The impact of SGLT2 inhibitors on uric acid levels in patients with heart failure with preserved ejection fraction: integration of mechanisms and clinical evidence

Abstract

Objective:

Hyperuricemia (HUA) is a prevalent comorbidity in patients with heart failure with preserved ejection fraction (HFpEF) and is closely associated with disease progression and adverse outcomes. Sodium-glucose cotransporter 2 (SGLT2) inhibitors can rapidly and sustainably reduce serum uric acid (SUA) levels and lower the incidence of HUA-related clinical events; however, their underlying mechanisms remain unclear. This study employed network pharmacology and molecular docking to systematically investigate the potential mechanisms of SGLT2 inhibitors in HFpEF patients with concomitant hyperuricemia, focusing on uric acid metabolism modulation and inflammation-related pathways. Additionally, a machine learning approach was applied to retrospectively analyze the effects of SGLT2 inhibitors on SUA levels, aiming to clarify the therapy’s biological basis and clinical relevance.

Methods:

The SMILES structures of eleven SGLT2 inhibitors were retrieved from PubChem. Potential drug targets were predicted via the SwissTargetPrediction platform and intersected with HFpEF- and HUA-related targets obtained from OMIM and GeneCards to identify common targets. Protein–protein interaction (PPI) networks were constructed using STRING to identify hub genes, followed by Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses using DAVID. Molecular docking with AutoDock Vina assessed binding affinities. Clinical data of HFpEF patients were collected, and a classification model based on 34 clinical variables was developed using the FLAML automated machine learning framework to predict SGLT2 inhibitor treatment status and changes in SUA. Model interpretability and feature importance were performed using SHAP analysis.

Results:

Eleven SGLT2 inhibitors acted on 32 targets associated with SUA reduction in HFpEF patients. Key pathways included tumor necrosis factor (TNF), advanced glycation end product–receptor for advanced glycation end product (AGE–RAGE), interleukin-17 (IL-17), and Kaposi’s sarcoma-associated herpesvirus (KSHV) signaling. Molecular docking demonstrated favorable binding of SGLT2 inhibitors to core targets, including GAPDH, CASP3, ICAM1, ACE, DPP4, and AGTR1, with canagliflozin showing the strongest predicted binding affinity. Clinical data analyses identified uric acid and creatinine as the most important predictors in the machine learning model, reflecting their importance in model predictions.

Conclusion:

By integrating network pharmacology and retrospective clinical analysis, this study elucidates the potential uric acid–lowering mechanisms of SGLT2 inhibitors in HFpEF patients. The identified targets and pathways provide mechanistic insights into SGLT2 inhibitor pharmacology and support their therapeutic potential for managing hyperuricemia.

1 Introduction

Heart failure (HF) is the terminal stage of various cardiac diseases and has emerged as a major global public health burden (1, 2). Among its subtypes, heart failure with preserved ejection fraction (HFpEF) accounts for approximately 50% of all HF cases, with both incidence and prevalence continuing to rise worldwide (3). In recent years, hyperuricemia (HUA), a common metabolic disorder, has garnered increasing attention due to its detrimental effects on cardiovascular health (4–7). Substantial epidemiological evidence demonstrates that HUA is not only an independent risk factor for hypertension, coronary artery disease, and chronic kidney disease but is also closely associated with the onset and progression of HFpEF (8). More than 50% of HFpEF patients present with elevated baseline serum uric acid (SUA) levels (9–11), which correlate with disease severity and adverse prognosis. SUA, the final product of purine nucleotide metabolism catalyzed by xanthine oxidase, reflects both enzyme activity and the balance between purine intake and renal excretion. Elevated SUA contributes to endothelial dysfunction, impaired myocardial energetics, increased blood pressure, and reduced renal function, thereby exacerbating the progression of HF (12, 13). These pathophysiological processes further contribute to the development of HFpEF.

Previous studies have demonstrated that sodium–glucose cotransporter 2 (SGLT2) inhibitors reduce the risk of heart failure hospitalization and improve cardiovascular outcomes, significantly enhancing patients’ quality of life (12, 14). Large-scale randomized clinical trials have consistently demonstrated the cardiovascular benefits of individual SGLT2 inhibitors. Empagliflozin has been shown to reduce the risk of heart failure hospitalization and cardiovascular events across a broad range of ejection fraction phenotypes (15). Similarly, dapagliflozin has demonstrated significant reductions in worsening heart failure or cardiovascular death in patients with heart failure (16). In addition, ertugliflozin has shown potential cardioprotective effects, further supporting the class-wide benefit of SGLT2 inhibitors in heart failure management (17). Consequently, SGLT2 inhibitors have emerged as a promising therapeutic option for HFpEF (18, 19). Consequently, SGLT2 inhibitors have emerged as a promising therapeutic option for HFpEF (18, 19). Encouragingly, these agents have also been shown to decrease serum uric acid levels; however, their underlying mechanisms and clinical significance in HFpEF remain unclear (20). In this study, a network pharmacology approach was applied to construct a comprehensive drug–target–disease interaction network to systematically elucidate the molecular mechanisms of SGLT2 inhibitors and address the limitations of conventional single-target studies (21). Protein–protein interaction (PPI) networks were established to identify potential target proteins, followed by Gene Ontology (GO) functional annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses to explore relevant biological processes. Furthermore, a machine learning–based analysis of clinical data was conducted to evaluate the effect of SGLT2 inhibitors on SUA levels in patients with HFpEF, providing clinical validation for the network pharmacology results. This multidimensional and interdisciplinary strategy provides comprehensive theoretical and empirical evidence for the comorbidity management of HFpEF and HUA.

