Nonlinear Association Between Serum Uric Acid and Incident Stroke in Patients with Diabetes: A Two-Centre Cohort Study

Abstract

Background

Serum uric acid (SUA) has been linked to vascular injury, but its impact on stroke risk in diabetic patients—a population inherently prone to cerebrovascular events—remains uncertain, and whether a threshold effect exists is unclear.

Methods

We included 3718 diabetic patients from two clinical centers. Multivariable-adjusted Cox proportional hazards models with stepwise covariate adjustment were used to assess associations between SUA and overall stroke and its subtypes. Kaplan–Meier curves estimated cumulative incidence across SUA quartiles. Restricted cubic spline regression examined dose–response relationships and potential threshold effects, and sex-stratified analyses were also performed.

Results

Elevated SUA was significantly associated with increased risks of stroke and subtypes. The highest SUA quartile showed a significantly higher cumulative risk. RCS revealed threshold effects: risks of stroke, ischemic stroke, and hemorrhagic stroke increased when SUA exceeded 335, 330, and 340 μmol/L, respectively. Sex-stratified analyses yielded consistent findings, with thresholds at 335 μmol/L for males and 350 μmol/L for females.

Conclusion

In conclusion, this observational study confirms a significant association between SUA and stroke risk in diabetic patients with a threshold effect, suggesting potential clinical value for monitoring and urate-lowering therapy. However, sex-specific threshold findings should be interpreted cautiously as hypothesis-generating, and prospective randomized controlled trials are needed for validation.

Introduction

Stroke is a leading cause of death worldwide, accounting for approximately the second most common cause of global mortality.^1–3 Based on pathophysiological characteristics, stroke is classified into ischemic and hemorrhagic subtypes.^4,5 Ischemic stroke (IS) is associated with a high risk of long-term disability, whereas hemorrhagic stroke (HS) tends to have a higher case-fatality rate.^6–8 Hypertension and diabetes mellitus (DM) are well-established major risk factors for stroke.^9,10 With improvements in living standards, increased life expectancy, and evolving lifestyle patterns, the prevalence of hypertension and diabetes has been rising steadily, which is anticipated to drive a parallel increase in the future incidence of stroke, thereby imposing a substantial burden on stroke prevention and management.^2,11

DM, a highly prevalent chronic metabolic disorder particularly common among middle-aged and older adults, is characterized by hyperglycemia that subsequently leads to vascular damage, renal impairment, and neurological decline, contributing to a wide range of complications, including cardiovascular diseases (CVD) such as stroke and coronary heart disease (CHD), as well as chronic kidney disease and neuropathy.^12–14 Moreover, diabetes is not only a major contributor to stroke but also a leading cause of death; therefore, stroke prevention in diabetic patients warrants particular attention.^15,16 Traditional approaches to stroke prevention in diabetic populations have predominantly focused on glycemic control, blood pressure (BP) management, and lipid-lowering therapy when indicated.^17–20 However, a considerable proportion of patients still experience stroke events despite achieving satisfactory control of these conventional risk factors. This residual risk suggests that other, as yet unidentified, factors may contribute to the occurrence of stroke in patients with diabetes.

Recently, serum uric acid (SUA) has emerged as a potential player in vascular injury and CVD.^21–23 Experimental studies have demonstrated that elevated SUA levels can trigger inflammatory responses, impair vascular endothelial function, and promote lipid accumulation, which may significantly influence plaque formation, vascular damage, and even plaque rupture.^21,22,24 Consistent with these mechanistic insights, accumulating clinical evidence has indicated that elevated SUA is associated with an increased risk of CVD, with Mendelian randomization analyses further supporting a causal relationship.^23,25,26 In addition, SUA has been shown to significantly improve the predictive performance of cardiovascular risk models, underscoring its robust association with vascular injury.²⁷ Notably, a recent study also confirmed that SUA is associated with incident diabetes, a process mediated by inflammation, providing further support for the role of SUA in the development and progression of diabetes and its vascular complications.²⁸

Clinically, several consensus documents have proposed SUA cut‑offs for cardiovascular risk stratification—for instance, thresholds around 360 µmol/L (6 mg/dL) have been suggested in hyperuricemia and gout management guidelines.^29–31 However, data specifically defining such thresholds for stroke risk in diabetic patients remain limited. Therefore, the present study aimed to explore whether a similar threshold effect exists in this high‑risk population. Given that renal function is a major determinant of SUA levels and a potent confounder of the association between SUA and vascular outcomes, we included estimated glomerular filtration rate (eGFR) as a covariate in our multivariable Cox models to adjust for its potential confounding effect.

To address this, we leveraged data from two independent centers to investigate the association between baseline SUA and subsequent stroke events—both overall and by subtype—among diabetic patients, and to examine whether this relationship exhibits a nonlinear pattern. Our findings are expected to broaden the framework for stroke prevention in diabetes and may inform future research on the potential role of urate-lowering strategies in this setting.

