A systematic search was executed through comprehensively searching in online databases PubMed and Web of Science for studies published between January 2007 and December 2024. The searches were conducted by information specialist MJ Zhao. The following keywords were used in the search: (flavonoids) and (hypouricemic OR hyperuricemia OR hyperuricemia or gout or hyperuricemia) AND (mice). Initial studies were imported to the EndNote 20 (Clarivate Analytics, Philadelphia, PA, USA) software. After removing duplicate studies, two independent researchers (MJ Zhao and Q Xiao) screened the titles and abstracts of the remaining articles. Studies that did not meet the inclusion criteria were excluded. When necessary, the full texts of the remaining publications were evaluated. Any disagreements regarding the inclusion or exclusion of studies were resolved through discussion among the authors until consensus was reached.

The criteria for included animal studies were as follows:

(1) original research that presented original data;

(2) study performed in male mice weighing 18 to 30 g;

(3) disease of interest (hyperuricemia induced by potassium oxonate);

(4) intervention of interest (flavonoids that were pure natural products);

The criteria for excluded animal studies were as follows:

(1) in vitro or ex vivo models;

(2) complex models of HUA and other disorders;

(3) studies not using one of pure compound as an intervention group;

(4) not interested intervention;

(5) reviews, conference abstracts, editorials, clinical trials, case reports and incomplete articles;

(6) not an English-language publication.

Two authors (JJ Xu and Q Xiao) independently extracted study characteristics The extracted information included the first author’s name, year of publication, animal details (species, weight, sex, strain, control group), and reagent characteristics (flavonoid type, sample size, study duration, intervention dose, potassium oxonate dose). Standard deviations (SD) and means for intervention and control groups were also recorded. If study data were only available as graphs, Origin 2025 (OriginLab Corporation, Northampton, MA, USA) software was used for extraction. Any discrepancies in measurements were resolved through discussion between the two authors.

The quality assessment was conducted independently by two researchers (Q Xiao and MJ Zhao) to ensure inter-rater reliability and to minimize the potential bias. Any discrepancies in the risk of bias assessment were resolved through consensus discussions, ensuring a standardized and objective evaluation of study quality. The methodological rigor of the included studies was critically appraised using SYRCLE’s Risk of Bias tool, a validated instrument specifically designed for assessing the risk of bias in animal intervention studies.

XOD was selected as the core target for molecular docking analysis. The 3D structures of the active compounds were retrieved from PubChem, and the crystal structure of XOD (PDB ID: 3NVY) was downloaded from the PDB database (RCSB Protein Data Bank, New Brunswick, NJ, USA). Solvent molecules were removed using PyMOL, and the target protein was prepared for hydrogenation and charge assignment using MGLTools 1.5.7 (The Scripps Research Institute, La Jolla, CA, USA). Both the protein and the 18 compounds were saved as “pdbqt” files, and the docking box was configured. Molecular docking was performed using Autodock suite 4.2.6 (The Scripps Research Institute, La Jolla, CA, USA), with binding affinities calculated in kcal/mol. Binding affinities less than –5.00 kcal/mol were considered indicate strong binding interactions.

To assess complex stability, 200 ns atomistic MD simulations were conducted with GROMACS 2020.6 (GROMACS Development Team, Stockholm, Sweden) for the three highest-affinity representatives (chalcones, flavones, flavonols) bound to XOD. The Amber99SB-ILDN force-field was used for the protein; ligands were parameterised with GAFF2 and AM1-BCC charges. Each system was solvated in a cubic TIP3P water box (minimum 1.2 nm buffer) and neutralised with Na⁺/Cl^– ions to 0.15 M. Following energy minimisation (steepest descent), 1-ns NVT (300 K, V-rescale) and 1-ns NPT (1 bar, Parrinello–Rahman) equilibrations were performed with position restraints on heavy atoms. Production runs used 2-fs time steps, LINCS constraints, and PME electrostatics. Trajectories were saved every 10 ps for subsequent analysis of root-mean-square deviation (RMSD), root-mean-square fluctuation (RMSF), solvent-accessible surface area (SASA), and hydrogen-bond occupancy.

Standardised mean differences (SMD, Hedges’ g) with 95% confidence intervals (CI) were pooled under a random-effects model (DerSimonian–Laird). Heterogeneity was quantified with I² and $\tau$²; sources were explored by subgroup (flavonoid subclass, mouse strain) and meta-regression (dose, duration). Publication bias was appraised with funnel plots, Egger’s regression, and the trim-and-fill method. All analyses were conducted in R 4.3.1 (metafor package; R Foundation for Statistical Computing, Vienna, Austria); two-tailed p 0.05 indicated significance.