2 Materials and methods

2.1 Network pharmacology analysis

2.1.1 Acquisition and integration of disease-associated targets

Potential disease-related targets associated with “heart failure with preserved ejection fraction (HFpEF)” and “hyperuricemia” were systematically retrieved from the OMIM (https://www.omim.org/) and GeneCards (https://www.genecards.org/) databases. The retrieved datasets were de-duplicated and integrated to construct a comprehensive repository of disease-associated targets for subsequent analyses.

2.1.2 Prediction and screening of SGLT2 inhibitor targets

Eleven clinically approved SGLT2 inhibitors—canagliflozin, dapagliflozin, empagliflozin, luseogliflozin, tofogliflozin, ipragliflozin, ertugliflozin, enavogliflozin, henagliflozin, sotagliflozin, and bexagliflozin—were included for target prediction. Their SMILES structures were obtained from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/) and imported into the SwissTargetPrediction platform (http://www.swisstargetprediction.ch/) to predict potential protein targets. Targets with a prediction probability greater than zero were retained. The intersection between drug-related and disease-associated targets was identified and visualized as Venn diagrams using the BioLadder online bioinformatics platform (https://www.bioladder.cn/web/), thereby identifying potential common targets for further analysis.

2.1.3 Construction of the protein–protein interaction network and identification of core targets

The predicted common targets were uploaded to the STRING database (https://string-db.org/) with the organism restricted to Homo sapiens and a minimum interaction confidence score of 0.4. The interaction network was visualized using Cytoscape (version 3.10.0). Topological parameters, including degree, betweenness centrality, and closeness centrality, were analyzed using the CentiScaPe 2.2 plugin. Nodes exhibiting all three parameters above their mean values were defined as core targets and subsequently subjected to functional enrichment analysis.

Core targets were identified based on topological parameters, including degree, betweenness centrality, and closeness centrality, with thresholds defined as values exceeding the corresponding mean levels within the network. This approach is widely used in network pharmacology to prioritize relatively central nodes that may play key regulatory roles.

However, we acknowledge that the use of mean-based thresholds may be relatively permissive and could introduce potential noise. Therefore, the identified core targets should be interpreted as candidate targets rather than definitive functional regulators. Future studies incorporating stricter selection criteria or sensitivity analyses are warranted to further validate the robustness of these findings.

2.1.4 Construction and topological analysis of the drug–target–disease network

A comprehensive “drug–target–disease” interaction network was constructed using Cytoscape (version 3.10.1). Distinct node types (drug, target, disease) were assigned unique attributes. Network parameters, including density, average clustering coefficient, and degree distribution, were calculated to evaluate the network’s structural properties and stability, thereby elucidating the mechanisms underlying drug–target–disease interactions.

2.1.5 Functional enrichment analysis

The identified core targets were submitted to the DAVID database (https://davidbioinformatics.nih.gov/) for Gene Ontology (GO) annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses. GO terms and pathways with both -values and -values < 0.05 were considered statistically significant. The top 10 enriched GO terms and top 20 KEGG pathways were visualized using the Bioinformatics online platform (https://www.bioinformatics.com.cn/) to highlight the major biological processes and signaling pathways involved.

2.1.6 Molecular docking validation

Molecular docking was performed between the active SGLT2 inhibitor compounds and the core target proteins identified from the network analysis. Three-dimensional structures of the ligands were obtained from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/), and protein crystal structures were downloaded from the Protein Data Bank (PDB, https://www.rcsb.org/). Protein preprocessing was conducted using AutoDockTools (version 1.5.7) by removing water molecules, adding polar hydrogens, and assigning Gasteiger charges. Docking simulations were executed in AutoDock Vina, with the grid box centered on the active binding pocket. The lowest binding energy of each ligand–receptor complex was recorded. PyMOL (version 2.4) was used for three-dimensional visualization of binding conformations, hydrogen-bond interactions, key amino acid residues, and bond lengths.

