Beyond glycemic control: differential effects of empagliflozin and sitagliptin on insulin sensitivity and a shared increase in adropin in type 2 diabetes Authors: Otbah Salim Taha, Golnaz Azami, Lotfollah Saed, Mohammad Saeed Hakhamaneshi, Heresh Moridi Published: 2026 ABSTRACT AbstractBackgroundType 2 diabetes mellitus (T2DM) is characterized by metabolic and inflammatory disturbances beyond hyperglycemia. Hepatokines such as adropin have emerged as regulators of insulin resistance and vascular function, yet the comparative metabolic effects of sodium–glucose cotransporter 2 (SGLT2) and dipeptidyl peptidase 4 (DPP-4) inhibitors on adropin remain unclear. This study compared the effects of empagliflozin versus sitagliptin, each added to metformin, on serum adropin, insulin resistance, glycemic control, lipid profile, and inflammation in adults with T2DM.MethodIn this single-center, randomized, open-label, parallel-group superiority trial with blinded outcome assessment and blinded statistical analysis, 100 adults with inadequately controlled type 2 diabetes mellitus (HbA1c ≥ 7.5%) receiving stable metformin therapy were allocated in a 1:1 ratio to receive empagliflozin 10 mg once daily or sitagliptin 100 mg once daily for a 12-week intervention period. The co-primary outcomes were changes in circulating adropin concentrations and insulin resistance, assessed by the homeostasis model assessment of insulin resistance (HOMA-IR). Secondary outcomes included changes in HbA1c, fasting insulin, lipid parameters, body weight, and tumor necrosis factor-α (TNF-α). All analyses were conducted according to the intention-to-treat principle.ResultNinety‑four participants completed the intervention; all were included in analyses. Serum adropin increased from 291 ± 133 to 362 ± 124 pg/mL with empagliflozin and from 299 ± 98 to 358 ± 126 pg/mL with sitagliptin (time effect: F = 19.67, p < .001). HOMA‑IR declined from 10.08 ± 4.44 to 6.65 ± 2.50 with empagliflozin and from 9.28 ± 2.97 to 7.15 ± 1.80 with sitagliptin (interaction: F = 4.85, p = .032, partial η² = 0.09). HbA1c decreased from 8.10 ± 0.53 to 7.04 ± 1.18 with empagliflozin and from 8.29 ± 0.66 to 7.62 ± 0.68 with sitagliptin (interaction: F = 4.30, p = .043, η² = 0.081). TNF‑α fell from 44.12 ± 21.52 to 30.78 ± 15.93 in empagliflozin and from 43.93 ± 21.33 to 37.65 ± 17.96 in sitagliptin (time effect: F = 37.09, p < .001). Empagliflozin produced greater reductions in fasting insulin (− 3.43 ± 2.34 µIU/mL vs. − 2.13 ± 1.17 µIU/mL, interaction p = .036), triglycerides (− 42.5 ± 23.9 mg/dL vs. − 21.9 ± 14.5 mg/dL, interaction p = .005), and a larger HDL‑C increase (+ 5.8 ± 3.1 mg/dL vs. + 3.3 ± 2.5 mg/dL, interaction p = .017). Body weight and BMI decreased similarly in both groups (time effect p < .001, no interaction). No serious adverse events occurred.ConclusionBoth empagliflozin and sitagliptin improved metabolic and inflammatory markers and were associated with comparable increases in circulating adropin. Empagliflozin conferred broader metabolic benefits, particularly in insulin resistance, glycemic control, and lipid profile. The parallel rise in adropin across treatment groups highlights its potential role as a treatment-responsive biomarker rather than a drug-specific effect.Trial registrationThis trial was prospectively registered with the Iranian Registry of Clinical Trials (IRCT ID: IRCT20160625028627N8) on May 27, 2025 (Trial ID: 83720). The complete trial record is accessible at https://irct.behdasht.gov.ir/user/trial/83720/view.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12902-026-02320-z.Keywords: Diabetes mellitus, Type 2 / drug therapy, Empagliflozin / therapeutic use, Sitagliptin Phosphate / therapeutic use, Insulin Resistance / drug effects, Adropin / blood Introduction T2DM is a leading global metabolic disorder characterized by chronic hyperglycemia resulting from insulin resistance and progressive β-cell dysfunction. It affects more than 422 million people worldwide and accounts for roughly 1.5 million deaths annually, largely from vascular complications of poor glycemic control [1]. The accelerating trends of urbanization, sedentary lifestyles, population aging, and obesity have markedly increased its prevalence across both developed and developing regions [2]. In Iran, the burden of T2DM is particularly concerning. The prevalence among adults has reached approximately 10.8%, and projections indicate that the number of affected individuals will rise from 4.5 million in 2011 to over 9 million by 2030 [3]. Moreover, the disease tends to appear at a younger age—typically between 40 and 50 years—compared with more than 65 years in many high-income countries [4]. Early-onset diabetes is associated with a longer disease duration, accelerated development of complications, and higher mortality, thereby contributing to substantial diabetes-related disability-adjusted life years [5]. These demographic patterns highlight the urgent need for context-specific metabolic research in Iran, where distinct genetic, dietary, and environmental factors may influence both disease pathogenesis and therapeutic response. At the mechanistic level, T2DM arises from two interdependent abnormalities: impaired pancreatic β-cell insulin secretion and peripheral insulin resistance in skeletal muscle, adipose tissue, and liver [6]. Insulin resistance disrupts glucose and lipid metabolism, leading to persistent hyperglycemia, hyperinsulinemia, and atherogenic dyslipidemia, which in turn promote cardiovascular, renal, and neurological complications [7]. Lifestyle modification remains the cornerstone of T2DM management; however, in most patients, sustained glycemic control cannot be achieved by lifestyle measures alone. Metformin, introduced more than half a century ago, has long been the preferred first-line pharmacologic therapy due to its robust efficacy, low cost, favorable safety profile, and evidence for cardiovascular benefit [8]. Landmark studies established its ability to lower blood glucose effectively with minimal risk of hypoglycemia, while later work suggested additional organ-protective properties beyond glycemic regulation [9]. Nevertheless, as the global burden of T2DM continues to rise, therapeutic algorithms have evolved. Growing recognition of diabetes as a multisystem metabolic and cardiovascular disorder has exposed the limitations of a “one-size-fits-all” approach centered on metformin alone. Over time, many patients experience “secondary failure,” with progressive β-cell decline and persistent insulin resistance despite continued metformin use. Moreover, contemporary guidelines now advocate individualized therapy—prioritizing agents that not only improve glycemia but also confer cardiovascular and renal protection in those at higher risk [10]. These developments have propelled newer drug classes such as SGLT2 inhibitors and DPP-4 inhibitors into prominence as essential add-on therapies to metformin. SGLT2 inhibitors, exemplified by empagliflozin, lower plasma glucose by inhibiting renal glucose reabsorption in the proximal tubules, thereby promoting glycosuria independently of insulin secretion [11]. Their benefits extend well beyond glycemic control: multiple cardiovascular outcome trials have shown reductions in hospitalization for heart failure and slowing of renal disease progression [12, 13]. They also modestly reduce body weight and blood pressure—effects particularly valuable in insulin-resistant, overweight individuals. In contrast, DPP-4 inhibitors such as sitagliptin improve glucose homeostasis by preventing enzymatic degradation of incretin hormones—glucagon-like peptide-1 (GLP-1) and glucose-dependent insulinotropic peptide (GIP)—thereby enhancing glucose-dependent insulin secretion and suppressing glucagon release [14]. These agents are weight-neutral, well tolerated, and carry minimal hypoglycemia risk, making them suitable for a broad range of patients [15]. Although their cardiovascular benefits are generally less striking than those of SGLT2 inhibitors, they remain a key therapeutic option, especially where renal impairment or intolerance to SGLT2 inhibitors limits treatment choices [16]. Beyond their classical mechanisms, growing interest has emerged in exploring whether antidiabetic agents exert pleiotropic effects on novel endocrine and inflammatory pathways. In this context, the liver is increasingly recognized as an endocrine organ that secretes hepatokines—proteins regulating systemic metabolism and insulin sensitivity [17]. Among these, adropin, a 76-amino-acid peptide encoded by the ENHO gene, has attracted attention as a promising metabolic regulator [18]. Adropin plays a crucial role in energy homeostasis, lipid handling, and endothelial function, and its circulating levels are inversely associated with obesity, insulin resistance, and dyslipidemia [19–22]. Experimental data further indicate that adropin improves insulin signaling via AMPK and PI3K/Akt activation and exerts anti-inflammatory effects through modulation of eNOS and ERK pathways [23]. Despite this growing evidence base, clinical data on how established glucose-lowering agents influence adropin remain scarce. Understanding whether SGLT2 or DPP-4 inhibitors differentially modulate adropin levels—and how these changes relate to insulin resistance, inflammation, and cardiometabolic risk—could illuminate novel mechanistic pathways and identify potential biomarkers of therapeutic efficacy. Therefore, the present randomized controlled trial was designed to test the hypothesis that empagliflozin and sitagliptin, when used as add-on therapy to metformin, exert differential metabolic effects in patients with type 2 diabetes. Specifically, we sought to determine whether empagliflozin confers greater improvements in insulin resistance and cardiometabolic risk factors compared with sitagliptin, while exploring whether adropin regulation represents a shared or drug-specific hepatokine response to these therapies. To address this objective, we compared changes in serum adropin and TNF-α, alongside established measures of glycemic control, lipid profile, body weight, BMI, and insulin resistance. Considering the earlier disease onset and distinct metabolic profile of Iranian patients, this comparative approach aims to provide clinically and regionally relevant insights into the metabolic and mechanistic effects of these widely used glucose-lowering agents. Materials and methods Study setting This investigation took place at an urban endocrinology outpatient unit affiliated with a teaching hospital in Sanandaj, Iran. The facility provides both primary and secondary endocrine care and includes a dedicated diabetes clinic that operates six days per week (closed on Fridays). On average, the clinic accommodates 20–30 patients daily, all managed through a multidisciplinary healthcare team comprising endocrinologists, nurses, and nutrition specialists. Notably, this center represents the only specialized diabetes care service available to individuals with T2DM in the Sanandaj region. Study design and participants This was a single-center, randomized, open-label, parallel-group superiority trial with blinded outcome assessment and blinded statistical analysis. Participants were randomly assigned in a 1:1 ratio to receive either empagliflozin or sitagliptin as an adjunct to their ongoing metformin therapy. Eligible participants were Iranian adults aged 18 years or older who had a confirmed diagnosis of T2DM for a duration of at least six months. Inclusion also required a BMI of at least 25 kg/m², ongoing metformin monotherapy maintained for a minimum of three months, and aHbA1c level of 7.5% or