Database searching retrieved 319 records; 76 duplicates were removed. After title/abstract screening, 222 articles were excluded (171 non-primary studies, non-matching models, or rat protocols; 51 non-pure compounds or irrelevant interventions). Twenty-one studies met the inclusion criteria and were entered into quantitative synthesis (Fig. 1).

Fig. 1.

PRISMA flowchart of literature search and study selection (21 studies included). PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses.

The 21 articles comprised 550 male mice. Detailed intervention parameters and animal characteristics are summarised in Table 1 (Ref. [15, 16, 17, 18, 19, 20, 21, 22, 23, 24]). Nine strains were represented: Kun-Ming (n = 9 studies), ICR (6), C57BL/6 (3), Albino Swiss (2) and BALB/c (1). Hyperuricemia was induced by potassium oxonate (PO) at 250–300 mg kg^-1 day^-1 for 1–28 days. Flavonoid interventions included chalcones (6 datasets), flavonols (18), flavones (7) and others (8); doses ranged 0.5–700 mg kg^-1 day^-1. All studies employed allopurinol (5–10 mg kg^-1) as positive control. Chemical structures and subclass assignments of the investigated flavonoids are provided in Table 2.

Table 2. Study characteristics of the flavonoids on hyperuricemia mouse model.

Subgroup		No.	Name	R1	R2	R3	R4	R5	R6
Chalcones		1	3,5,2′,4′-tetrahydroxychalcone	-H	-OH	-H	-OH
		2	Okanin	-OH	-OH	-OH	-H
		3	Phloretin
Flavones		4	Apigenin	-H	-OH	-H	-H	-OH
		5	Baicalein	-OH	-OH	-H	-H	-H
		6	Luteolin	-H	-OH	-H	-OH	-OH
		7	Luteolin-4′-O-glucoside	-H	-OH	-H	-OH	-OGlc
		8	Scutellarin	-OH	-OGlu A	-H	-OH	-H
		9	Vitexin	-H	-OH	-OGlc	-OH	-H
Flavonols		10	Astilbin	-ORha	-H	-OH	-OH	-H	-OH
		11	Hesperidin	-H	-H	-OH	-OCH3	-H	-OGlc-(6$\to$1)-Rha
		12	Kaempferol	-OH	-H	-H	-OH	-H	-OH
		13	Kaempferol-3-O-sophoroside	-Osophoroside	-H	-H	-OH	-H	-OH
		14	Sodium kaempferol-3′-sulfonate	-OH	-H	-H	-OH	-SO3Na	-OH
		15	Morin	-OH	-OH	-H	-OH	-H	-OH
		16	Quercetin	-OGlc	-H	-OH	-OH	-H	-OH
		17	Myricetin	-OH	-H	-OH	-OH	-OH	-OH
		18	Rutin	-OGlc-(6$\to$1)-Rha	-H	-OH	-OH	-H	-OH
		19	6,8,3′,4′-tetrahydroxyflavanone-7-C-$\beta$-D-glucopyranoside
		20	6,8,4′-trihydroxyflavanone-7-C-$\beta$-D-glucopyranoside
Others	Flavan-3-ol	21	(-)-2,3-cis-3,4-cis-3,3′4,4′,7,8-hexahydroxyflavan	-OH
		22	(-)-2,3-cis-3,4-cis-4′-methoxy-3,3′,4,7,8-pentahydroxyflavan	-OMe
	Isoflavones	23	Genistein
	Anthocyanidins	24	Anthocyanin
		25	Epigallocatechin-3-gallate
		26	Epiphyllocoumarin-3-O-$\beta$-D-allopyranoside
		27	2′,4′-dihydroxychalcone-(4-O-7′′)-8′′,4′′′-dihydroxyflavanone
		28	Puerarin

As the final product of purine metabolism, UA is crucial for diagnosing HUA and understanding the pathogenesis of gout and rheumatoid arthritis [25]. Meta-analysis of 21 comparisons revealed that revealed that flavonoid interventions significantly reduced serum UA levels compared to control groups (SMD = –2.22, 95% CI [–2.80, –1.64], I² = 79.7%, p 0.001; Fig. 2A).

Fig. 2.

Forest plots of flavonoids effects on (A) serum uric acid and (B) xanthine oxidase activity versus control. Pooled weighted mean differences (random-effects) are shown with 95% CIs. flavonoid intervention significantly reduced both outcomes (A: WMD = –2.22, 95% CI –2.80 to –1.64, I² = 79.7%, p 0.001; B: WMD = –1.79, 95% CI –2.50 to –1.08, I² = 76.5%, p 0.001). Marker size reflects study weight; vertical dashed line denotes no effect. DL, DerSimonian–Laird; UA, uric acid; XOD, xanthine oxidase; WMD, Weighted Mean Difference.