2.2 Clinical data analysis based on machine learning

2.2.1 Inclusion criteria

Patients diagnosed with heart failure with preserved ejection fraction (HFpEF) who were hospitalized in the Department of Cardiology, Zhejiang Provincial People’s Hospital Bijie Hospital, between November 2021 and December 2022, were included in this study. The inclusion and exclusion criteria were formulated in accordance with relevant clinical guidelines and research objectives, encompassing age range, diagnostic standards, comorbidities, and prior treatment history, to ensure both representativeness and data completeness. This study was approved by the Ethics Committee of Zhejiang Provincial People’s Hospital Bijie Hospital (Approval No. [2023-12-24(2)]) and conducted in compliance with the Declaration of Helsinki. Ethical approval was not additionally required because only secondary, de-identified, and publicly available data were used.

2.2.2 Data collection

Comprehensive multidimensional clinical data were collected, including 34 variables covering hematological, biochemical, inflammatory, echocardiographic, and vital-sign parameters. Specific indicators comprised hemoglobin, platelet count, uric acid, creatinine, urea, glucose, triglycerides, high- and low-density lipoprotein cholesterol, glycated hemoglobin, troponin, myoglobin, creatine kinase-MB, C-reactive protein, high-sensitivity C-reactive protein, interleukin-6, left ventricular diastolic function, right ventricular outflow tract diameter, left ventricular end-diastolic and end-systolic diameters, right atrial and ventricular diameters, left atrial diameter, left ventricular ejection fraction, interventricular septal and posterior wall thickness, as well as heart rate, systolic, and diastolic blood pressure at both admission and discharge.

2.2.3 Data organization and preprocessing

A total of 398 patients were enrolled and categorized into treatment and control groups according to SGLT2 inhibitor therapy status, enabling comprehensive evaluation of therapeutic efficacy. Among them, 34 patients with complete pre- and post-treatment records were selected, yielding 68 paired samples for individual-level effect analysis. Missing data were handled using a complete-case analysis strategy. For continuous variables, records with missing values were excluded from statistical analyses. Categorical variables with missing values were excluded from corresponding analyses. Categorical variables were converted to numerical features via one-hot encoding, and continuous variables were standardized to eliminate dimensional disparities and ensure data consistency.

2.2.4 Model construction and evaluation

Automated machine learning (AutoML) was implemented using FLAML 2.3.4 (https://microsoft.github.io/FLAML/) (22) to develop classification models. Five-fold stratified cross-validation was employed to maintain class balance between training and validation sets, thereby enhancing model robustness and generalizability. During preprocessing, categorical variables were converted to categorical data types and continuous variables to floating-point types to improve compatibility and computational efficiency. FLAML automatically searched for the optimal learning algorithm and hyperparameter configuration using the area under the receiver operating characteristic curve (ROC AUC) on the validation set as the optimization metric. Each fold was trained under a fixed time budget with a unified random seed to ensure reproducibility. Two independent exploratory classification tasks were conducted: (1) identifying patterns associated with SGLT2 inhibitor treatment status, and (2) distinguishing pre- and post-treatment samples for hypothesis-generating analysis rather than confirmatory treatment effect estimation. Model performance was evaluated based on the mean ROC AUC across five folds.

2.2.5 Model interpretation and target analysis

The SHAP (SHapley Additive exPlanations) method (23), grounded in cooperative game theory, was employed to quantify the contribution of each input feature to model predictions, thereby elucidating the underlying decision-making mechanisms. This interpretability analysis identified key predictive features exerting substantial influence, offering insights into potential biological markers or therapeutic targets and supporting subsequent mechanistic and clinical strategy development.

3 Results

3.1 Network pharmacology analysis results

3.1.1 Disease-related target screening

A total of 2,095 targets associated with HFpEF and 1,381 targets associated with hyperuricemia (HUA) were retrieved from the OMIM and GeneCards databases. After further filtering, 32 overlapping targets associated with SGLT2 inhibitors, HFpEF, and HUA were identified (Figure 1). The overlap was visualized using a Venn diagram, with distinct colors representing the target sets corresponding to SGLT2 inhibitors, HFpEF, and HUA.

Figure 1

3.1.2 Protein–protein interaction (PPI) network analysis

The 32 overlapping targets were imported into the STRING database to obtain protein–protein interaction information and construct a PPI network. In the visualization, node size and color intensity correspond to node degree, while edge thickness reflects interaction confidence (Figure 2). Key hub targets, including GAPDH, CASP3, ICAM1, ACE, DPP4, and AGTR1, were identified, with detailed degree values provided in Table 1.

Figure 2

A drug–target–disease regulatory network was subsequently constructed using Cytoscape (v3.10.1), comprising 44 nodes (32 overlapping targets and 11 SGLT2 inhibitors) and 163 interactions. In Figure 3, red nodes represent SGLT2 inhibitors, and yellow nodes indicate disease-related targets. These results suggest that SGLT2 inhibitors may exert a multi-target synergistic effect in the treatment of HFpEF concomitant with HUA.