Materials and Methods

Study Population Selection

A total of 5,617 patients with DM, including both type 1 and type 2 diabetes (pooled together for analysis) were initially enrolled from the two medical center–based diabetes cohorts, of whom 5,006 had complete follow-up data. During the screening process, 714 patients were excluded due to a history of stroke at baseline, and 255 were excluded owing to missing SUA data. In addition, we sequentially excluded those who were taking medications that could affect UA metabolism at baseline, as well as patients with severe heart failure, cancer, or severe renal insufficiency. After applying these stepwise exclusion criteria, a total of 3,718 patients were ultimately included in the final analysis (Figure 1).

Figure 1.

Screening of the Study Population.

Prior to the enrollment of any participant, the study protocol was thoroughly reviewed and approved by the institutional review boards of the two participating centers: The Affiliated Huaian No.1 People’s Hospital of Nanjing Medical University (KY-2024-066-01) and Changzhi People’s Hospital (N-CZ2019-0217). All procedures were performed in accordance with the ethical principles laid down in the Declaration of Helsinki. Every patient provided written informed consent after receiving a comprehensive explanation of the study’s purpose and procedures.

Covariate Collection

Data for this study were derived from cohort databases established at two independent centers. Information on covariates was collected through a combination of electronic medical records, Medicare/insurance claims data, routine outpatient visits, scheduled in-person interviews, and telephone follow-ups. Baseline demographic characteristics, including sex, age, height, weight, and body mass index (BMI), were recorded for each participant. BP measurements, comprising both systolic and diastolic blood pressure (DBP), as well as fasting plasma glucose (FPG) levels, were obtained by trained nurses using standardized protocols and calibrated equipment.

Laboratory parameters were centrally measured and included alanine aminotransferase (ALT), aspartate aminotransferase (AST), total cholesterol (TC), triglycerides, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), hemoglobin A1c (HbA1c), and SUA. All biochemical assays were performed in accredited clinical laboratories according to routine procedures.

Information on pre-existing medical conditions was collected, with particular emphasis on documented histories of hypertension, CHD, and hyperlipidemia. Detailed definitions of the diseases are provided in the Supplementary Materials. Medication use was systematically ascertained, focusing on lipid-lowering agents and glucose-lowering therapies, including both oral antidiabetic agents and insulin.

Study Outcome

The primary outcome of this study was the first incident stroke event, including both ischemic and hemorrhagic subtypes. All suspected stroke cases were diagnosed based on neuroimaging findings (cranial computed tomography [CT] or magnetic resonance imaging [MRI]) in conjunction with characteristic clinical manifestations, and the final diagnosis was adjudicated by two qualified neurologists; any disagreement was resolved through consensus. During the follow-up period, 57 deaths from any cause were recorded. Follow-up commenced on the date of enrollment (baseline). The end of follow-up for each participant was defined as the earliest occurrence of: (1) first documented stroke, (2) death from any cause, or (3) the pre-specified administrative cutoff date (May 31, 2026). For participants who remained stroke-free and alive, data were censored at the date of last contact or at the administrative cutoff, whichever occurred first.

Statistical Analysis

Based on SUA quartile stratification, the study population was divided into four groups, and cross‑group comparisons were made for all baseline variables. Depending on their distribution, continuous variables were expressed as means ± SD or medians (IQR), while categorical data were reported as frequencies with percentages.

As a preliminary step, bar charts were used to display the crude incidence rates for overall stroke, IS, and HS across the quartile groups. First, we made the proportional hazards assumption, which was fully consistent with the equal proportion risk (Figure S1). To formally test the SUA‑stroke association, a series of Cox regression models with progressive covariate adjustment (four models) were employed. KM methodology was adopted to compare the cumulative event probabilities among the quartile categories. RCS modeling was subsequently performed to depict the dose‑response curve and to pinpoint any nonlinear threshold points. Once such a point was identified, we applied a piecewise Cox approach to estimate the risk differential above and below that level.

Subgroup exploration was carried out according to sex and SUA normality, repeating the RCS and threshold estimation in each subset to assess whether the observed pattern varied across these factors. Detailed information on the statistical analyses is available in the Supplementary Materials.

R (version 4.4) served as the analytical platform, with statistical significance predefined as a two‑sided alpha of 0.05.

Results

Basic Characteristics of Participants

Overall, 3,718 diabetic patients were enrolled, with a median follow-up of 4.67 years. Over this period, 892 first-ever stroke events occurred—607 classified as IS and 285 as HS (Table S1).

Based on SUA quartile cutoffs, the cohort was divided into four groups, and Table 1 summarizes their baseline features. Patients in the top SUA quartiles tended to be male, had a higher BMI, and showed greater proportions of current smokers and alcohol drinkers compared with those in the lower quartiles. SBP and DBP also rose progressively with increasing SUA quartiles.

Table 1.