XOD is the rate-limiting enzyme in UA production; its inhibition offers a direct index of hypouricemic efficacy [26]. Pooled data from 21 studies demonstrated that flavonoids interventions significantly reduced serum XOD activity compared to control groups (SMD = –1.79, 95% CI [–2.50, –1.08], I² = 76.5%, p 0.001; Fig. 2B). These results indicated that flavonoids can effectively lower UA synthesis by reducing XOD activity.

Subgroup analyses classified flavonoids into five categories—flavonols, flavanones, flavones, chalcones, and others—and consistently revealed significant serum-UA reductions relative to controls (random-effects; 21 studies). Flavonols: SMD = –1.99, 95% CI [–2.77, –1.22], p 0.001, I² = 72.0%; Flavones: SMD = –2.99, 95% CI [–3.76, –2.22], p 0.001, I² = 11.7%; Chalcones: SMD = –3.63, 95% CI [–5.68, –1.58], p = 0.001, I² = 88.0%; Others: SMD = –1.27, 95% CI [–2.52, –0.03], p = 0.046, I² = 84.6%. Parallel decreases in XOD activity were observed for all subclasses: Flavonols: SMD = –2.14, 95% CI [–3.26, –1.03], p 0.001, I² = 74.9%; Flavones: SMD = –2.59, 95% CI [–5.09, –0.08], p = 0.043, I² = 84.3%; Chalcones: SMD = –1.83, 95% CI [–3.48, –0.18], p = 0.030, I² = 86.6%; Others: SMD = –0.32, 95% CI [–1.34, 0.68], p = 0.525, I² = 40.6%.

When data were stratified by mouse strain, serum UA was significantly lowered in ICR (SMD = –2.32, 95% CI [–3.23, –1.42], p 0.001, I² = 72.6%), Kun-Ming (SMD = –3.27, 95% CI [–4.88, –1.65], p 0.001, I² = 91.2%), Albino Swiss (SMD = –2.56, 95% CI [–3.50, –1.63], p 0.001, I² = 0%) and C57BL/6 (SMD = –1.19, 95% CI [–2.28, –0.12], p = 0.030, I² = 62.5%) mice, whereas BALB/c mice exhibited a non–significant change (SMD = –0.61, 95% CI [–1.30, 0.09], p = 0.088, I² = 0%). Comparable patterns were seen for XOD activity: ICR SMD = –1.77, 95% CI [–2.87, –0.67], p = 0.002, I² = 70.0%; Kun-Ming: SMD = –2.87, 95% CI [–3.72, –2.02], p 0.001, I² = 40.8%; Albino Swiss: SMD = –8.56, 95% CI [–12.16, –4.95], p 0.001, I² = 0%), with the exception of BALB/c (SMD = 0.02, 95% CI [–0.66, 0.71], p = 0.947, I² = 0%). Full details are presented in Table 3.

As illustrated in Table 4 (Ref. [15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37]), the baseline characteristics that might influence outcome indicators were inadequately described in the majority of studies (19/21). Specific methodological facets—including sequence generation, allocation concealment, random outcome assessment, and blinding of caregivers or investigators—were either sparsely reported or not documented at all, leading to an “unclear” risk-of-bias classification for these domains. Nevertheless, 17 investigations (80.9%) explicitly stated that experimental animals were housed in identical macro-environmental conditions (temperature, humidity, light–dark cycle, and bedding), allowing us to assign a “low risk” rating to the item of random housing. Furthermore, all included publications provided evidence that baseline body weight, serum uric acid, and xanthine oxidase activity did not differ between treatment and control arms; consequently, the risk of confounding attrition was considered “low” across the entire data set. Incomplete outcome data were adequately addressed in every study (low risk), and neither selective reporting nor other sources of bias (e.g., early termination, unit-of-analysis errors) were detected (low risk for all trials).

As shown in Fig. 3, funnel-plot appraisal of serum UA and XOD outcomes revealed marked asymmetry, with numerous small-effect studies lying outside the pseudo-confidence limits, strongly suggestive of a small-study effect. Egger’s linear regression corroborated this impression, yielding statistically significant intercepts for both UA (t = 3.11, p = 0.007) and XOD (t = 2.89, p = 0.012), thereby confirming the presence of publication bias. Such bias could, in principle, inflate or deflate the pooled effect estimates and compromise the validity of any therapeutic inference. To gauge the robustness of the summary statistics we therefore implemented the trim-and-fill algorithm (Lönnér estimator) under a random-effects model. After two iterative cycles no additional trial was imputed and the adjusted pooled SMD for UA remained –2.20 (95% CI –2.79 to –1.63), virtually identical to the original –2.22, indicating that the asymmetry had negligible influence on the overall estimate. Complementary Harbord and Peters tests likewise returned non-significant bias coefficients (both p $>$ 0.10), reinforcing the conclusion that the meta-analytic summary is stable despite the observed funnel-plot distortion.