higher. No eligibility restrictions were applied based on prior healthcare utilization or visit frequency. Eligible patients were expected to demonstrate consistent follow-up—defined as attending at least two clinic visits per year—and to have remained free from any acute illness within the preceding six months. Individuals were excluded if they presented with diabetes-related complications that could interfere with study participation or clinical care, cognitive impairment, or uncontrolled hypertension (systolic ≥ 180 mmHg or diastolic ≥ 110 mmHg). Additional exclusion criteria encompassed hemolytic anemia; significant hepatic, renal, or cardiac disease; malignancy; clinically evident non-alcoholic fatty liver disease (NAFLD), defined by a prior documented diagnosis and/or hepatic steatosis on abdominal ultrasonography in the absence of significant alcohol consumption, or endocrine abnormalities such as polycystic ovary syndrome. Women who were pregnant or breastfeeding were not eligible. Patients using insulin, injectable antidiabetic agents, antioxidant, anti-inflammatory, or weight-reducing medications were also excluded, as were those with active daily smoking (defined as regular cigarette smoking within the preceding 6 months) or substance use, poor medication adherence, planned surgical procedures, or concurrent involvement in another clinical trial. Furthermore, individuals with an estimated glomerular filtration rate (eGFR) below 50 mL/min/1.73 m² were excluded from participation. Sample size and sampling procedure The required sample size was determined using G*Power software (version 3.1.9.4), applying the repeated-measures ANOVA framework with a within–between interaction design. Calculations were based on a two-tailed significance level (α = 0.05) and a statistical power of 95%. Because the trial specified two co-primary outcomes (serum adropin and HOMA-IR), separate a priori sample size estimations were performed for each of these endpoints using effect size assumptions derived from previously published interventional and observational studies in patients with type 2 diabetes. For HOMA-IR, the calculation assumed a moderate between-group effect size (Cohen’s f = 0.25), corresponding to clinically meaningful changes in insulin resistance reported in prior SGLT2 inhibitor trials, with a within-subject correlation of 0.5 across repeated measurements. This yielded a minimum required sample of 72 participants. For serum adropin, a slightly larger effect size assumption (Cohen’s f = 0.28) was applied based on reported variability and responsiveness of circulating adropin concentrations in metabolic intervention studies, resulting in a required sample of 78 participants. To ensure adequate power across both co-primary outcomes, the larger of the calculated sample sizes was selected as the minimum required sample prior to adjustment for attrition. To mitigate risk of under-powering due to participant attrition, a prospective dropout rate of approximately 30% was conservatively assumed during design, reflecting the real-world context of an open-label outpatient trial involving multimorbid adults with type 2 diabetes, frequent follow-up visits, and potential intolerance or safety-related discontinuation of oral hypoglycemic agents. This value was chosen to ensure sufficient power under worst-case retention scenarios typical of community-based studies in the region rather than controlled inpatient trials. After attrition adjustment, a total sample size of 100 participants (50 per group: empagliflozin + metformin vs. sitagliptin + metformin) was prespecified. No additional 5% correction was applied beyond the 30% conservative estimate; prior text referencing a combined 35% attrition reflected a typographical inconsistency and has been corrected. Actual trial attrition was substantially lower than anticipated (6%), resulting in a final analyzed sample that exceeded the minimum required for adequate statistical power. This lower-than-expected dropout validated the robustness of the original conservative assumption and enhanced the precision of treatment effect estimates without altering the prespecified analytical framework. No interim analyses, adaptive re-estimation, or stopping rules were planned or conducted. Recruitment, screening, and randomization After obtaining approval from the Research Ethics Committee and Research Council of Kurdistan University of Medical Sciences and registering the trial in the Iranian Registry of Clinical Trials (IRCT), recruitment took place at the Diabetes Specialty Center, Tohid Hospital (Sanandaj, Iran). Enrollment followed a two-stage process. First, an informational notice describing the study was posted on the clinic bulletin board, inviting patients with type 2 diabetes to express voluntary interest. Preliminary demographic and contact details were collected, forming an initial list of prospective participants. In the second stage, medical charts of these volunteers were reviewed by the research team to verify eligibility according to predefined inclusion and exclusion criteria. Individuals meeting all requirements were subsequently invited for an on-site orientation, during which the study objectives, procedures, and potential risks were fully explained. Written informed consent was obtained prior to performing baseline clinical evaluations and randomization. Participants were randomly assigned in a 1:1 allocation ratio to receive either empagliflozin + metformin or sitagliptin + metformin. The randomization sequence was generated by an independent statistician using R software, employing a stratified permuted-block approach with randomly varying block sizes of 4 and 6 to maintain balance between groups. Stratification variables included age (18–44, 45–60, 61–65, > 65 years), sex, duration of diabetes (< 5 vs. ≥ 5 years), and body-mass index (25.0–29.9, 30.0–34.9, ≥ 35.0 kg/m²). To preserve allocation concealment, the sequence was embedded within sequentially numbered, opaque, sealed envelopes prepared by the statistician. Envelopes were constructed from thick, non-transparent paper, sealed with tamper-evident adhesive, and signed across the seal. They were stored in a locked cabinet under the supervision of an independent study coordinator and were opened sequentially and irreversibly only after confirmation of participant eligibility and completion of baseline assessments. Site investigators oversaw recruitment, while an independent study coordinator—not involved in intervention delivery, clinical assessments, or data analysis—implemented group assignment. Because the two medications required distinct dosing regimens, blinding of participants and treating clinicians was not feasible; therefore, the trial was conducted as an open-label, randomized, parallel-group study with blinded outcome assessment. Nonetheless, outcome assessors, data collectors, and statisticians remained blinded throughout the study to minimize assessment and analytical bias. Blinding was maintained by separating clinical care from outcome evaluation and statistical analysis; assessors and analysts had no access to treatment allocation lists, and all study data were labeled using anonymized, group-neutral codes (e.g., Group A and Group B). To safeguard blinding integrity, outcome assessors and statisticians were not involved in participant recruitment, intervention delivery, or follow-up scheduling, and were instructed not to inquire about treatment assignment. Any accidental unblinding events were to be documented and reported; no such events occurred during the study. Group identities were replaced with coded identifiers, and access to allocation data was strictly restricted to the independent study coordinator. Emergency unblinding was authorized solely in the case of a serious adverse event, in accordance with predefined safety protocols. Given the objective nature of the primary outcomes (biochemical and laboratory-based measures), formal post-hoc assessment of blinding success (e.g., blinding indices) was not performed. Patient and public involvement This randomized controlled trial was conducted in a single primary and secondary outpatient diabetes center with a high patient load and an experienced multidisciplinary team. All interventions were supervised by board-certified endocrinologists and implemented by trained clinical staff, each of whom received protocol-specific instruction before participant enrollment began. Because this was a pharmacological study requiring specialist supervision and regulatory oversight, neither patients nor members of the public were involved in the design, conduct, analysis, or dissemination of the research. Participants were involved only as volunteers after providing written informed consent. In accordance with ethical transparency, all participants will receive a plain-language summary of the study results upon trial completion. Intervention and comparator Group A – empagliflozin plus metformin Participants in this group received empagliflozin 10 mg once daily (oral), in addition to metformin. All participants were required to be on a pre-existing, ongoing metformin monotherapy regimen for at least 3 months prior to enrollment. Metformin was not newly initiated in any participant. For participants who were not yet receiving their maximally tolerated metformin dose at screening, a standardized pre-randomization dose optimization protocol was applied, starting at their current dose and incrementally increased by 500 mg/week, as tolerated, up to a maximum of 2,000 mg/day. This optimization phase was completed before or at the time of randomization, with the explicit goal of achieving a stable and fixed metformin dose prior to the 12-week efficacy follow-up period. This pre-randomization metformin optimization was explicitly designed to eliminate post-randomization dose variation and thereby prevent metformin titration from confounding treatment effects attributed to the randomized add-on therapies. No protocol-driven metformin dose adjustments were permitted during the post-randomization follow-up period, unless required for safety reasons. Renal function was evaluated before treatment initiation and re-assessed periodically during the 12-week intervention period. Consistent with international safety guidelines, metformin was discontinued if eGFR declined below 30 mL/min/1.73 m². Group B – sitagliptin plus metformin Participants assigned to the comparator arm received sitagliptin 100 mg once daily in combination with metformin. Eligibility required a pre-existing, stable metformin regimen for ≥ 3 months prior to enrollment. Metformin dose optimization, when required, followed the identical pre-randomization protocol used in the empagliflozin group and was completed before the start of the 12-week intervention period. No participant underwent metformin dose escalation during follow-up, except for safety-mandated discontinuation. Renal monitoring followed identical procedures in both groups. This identical pre-randomization metformin stabilization strategy ensured comparable and fixed background therapy across treatment arms, minimizing the risk that subsequent metabolic changes could be attributed to differential metformin exposure. All medications were administered orally, and the use of other glucose-lowering drugs, anti-inflammatory agents, or weight-reducing supplements was strictly prohibited during the trial. Participants were instructed to maintain their habitual diet and physical activity patterns, and no rescue medication was provided. Participants with persistent severe hyperglycemia (fasting plasma glucose > 300 mg/dL or HbA1c > 10%) were to be withdrawn from the