Baseline Characteristics of the Four Groups Stratified by SUA Quartiles

Characteristic	Overall	Q1	Q2	Q3	Q4	P value
Numbers	3718	929	929	930	930
Sex						<0.001
Male	2064 (55.51%)	388 (41.77%)	480 (51.67%)	524 (56.34%)	672 (72.26%)
Female	1654 (44.49%)	541 (58.23%)	449 (48.33%)	406 (43.66%)	258 (27.74%)
Age	56.59±8.34	56.46±8.47	56.64±7.93	56.50±8.50	56.76±8.45	0.868
BMI (kg/m2)	25.99±3.99	24.61±3.91	25.26±3.97	26.59±3.80	27.50±3.61	<0.001
Current smoking (%)	767 (20.63%)	187 (20.13%)	167 (17.98%)	182 (19.57%)	231 (24.84%)	0.002
Current drinking (%)	886 (23.83%)	178 (19.16%)	190 (20.45%)	218 (23.44%)	300 (32.26%)	<0.001
SBP (mmHg)	143.00±18.34	141.12±17.84	142.82±18.48	143.92±18.25	144.14±18.67	0.001
DBP (mmHg)	86.75±13.30	85.02±12.35	86.06±13.09	86.73±13.31	89.19±14.05	<0.001
Laboratory tests
ALT (U/L)	18.17 (12.12–28.28)	14.14(10.43–19.19)	17.17(11.21–25.65)	20.20(13.13–31.41)	23.23(16.05–38.38)	<0.001
AST (U/L)	18.77(15.30–24.48)	17.34(14.28–21.95)	18.36(15.30–24.48)	19.38(15.51–25.50)	20.40(16.73–28.56)	<0.001
TC (mmol/L)	4.13±0.94	4.00±0.87	4.11±0.90	4.12±0.94	4.28±1.03	<0.001
Triglycerides (mmol/L)	0.66 (0.55–1.63)	0.60 (0.51–0.72)	0.63 (0.54–1.09)	0.67 (0.56–1.75)	1.22 (0.65–2.18)	<0.001
LDL-C (mmol/L)	2.76±0.79	2.80±0.77	2.72±0.79	2.68±0.77	2.86±0.82	<0.001
HDL-C (mmol/L)	1.16±0.29	1.26±0.28	1.24±0.28	1.12±0.26	1.02±0.26	<0.001
FPG (mmol/L)	5.73±0.88	5.78±1.03	5.62±0.76	5.71±0.85	5.78±0.85	<0.001
HbA1c (%)	6.38±0.47	5.76±0.32	6.26±0.10	6.57±0.10	6.93±0.14	<0.001
SUA (umol/L)	344.97±86.68	242.90±28.90	308.54±17.13	368.74±19.10	459.53±55.85	<0.001
Medical history
Hypertension (%)	2118 (56.97%)	485 (52.21%)	448 (48.22%)	488 (52.47%)	697 (74.95%)	<0.001
CHD (%)	1149 (30.90%)	242 (26.05%)	299 (32.19%)	285 (30.65%)	323 (34.73%)	<0.001
Hyperlipidemia (%)	1847 (49.68%)	417 (44.89%)	430 (46.29%)	429 (46.13%)	571 (61.40%)	<0.001
Medications
Lipid-lowering drugs (%)	593 (15.95%)	95 (10.23%)	135 (14.53%)	128 (13.76%)	235 (25.27%)	<0.001
Oral hypoglycemic agents (%)	3120 (83.92%)	756 (81.38%)	812 (87.41%)	797 (85.70%)	755 (81.18%)	<0.001
Insulin (%)	785 (21.11%)	196 (21.10%)	167 (17.98%)	184 (19.78%)	238 (25.59%)	<0.001

Notes: Data are presented as mean ± standard deviation, median (interquartile range), or as numbers, and percentages.

Abbreviations: SUA, serum uric acid; BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; ALT, alanine transaminase; AST, aspartate transaminase; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; TC, total cholesterol; FPG, fasting plasma glucose; HbA1c, Hemoglobin A1c; CHD, Coronary heart disease.

In terms of laboratory data, higher SUA strata were associated with elevated ALT, AST, TC, triglycerides, LDL-C, FPG, and HbA1c, alongside lower HDL-C levels. Regarding pre-existing conditions, the frequencies of hypertension, CHD, and hyperlipidemia were consistently higher in patients with elevated SUA. These individuals were also more frequently prescribed lipid-lowering agents and insulin.

Impact of SUA on Stroke and Its Subtypes in Patients with Diabetes

To preliminarily explore the potential association between SUA and the risk of stroke and its subtypes, we first compared the crude incidence rates across the four SUA groups. The results showed that, for overall stroke as well as for IS and HS, the incidence rates were consistently highest in the Q4 group (the highest SUA quartile) (Figure 2). The incidence rates per 1,000 person-years were similarly consistent across groups (Table S2).

Figure 2.

Comparison of cumulative event proportions across serum uric acid quartile groups, (A) Stroke; (B) Ischemic stroke; (C) Hemorrhagic stroke.