Fig. 3.

Funnel plots and Egger’s regression-based publication bias tests for the 21 included studies. (A) Funnel plot for serum UA; (B) funnel plot for XOD activity. Open circles represent individual comparisons; diagonal dashed lines indicate the 95% pseudo-confidence limits. Egger’s linear regression intercept is displayed in the lower left corner of each panel, confirming small-study effects for both outcomes. (C) Egger’s linear regression plot for serum uric acid (UA); (D) Egger’s linear regression plot for xanthine oxidase (XOD). Blue scatter points in the plots represent individual studies included in the meta-analysis; pink diagonal lines denote the Egger’s linear regression fitting lines, and black horizontal lines indicate the reference lines for pooled effect sizes. The red vertical lines with error bars correspond to the 95% confidence intervals (95% CI for intercept) of the intercepts.

We investigated the interactions between drugs and proteins via molecular docking, and in this study, 28 flavonoid compounds were subjected to molecular docking with xanthine oxidase (XOD), as shown in Fig. 4 among chalcones-type flavonoids, 3,5,2^′,4^′-tetrahydroxychalcone, Okanin, and Phloretin all exhibited a binding affinity of –8.9 kcal/mol; for flavones-type flavonoids, the binding affinities of Apigenin, Baicalein, Luteolin, Luteolin-4^′-O-glucoside, Scutellarin, and Vitexin were –9.2, –9.8, –9.5, –10.3, –10.2, and –8.8 kcal/mol, respectively; regarding flavonols-type flavonoids, Astilbin, Hesperidin, Sodium kaempferol-3^′-sulfonate, Morin, Quercetin, Myricetin, Rutin, 6,8,3^′,4^′-tetrahydroxyflavanone-7-C-$\beta$-D-glucopyranoside, and 6,8,4^′-trihydroxyflavanone -7-C-$\beta$-D-glucopyranoside had binding affinities of –9.4, –10.6, –10.4, –8.6, –9.9, –9.5, –9.8, –8.4, and –8.8 kcal/mol, respectively; and for other flavonoid compounds, (-)-2,3-cis-3,4-cis-3,3^′,4,4^′,7,8-hexahydroxyflavan, (-)-2,3-cis-3,4-cis-4^′-methoxy-3,3^′,4,7,8-pentahydroxyflavan, Genistein, Anthocyanin, Epigallocatechin-3-gallate, Epiphyllocoumarin-3-O-$\beta$-D-allopyranoside, 2^′,4^′-dihydroxychalcone-(4-O-7^′′)-8^′′,4^′′′-dihydroxyflavanone, and Puerarin showed binding affinities of –8.7, –8.9, –9.1, –8.9, –9.9, –10.8, –10.9, and –10.1 kcal/mol, respectively, with detailed information presented in Table 5. Molecular docking results indicated that flavonoid compounds interact tightly with XOD, providing strong evidence for our research.

Fig. 4.

Visualization of molecular docking between active ingredients of flavonoids and XOD. (A1–A3) show the visualization of molecular docking between chalcones-type flavonoids and XOD; (B1–B6) represent the visualization of molecular docking between flavone-type flavonoids and XOD; (C1–C11) display the visualization of molecular docking between flavonols-type flavonoids and XOD; and (C12–C19) present the visualization of molecular docking between other types of flavonoids and XOD.