Figure 3

3.1.3 Functional enrichment analysis

GO annotation was performed for biological process (BP), cellular component (CC), and molecular function (MF) categories. Significantly enriched terms included responses to lipopolysaccharide, regulation of apoptotic processes, membrane regions, and various enzyme activities (Figure 4). KEGG pathway enrichment analysis revealed that core targets were significantly enriched in signaling pathways closely associated with cardiovascular disease and metabolic dysregulation, including the TNF, AGE–RAGE, and IL-17 pathways (Figure 5).

Figure 4

Figure 5

3.1.4 Molecular docking analysis

Representative active components of the 11 SGLT2 inhibitors were docked with hub targets (GAPDH, CASP3, ICAM1, ACE, DPP4, and AGTR1) using AutoDock Vina. Lower binding energies indicate greater stability of the ligand–receptor complex; binding energies below  kcal/mol indicate favorable binding, whereas energies below  kcal/mol indicate strong affinity.

All active components exhibited binding energies below  kcal/mol, suggesting favorable predicted interactions with the core targets in silico (Table 2). Representative docking conformations indicated that hydrogen bonding and hydrophobic interactions may contribute to ligand–target stabilization. These results represent computational predictions rather than experimentally validated molecular interactions and should be interpreted as hypothesis-generating.

Representative low-energy conformations were visualized using PyMOL to illustrate key hydrogen bond interactions and binding residues (Figure 6).

Figure 6

Further experimental validation, such as biochemical binding assays or functional studies, is required to confirm the biological relevance of these predicted interactions.

3.2 Results of machine learning model analysis

3.2.1 Clinical data statistical analysis

Multiple clinical indicators were compared between the SGLT2 inhibitor treatment group and the control group. Several variables exhibited statistically significant differences between the two groups (). Specifically, hemoglobin levels were significantly higher in the treatment group compared with the control group (139.19 vs. 130.78 g/L, ), urea levels were lower (6.39 vs. 7.68 mmol/L, ), left ventricular ejection fraction increased (59.14% vs. 57.00%, ), and creatinine levels decreased (81.00 vs. 135.81 μmol/L, ). Additionally, creatine kinase isoenzyme (16.69 vs. 15.24 U/L, ), C-reactive protein (18.34 vs. 32.33 mg/L, ), and blood glucose (11.43 vs. 10.24 mmol/L, ) also differed significantly. Most other indicators, including right atrial diameter, troponin, platelet count, and interleukin-6, showed no significant differences (). These results provide key feature variables for subsequent predictive model development (Table 3).

Within the SGLT2 inhibitor treatment group, paired-sample -tests were performed for 34 patients comparing pre- and post-intervention clinical indicators. A significant increase was observed in right ventricular outflow tract diameter (29.09 mm vs. 27.15 mm, ), whereas other indicators did not change significantly. Although creatinine (98.91 vs. 90.94 μmol/L, ) and uric acid (356.62 vs. 391.53 μmol/L, ) fluctuated, these changes were not statistically significant. Hemoglobin levels remained stable (138.97 vs. 138.94 g/L, ). These findings suggest that short-term intervention may benefit certain cardiac structural parameters but has limited impact on overall biochemical indicators. Detailed statistical results are provided in Table 3.

3.2.2 Machine learning model construction and performance evaluation

Using the 34 clinical features described above, the FLAML automated machine learning framework was employed to develop multiple classification models to explore associations between multidimensional clinical features, SGLT2 inhibitor treatment status, and serum uric acid variation. These analyses were conducted for hypothesis generation rather than causal inference. Through five-fold cross-validation, the model predicting treatment status achieved an average AUC of 0.92 on the validation set, indicating good discriminatory performance within this retrospective dataset, which should be interpreted cautiously given the limited sample size. (Figure 7). The model predicting changes in serum uric acid levels demonstrated even better performance, with an average AUC of 0.94 (Figure 8), indicating strong discriminative capability.

Figure 7

Figure 8

3.2.3 SHAP-based interpretability analysis

To enhance clinical interpretability, SHAP (SHapley Additive exPlanations) was applied to quantify the contribution of each clinical feature to model predictions (23). As shown in Figure 9, serum creatinine and uric acid were the most important variables for predicting SGLT2 inhibitor treatment. Higher creatinine levels corresponded to elevated SHAP values, indicating that the model was more likely to classify patients with impaired renal function as treated, consistent with the renal protective indication of SGLT2 inhibitors. Elevated uric acid levels also increased the predicted probability of treatment, suggesting that these variables were strongly associated with treatment allocation patterns in this cohort, without implying a causal treatment effect.

Figure 9

For the model predicting changes in serum uric acid, Figure 10 highlights right ventricular outflow tract diameter, uric acid, and interleukin-6 as key contributing factors. Uric acid exhibited a strong positive SHAP contribution, highlighting its importance within the model’s predictive structure. Creatinine also contributed positively, indicating that the models effectively captured renal function-related changes. Additionally, inflammatory marker interleukin-6, metabolic indicator glucose, and cardiac structural parameter left ventricular end-diastolic diameter influenced model predictions, demonstrating the model’s capacity to integrate multi-dimensional clinical information, including inflammation, metabolism, and cardiac structure.