We then applied Cox proportional hazards regression models to evaluate the association between SUA and stroke risk. In the unadjusted model 1, each 1-standard deviation (SD) increment in SUA was associated with a 58.1% higher risk of stroke (HR: 1.581, 95% CI: 1.485–1.682). After full adjustment for potential confounders in model 4, the association remained robust, with a 54.2% increased risk per 1-SD increment (HR: 1.542, 95% CI: 1.456–1.632) (Table 2). When SUA was analyzed as a categorical variable (using Q1 as the reference), the risks of stroke in Q2, Q3, and Q4 were progressively higher, with HRs of 1.652, 2.205, and 3.042, respectively (Table 2). Furthermore, to account for the competing risk of death, we performed Fine-Gray competing-risk models. The results showed that, during follow-up, patients in Q2, Q3, and Q4 had progressively higher cumulative stroke incidence compared with those in Q1, with Q4 exhibiting the highest risk (Table S3 and Figure 3).

Table 2.

Association Between Elevated SUA and Increased Stroke Risk in Patients with DM

Stroke	Model 1	Model 2	Model 3	Model 4
	HR (95% CI) P	HR (95% CI) P	HR (95% CI) P	HR (95% CI) P
SUA (per 1SD increase)	1.581 [1.485, 1.682] <0.001	1.572 [1.474, 1.676] <0.001	1.570 [1.471, 1.676] <0.001	1.542 [1.456, 1.632] <0.001
Quartiles of SUA
Q 1	Reference	Reference	Reference	Reference
Q 2	1.696 [1.340, 2.146] <0.001	1.679 [1.327, 2.123] <0.001	1.663 [1.315, 2.102] <0.001	1.652 [1.309, 2.084] <0.001
Q 3	2.264 [1.812, 2.828] <0.001	2.251 [1.802, 2.811] <0.001	2.249 [1.803, 2.806] <0.001	2.205 [1.776, 2.738] <0.001
Q 4	3.162 [2.531, 3.951] <0.001	3.083 [2.461, 3.862] <0.001	3.076 [2.491, 3.798] <0.001	3.042 [2.422, 3.821] <0.001
P for trend	<0.001	<0.001	<0.001	<0.001

Notes: Model 1: no covariates were adjusted. Model 2: age, sex, BMI, smoking status, drinking status, SBP, and DBP were adjusted. Model 3: Model 2 plus adjustment for ALT, AST, TC, TG, HDL-C, LDL-C, eGFR HbA1c, and FPG. Model 4: Model 3 plus adjustment for CHD, Dyslipidemia, Hypertension, Oral hypoglycemic agents, and Insulin.

Abbreviations: SUA, serum uric acid; DM, diabetes mellitus; HR, hazard ratio; CI, confidence interval.

Figure 3.

Cumulative incidence of stroke according to serum uric acid quartiles using competing risk analysis.

Similar patterns were observed for both IS and HS. In the fully adjusted model 4, each 1-SD increment in SUA was associated with a 42.8% higher risk of IS (HR: 1.428, 95% CI: 1.330–1.533) and a 77.2% higher risk of HS (HR: 1.772, 95% CI: 1.610–1.951) (Table 3). When analyzed categorically, the risks for both subtypes increased progressively from Q2 to Q4 compared with Q1. Specifically, participants in Q4 had a 157.6% higher risk of IS (HR: 2.576, 95% CI: 1.972–3.364) and a 355.8% higher risk of HS (HR: 4.558, 95% CI: 2.910–7.141) (Table 3). Fine-Gray models similarly showed that the Q4 group had the highest cumulative incidence for both subtypes (Table S4 and Figure 4).

Table 3.

Association Between Elevated SUA and Increased Risk of Stroke Subtypes in Patients with DM

Stroke Subtype	Model 1	Model 2	Model 3	Model 4
	HR (95% CI) P	HR (95% CI) P	HR (95% CI) P	HR (95% CI) P
Ischemic stroke
SUA (per 1SD increase)	1.446 [1.339, 1.562] <0.001	1.438 [1.328, 1.557] <0.001	1.434 [1.322, 1.555] <0.001	1.428 [1.330, 1.533] <0.001
Quartiles of SUA
Q 1	Reference	Reference	Reference	Reference
Q 2	1.488 [1.131, 1.958] 0.001	1.472 [1.119, 1.935] 0.006	1.461 [1.112, 1.920] 0.007	1.455 [1.110, 1.908] 0.007
Q 3	1.913 [1.474, 2.482] <0.001	1.898 [1.463, 2.463] <0.001	1.890 [1.459, 2.449] <0.001	1.857 [1.443, 2.391] <0.001
Q 4	2.647 [2.041, 3.433] <0.001	2.604 [2.037, 3.328] <0.001	2.592 [1.991, 3.374] <0.001	2.576 [1.972, 3.364] <0.001
P for trend	<0.001	<0.001	<0.001	<0.001
Hemorrhagic stroke
SUA (per 1SD increase)	1.873 [1.675, 2.094] <0.001	1.864 [1.678, 2.070] <0.001	1.863 [1.671, 2.078] <0.001	1.772 [1.610, 1.951] <0.001
Quartiles of SUA
Q 1	Reference	Reference	Reference	Reference
Q 2	2.378 [1.493, 3.786] <0.001	2.359 [1.483, 3.751] <0.001	2.325 [1.463, 3.695] <0.001	2.289 [1.445, 3.625] <0.001
Q 3	3.395 [2.188, 5.267] <0.001	3.395 [2.189, 5.264] <0.001	3.419 [2.209, 5.289] <0.001	3.327 [2.167, 5.108] <0.001
Q 4	4.822 [3.104, 7.491] <0.001	4.688 [3.004, 7.315] <0.001	4.560 [2.991, 6.951] <0.001	4.558 [2.910, 7.141] <0.001
P for trend	<0.001	<0.001	<0.001	<0.001