study and referred for standard clinical management to ensure safety. Treatment adherence was monitored through scheduled clinic visits, pill counts, and patient self-reports. To preserve protocol fidelity, all clinical staff completed standardized training before the study began. Concomitant non-diabetic medications (e.g., antihypertensives, statins) were permitted provided that their dosages remained unchanged throughout the study. Participants were followed for 12 weeks, with outcome assessments conducted at baseline and post-intervention. All trial procedures and monitoring frameworks were predefined and documented in the approved study protocol, which is publicly available in the Iranian Registry of Clinical Trials (IRCT20160625028627N8) and upon reasonable request from the corresponding author. Study outcomes The primary outcomes were changes in serum adropin concentration and insulin resistance, evaluated using the HOMA-IR from baseline to week 12. These outcomes were defined a priori as two co-primary endpoints addressing distinct and non-overlapping scientific objectives: adropin as an exploratory mechanistic biomarker of hepatokine regulation, and HOMA-IR as a clinically relevant metabolic endpoint reflecting insulin sensitivity. Importantly, these outcomes were not intended to jointly test a single confirmatory hypothesis or to support a composite efficacy claim, but rather to address complementary mechanistic and clinical questions within the same randomized framework. Because the two co-primary outcomes were biologically distinct, prespecified, and interpreted independently rather than jointly, and because neither outcome was contingent on the significance of the other, no formal multiplicity adjustment or hierarchical testing procedure was applied for the co-primary endpoints, consistent with methodological guidance for trials addressing parallel mechanistic and clinical objectives. HOMA-IR was selected as a validated, widely used surrogate index of insulin sensitivity appropriate for randomized clinical trials in type 2 diabetes, given its feasibility in outpatient settings and its capacity to detect between-group differences when applied longitudinally under standardized conditions. However, because HOMA-IR relies on fasting insulin concentrations, it is inherently sensitive to analytical variability in insulin assays. For this reason, HOMA-IR findings were interpreted as reflective of relative changes in insulin sensitivity rather than as definitive measures of insulin action. These variables were also selected to examine the effects of empagliflozin and sitagliptin on hepatokine regulation and insulin sensitivity, extending beyond their glucose-lowering actions. The secondary outcomes included changes in TNF-α, fasting blood glucose (FBG), HbA1c, fasting insulin, triglycerides (TG), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), body weight, and body-mass index (BMI) over the same 12-week period. These secondary outcomes were prespecified as exploratory and hypothesis-generating, intended to contextualize the primary findings rather than to support confirmatory inference. Accordingly, no formal false discovery rate (FDR) control was applied, and results were interpreted with appropriate caution. All outcomes were assessed at baseline (pre-randomization) and again after 12 weeks of treatment. Anthropometric assessments Anthropometric data, including height, weight, and BMI, were recorded by trained staff using a calibrated digital scale (BT-ETS002) under standardized conditions. Participants were measured without shoes or heavy clothing, and BMI was computed as weight (kg) / height (m²), categorized according to World Health Organization (WHO) criteria. Blood sampling and laboratory analyses Fasting venous blood samples (10 mL) were obtained at both assessment points between 8:00 and 9:00 a.m. following a 10–12-hour overnight fast, prior to morning medication intake. An additional 2 mL of blood was collected in K₂EDTA tubes for HbA1c measurement. All collections were performed by certified phlebotomists using sterile venipuncture techniques. Samples were allowed to clot at room temperature for 30 min and centrifuged at 3,000 × g for 10 min at 4 °C. The separated serum was aliquoted and stored at − 80 °C until analysis. Routine biochemical variables—including FBG, HbA1c, and the lipid profile—were determined using standard enzymatic colorimetric assays. Plasma glucose concentrations were measured and recorded consistently in milligrams per deciliter (mg/dL) across all participants and time points, in accordance with standard local laboratory practice. Serum insulin, adropin, and TNF-α concentrations were quantified via enzyme-linked immunosorbent assay (ELISA) using commercially validated kits (Zellbio GmbH, Germany). The adropin assay demonstrated a sensitivity of 0.1 ng/mL, with intra- and inter-assay coefficients of variation below 10%, ensuring analytical precision. All biochemical analyses were performed in the central clinical biochemistry laboratory of Tohid Hospital using automated, calibrated analyzers under stringent internal quality-control procedures. HOMA-IR was calculated using glucose values expressed in mg/dL as follows: If glucose were expressed in mmol/L, the appropriate constant (22.5) would apply; however, since all analyses were conducted using mg/dL units, the divisor 405 was used consistently throughout the study. The selection of primary and secondary endpoints was driven by their biological relevance, clinical interpretability, and feasibility within a randomized outpatient trial, rather than by the absence of a formal core outcome set. Serum adropin was selected as a primary outcome based on emerging evidence implicating this hepatokine in systemic energy homeostasis, insulin sensitivity, and inflammatory regulation, positioning it as a mechanistically informative biomarker to explore potential drug class–specific effects beyond glycemic control. HOMA-IR was selected as a complementary co-primary endpoint due to its established use as a validated surrogate measure of insulin resistance in type 2 diabetes trials, allowing clinically meaningful assessment of longitudinal changes in insulin sensitivity under standardized conditions. Secondary metabolic and inflammatory markers were prespecified to contextualize the primary outcomes and to capture broader cardiometabolic effects commonly evaluated in diabetes research, rather than to serve as confirmatory efficacy endpoints. Protocol amendments Two amendments were made to the original study protocol after trial initiation. First, tumor necrosis factor-alpha (TNF-α) was added as a secondary outcome during the early recruitment phase, prior to database lock and before any interim or final data analyses were performed. This amendment was introduced to broaden the evaluation of inflammatory pathways relevant to insulin resistance, in light of accumulating external evidence identifying TNF-α as a key biomarker in the pathophysiology of T2DM. Importantly, this modification did not alter the prespecified primary outcomes, study hypotheses, or statistical analysis plan. The amended protocol was formally approved by the Research Ethics Committee and was submitted for update in the trial registry to ensure transparency and alignment with the final study conduct. Second, the planned sample size was increased from 74 to 100 participants during the recruitment phase, to ensure adequate statistical power and to offset potential participant attrition. This modification was implemented before completion of enrollment and obtained formal approval from the Research Ethics Committee as part of an updated protocol submission. Harms Adverse events (AEs) were defined as any unfavorable or unintended medical occurrence following the administration of a study intervention, while serious adverse events (SAEs) referred to outcomes resulting in death, life-threatening conditions, hospitalization or its prolongation, persistent disability, or requiring medical or surgical intervention. Potential harms were monitored systematically at baseline, week 12, and interim study visits through clinical assessments, laboratory investigations, and participant self-reports. All events were documented in case report forms (CRFs) and subsequently reviewed by the site endocrinologist, who remained blinded to treatment allocation. Each event’s severity was classified using standardized clinical grading criteria, and causality was evaluated by the endocrinologist in accordance with the study protocol and established pharmacovigilance guidelines. Statistical methods All statistical analyses were performed using IBM SPSS Statistics software (version 26; IBM Corp., Armonk, NY, USA). Statistical significance was defined as two-tailed p < .05. Continuous variables are reported as mean ± standard deviation (SD), whereas categorical variables are presented as frequencies and percentages. Analyses were conducted in two sequential phases. First, baseline characteristics were compared between treatment groups. Depending on data distribution, Student’s t-test or the Mann–Whitney U test was used for continuous variables, and the Chi-square or Fisher’s exact test for categorical variables. Normality of continuous variables at baseline was formally assessed using the Shapiro–Wilk test, and homogeneity of variances was evaluated using Levene’s test. The primary efficacy analysis adhered to the intention-to-treat principle and employed a mixed-design repeated-measures analysis of variance (ANOVA), with time (baseline and 12 weeks) as the within-subject factor and treatment group as the between-subject factor. Normality was evaluated at two levels: (1) distribution of raw outcome variables at each time point using the Shapiro–Wilk test, and (2) distribution of studentized residuals from the repeated-measures ANOVA models. Several outcome variables demonstrated statistically significant deviations from normality at the raw data level; however, residual diagnostics indicated approximate normality, symmetry, and absence of influential outliers across models. Repeated-measures ANOVA was retained based on its documented robustness to moderate violations of normality in balanced designs with comparable group sizes and continuous outcomes, particularly when the primary inferential focus is on interaction effects rather than individual cell means. No outcome exhibited extreme skewness or kurtosis requiring transformation, and nonparametric alternatives for mixed designs were deemed less appropriate for modeling group × time interactions with effect size estimation. Sphericity was formally tested using Mauchly’s test, and Greenhouse–Geisser corrections were applied when violations were detected. Effect sizes were expressed as partial eta-squared (η²), with 95% confidence intervals reported for primary and secondary outcomes to enhance interpretability and clinical relevance. Missing data were limited in extent and occurred sporadically across follow-up measurements, primarily due to missed visits or incomplete laboratory sampling. The overall proportion of missing values across repeated-measures outcomes was < 10%, with no outcome exhibiting > 15% missingness at any time point. Patterns of missingness were examined descriptively and did not suggest differential dropout between treatment groups. Under the assumption that data were missing at random (MAR), missing values were handled using the expectation–maximization (EM) algorithm implemented in IBM SPSS Statistics (Analyze → Missing Value Analysis → EM). EM estimation was applied at the variable level for repeated-measures outcomes included in the