Figure 10

4 Discussion

4.1 HFpEF and HUA comorbidity mechanisms

Heart failure with preserved ejection fraction (HFpEF) and hyperuricemia (HUA) are interrelated conditions that mutually influence prognosis. Our network pharmacology analysis identified 1,381 common targets associated with HFpEF and HUA, enriched in oxidative stress, apoptosis, and inflammation-related pathways, providing a molecular basis for the comorbidity.

GAPDH participates in glycolysis and also functions as a cellular sensor, mediating mitochondrial signaling and apoptosis (24–27). Caspase-3, as a key executioner of apoptosis, mediates cardiomyocyte death and contributes to cardiac dysfunction (20, 28). ICAM1 is involved in inflammatory responses and immune regulation, linking HFpEF and HUA pathophysiology (29, 30). ACE, a key enzyme of RAAS, may promote myocardial remodeling and vascular stiffness in HFpEF (31). DPP4 modulates inflammation and fibrosis via cleavage of multiple substrates, indirectly affecting cardiovascular health (32–34). AGTR1 influences cardiac immune microenvironment and participates in HFpEF development (35).

Approximately 49% of HFpEF patients have hyperuricemia (female SUA > 5.7 mg/dL, male SUA > 7.0 mg/dL). Elevated serum uric acid is associated with advanced HF severity. These data suggest that HUA contributes to HFpEF risk and progression.

4.2 Mechanistic insights of SGLT2 inhibitors

SGLT2 inhibitors may modulate uric acid metabolism and mitigate inflammation induced by urate crystals (12). Molecular docking demonstrated high binding affinities of 11 SGLT2 inhibitors to the core targets (GAPDH, CASP3, ICAM1, ACE, DPP4, AGTR1), with canagliflozin potentially being the most effective component (36, 37).

In addition to GAPDH-mediated anti-inflammatory effects, SGLT2 inhibitors have been reported to decrease ICAM-1 expression in both animal models and patients with diabetes. This reduction likely mitigates endothelial inflammation, protects the vascular system, and contributes to the cardiovascular benefits observed in HFpEF patients. Thus, modulation of ICAM-1 may represent an additional mechanism by which SGLT2 inhibitors exert multi-targeted protective effects.

Clinical studies show that SGLT2 inhibitors reduce serum uric acid in T2DM patients, e.g., empagliflozin decreased SUA by 46 μmol/L (37), dapagliflozin by 39.798 μmol/L (38), ipragliflozin by 11.89-52.3 μmol/L (39), canagliflozin by 23.3 μmol/L (40), and tofogliflozin by 17.226 ± 6.237 μmol/L (41).

4.3 Clinical findings

In our retrospective cohort, SGLT2 inhibitor-treated patients showed improvements in hemoglobin, urea, left ventricular ejection fraction, and serum creatinine compared with controls. Prior meta-analyses of randomized trials have demonstrated that SGLT2 inhibitors improve composite outcomes such as heart failure hospitalization or cardiovascular death in patients with HFpEF, though effects on specific biomarkers vary across studies (42, 43). However, no statistically significant reduction in serum uric acid was observed in our cohort. This finding may differ from prior reports and could be attributable to the limited sample size, retrospective design, and population heterogeneity. Therefore, our clinical findings should be interpreted as exploratory and hypothesis-generating rather than confirmatory. Paired-sample analysis suggested potential right heart remodeling, reflected by an increase in right ventricular outflow tract diameter, supporting the need for longitudinal evaluation.

4.4 Machine learning-based analysis

Using the FLAML automated machine learning framework, predictive models achieved high accuracy for medication status and SUA variation (AUC 0.85 and 0.92). SHAP analysis identified creatinine and uric acid as the most influential predictors, highlighting their importance in SGLT2 inhibitor pharmacodynamics. Additional predictors included interleukin-6 and cardiac structural indices (e.g., RVOT diameter, LVEDD), suggesting multi-pathway benefits through anti-inflammatory effects and cardiac remodeling (44).

4.5 Limitations and future directions

Limitations of this study include a relatively small sample size, particularly for paired pre- and post-intervention data, and a short observation period precluding assessment of long-term efficacy and cardiac remodeling. Additionally, due to the limited paired sample size and retrospective design, multivariable adjusted modeling could be statistically unstable and prone to overfitting. Therefore, the clinical findings are presented as descriptive rather than inferential. Future studies should involve multicenter, large-scale trials with long-term follow-up, complemented by functional validation (e.g., gene knockdown, overexpression, or pathway inhibition) to further elucidate the molecular mechanisms of SGLT2 inhibitors.