Notes: Model 1: no covariates were adjusted. Model 2: age, sex, BMI, smoking status, drinking status, SBP, and DBP were adjusted. Model 3: Model 2 plus adjustment for ALT, AST, TC, TG, HDL-C, LDL-C, eGFR HbA1c, and FPG. Model 4: Model 3 plus adjustment for CHD, Dyslipidemia, Hypertension, Oral hypoglycemic agents, and Insulin.

Abbreviations: SUA, serum uric acid; DM, diabetes mellitus; HR, hazard ratio; CI, confidence interval.

Figure 4.

Cumulative incidence function curves of stroke subtypes according to SUA quartiles (A) Ischemic stroke; (B) Hemorrhagic stroke.

Threshold Effect of SUA on the Risk of Stroke and Its Subtypes in Diabetic Patients

To further explore the dose-response relationship between SUA and the risk of stroke and its subtypes in patients with diabetes, we performed RCS regression analyses. The results revealed a significant nonlinear dose-response association between SUA and the risks of overall stroke, IS, and HSe (all P for nonlinear < 0.05), with a notable threshold effect identified for each outcome (Figure 5). Specifically, the risk of overall stroke began to increase significantly when SUA exceeded 335 µmol/L (Figure 5). For stroke subtypes, the corresponding threshold points were 330 µmol/L for IS and 340 µmol/L for HS (Figure 5).

Figure 5.

Dose–response relationship and threshold effect between serum uric acid and the risk of stroke and its subtypes (A) Stroke; (B) Ischemic stroke; (C) Hemorrhagic stroke The solid line represents the fitted hazard ratio; the dashed lines represent the 95% confidence intervals. The red dashed vertical line indicates the reference point where HR=1, corresponding to the threshold value.

Based on these identified threshold values, we conducted a two-stage piecewise Cox regression analysis to compare the risks below and above the thresholds. Compared with patients with SUA ≤ 335 µmol/L, those with levels > 335 µmol/L had a 92.9% higher risk of overall stroke (HR: 1.929, 95% CI: 1.660–2.241) (Table 4). For stroke subtypes, relative to patients below the respective threshold points, those above the thresholds had a 72.4% higher risk of IS (HR: 1.724, 95% CI: 1.435–2.070) and a 155.2% higher risk of HS (HR: 2.552, 95% CI: 1.942–3.355) (Table 4). Collectively, these findings underscore a clear threshold effect of SUA on stroke and subtype risks in diabetic patients, indicating that once SUA levels exceed a certain limit, the associated stroke risk increases substantially.

Table 4.

Turning Points Identified in the Nonlinear Relationship Between SUA and the Increased Risk of Stroke and Its Subtypes in Patients with DM

Exposure	Model 1 HR (95% CI) P	Model 2 HR (95% CI) P	Model 3 HR (95% CI) P	Model 4 HR (95% CI) P
Stroke
Turning point (umol/L)	335	335	335	335
<= 335	Reference	Reference	Reference	Reference
> 335	1.998 [1.738, 2.296] <0.001	1.989 [1.719, 2.301] <0.001	1.953 [1.683, 2.266] <0.001	1.929 [1.660, 2.241] <0.001
Ischemic stroke
Turning point (umol/L)	330	330	330	330
<= 330	Reference	Reference	Reference	Reference
> 330	1.782 [1.503, 2.113] <0.001	1.775 [1.484, 2.122] <0.001	1.745 [1.455, 2.093] <0.001	1.724 [1.435, 2.070] <0.001
Hemorrhagic stroke
Turning point (umol/L)	340	340	340	340
<= 340	Reference	Reference	Reference	Reference
> 340	2.674 [2.074, 3.446] <0.001	2.664 [2.046, 3.469] <0.001	2.603 [1.986, 3.411] <0.001	2.552 [1.942, 3.355] <0.001

Notes: Model 1: no covariates were adjusted. Model 2: age, sex, BMI, smoking status, drinking status, SBP, and DBP were adjusted. Model 3: Model 2 plus adjustment for ALT, AST, TC, TG, HDL-C, LDL-C, eGFR HbA1c, and FPG. Model 4: Model 3 plus adjustment for CHD, Dyslipidemia, Hypertension, Oral hypoglycemic agents, and Insulin.

Abbreviations: SUA, serum uric acid; DM, diabetes mellitus; HR, hazard ratio; CI, confidence interval.

Subgroup Analysis: Association Between SUA and Stroke Risk by Sex and Uric Acid Status

To further examine whether SUA normality status similarly affects the risk of stroke and its subtypes, we categorized patients into normouricemic and hyperuricemic groups (SUA> 420 µmol/L) according to current clinical guidelines. The results showed that patients with hyperuricemia had a 2.0-fold higher risk of overall stroke (HR: 2.000) compared with the normouricemic group; for stroke subtypes, the corresponding HRs were 1.91 for IS and 2.20 for HS (Table 5).