primary and secondary ANOVA models, using all available observed data without auxiliary variables. This approach yields maximum-likelihood estimates under MAR assumptions and was selected to preserve statistical power and reduce bias relative to complete-case analysis. The primary interpretive focus was placed on the group × time interaction, representing the differential treatment-related changes over the 12-week study period. Main effects of time were interpreted as within-group treatment responses, whereas main effects of group were considered secondary and descriptive. To explore potential mechanistic relationships between hepatokine regulation and metabolic or inflammatory changes, prespecified exploratory Spearman correlation analyses were conducted between serum adropin concentrations and key biomarkers, including HOMA-IR, TNF-α, triglycerides, and HDL-C, at week 12. These correlation analyses were defined a priori in the statistical analysis plan as hypothesis-generating procedures intended to provide pathophysiological insight rather than to support confirmatory inference. Accordingly, results were interpreted cautiously and not used to modify or reinforce conclusions drawn from the primary or secondary outcome analyses. Spearman’s rank correlation was selected due to non-normal distributions of several variables and the exploratory nature of the analysis. No formal adjustment for multiplicity was applied to the two co-primary outcomes. This approach was prespecified and reflects the distinct conceptual roles of the endpoints: HOMA-IR served as a hypothesis-testing metabolic outcome, whereas adropin was evaluated as a mechanistic biomarker to characterize treatment-associated hepatokine responses. Accordingly, conclusions regarding adropin were restricted to equivalence of response rather than treatment superiority. Secondary outcomes were analyzed descriptively with appropriate control of Type I error through cautious interpretation rather than formal correction. Results Participants flow A total of 142 patients with T2DM were screened and invited to participate in the study. Of these, 27 patients did not meet the inclusion and exclusion criteria (15 due to uncontrolled hypertension, 8 with abnormal renal function, and 4 for concurrent participation in another clinical trial). In addition, 15 patients declined to participate—mainly because of time constraints (n = 9), fear of potential side effects (n = 4), or personal reasons (n = 2). Among the remaining 100 eligible participants, all provided written informed consent and were randomly assigned in a 1:1 ratio to one of two treatment arms: empagliflozin + metformin (n = 50; empagliflozin 10 mg/day) or sitagliptin + metformin (n = 50; sitagliptin 100 mg/day). During the 12-week follow-up period, six participants (6.0%) discontinued the study. In the sitagliptin group, four participants withdrew— one due to adverse events, and three following dose modification. In the empagliflozin group, two participants withdrew—one due to dose modification and one because of adverse events. Consequently, 94 participants completed the study (48 in the empagliflozin group and 46 in the sitagliptin group). All randomized participants were included in the intention-to-treat (ITT) analysis. A CONSORT-compliant participant flow diagram depicting the screening, randomization, follow-up, and analysis phases is shown in Fig. 1. The CONSORT 2025 checklist, outlining the essential items for transparent reporting of randomized controlled trials, is provided in Supplementary File 1. CONSORT 2025 flow diagram. Citation: Hopewell S, Chan AW, Collins GS, Hróbjartsson A, Moher D, Schulz KF, et al. CONSORT 2025 Statement: updated guideline for reporting randomised trials. BMJ. 2025; 388:e081123. https://dx.doi.org/10.1136/bmj-2024-081123 © 2025 Hopewell et al. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited Recruitment and follow up Participant enrollment took place between July and September 2025. Each participant was prospectively followed for 12 weeks after randomization. No interim analyses were performed, and the trial proceeded to completion without early termination. Adherence and intervention delivery All participants in both treatment groups received their allocated interventions as per protocol. Based on pill counts and participant self-reports, adherence exceeded 90% in both study arms. No treatment crossovers or protocol deviations were observed. Adverse event Adverse events were reported in a limited number of participants during the 12-week intervention period. In the empagliflozin group, one participant (2.0%) experienced mild polyuria and dehydration, which led to treatment discontinuation. In the sitagliptin group, one participant (2.0%) developed gastrointestinal discomfort, also resulting in early withdrawal from the study. All other reported symptoms were transient, mild in nature, and did not require interruption of therapy. No serious adverse events, hospitalizations, or deaths occurred throughout the study. Overall, both treatment regimens were well tolerated, with no clinically significant safety concerns identified. Sociodemographic, clinical, and laboratory characteristics of the respondents The study included 100 participants with a median age of 54 years (IQR: 14), of whom the majority were female (57%). The median duration of diabetes was 6 years (IQR: 4), indicating a relatively long-standing disease history. Participants were prescribed a median of five medications (IQR: 4) daily, and the median daily metformin dose was 1882 mg (IQR: 1000), consistent with standard therapeutic regimens for glycemic control in type 2 diabetes. Regarding lifestyle characteristics, one-quarter of the cohort (25%) reported a history of smoking or occasional non-daily smoking, while three-quarters (75%) reported no history of smoking. No participant met the study definition of active daily smoking at enrollment, consistent with the exclusion criteria. In terms of comorbidities, hypertension and hyperlipidemia were the most prevalent chronic conditions. Specifically, 41% of participants presented with both disorders concurrently, 22% had hypertension alone, and 16% had isolated hyperlipidemia. Only 14% had no recorded comorbidity, while a minority reported asthma (4%), chronic obstructive airway disease (2%), or gout (1%). These data reflect a high burden of cardiometabolic multimorbidity typical of middle-aged diabetic populations. The pattern of concomitant pharmacological therapy corresponded well with the observed disease profile. Nearly half of the participants (42%) were on a combination of antihypertensive, anticoagulant, and lipid-lowering agents, while 23% were prescribed antihypertensive therapy alone, and 16% were on lipid-lowering agents only. Smaller proportions used antihypertensive plus anticoagulant agents (14%), anticoagulant agents alone (3%), or lipid plus anticoagulant agents (2%). The relatively high proportion of anticoagulant use likely reflects an elevated cardiovascular risk burden within this cohort, in which coexisting hypertension and dyslipidemia were common. Concerning self-management behaviors, 43% of participants reported performing routine self-blood glucose monitoring (SBGM), reflecting moderate adherence to recommended diabetes self-care practices. The use of complementary and alternative medicine was reported by 33% of respondents, most commonly nutritional supplements (33%), followed by herbal products (5%), while 67% did not use such therapies. Laboratory analyses revealed evidence of suboptimal metabolic control. The median fasting blood glucose was 182 mg/dL (IQR: 31), and the median HbA1c was 8.1% (IQR: 1), both exceeding the recommended glycemic targets for adults with diabetes. The median BMI was 28.74 kg/m² (IQR: 2), indicating a predominance of overweight and obese individuals in the sample. The lipid profile showed elevated TG (median 184 mg/dL, IQR: 49) and moderate TC (mean 170.37 ± 31.96 mg/dL), with median LDL of 101.5 mg/dL (IQR: 46) and median HDL of 47.5 mg/dL (IQR: 18). This pattern is consistent with atherogenic dyslipidemia characteristic of insulin-resistant states. Regarding biochemical markers, the median fasting insulin level was 21.25 µIU/mL (IQR: 8) and the median HOMA-IR was 9 (IQR: 4), suggesting moderate insulin resistance across participants. The median adropin concentration was 299.01 pg/mL (IQR: 137), whereas the median TNF-α level was 42.26 pg/mL (IQR: 21), reflecting a moderate degree of systemic inflammation and metabolic dysregulation at baseline. Taken together, these findings describe a cohort of middle-aged adults with long-standing T2DM, marked by excess weight, poor glycemic control, and prevalent cardiometabolic comorbidities. Their pharmacological profiles appropriately reflected disease complexity, and the laboratory data underscore the ongoing need for intensified metabolic management to reduce long-term cardiovascular and inflammatory risk (Table 1). Table 1Sociodemographic, clinical, and laboratory characteristics of the participantsCharacteristicsMean ± SD/Median ± IQR/ n (%)Age (years), median ± IQR (years)54 ± 14Gender, n (%) • Male43 (43.0) • Female57 (57.0)Duration of diabetes (years), median ± IQR (years)6 ± 4Daily metformin dose (mg), median ± IQR (years)1882 ± 1000Number of prescribed medications, median ± IQR (years)5 ± 4 Smoking status • Smoker25 (25.0) • Non-smoker75 (75.0)Comorbid conditions, n (%) • Hypertension (HTN)22 (22.0) • Hyperlipidemia (HLP)16 (16.0) • HTN + HLP41 (41.0) • Asthma4 (4.0) • Chronic obstructive airway disease2 (2.0) • Gout1 (1.0) • No comorbidity14 (14.0)Concurrent medication use, n (%) • Use of antihypertensive agent(s)23 (23.0) • Use of lipid-lowering agent(s)16 (16.0) • Use of anticoagulant agent(s)3 (3.0) • Use of antihypertensive + anticoagulant agent(s)14 (14.0) • Use of antihypertensive + anticoagulant + lipid-lowering agent(s)42 (42.0) • Use of lipid-lowering agent(s) + anticoagulant agent(s)2 (2.0)Routine self-blood glucose monitoring, n (%) • Yes43 (43.0) • No57 (57.0)Use of complementary or alternative medicine, n (%) • Yes33 (33.0) • No67 (67.0)Type of complementary medicine used, n (%) • Nutritional supplements33 (33.0) • Herbal leaves/roots5 (5.0) • None62 (62.0) Clinical and Laboratory Parameters • BMI (kg/m²), median ± IQR (years)28.74 ± 2 • Fasting blood glucose (mg/dL), median ± IQR (years)182 ± 31 • HbA1c (%), median ± IQR (years)8.10 ± 1 • Triglyceride (mg/dL), median ± IQR (years)184 ± 49 • Total cholesterol (mg/dL), mean ± SD170.37 ± 31.96 • LDL cholesterol (mg/dL), median ± IQR (years)101.50 ± 46 • HDL cholesterol (mg/dL), median ± IQR (years)47.50 ± 18 • Serum insulin (µIU/mL), median ± IQR (years)21.25 ± 8 • HOMA-IR, median ± IQR (years)9 ± 4 • Adropin (pg/mL), median ± IQR (years)299.01 ± 137 • TNF-α (pg/mL), median ± IQR (years)42.26 ± 21Note: SD = Standard deviation; HTN = Hypertension; HLP = Hyperlipidemia; LDL = Low-density lipoprotein; HDL = High-density lipoprotein; TNF-α = Tumor necrosis factor alpha; SBGM = Self-blood glucose monitoring. Smoking status reflects self-reported history or occasional non-daily use; individuals meeting criteria for active daily smoking were excluded during screening Sociodemographic, clinical, and laboratory characteristics of the participants Note: SD = Standard deviation; HTN = Hypertension; HLP = Hyperlipidemia; LDL = Low-density lipoprotein; HDL = High-density lipoprotein; TNF-α = Tumor necrosis factor alpha; SBGM = Self-blood glucose monitoring. Smoking status reflects self-reported history or occasional non-daily use; individuals meeting criteria for active daily smoking were excluded during screening