5 Conclusion

This study systematically explored the potential mechanisms and clinical effects of SGLT2 inhibitors in HFpEF patients with hyperuricemia using network pharmacology, clinical statistics, and machine learning. A total of 32 core target genes, including GAPDH, CASP3, and ICAM1, were identified, involved in oxidative stress, apoptosis, and inflammatory pathways. Functional enrichment and molecular docking analyses suggest multi-target effects, particularly on MAPK1 and other cardiovascular and metabolic pathways.

Clinically, SGLT2 inhibitor treatment was associated with exploratory improvements in hemoglobin, urea, left ventricular ejection fraction, and serum creatinine, but no statistically significant reduction in serum uric acid was observed. Machine learning models demonstrated exploratory predictive potential, with creatinine and uric acid as key predictive variables. These findings provide a preliminary mechanistic framework for SGLT2 inhibitors in HFpEF-HUA comorbidity and generate hypotheses for future research. Larger, prospective, multicenter studies with long-term follow-up and experimental validation are warranted to clarify the efficacy and mechanisms of SGLT2 inhibitors in this patient population.

References

• 1.

  BorlaugBAReddyYNBraunASorimachiHOmarMPopovicD, et al. Cardiac and metabolic effects of dapagliflozin in heart failure with preserved ejection fraction: the CAMEO-DAPA trial. Circulation. (2023) 148:834–44. 10.1161/CIRCULATIONAHA.123.065134

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 2.

  ViraniSSAlonsoABenjaminEJBittencourtMSCallawayCWCarsonAP, et al. Heart disease and stroke statistics—2020 update: a report from the American Heart Association. Circulation. (2020) 141:e139–e596. 10.1161/CIR.0000000000000757

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 3.

  KosiborodMNBhattASClaggettBLVaduganathanMKulacIJLamCS, et al. Effect of dapagliflozin on health status in patients with preserved or mildly reduced ejection fraction. J Am Coll Cardiol. (2023) 81:460–73. 10.1016/j.jacc.2022.11.006

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 4.

  CampbellPRuttenFHLeeMMHawkinsNMPetrieMC. Heart failure with preserved ejection fraction: everything the clinician needs to know. Lancet. (2024) 403:1083–92. 10.1016/S0140-6736(23)02756-3

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 5.

  OwanTEHodgeDOHergesRMJacobsenSJRogerVLRedfieldMM. Trends in prevalence and outcome of heart failure with preserved ejection fraction. N Engl J Med. (2006) 355:251–9. 10.1056/NEJMoa052256

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 6.

  PackerM. Hyperuricemia and gout reduction by SGLT2 inhibitors in diabetes and heart failure: JACC review topic of the week. J Am Coll Cardiol. (2024) 83:371–81. 10.1016/j.jacc.2023.10.030

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 7.

  ShahSJBorlaugBAKitzmanDWMcCullochADBlaxallBCAgarwalR, et al. Research priorities for heart failure with preserved ejection fraction: national heart, lung, and blood institute working group summary. Circulation. (2020) 141:1001–26. 10.1161/CIRCULATIONAHA.119.041886

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 8.

  LiLChangYLiFYinY. Relationship between serum uric acid levels and uric acid lowering therapy with the prognosis of patients with heart failure with preserved ejection fraction: a meta-analysis. Front Cardiovasc Med. (2024) 11:1403242. 10.3389/fcvm.2024.1403242

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 9.

  AmbrosioGLeiroMGCLundLHCoiroSCardonaAFilippatosG, et al. Serum uric acid and outcomes in patients with chronic heart failure through the whole spectrum of ejection fraction phenotypes: analysis of the ESC-EORP heart failure long-term (HF LT) registry. Eur J Intern Med. (2021) 89:65–75. 10.1016/j.ejim.2021.04.001

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 10.

  CarnicelliAPClareRChiswellKLytleBBjursellMPerlS, et al. Comparison of characteristics and outcomes of patients with heart failure with preserved ejection fraction with versus without hyperuricemia or gout. Am J Cardiol. (2020a) 127:64–72. 10.1016/j.amjcard.2020.04.026

  • CrossRef
  • Google Scholar
  • View reference in article

• 11.

  CarnicelliAPSunJ-LAlhantiBBjursellMPerlSLytleB, et al. Elevated uric acid prevalence and clinical outcomes in patients with heart failure with preserved ejection fraction: insights from relax. Am J Med. (2020b) 133:e716–21. 10.1016/j.amjmed.2020.03.054

  • CrossRef
  • Google Scholar
  • View reference in article

• 12.

  DoehnerWAnkerSDButlerJZannadFFilippatosGCoatsAJ, et al. Uric acid and SGLT2 inhibition with empagliflozin in heart failure with preserved ejection fraction: the EMPEROR-preserved trial. Heart Fail. (2024) 12:2057–70. 10.1016/j.jchf.2024.08.020

  • CrossRef
  • Google Scholar
  • View reference in article

• 13.