Table 5.

Stroke Risk and Its Subtypes in Diabetic Patients Stratified by SUA Normality

Exposure	Model 1 HR (95% CI) P	Model 2 HR (95% CI) P	Model 3 HR (95% CI) P	Model 4 HR (95% CI) P
Stroke
Normouricemic group (3087)	Reference	Reference	Reference	Reference
Hyperuricemic group (631)	2.202 [1.839, 2.637] <0.001	2.074 [1.725, 2.494] <0.001	2.074 [1.746, 2.465] <0.001	2.000 [1.647, 2.428] <0.001
Ischemic stroke
Normouricemic group (3087)	Reference	Reference	Reference	Reference
Hyperuricemic group (631)	2.062 [1.657, 2.566] <0.001	1.979 [1.581, 2.476] <0.001	1.970 [1.597, 2.431] <0.001	1.910 [1.508, 2.420] <0.001
Hemorrhagic stroke
Normouricemic group (3087)	Reference	Reference	Reference	Reference
Hyperuricemic group (631)	2.491 [1.813, 3.423] <0.001	2.279 [1.646, 3.153] <0.001	2.274 [1.680, 3.077] <0.001	2.201 [1.564, 3.098] <0.001

Notes: Model 1: no covariates were adjusted. Model 2: age, sex, BMI, smoking status, drinking status, SBP, and DBP were adjusted. Model 3: Model 2 plus adjustment for ALT, AST, TC, TG, HDL-C, LDL-C, eGFR HbA1c, and FPG. Model 4: Model 3 plus adjustment for CHD, Dyslipidemia, Hypertension, Oral hypoglycemic agents, and Insulin.

Abbreviations: SUA, serum uric acid; HR, hazard ratio; CI, confidence interval.

Given the known physiological differences in SUA levels between sexes, we further performed sex-stratified RCS analyses. Although the interaction test between SUA and sex did not reach statistical significance (P for Interaction=0.412), the thresholds derived from the RCS models still differed between sexes. In male patients, the risk of overall stroke began to increase significantly when SUA exceeded 335 µmol/L; for IS and HS, the respective threshold points were 335 µmol/L and 360 µmol/L (Figure 6). In female patients, the risk thresholds for overall stroke, IS, and HS were 350 µmol/L, 355 µmol/L, and 330 µmol/L, respectively (Figure 6). Based on these sex-specific threshold values, we conducted two-stage piecewise analyses, which yielded results consistent with the primary analysis—risks of stroke and its subtypes increased substantially above each corresponding threshold (Table 6). In addition, we performed an E-value analysis to assess unmeasured confounding. The E-value for the point estimate indicated that unmeasured confounding would need to be of substantial magnitude to nullify our findings (Table S5).

Figure 6.

Sex-stratified dose–response relationship and threshold effect between serum uric acid and the risk of stroke and its subtypes (A) Stroke; (B) Ischemic stroke; (C) Hemorrhagic stroke.

Table 6.

Sex-Stratified Threshold Effect of SUA on Increased Risk of Stroke and Its Subtypes in Patients with DM

Exposure	Model 1 HR (95% CI) P	Model 2 HR (95% CI) P	Model 3 HR (95% CI) P	Model 4 HR (95% CI) P
Stroke
Male
Turning point (umol/L)	335	335	335	335
<= 335	Reference	Reference	Reference	Reference
> 335	2.170 [1.719, 2.739] <0.001	2.166 [1.740, 2.696] <0.001	2.163 [1.713, 2.731] <0.001	2.135 [1.702, 2.678] <0.001
Female
Turning point (umol/L)	350	350	350	350
<= 350	Reference	Reference	Reference	Reference
> 350	2.311 [1.915, 2.788] <0.001	2.227 [1.841, 2.692] <0.001	2.159 [1.780, 2.618] <0.001	2.147 [1.793, 2.571] <0.001
Ischemic stroke
Male
Turning point (umol/L)	335	335	335	335
<= 335	Reference	Reference	Reference	Reference
> 335	2.001 [1.509, 2.652] <0.001	1.997 [1.518, 2.627] <0.001	1.993 [1.503, 2.643] <0.001	1.987 [1.525, 2.589] <0.001
Female
Turning point (umol/L)	355	355	355	355
<= 355	Reference	Reference	Reference	Reference
> 355	2.043 [1.637, 2.550] <0.001	2.011 [1.607, 2.516] <0.001	1.952 [1.554, 2.453] <0.001	1.937 [1.566, 2.397] <0.001
Hemorrhagic stroke
Male
Turning point (umol/L)	360	360	360	360
<= 360	Reference	Reference	Reference	Reference
> 360	3.209 [2.143, 4.806] <0.001	3.191 [2.176, 4.680] <0.001	3.159 [2.111, 4.727] <0.001	3.008 [2.030, 4.457] <0.001
Female
Turning point (umol/L)	330	330	330	330
<= 330	Reference	Reference	Reference	Reference
> 330	4.297 [2.774, 6.654] <0.001	4.049 [2.604, 6.298] <0.001	3.957 [2.535, 6.179] <0.001	3.643 [2.386, 5.561] <0.001

Notes: Model 1: no covariates were adjusted. Model 2: age, sex, BMI, smoking status, drinking status, SBP, and DBP were adjusted.Model 3: Model 2 plus adjustment for ALT, AST, TC, TG, HDL-C, LDL-C, eGFR HbA1c, and FPG. Model 4: Model 3 plus adjustment for CHD, Dyslipidemia, Hypertension, Oral hypoglycemic agents, and Insulin.