Comparison of baseline characteristics of participants across treatment groups Baseline comparisons between the empagliflozin and sitagliptin groups revealed no significant differences in sociodemographic or clinical variables, confirming that randomization successfully balanced participant characteristics (Table 2). Age, gender, diabetes duration, metformin dose, number of medications, complementary medicine use, and concomitant drug categories were comparable (all p > .05). Smoking history or occasional use did not differ significantly between treatment groups (p > .05) and was therefore not included as a covariate in outcome analyses. Similarly, anthropometric and metabolic indices—including BMI, HbA1c, lipid profile, insulin, HOMA-IR, adropin, and TNF-α—did not differ significantly between groups. Table 2Comparison of baseline sociodemographic, clinical, and laboratory measures by treatment groups (N = 100)VariableEmpagliflozin (n = 50)Sitagliptin (n = 50)Mann Whitney Ua/ Fisher exact testb/t-testc/Kruskal-Wallis Hdz/ χ2/tp-valueAge (years)a, median ± IQR (years)53 ± 1154.50 ± 14–0.8310.406Genderb, n (%)0.410.84 • Male22 (44%)21 (42%) • Female28 (56%)29 (58%)Smoking statusb, n (%)0.480.645 • Yes14 (28%)11 (22%) • No36 (72%)39 (78%)Duration of diabetes (years) a, median ± IQR (years)6.40 ± 46 ± 3–1.1910.234Metformin dose (mg/day) a, median ± IQR (years)2000 ± 10001882 ± 500–1.2750.202No. of medicationsa, median ± IQR (years)5 ± 3.35 ± 4–0.0310.975Use of complementary Medicineb, n (%)0.450.83 • Yes17 (34%)16 (32%) • No33 (66%)34 (68%)Type of complementary Medicined, n (%)0.0950.758 • Herbal leaves/roots2 (4%)3 (6%) • Nutritional supplements18 (36%)15 (30%) • None30 (60%)32 (64%)Concurrent drug categoryd, n (%)2.3660.124 • Use of antihypertensive agent(s)10 (20%)13 (26%) • Use of lipid-lowering agent(s)7 (14%)9 (18%)• Use of anticoagulant agent(s)1(2%)2 (4%) • Use of antihypertensive + anticoagulant agent(s)6 (12%)8 (16%) • Use of antihypertensive + anticoagulant + lipid-lowering agent(s)24 (48%)18 (36%) • Use of lipid-lowering agent(s) + anticoagulant agent(s)2 (4%)-Blood Glucose Monitoringb, n (%)0.450.832 • Yes22 (44%)21 (42%) • No28 (56%)29 (58%)BMI (kg/m²) a, median ± IQR (years)28.71 ± 228.72 ± 2–1.1960.232FBG (mg/dL) a, median ± IQR (years)175.50 ± 33188.50 ± 29–2.6550.008HbA1c (%)a, median ± IQR (years)8.10 ± 18.10 ± 1–1.2610.207Triglyceride (mg/dL) a, median ± IQR (years)185 ± 37184 ± 63–0.2140.831Total cholesterol (mg/dL)c, Mean ± SD172.09 ± 27.16168.65 ± 36.320.5380.592LDL-C (mg/dL) a, median ± IQR (years)104 ± 5599 ± 28–1.6740.094HDL-C (mg/dL) a, median ± IQR (years)46 ± 2148 ± 15–0.8000.424Serum insulin (µIU/mL) a, median ± IQR (years)21.35 ± 820.55 ± 6–1.4280.153HOMA-IRa, median ± IQR (years)9 ± 59.01 ± 4–0.7170.473Adropin (pg/mL) a, median ± IQR (years)289.50 ± 179307.52 ± 84–0.8100.418TNF-α (pg/mL) a, median ± IQR (years)42.39 ± 2041.99 ± 23–0.4270.669Note. Values are presented as median ± interquartile range (IQR) or mean ± standard deviation (SD) for continuous variables, and as frequency (percentage) for categorical variables. Between-group comparisons were conducted using the Mann–Whitney U test for non-normally distributed continuous variables (a), independent samples t-test for normally distributed variables (c), Fisher’s exact test for binary categorical variables (b), and Kruskal–Wallis H test for multi-category variables (d). BMI = body mass index; FBG = fasting blood glucose; HbA1c = glycated hemoglobin; LDL-C = low-density lipoprotein cholesterol; HDL-C = high-density lipoprotein cholesterol; HOMA-IR = homeostatic model assessment of insulin resistance; TNF-α = tumor necrosis factor alpha. Smoking status reflects self-reported history or occasional non-daily use; individuals meeting criteria for active daily smoking were excluded during screening Comparison of baseline sociodemographic, clinical, and laboratory measures by treatment groups (N = 100) Note. Values are presented as median ± interquartile range (IQR) or mean ± standard deviation (SD) for continuous variables, and as frequency (percentage) for categorical variables. Between-group comparisons were conducted using the Mann–Whitney U test for non-normally distributed continuous variables (a), independent samples t-test for normally distributed variables (c), Fisher’s exact test for binary categorical variables (b), and Kruskal–Wallis H test for multi-category variables (d). BMI = body mass index; FBG = fasting blood glucose; HbA1c = glycated hemoglobin; LDL-C = low-density lipoprotein cholesterol; HDL-C = high-density lipoprotein cholesterol; HOMA-IR = homeostatic model assessment of insulin resistance; TNF-α = tumor necrosis factor alpha. Smoking status reflects self-reported history or occasional non-daily use; individuals meeting criteria for active daily smoking were excluded during screening Although FBG was statistically higher in the sitagliptin group (188.5 vs. 175.5 mg/dL; p = .008), the absolute difference was small and clinically negligible, falling within the expected range of biological and analytical variation. This minor difference is unlikely to influence treatment outcomes. Overall, both treatment arms were considered well matched at baseline, demonstrating effective randomization and group comparability before intervention. Clinical and laboratory measures. ITT analysis was conducted to evaluate both primary and secondary outcomes across two time points: baseline (T1) and 12 weeks (T2). Missing data were handled using an EM algorithm under a missing-at-random assumption, preserving statistical power and reducing bias. Although the distribution of several outcome variables deviated from normality, the decision to apply General Linear Model (GLM) Repeated Measures ANOVA was supported by the robustness of this method under moderate violations of normality, especially with balanced group sizes and sufficient sample size (n = 50 per arm). Additionally, residuals were examined for approximate normality, and sphericity was tested and corrected where necessary using Greenhouse-Geisser estimates (Table 3). Table 3Comparison of clinical and biochemical parameters between empagliflozin and sitagliptin groups at baseline and 12 weeks using GLM repeated measures ANOVAModelEmpagliflozin Mean (SD)Sitagliptin Mean (SD)F valueP valueEffect size (Eta²)T1T2T1T2 Adropin (pg/mL) 291.13 (132.58)362.40 (124.24)298.92 (98.01)358.10 (126.22)Group————0.0090.9230.000Time————19.674< 0.0010.286Group × Time————0.1290.7210.003 HOMA-IR 10.08 (4.44)6.65 (2.50)9.28 (2.97)7.15 (1.80)GroupTime————————0.063113.8290.802< 0.0010.0010.699Group × Time————4.8450.0320.090 HbA1c (%) 8.10 (0.53)7.04 (1.18)8.29 (0.66)7.62 (0.68)Group————9.1500.0040.157Time————98.965< 0.0010.669Group × Time————4.3000.0430.081 TNF-α (pg/mL) 44.12 (21.52)30.78 (15.93)43.93 (21.33)37.65 (17.96)Group————0.8020.3750.016Time————37.094< 0.0010.431Group × Time————3.6890.0610.070Triglycerides (mg/dL)183.24 (42.01)156.34 (30.14)179.71 (57.43)165.41 (43.98)Group————0.0830.7750.002Time————101.219< 0.0010.674Group × Time————8.5950.0050.149Total Cholesterol (mg/dL)172.10 (27.16)167.32 (24.51)168.65 (36.32)165.55 (26.42)Group————0.1910.6640.004Time————5.2840.0260.097Group × Time————0.2480.6210.005 LDL-C (mg/dL) 114.88 (26.43)111.15 (19.88)107.26 (24.83)106.15 (35.10)Group————1.8620.1790.037Time————1.3850.2450.027Group × Time————0.3640.5490.007 HDL-C (mg/dL)- in general 47.53 (11.32)53.78 (8.40)49.01 (9.47)51.24 (6.86)Group————0.1300.7200.003Time————14.641< 0.0010.230Group × Time————6.1200.0170.111Male (n = 22)48.86 (12.02)54.05 (9.75)50.06 (11.01)51.66 (7.35)Female (n = 26)46.35 (11.25)53.65 (7.63)48.19 (8.51)50.78 (6.80) Interaction Effects ————Group × Gender————0.0010.9790.000Time × Gender————0.4430.5090.010Group × Time————5.9040.0190.114Group × Time × Gender————0.1130.7390.002 Fasting Insulin (µIU/mL) 22.95 (8.99)17.36 (5.91)20.20 (5.44)17.06 (4.02)Group————1.4410.2360.029Time————68.611< 0.0010.583Group × Time————4.6310.0360.086 Body Weight (kg) 83.9955 (8.56)80.9775 (6.72)84.7716 (10.58)82.8966 (10.06)Group————0.6000.4420.012Time————48.562< 0.0010.498Group × Time————2.1890.1450.043 BMI (kg/m²) 28.97 (2.28)27.98 (2.22)28.53 (1.94)27.90 (1.85)Group————0.3950.5330.008Time————46.180< 0.0010.485Group × Time————1.9080.1730.037Note. Group refers to the test of between-subject effects comparing Empagliflozin and Sitagliptin; Time reflects the within-subject effect across baseline (T1) and 12-week (T2) measurements; Group × Time indicates the interaction effect between treatment group and time. Values are presented as mean (standard deviation). T1 = baseline; T2 = 12-week follow-up; SD = standard deviation; BMI = body mass index; HOMA-IR = homeostatic model assessment of insulin resistance; HbA1c = glycated hemoglobin; LDL-C = low-density lipoprotein cholesterol; HDL-C = high-density lipoprotein cholesterol; TNF-α = tumor necrosis factor alpha. Effect sizes (partial eta squared) are interpreted as follows: 0.01 = small, 0.06 = moderate, and ≥ 0.14 = large magnitude of effect [24] undefined Comparison of clinical and biochemical parameters between empagliflozin and sitagliptin groups at baseline and 12 weeks using GLM repeated measures ANOVA Group Time — — — — — — — — 0.063 113.829 0.802 < 0.001 0.001 0.699 Note. Group refers to the test of between-subject effects comparing Empagliflozin and Sitagliptin; Time reflects the within-subject effect across baseline (T1) and 12-week (T2) measurements; Group × Time indicates the interaction effect between treatment group and time. Values are presented as mean (standard deviation). T1 = baseline; T2 = 12-week follow-up; SD = standard deviation; BMI = body mass index; HOMA-IR = homeostatic model assessment of insulin resistance; HbA1c = glycated hemoglobin; LDL-C = low-density lipoprotein cholesterol; HDL-C = high-density lipoprotein cholesterol; TNF-α = tumor necrosis factor alpha. Effect sizes (partial eta squared) are interpreted as follows: 0.01 = small, 0.06 = moderate, and ≥ 0.14 = large magnitude of effect [24] undefined Primary outcomes Adropin levels increased significantly over time in both groups (F = 19.674, P < .001, η² = 0.286), rising from 291.13 ± 132.58 to 362.40 ± 124.24 pg/mL in the Empagliflozin group and from 298.92 ± 98.01 to 358.10 ± 126.22 pg/mL in the Sitagliptin group. However, the group-by-time interaction was not significant (F = 0.129, P = .721, η² = 0.003), indicating similar trajectories of change. The main effect of group was also non-significant (F = 0.009, P = .923, η² = 0.000). These findings indicate that both empagliflozin and sitagliptin are associated with increased circulating adropin over time, with no evidence of differential regulation between treatments. This pattern suggests that adropin may reflect a shared downstream metabolic response to improved glycemic control rather than a drug-specific mechanism. HOMA-IR showed a significant main effect of time (F = 113.829, P < .001, η² = 0.699), reflecting improved insulin sensitivity in both groups. Values declined from 10.08 ± 4.44 to 6.65 ± 2.50 in the Empagliflozin group and from 9.28 ± 2.97 to 7.15 ± 1.80 in the Sitagliptin group. A significant group-by-time interaction was observed (F = 4.845, P = .032, η² = 0.090), suggesting a greater reduction in the Empagliflozin arm. The main effect of group was not significant (F = 0.063, P = .802, η² = 0.001). Empagliflozin may be more effective in reducing insulin resistance, which is critical for patients with metabolic syndrome or early-stage type 2 diabetes. Figures 2 and 3 illustrate the comparative changes in the adjusted mean values of the primary outcomes between the two treatment groups over the study period. Accordingly, the