To quantitatively compare the stability of XOD-flavones, XOD-flavonols and XOD-chalcones complexes, 200-ns GROMACS trajectories were analysed by RMSD, RMSF, SASA and hydrogen-bond metrics. RMSD records (Fig. 5A–C) show that the flavones adduct equilibrated within 20 ns and maintained a mean value of 0.22 $\pm$ 0.02 nm over the final 50 ns—the lowest and least variable among the three systems—whereas the flavonols complex averaged 0.27 $\pm$ 0.03 nm with wider oscillations, and the chalcones complex exhibited a continuous upward drift to 0.34 $\pm$ 0.04 nm, indicating the weakest stability. RMSF profiles (Fig. 5D–F) reveal that flavones preserved the most rigid backbone (mean residue fluctuation 0.25 $\pm$ 0.08 nm, no peak $>$0.5 nm) with highly overlapping chains, while chalcones displayed enhanced flexibility (mean RMSF 0.58 $\pm$ 0.15 nm) and divergent inter-chain motions, and flavonols presented anomalously high local mobility (mean 0.42 $\pm$ 0.11 nm, several regions $>$2 nm). SASA values (Fig. 5G–I) remained compact for flavones (860 $\pm$ 10 nm²) with minimal drift, whereas chalcones showed a progressive increase from 860 to 920 nm², implying gradual structural loosening, and flavonols exhibited large-amplitude oscillations (840–920 nm²) consistent with intermittent pocket opening. Hydrogen-bond analysis (Fig. 5J–L) identified one persistent bond (78% occupancy) for flavones throughout the 200 ns trajectory, while flavonols formed 0–8 transient hydrogen bonds that fluctuated markedly, and chalcones lacked any enduring interaction. Collectively, these quantitative data establish the stability ranking flavones $>$ flavonols $>$ chalcones, corroborating the higher binding affinity observed in docking studies and providing atomistic evidence for the superior inhibitory potential of flavones-type flavonoids against xanthine oxidase.

Fig. 5.

Molecular dynamics simulation between active ingredients of flavonoids and XOD. The figure presents a comprehensive analysis of the molecular dynamics simulation trajectory over 200 ns. (A–C) RMSD of the protein backbone and ligand atoms over time. (D–F) RMSF per residue for chain A and chain B of the XOD dimer. (G–I) Evolution of key interaction parameters (hydrogen bonds, hydrophobic contacts) throughout the simulation. (J–L) Analysis of binding free energy contributions and the interaction frequency of key residues. The simulation results demonstrate the stability and binding mode of flavonoids within the active site of XOD. RMSD, root-mean-square deviation; RMSF, root-mean-square fluctuation.

In the included studies, most flavonoid subclasses reduced xanthine production by inhibiting XOD activity, resulting in decreasing UA synthesis [16, 20, 27, 28, 29, 30, 31, 32]. Chalcones exhibited a dose-dependent effect, with higher concentrations yielding better UA-lowering effect [38]. XOD is present in both serum and liver, but changes in UA levels were not parallel to changes in hepatic XOD levels [15]. Another study showed that serum XOD was correlated with uric acid levels, but not with hepatic XOD levels. Conversely, serum XOD levels were correlated with UA levels but not with hepatic XOD levels [18].

Rutin, Morin, Astilbin, anthocyanins significantly downregulated the mRNA and protein levels of GLUT9 and URAT1, and Morin, Astilbin downregulated mRNA and protein levels of OCT1, OCT2 in HUA mice. Anthocyanins, Phloretin downregulated the expression of hepatic XOD, caspase-1, TNF-$\alpha$ and IL-1$\beta$ [18, 21, 23, 33, 34]. Moreover, astilbin also regulated NLRP3/NF-$\kappa$B pathway to suppress oxidative stress [23]. Baicalein alleviated oxidative stress in HUA mice [24]. Luteolin and luteolin-4^′-O-glucoside decreased the levels of IL-1$\beta$ and TNF-$\alpha$ to improve the symptoms of inflammation [35]. Phloretin suppressed the NLPR3 pathway under LPS or UA stimulation in HK-2 cells [36]. Scutellarin might regulated CCN1 on NLRP3 inflammasome activation [37]. An overview of the molecular mechanisms underlying the regulation of hyperuricemia by flavonoids is illustrated in Fig. 6.

Fig. 6.

An overview of the molecular mechanisms underlying the regulation of hyperuricemia by flavonoids. Created in BioRender. Panner Selvam, M. (2025) https://BioRender.com/mg1p7ut. The schematic diagram illustrates the multi-targeted mechanisms by which various flavonoids (e.g., Luteolin, Morin) ameliorate hyperuricemia. Key actions include the direct inhibition of XOD activity, reduction of oxidative stress, and suppression of inflammatory signaling pathways (e.g., NLRP3/NF-$\kappa$B, TNF-$\alpha$, IL-1$\beta$). Flavonoids also modulate the expression of urate transporters (e.g., GLUT9, URAT1, OCT1/2) and influence critical cellular processes such as autophagy and cytoskeleton integrity, primarily within the liver and gut. NLRP3, NOD-like receptor pyrin domain-containing 3; NF-$\kappa$B, nuclear factor kappa-B; TNF-$\alpha$, tumor necrosis factor-alpha; IL-1$\beta$, interleukin-1$\beta$; GLUT9, glucose transporter 9; URAT1, uric acid transporter 1; OCT1/2, Organic cation transporter 1/2.