Abbreviations: SUA, serum uric acid; DM, diabetes mellitus; HR, hazard ratio; CI, confidence interval.

Collectively, these findings further confirm that the effect of SUA on stroke and subtype risk is not modified by sex. However, it is noteworthy that the threshold points may differ between male and female patients, and these sex-specific thresholds may have important clinical implications for risk stratification in stroke prevention among diabetic patients of different sexes.

Discussion

Despite accumulating evidence linking SUA to CVD, the association between SUA and stroke risk—along with its subtypes, the presence of a threshold effect, and potential sex differences—remains largely unexplored in patients with diabetes, a population at particularly high risk of stroke. To address this gap, we conducted the present study using cohort data from two independent centers to evaluate the predictive value of SUA for stroke and its subtypes in diabetic patients. After rigorous adjustment for potential confounders in a series of incrementally adjusted models, our analysis revealed that higher SUA levels were independently associated with greater risks of both overall stroke and its major subtypes. Moreover, all three outcomes displayed a consistent nonlinear pattern, with clear threshold points beyond which the risk escalated sharply. Specifically, the risk of overall stroke, IS, and HS began to increase significantly when SUA exceeded 335 µmol/L, 330 µmol/L, and 340 µmol/L, respectively. Sex-stratified analyses further revealed clinically meaningful differences: in male patients, the threshold values for overall stroke, IS, and HS were 335 µmol/L, 335 µmol/L, and 360 µmol/L, respectively; in female patients, the corresponding thresholds were 350 µmol/L, 355 µmol/L, and 330 µmol/L. At a broader level, our results suggest that keeping SUA concentrations below these cutoff levels may be associated with a lower risk of future stroke and its subtypes. These findings expand the existing framework for cerebrovascular prevention in diabetic populations and provide preliminary evidence that urate‑lowering strategies warrant further investigation in this setting.

DM is a common and progressive chronic metabolic disorder, whose prevalence has been steadily increasing worldwide.^32,33 It is characterized by multiple metabolic disturbances, including dysglycemia and dyslipidemia, and is closely associated with a broad spectrum of cardiovascular and cerebrovascular diseases.^34,35 Chronic hyperglycemia in diabetic patients can induce vascular endothelial injury, lipid deposition, and atherosclerosis, thereby creating a favorable pathophysiological milieu for the development of stroke.^36,37 Of note, both diabetes and stroke predominantly affect middle-aged and older adults, and their coexistence not only severely impairs patients’ quality of life and increases caregiving burdens but also leads to substantial healthcare resource utilization and economic expenditure. More importantly, the combination of diabetes and stroke is associated with a significantly elevated risk of all-cause mortality.^38–40 Consequently, stroke prevention should be given high priority in the management of diabetic patients. Early, close monitoring and periodic systematic screening and risk assessment for stroke are essential to reduce the stroke-related morbidity and mortality in this high-risk population.

UA, the final product of purine metabolism, has attracted growing attention regarding its role in CVD in recent years.^26,41–43 A study investigating visit‑to‑visit variability in SUA reported that elevated SUA levels, irrespective of their fluctuation patterns, were consistently associated with an increased risk of CVD, with inflammatory pathways playing a significant mediating role.⁴¹ Moreover, Mendelian randomization analyses have further supported a causal relationship between SUA and CVD, suggesting that uric acid may directly contribute to the development of cardiovascular events.²⁶ In parallel, studies on hyperuricemia have reached concordant conclusions, indicating that hyperuricemia is not only a risk factor for hypertension but also an important risk marker for CVD.²⁹ In the context of stroke, a Chinese population‑based study demonstrated that cumulative exposure to elevated SUA significantly increased the risk of stroke, providing key epidemiological evidence for the SUA–stroke association.⁴⁴ More importantly, a recently published study further revealed that elevated SUA was significantly associated with an increased risk of incident diabetes in patients with CHD, which offers strong indirect evidence supporting our investigation into the relationship between SUA and stroke in diabetic patients.²⁸

Against this background, the present study further focused on patients with diabetes—a population at particularly high risk of stroke—and systematically elucidated the specific impact of SUA on stroke risk, as well as its potential therapeutic thresholds. Our findings are consistent with previous literature and not only expand the current understanding of stroke risk factors in diabetic patients, but also provide a novel theoretical basis and a potential entry point for clinical practice regarding the value of urate‑lowering therapy and its potential threshold in stroke prevention among individuals with diabetes.