significant group × time interaction for HOMA-IR supports a differential metabolic effect favoring empagliflozin, whereas the absence of such an interaction for adropin indicates comparable hepatokine responses across treatments rather than a lack of treatment efficacy. Comparisons of adjusted means of Adropin levels between two groups across study period Comparisons of adjusted means of Homa-IR between two groups across study period Secondary outcomes HbA1c (%) decreased significantly over time (F = 98.965, P < .001, η² = 0.669), with a sharper reduction in the Empagliflozin group (from 8.10 ± 0.53 to 7.04 ± 1.18) compared to Sitagliptin (from 8.29 ± 0.66 to 7.62 ± 0.68). Both the group effect (F = 9.150, P = .004, η² = 0.157) and the group-by-time interaction (F = 4.300, P = .043, η² = 0.081) were statistically significant. Empagliflozin demonstrated superior glycemic control, making it a strong candidate for patients needing tighter HbA1c management. TNF-α levels declined significantly over time (F = 37.094, P < .001, η² = 0.431), with a greater reduction in the Empagliflozin group (from 44.12 ± 21.52 to 30.78 ± 15.93) than Sitagliptin (from 43.93 ± 21.33 to 37.65 ± 17.96). However, the group-by-time interaction did not reach conventional statistical significance (F = 3.689, P = .061, η² = 0.070), and the main effect of treatment group was also not significant (F = 0.802, P = .375, η² = 0.016). Accordingly, no statistically significant differential treatment effect on TNF-α change over time was demonstrated. TG showed a significant time effect (F = 101.219, P < .001, η² = 0.674) and a significant group-by-time interaction (F = 8.595, P = .005, η² = 0.149), with greater reductions in the Empagliflozin group. The group effect was not significant (F = 0.083, P = .775, η² = 0.002). Empagliflozin may be more effective in improving lipid metabolism, particularly in patients with elevated triglycerides. TC decreased modestly over time (F = 5.284, P = .026, η² = 0.097), but neither the group effect (F = 0.191, P = .664, η² = 0.004) nor the interaction (F = 0.248, P = .621, η² = 0.005) reached significance. Both treatments contributed to modest cholesterol reduction, though not enough to differentiate their impact clinically. LDL-C and HDL-C demonstrated contrasting patterns in response to treatment. LDL-C changes were not statistically significant across time (F (1, 46) = 1.385, p = .245, η² = 0.027), between groups (F (1, 46) = 1.862, p = .179, η² = 0.037), or in the group × time interaction (F (1, 46) = 0.364, p = .549, η² = 0.007), indicating minimal treatment-related impact on LDL-C levels. In contrast, HDL-C showed a significant main effect of time (F (1, 46) = 14.641, p < .001, η² = 0.230), reflecting a consistent increase from baseline to 12 weeks across both treatment groups. The group × time interaction was also significant (F (1, 46) = 6.120, p = .017, η² = 0.111), suggesting that the magnitude of HDL-C improvement differed between Empagliflozin and Sitagliptin. Specifically, Empagliflozin was associated with a more pronounced elevation in HDL-C over time, supporting its potential advantage in modulating lipid metabolism and offering improvements in lipid parameters that are biologically linked to cardiometabolic risk, without direct evidence of cardiovascular protection. Subgroup analysis was conducted for HDL-C based on its known sex-related biological variability and its significant time and interaction effects in the overall model. Other lipid markers did not demonstrate comparable interaction patterns, and HDL-C was prioritized to maintain analytical focus and reduce Type I error. Although HDL-C levels were slightly higher in males than females at both time points, the group × gender interaction was not significant (F (1, 46) = 0.001, p = .979, η² = 0.000), nor was the time × gender interaction (F (1, 46) = 0.443, p = .509, η² = 0.010). The three-way interaction (group × time × gender) was also non-significant (F (1, 46) = 0.113, p = .739, η² = 0.002), indicating that the pattern of HDL-C change over time did not differ meaningfully by sex or treatment group. Clinically, these findings suggest that both males and females benefited from HDL-C elevation over time, with Empagliflozin showing a modestly superior effect. The absence of significant gender interactions supports the generalizability of this lipid benefit across sexes. Fasting Insulin declined significantly over time (F = 68.611, P < .001, η² = 0.583), with a significant interaction effect (F = 4.631, P = .036, η² = 0.086), indicating a more pronounced reduction in the Empagliflozin group. The group effect was not significant (F = 1.441, P = .236, η² = 0.029). Empagliflozin appears to reduce pancreatic insulin demand more effectively, which may help preserve β-cell function over time. Body Weight and BMI both decreased significantly over time (F = 48.562, P < .001, η² = 0.498 and F = 46.180, P < .001, η² = 0.485, respectively), but neither showed significant group effects (F = 0.600, P = .442, η² = 0.012 and F = 0.395, P = .533, η² = 0.008) or interactions (F = 2.189, P = .145, η² = 0.043 and F = 1.908, P = .173, η² = 0.037). Both treatments supported modest weight and BMI reductions, reinforcing their utility in weight-sensitive diabetes management. Correlation between adropin and metabolic/inflammatory markers (Prespecified exploratory analysis) Spearman correlation analysis revealed a significant inverse association between adropin and TNF-α at 12 weeks (r = − .245, p = .014), indicating that higher adropin levels were linked to lower inflammatory activity (Table 4). No significant correlations were observed between adropin and HOMA-IR (r = .048, p = .635), triglycerides (r = − .065, p = .521), or HDL-C (r = − .091, p = .367). Table 4Correlations between serum adropin and metabolic/inflammatory parameters at week 12VariablesTNF-αHOMA-IRTriglyceridesHDL-C Adropin −0.245 *0.048−0.065−0.091p value(0.014)(0.635)(0.521)(0.367)Note. Values represent Spearman’s ρ (rank-order correlation coefficients)* Correlation is significant at the 0.05 level (two-tailed) Correlations between serum adropin and metabolic/inflammatory parameters at week 12 Note. Values represent Spearman’s ρ (rank-order correlation coefficients) * Correlation is significant at the 0.05 level (two-tailed) Metformin dosing stability during follow-up To formally assess whether metformin dosing could confound treatment effects, metformin dose stability was evaluated longitudinally across the 12-week follow-up period. At baseline, the median daily metformin dose was 1882 mg (IQR: 1000) in the overall cohort and did not differ significantly between the empagliflozin and sitagliptin groups (p = .202; Table 2). Repeated-measures analysis demonstrated no significant main effect of time on metformin dose (p > .99), indicating complete dose stability from baseline to week 12. There was also no significant between-group effect (F = 2.231, p = .142, partial η² = 0.044) and no group × time interaction, confirming that metformin dosing remained comparable between treatment arms throughout the study period. Mean metformin doses were identical at baseline and week 12 within each group (empagliflozin: 1930 mg/day at both time points; sitagliptin: 1790 mg/day at both time points), and no participant required metformin dose escalation or protocol-driven adjustment after randomization. Taken together, these findings confirm that metformin exposure was stable and balanced across groups, minimizing the likelihood that observed differences in glycemic control, insulin resistance, lipid parameters, or inflammatory markers were attributable to metformin dose variation rather than the randomized add-on therapies. Collectively, these findings demonstrate that metformin exposure was stable, balanced, and non-differential between groups, effectively excluding metformin titration as a confounder of the observed between-group differences in glycemic, insulin resistance, lipid, or inflammatory outcomes. Discussion Despite widespread use of SGLT2 and DPP-4 inhibitors, their comparative effects on emerging hepatokines such as adropin remain poorly understood. By contrast, several randomized head-to-head trials have systematically compared SGLT2 and DPP-4 inhibitors with respect to conventional metabolic markers, including glycemic control, body weight, insulin resistance surrogates, and lipid parameters. This trial demonstrates that both drug classes are associated with increased circulating adropin over time, with no evidence of a treatment-specific or superior effect of empagliflozin on this endpoint, while exerting distinct effects on insulin resistance and other metabolic parameters. This randomized controlled trial evaluated the metabolic, glycemic, and inflammatory effects of empagliflozin versus sitagliptin, each added to metformin in adults with T2DM. Both regimens improved key metabolic parameters—body weight, insulin resistance, glycemic indices, and lipid profile—while inflammatory markers improved within both groups without a significant between-group difference—over a 12-week intervention period. These findings reflect short-term metabolic responses and should not be interpreted as evidence of durability or cardiovascular risk reduction, as the present trial was not designed or powered to assess clinical cardiovascular outcomes. Empagliflozin produced significantly larger improvements in HOMA-IR, fasting insulin, HbA1c, and triglycerides, together with a greater rise in HDL-C. Importantly, these between-group differences cannot be attributed to background metformin therapy, as metformin dosing was optimized before randomization and remained completely stable and comparable between groups throughout follow-up, as confirmed by longitudinal analysis. In contrast, adropin increased to a similar extent in both treatment arms, and no group or group-by-time effect was observed, explicitly precluding any claim of empagliflozin superiority for adropin regulation. Although TNF-α declined significantly over time in both arms, the between-group difference did not reach statistical significance, indicating that the observed numerical difference should be interpreted descriptively rather than as evidence of a differential anti-inflammatory effect. These findings should be interpreted in the context of the trial’s open-label design and reliance on instructed—but not objectively monitored—lifestyle stability, as discussed below [13, 25, 26]. Importantly, the presence of two co-primary outcomes with divergent interaction results does not reflect statistical inconsistency but rather biological differentiation. HOMA-IR was selected to test a clinically oriented hypothesis regarding insulin sensitivity, whereas adropin was included to explore hepatokine responsiveness to glucose-lowering therapy. The absence of a differential adropin response therefore suggests a class-independent mechanism of hepatokine modulation, and should be interpreted as biological equivalence between treatments for this endpoint rather than evidence of differential efficacy. This distinction mitigates concerns regarding multiplicity and supports independent interpretation of the co-primary endpoints. All secondary metabolic and inflammatory outcomes should therefore be interpreted as exploratory, supportive signals rather than confirmatory evidence, with primary inference anchored exclusively to the prespecified co-primary endpoints. Furthermore, both treatments were well tolerated, with only mild polyuria and gastrointestinal discomfort reported. The absence of serious adverse events reinforces the established safety profiles of both drug classes, satisfying