  DuLZongYLiHWangQXieLYangB, et al. Hyperuricemia and its related diseases: mechanisms and advances in therapy. Signal Transduct Targeted Ther. (2024) 9:212. 10.1038/s41392-024-01916-y

  • CrossRef
  • Google Scholar
  • View reference in article

• 14.

  ButtJHDochertyKFClaggettBLDesaiASPeterssonMLangkildeAM, et al. Association of dapagliflozin use with clinical outcomes and the introduction of uric acid–lowering therapy and colchicine in patients with heart failure with and without gout: a patient-level pooled meta-analysis of DAPA-HF and DELIVER. JAMA Cardiol. (2023) 8:386–93. 10.1001/jamacardio.2022.5608

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 15.

  FramptonJE. Empagliflozin: a review in symptomatic chronic heart failure: JE frampton et al.. Drugs. (2022) 82:1591–602. 10.1007/s40265-022-01778-0

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 16.

  KhanRHQureshiHAhmedZHassanAAliSAKumarS, et al. Dapagliflozin’s role in heart failure: an overview of clinical trial evidence. Cureus. (2024) 16:e70727. 10.7759/cureus.70727

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 17.

  PescariuSAElagezANallapatiBBratosinFBucurANegruA, et al. Examining the impact of ertugliflozin on cardiovascular outcomes in patients with diabetes and metabolic syndrome: a systematic review of clinical trials. Pharmaceuticals. (2024) 17:929. 10.3390/ph17070929

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 18.

  ButtJHKondoTJhundPSComin-ColetJde BoerRADesaiAS, et al. Dapagliflozin, atrial fibrillation, and heart failure with mildly reduced or preserved ejection fraction in deliver. J Am Coll Cardiol. (2022) 80:1705–17. 10.1016/j.jacc.2022.08.718

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 19.

  ZhouJWangHLiL. Chinese expert consensus on the diagnosis and treatment of heart failure with preserved ejection fraction (hfpef) 2023. Chin Circ J. (2023) 38:375–93. (in Chinese)

  • Google Scholar

• 20.

  KitzmanDWVoorsAAMentzRJLewisGDPerlSMyteR, et al. Verinurad plus allopurinol for heart failure with preserved ejection fraction: the amethyst randomized clinical trial. JAMA Cardiol. (2024) 9:892–900. 10.1001/jamacardio.2024.2435

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 21.

  HopkinsAL. Network pharmacology: the next paradigm in drug discovery. Nat Chem Biol. (2007) 4:682–90. 10.1038/nchembio.118

  • CrossRef
  • Google Scholar

• 22.

  WangCWuQWeimerMZhuE. Flaml: a fast and lightweight automl library. Proc Mach Learn Syst. (2021a) 3:434–47. 10.48550/arXiv.2101.05374

  • CrossRef
  • Google Scholar

• 23.

  LundbergSMNairBVavilalaMSHoribeMEissesMJAdamsT, et al. Explainable machine-learning predictions for the prevention of hypoxaemia during surgery. Nat Biomed Eng. (2018) 2:749–60. 10.1038/s41551-018-0304-0

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 24.

  Del ReDPAmgalanDLinkermannALiuQKitsisRN. Fundamental mechanisms of regulated cell death and implications for heart disease. Physiol Rev. (2019) 99:1765–817. 10.1152/physrev.00022.2018

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 25.

  EleftheriadisTPissasGTsogkaKNikolaouELiakopoulosVStefanidisI. A unifying model of glucotoxicity in human renal proximal tubular epithelial cells and the effect of the SGLT2 inhibitor dapagliflozin. Int Urol Nephrol. (2020) 52:1179–89. 10.1007/s11255-020-02481-3

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 26.

  MolinaCEJacquetEPonienPMuñoz-GuijosaCBaczkóIMaierLS, et al. Identification of optimal reference genes for transcriptomic analyses in normal and diseased human heart. Cardiovasc Res. (2018) 114:247–58. 10.1093/cvr/cvx182

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 27.

  TristanCShahaniNSedlakTWSawaA. The diverse functions of GAPDH: views from different subcellular compartments. Cell Signal. (2011) 23:317–23. 10.1016/j.cellsig.2010.08.003

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 28.

  ZhaoXZhangD-QSongRWangRZhangG. The clinical significance of circulating glucose-regulated protein 78, caspase-3, and C/EBP homologous protein levels in patients with heart failure. Heliyon. (2023) 9:e13436. 10.1016/j.heliyon.2023.e13436

  • CrossRef
  • Google Scholar

• 29.

  CullerKZelnickLRMehtaRCDe SosaGRKansalMRaoPS, et al. The ICAM1 p. K56M variant and risk of heart failure in chronic kidney disease: the chronic renal insufficiency cohort study. Heart. (2025) 111:441–4. 10.1136/heartjnl-2024-325205

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 30.