The potential mechanisms linking elevated SUA to increased stroke risk in diabetic patients involve multiple pathways. The threshold effect is biologically plausible: uric acid acts as an antioxidant at physiological levels but shifts to a pro-oxidant state above its solubility limit, promoting endothelial dysfunction and vascular inflammation.^24,45,46 This biphasic switch, supported by recent cardiovascular studies, suggests that excessive SUA—rather than SUA per se—increases stroke risk, consistent with our threshold findings.^29,47 First, UA induces endothelial dysfunction by inhibiting nitric oxide production and activating the renin-angiotensin system, thereby impairing endothelium-dependent vasodilation and promoting endothelial apoptosis, which compromises vascular wall integrity.^48–50 Second, UA triggers inflammatory cascades, stimulating vascular smooth muscle and endothelial cells to release pro-inflammatory cytokines (eg, IL-1β, IL-6, TNF-α) and activating the NLRP3 inflammasome, which exacerbates chronic low-grade inflammation in the vessel wall and accelerates atherosclerotic plaque progression.^50–52 Then, UA enhances oxidative stress, increasing reactive oxygen species generation, which promotes oxidative modification of LDL-C to oxidized LDL, subsequently taken up by macrophages to form foam cells, leading to plaque destabilization.^53–55 Finally, in diabetic patients, hyperglycemia and insulin resistance significantly amplify these processes, establishing a vicious cycle of “hyperglycemia-hyperuricemia-vascular damage,” ultimately culminating in plaque rupture and thrombosis, thereby precipitating stroke events.^21,56,57

This study has several strengths. First, it was based on a relatively large cohort with sufficient sample size and adequate number of events from two independent centers, along with a relatively long follow-up duration, which provided reliable data for evaluating the association between SUA and stroke risk in patients with diabetes. Second, we adopted a multi-model adjustment strategy and analyzed SUA both as a continuous and a categorical variable, ensuring the robustness of our findings. Moreover, we systematically evaluated the threshold effects of SUA on the risk of stroke and its subtypes in diabetic patients, and further explored sex-specific threshold points, offering a novel perspective for individualized risk assessment in clinical practice.

Despite the strengths of this study, a few limitations merit consideration. First, the observational nature of the design means that causality cannot be established. While we carefully controlled for confounders, residual or unmeasured confounding may still exist. Our E-value analysis indicated that the results are reasonably robust against such confounding. However, because causality cannot be definitively determined from observational data, future prospective interventional or Mendelian randomization studies are needed to confirm the causal associations. Second, SUA levels were measured only at baseline, and repeated measures during follow-up were not available, which may have underestimated the true effect of long-term UA exposure on stroke risk. Third, the study population was predominantly Chinese patients with diabetes, and whether our findings are generalizable to other ethnicities or regions requires further validation. Fourth, information on diabetes duration was unavailable in our study, although it may also affect stroke risk; we therefore suggest that future studies incorporate this variable. Finally, we did not collect detailed information on dietary factors (eg, purine intake), the use of drugs for reducing uric acid during the follow-up period, or inflammatory biomarkers, which may have influenced the observed association between SUA and stroke risk.

Conclusion

This study identified a clear nonlinear threshold effect of SUA on the risk of stroke and its subtypes in patients with diabetes. The risk increased significantly when SUA exceeded 335 µmol/L for overall stroke, 330 µmol/L for IS, and 340 µmol/L for HS, respectively. Moreover, the threshold values differed between male and female patients. Given the observational nature of our study, these sex-specific thresholds should be interpreted as exploratory findings rather than established therapeutic targets. While routine monitoring of SUA may be of clinical interest, the value of sex-specific control targets requires confirmation in future prospective studies.

Funding Statement

There is no funding to report.

Institutional Review Board Statement

Prior to the enrollment of any participant, the study protocol was thoroughly reviewed and approved by the institutional review boards of the two participating centers: The Affiliated Huaian No.1 People’s Hospital of Nanjing Medical University (KY-2024-066-01) and Changzhi People’s Hospital (N-CZ2019-0217). All procedures were performed in accordance with the ethical principles laid down in the Declaration of Helsinki. Every patient provided written informed consent after receiving a comprehensive explanation of the study’s purpose and procedures.

Data Sharing Statement

The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request.

Author Contributions

Song Xu: Conceptualization, Methodology, Investigation, Data Curation, Formal Analysis, Writing – Original Draft, Writing – review and editing. Jiajia Yuan: Methodology, Data Curation, Investigation, Formal Analysis, Validation Writing – Original Draft. Yongtao Guo: Methodology, Data Curation, Writing – Review and Editing. Kun An: Methodology, Investigation, Data Curation, Formal Analysis, Writing – review and editing. Chaoqian Liu: Methodology, Data Curation, Investigation, Writing – review and editing. Hao Wang: Methodology, Formal Analysis, Writing – review and editing. Mingchao Li: Conceptualization, Methodology, Investigation, Supervision, Data Curation, Writing – Original Draft, Writing – review and editing. All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors report no conflicts of interest in this work.