the CONSORT recommendation to balance benefits and harms in pharmacological trials. Body weight and obesity management Weight and BMI declined significantly over time in both groups, indicating that short-term combination therapy with metformin is associated with modest weight reduction irrespective of the add-on agent. The absence of a significant between-group difference suggests that weight loss alone is unlikely to account for empagliflozin’s superior metabolic outcomes. From a clinical perspective, these results support the use of SGLT2 inhibitors as practical therapeutic options for overweight or obese patients, in whom even modest weight reductions may contribute to improvements in insulin sensitivity and overall cardiometabolic risk profiles [27–30]. However, as dietary intake and physical activity were not objectively assessed, subtle behavioral adaptations or expectancy effects cannot be entirely excluded as contributing factors, particularly for weight-sensitive metabolic endpoints. Glycemic control and clinical relevance Empagliflozin achieved greater reductions in HbA1c and FBG than sitagliptin, reinforcing evidence that SGLT2 inhibitors offer superior glycemic durability when added to metformin. These findings are consistent with prior head-to-head randomized trials comparing SGLT2 inhibitors with DPP-4 inhibitors, which have consistently reported larger HbA1c reductions with SGLT2 inhibition, particularly in patients with higher baseline hyperglycemia or insulin resistance. While these differences are consistent with known pharmacological mechanisms, the open-label nature of the study may have permitted differential behavioral responses that could modestly influence glycemic outcomes. Nevertheless, the magnitude of HbA1c reduction observed with empagliflozin aligns with prior blinded trials, supporting a true treatment effect [25, 26, 31–33]. Insulin resistance and β-Cell workload Both agents improved insulin resistance, but the empagliflozin group showed a larger decline in HOMA-IR and fasting insulin. Although few head-to-head trials have assessed insulin resistance directly, existing comparative studies using surrogate markers (fasting insulin, HOMA-IR, or weight-adjusted indices) generally favor SGLT2 inhibitors over DPP-4 inhibitors, supporting the direction and clinical plausibility of the present findings. Given that insulin resistance is sensitive to short-term changes in energy balance and activity, the absence of objective lifestyle monitoring represents a potential source of residual confounding. Even so, the direction and consistency of these findings with mechanistic and long-term outcome trials suggest that pharmacological effects predominated. This greater insulin-sensitizing effect can alleviate β-cell stress and delay progression toward insulin therapy. From a translational perspective, such early improvement in insulin sensitivity may predict better long-term metabolic flexibility and lower exogenous insulin requirements in high-risk patients. In clinical terms, choosing empagliflozin for patients with high baseline HOMA-IR or hyperinsulinemia may yield better long-term glycemic stability and lower exogenous insulin requirements [34–38]. Nevertheless, insulin sensitivity was assessed using HOMA-IR rather than the hyperinsulinemic–euglycemic clamp, and thus reflects an indirect estimate rather than a direct physiological measurement. In addition, HOMA-IR is dependent on fasting insulin measurements and is therefore sensitive to intra- and inter-assay variability. While clamp studies provide superior precision, their invasiveness and logistical demands limit feasibility in pragmatic clinical trials. In the present study, insulin concentrations were measured using the same commercially validated ELISA platform and analyzed in a single central laboratory under standardized quality-control procedures, minimizing analytical variability and supporting the internal consistency of longitudinal and between-group comparisons. The consistent improvements observed across HOMA-IR, fasting insulin, and glycemic indices support the internal validity of the insulin-sensitizing effects reported here. Lipid profile and cardiometabolic risk Empagliflozin induced a greater TG reduction and HDL-C rise, whereas TC and LDL-C changes were minimal. This lipid pattern mirrors prior head-to-head SGLT2 versus DPP-4 trials, in which SGLT2 inhibitors consistently improved triglycerides and HDL-C to a greater extent, while effects on LDL-C and total cholesterol were neutral or modest. The parallel reduction in TG and rise in HDL-C observed here suggests an improved atherogenic lipid balance and may reflect a more favorable cardiometabolic risk profile, although no inference regarding cardiovascular benefit can be made in the absence of outcome data. Thus, empagliflozin may be particularly beneficial in patients with diabetic dyslipidemia, offering metabolic control without requiring additional lipid-lowering therapy [7, 28, 31, 35]. Adropin and inflammation: mechanistic and translational meaning Adropin levels rose significantly over time in both treatment arms, with no statistically significant difference between empagliflozin and sitagliptin. Importantly, unlike glycemic, lipid, and weight outcomes—where multiple head-to-head SGLT2 versus DPP-4 trials exist—comparative data on hepatokines such as adropin are essentially absent. This finding suggests a class-independent hepatokine response associated with improved metabolic homeostasis and is consistent with the non-significant between-group interaction observed in the repeated-measures analysis, rather than a drug-specific mechanism. Human clinical evidence regarding adropin in metabolic disease remains heterogeneous. Observational studies and pooled analyses generally report lower circulating adropin in T2DM versus controls, although between-study heterogeneity and residual confounding limit causal inference [39]. Lower adropin has also been reported in overweight/obese humans in narrative syntheses of the field [40]. In some cohorts, adropin correlates inversely with indices of insulin resistance and cardiometabolic risk; however, other studies report weak or non-significant associations after adjustment for adiposity, age, and comorbidities, supporting an overall equivocal human literature [39]. Interventional human data are particularly limited. Small studies examining lifestyle modification or pharmacological interventions have reported variable changes in adropin, with some showing increases following weight loss or improved glycemic control, while others observed no meaningful change despite metabolic improvement. To date, no randomized controlled trials have directly compared the effects of different glucose-lowering drug classes on circulating adropin, leaving its treatment responsiveness and class specificity largely undefined. In this context, the present trial provides novel comparative clinical evidence that adropin increases over time with both SGLT2 inhibition and DPP-4 inhibition when added to metformin, without evidence of differential regulation over a short-term (12-week) treatment window. Whether this hepatokine response is sustained, attenuated, or amplified with longer exposure cannot be determined from the present data. This pattern supports the interpretation of adropin as a treatment-responsive but class-independent hepatokine, potentially reflecting improved systemic metabolic milieu rather than a pathway-specific drug effect. Notably, expectation or behavioral effects are unlikely to selectively influence adropin regulation, supporting its potential utility as a treatment-responsive biomarker less prone to subjective bias. In addition, given adropin’s reported associations with endothelial and metabolic pathways, its elevation may coincide with broader metabolic improvement without implying cardiovascular benefit or prognostic relevance in the absence of longitudinal outcome data rather than isolated glycemic change, consistent with prior experimental and translational observations [18, 23, 41, 42]. TNF-α decreased significantly over time in both groups, but the between-group difference was not significant. Accordingly, the present data do not support a differential anti-inflammatory effect between treatments, although complementary mechanisms may exist. SGLT2 inhibitors reduce oxidative stress and adipose inflammation, while DPP-4 inhibitors suppress NF-κB activation through incretin pathways. The modest inverse correlation between adropin and TNF-α observed in this study is consistent with prior human and experimental reports describing an association between circulating adropin levels and inflammatory markers; however, this relationship should be interpreted strictly as associative, and no causal inference can be drawn. Future mechanistic work should quantify whether the observed TNF-α decline correlates with adropin elevation, clarifying potential hepatokine-inflammation crosstalk. Clinically, either agent may be associated with short-term improvements in inflammatory profiles, and no differential conclusion can be drawn for adropin-mediated effects. However, given the short study duration, it remains uncertain whether these early inflammatory changes translate into sustained vascular benefit or long-term cardiometabolic risk reduction. Notably, although SGLT2 inhibition has demonstrated cardiovascular benefit in longer, outcome-driven trials, these effects are mediated through pathways that appear independent of adropin regulation [43–45]. Although the group-by-time interaction P value was marginally above the conventional threshold (P = .061), tests of interaction are typically underpowered in short-term trials, and such observations should be regarded as exploratory signals rather than confirmatory evidence. Furthermore, while systemic inflammation is mechanistically linked to atherosclerosis, the present findings should not be extrapolated to infer cardiovascular risk reduction, as short-term changes in inflammatory biomarkers do not reliably predict hard cardiovascular outcomes. Correlation analysis and exploratory associations The modest but significant inverse correlation between adropin and TNF-α indicates an association between higher circulating adropin levels and lower inflammatory marker concentrations in type 2 diabetes. This finding does not imply a causal or mechanistic relationship, but is consistent with prior observational and experimental literature reporting concurrent variation between adropin and inflammatory indices. Although mechanistic studies have linked adropin to endothelial and intracellular signaling pathways, the present clinical data are insufficient to determine whether adropin directly modulates inflammation or merely reflects a shared downstream metabolic state [23, 41]. Although correlations with insulin resistance and lipid indices were not significant, the overall pattern suggests that adropin elevation may be associated with a broader improvement in metabolic–inflammatory status, rather than serving as a direct determinant of glycemic or lipid changes. Importantly, the consistent and comparable adropin elevation across both therapies underscores its promise as a treatment-responsive biomarker rather than a discriminator between pharmacological classes. Future longitudinal studies should examine whether early adropin elevation observed over short-term interventions is associated with sustained metabolic improvement or cardiovascular outcomes over longer follow-up periods. Mechanistic integration Empagliflozin’s metabolic advantages stem from insulin-independent glucosuria, osmotic diuresis, and