, the potential influence of publication bias remains as a limitation.

From a clinical perspective, the ability of SGLT2 inhibitors to improve insulin resistance and β-cell function suggests that these agents can be valuable components of T2D management, potentially slowing disease progression and preserving pancreatic function. The observed benefits in insulin resistance and β-cell function may be the alternative mechanisms underlying the long-term glycemic control and reduced risk of diabetes-related complications after treatment with SGLT2 inhibitors. Future research should focus on elucidating the long-term effects of SGLT2 inhibitors on insulin resistance and β-cell function. Large-scale, long-duration RCTs are needed to confirm and extend our findings. Furthermore, studies should explore the molecular mechanisms underlying the beneficial effects of SGLT2 inhibitors, including their impact on glucotoxicity, lipotoxicity, oxidative stress, and inflammation. In particular, mechanistic research is needed to distinguish whether the observed improvements in β-cell function and insulin resistance arise from direct pharmacologic actions or are secondary to metabolic changes such as weight loss or reduced glucose toxicity.

CONCLUSION

This systematic review and meta-analysis suggest that SGLT2 inhibitors are associated with improvements in insulin resistance and β-cell function, although these findings should be interpreted with caution due to moderate certainty of evidence and observed heterogeneity. These findings suggest the potential of SGLT2 inhibitors to modify disease progression and to improve clinical outcomes. Further research is needed to fully elucidate the specific mechanisms of action and long-term benefits of these agents. As the prevalence of T2D continues to rise globally, the role of SGLT2 inhibitors in comprehensive diabetes management is likely to become increasingly important.

ACKNOWLEDGEMENTS

Administrative assistance was provided by Qi L of MSD China Holding Co., Ltd., Shanghai, China.