  InzucchiSEZinmanBFitchettDWannerCFerranniniESchumacherM, et al. How does empagliflozin reduce cardiovascular mortality? insights from a mediation analysis of the EMPA-REG OUTCOME trial. Diabetes Care. (2018) 41:356–63. 10.2337/dc17-1096

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 31.

  LytvynYKimuraKPeterNLaiVTseJChamL, et al. Renal and vascular effects of combined SGLT2 and angiotensin-converting enzyme inhibition. Circulation. (2022) 146:450–62. 10.1161/CIRCULATIONAHA.122.059150

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 32.

  AbdelzaherWYRofaeilRRAliDMEAttyaME. Protective effect of dipeptidyl peptidase-4 inhibitors in testicular torsion/detorsion in rats: a possible role of HIF-1 and nitric oxide. Naunyn Schmiedebergs Arch Pharmacol. (2020) 393:603–14. 10.1007/s00210-019-01765-5

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 33.

  CarluccioECoiroSAmbrosioG. Unraveling the relationship between serum uric acid levels and cardiovascular risk. Int J Cardiol. (2018) 253:174–5. 10.1016/j.ijcard.2017.11.035

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 34.

  WangXKeJZhuY-j.CaoBYinR-l.WangY, et al. Dipeptidyl peptidase-4 (DPP4) inhibitor sitagliptin alleviates liver inflammation of diabetic mice by acting as a ROS scavenger and inhibiting the NFB pathway. Cell Death Discov. (2021b) 7:236. 10.1038/s41420-021-00625-7

  • CrossRef
  • Google Scholar

• 35.

  LiGZhouQXieMZhaoBZhangKLuoY, et al. Identification of ageing-associated gene signatures in heart failure with preserved ejection fraction by integrated bioinformatics analysis and machine learning. Genes Dis. (2025) 12:101478. 10.1016/j.gendis.2024.101478

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 36.

  HuXYangYHuXJiaXLiuHWeiM, et al. Effects of sodium-glucose cotransporter 2 inhibitors on serum uric acid in patients with type 2 diabetes mellitus: a systematic review and network meta-analysis. Diabetes Obes Metab. (2022) 24:228–38. 10.1111/dom.14570

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 37.

  KadowakiTHanedaMInagakiNTerauchiYTaniguchiAKoiwaiK, et al. Empagliflozin monotherapy in Japanese patients with type 2 diabetes mellitus: a randomized, 12-week, double-blind, placebo-controlled, phase II trial. Adv Ther. (2014) 31:621–38. 10.1007/s12325-014-0126-8

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 38.

  JiLMaJLiHMansfieldTAT’joenCLIqbalN, et al. Dapagliflozin as monotherapy in drug-naive Asian patients with type 2 diabetes mellitus: a randomized, blinded, prospective phase III study. Clin Ther. (2014) 36:84–100. 10.1016/j.clinthera.2013.11.002

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 39.

  KashiwagiAKazutaKTakinamiYYoshidaSUtsunoANagaseI. Ipragliflozin improves glycemic control in Japanese patients with type 2 diabetes mellitus: the BRIGHTEN study: BRIGHTEN: double-b lind r andomized study of i pra g liflozin to sh ow its efficacy as monotherapy in T 2DM pati en ts. Diabetol Int. (2015) 6:8–18. 10.1007/s13340-014-0164-0

  • CrossRef
  • Google Scholar

• 40.

  LiJBadveSVZhouZRodgersADayROhR, et al. The effects of canagliflozin on gout in type 2 diabetes: a post-hoc analysis of the CANVAS program. Lancet Rheumatol. (2019) 1:e220–8. 10.1016/S2665-9913(19)30078-5

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 41.

  TerauchiYTamuraMSendaMGunjiRKakuK. Efficacy and safety of tofogliflozin in Japanese patients with type 2 diabetes mellitus with inadequate glycaemic control on insulin therapy (J-STEP/INS): results of a 16-week randomized, double-blind, placebo-controlled multicentre trial. Diabetes Obes Metab. (2017) 19:1397–407. 10.1111/dom.12957

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 42.

  YangDZhangYYanJLiuMAnF. SGLT-2 inhibitors on prognosis and health-related quality of life in patients with heart failure and preserved ejection fraction: a systematic review and meta-analysis. Front Cardiovasc Med. (2022) 9:942125. 10.3389/fcvm.2022.942125

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 43.

  ZhouHPengWLiFWangYWangBDingY, et al. Effect of sodium-glucose cotransporter 2 inhibitors for heart failure with preserved ejection fraction: a systematic review and meta-analysis of randomized clinical trials. Front Cardiovasc Med. (2022) 9:875327. 10.3389/fcvm.2022.875327

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 44.

  YangQGaoJWangT-YDingJ-CHuP-F. Integration of rna-seq and machine learning identifies hub genes for empagliflozin benefitable heart failure with reduced ejection fraction. J Inflamm Res. (2023) 16:4733–49. 10.2147/JIR.S429096

  • Pubmed Abstract
  • CrossRef
  • Google Scholar