a negative energy balance that shifts substrate utilization toward fatty-acid oxidation, thereby reducing glucotoxicity and hepatic lipogenesis. In contrast, sitagliptin improves glucose control mainly via incretin-mediated insulin secretion and glucagon suppression, which are glucose-dependent and may plateau in advanced insulin resistance [15, 25, 46–49]. Clinical and practical implications For clinicians managing metformin-treated T2DM with marked insulin resistance or hypertriglyceridemia, empagliflozin provides a more comprehensive metabolic benefit, improving both glucose and lipid abnormalities while maintaining a low hypoglycemia risk. Sitagliptin remains an appropriate option when SGLT2 inhibitors are contraindicated (e.g., in eGFR < 45 mL/min/1.73 m² or volume-depleted patients) [26, 31]. These recommendations should be interpreted alongside individual patient characteristics and with recognition of potential behavioral influences inherent to open-label designs. Importantly, the consistent adropin elevation across both therapies underscores its promise as a short-term, treatment-responsive biomarker of metabolic change. Whether adropin has value as a durable prognostic or cardiometabolic biomarker requires longer follow-up and outcome-driven validation. Future longitudinal studies should examine whether early adropin elevation during short-term glucose-lowering therapy predicts sustained metabolic improvement or cardiovascular benefit, positioning it within precision-medicine frameworks for diabetes care only after durability and outcome relevance are established. Strengths and limitations Strengths of this trial include randomized allocation, concealed sequence generation, blinded analysis, prespecified outcomes, and high adherence, aligning with CONSORT recommendations for pharmacological trials. The pragmatic recruitment of multimorbid T2DM patients enhances the real-world relevance of the findings. Moreover, participant retention was high, with an observed dropout rate substantially lower than anticipated, strengthening internal validity and statistical precision. An additional methodological strength is the strict pre-randomization stabilization and post-randomization maintenance of metformin dosing, which minimized background therapy confounding and strengthened causal attribution to the randomized add-on agents. Limitations include the 12-week study duration, which constrains conclusions regarding the durability of the observed metabolic and inflammatory changes and precludes inference about long-term cardiometabolic or cardiovascular outcomes, as such outcomes typically require months to years of follow-up in outcome-driven trials; the single-center design, which limits generalizability; open-label allocation of participants and treating clinicians, which may have introduced expectation or behavioral bias despite blinded outcome assessment; the absence of objective monitoring of dietary intake and physical activity, allowing potential unmeasured lifestyle influences on metabolic outcomes; the inclusion of multiple secondary endpoints, increasing the risk of type I error; although these secondary endpoints were prespecified and interpreted descriptively without formal multiplicity correction, this increases the potential for type I error and warrants cautious interpretation; reliance on ELISA assays for adropin and TNF-α, which lack external harmonization; and the presence of two co-primary outcomes without formal multiplicity adjustment, necessitating cautious interpretation of statistical significance across endpoints. In addition, insulin resistance was assessed using HOMA-IR rather than the hyperinsulinemic–euglycemic clamp—the gold-standard method for quantifying insulin sensitivity. Although HOMA-IR is a validated and widely applied surrogate in T2DM trials, it provides an indirect estimate and may be less sensitive to subtle physiological changes; moreover, its precision can be influenced by insulin assay variability, which should be considered when interpreting insulin-resistance estimates. Participants were instructed to maintain habitual dietary and physical activity patterns; however, adherence to these instructions was not objectively verified, and differential co-interventions cannot be fully excluded. Future trials incorporating accelerometry, dietary records, or controlled feeding designs would strengthen causal inference. Moreover, the study population comprised Iranian adults with moderate metabolic impairment recruited from a single tertiary endocrinology center, and stringent eligibility criteria (e.g., exclusion of insulin users, smokers, and individuals with NAFLD or significant comorbidities) may have yielded a more homogeneous and health-selected cohort. This restricted sample reduces representativeness and limits generalizability to multi-ethnic or more clinically diverse T2DM populations. Future multi-center studies across diverse settings are warranted to validate the reproducibility and external applicability of these results. The absence of cardiovascular endpoints or long-term follow-up precludes any inference regarding cardiovascular benefit, and all cardiometabolic interpretations should be understood as biomarker-based and hypothesis-generating only. A placebo control arm was not used in this trial because it would have been ethically inappropriate. All participants had been receiving metformin monotherapy for at least three months prior to enrollment and continued to exhibit inadequate glycemic control (HbA1c ≥ 7.5%). In such cases, withholding or delaying the addition of an evidence-based second-line agent solely for research purposes would expose patients to persistent hyperglycemia and its complications, thereby violating ethical principles of beneficence and clinical responsibility. Therefore, this study compared two approved active treatments—empagliflozin and sitagliptin—as adjuncts to metformin to evaluate their relative therapeutic benefits under standard-of-care conditions. Longer multicenter trials are warranted to determine whether the early metabolic and inflammatory improvements observed with SGLT2 inhibition translate into durable cardiovascular and renal protection. Mechanistic research should explore whether adropin mediates insulin-signaling pathways (e.g., AMPK/PI3K-Akt) and endothelial benefits [41]. Conclusions Over 12 weeks, both empagliflozin and sitagliptin improved metabolic control in metformin-treated T2DM, but empagliflozin achieved larger gains in insulin sensitivity and triglyceride-rich dyslipidemia, while both agents showed comparable improvements in inflammatory modulation. The trial’s safety profile, mechanistic insight, and hepatokine findings collectively reinforce SGLT2 inhibitors as a rational first add-on to metformin when therapeutic goals extend beyond glycemic control to broader cardiometabolic optimization. Electronic Supplementary Material Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements During the preparation of this manuscript, the authors utilized Ludwig.Guru, an AI-based linguistic search and editing tool, to enhance the clarity and precision of the language. This resource assisted in refining the grammar and academic style of the text. Abbreviations Type 2 diabetes Mellitus Sodium–Glucose Cotransporter-2 Inhibitor Dipeptidyl Peptidase-4 Inhibitor Total Cholesterol TG Low-density lipoprotein cholesterol High-density lipoprotein cholesterol Homeostatic Model Assessment of Insulin Resistance Tumor Necrosis Factor-alpha Body Mass Index AMP-activated protein kinase Phosphoinositide 3-kinase/Protein kinase B activation Estimated Glomerular Filtration Rate Iranian Registry of Clinical Trials Fasting Blood Glucose Glycated Hemoglobin Interquartile Range Golnaz Azami is an Assistant Professor in the field of Nursing with nine years of experience in nursing practice and education. Her academic and professional interests include nursing education, clinical training, evidence-based practice, and curriculum development. She has contributed to the advancement of nursing education through teaching, student supervision, and scholarly activities. Dr. Azami is committed to promoting high-quality education, professional development, and excellence in patient care. Author contributions Otbah Salim Taha: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Visualization, Writing – original draft. Golnaz Azami: Conceptualization, Methodology, Investigation, Project administration, Validation, Writing – review and editing. Lotfollah Saed: Investigation, Resources, Supervision, Writing – review and editing. Mohammad Saeed Hakhamaneshi: Conceptualization, Methodology, Supervision, Validation, Writing – review and editing. Heresh Moridi: Methodology, Supervision, Resources, Writing – review and editing, Funding acquisition, Final approval of the manuscript. Funding This study received no external funding or financial support. No pharmaceutical company or external organization supplied medications, materials, or services for the conduct of this trial. Data availability De-identified participant data, including the data dictionary and statistical code used for analyses, will be available from the corresponding author upon reasonable request. Data access will require submission of a written proposal and execution of a data-sharing agreement. At present, no public repository has been designated for this dataset. Declarations Ethics approval and consent to participate The study protocol and statistical analysis plan were reviewed and approved by the Research Ethics Committee of Kurdistan University of Medical Sciences, Sanandaj, Iran (Ethics approval code: IR.MUK.REC.1403.323). All procedures involving human participants were conducted in accordance with the ethical standards of the institutional research committee and adhered to the principles of the Declaration of Helsinki. Written and verbal informed consent was obtained from all participants prior to enrollment, and the consent procedure was formally approved by the Ethics Committee. Participants received a detailed explanation of the study’s purpose, procedures, and potential risks, and all questions were addressed by the research team. Participation was entirely voluntary, with the right to withdraw at any stage without penalty. Confidentiality and anonymity were rigorously maintained: personal identifiers were removed from the dataset, data were stored securely with access restricted to the research team, and results are reported in aggregate form to ensure that no individual participant can be identified. Use of AI and AI-assisted technologies During the preparation of this manuscript, the authors utilized Ludwig.Guru, an AI-based linguistic search and editing tool, to enhance the clarity and precision of the language. This resource assisted in refining the grammar and academic style of the text. Consent for publication Not applicable, as this manuscript does not contain any individual person’s data in any form (including images, videos, or case details). Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Otbah Salim Taha and Golnaz Azami contributed equally to this work. Contributor Information Mohammad Saeed Hakhamaneshi, Email: biohakha@gmail.com. Heresh Moridi, Email: hereshmoridi@yahoo.com. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement De-identified participant data, including the data dictionary and statistical code used for analyses, will be available from the corresponding author upon reasonable request. Data access will require submission of a written proposal and execution of a data-sharing agreement. At present, no public repository has been designated for this dataset.