Association Between Serum 25-Hydroxyvitamin D and Hyperuricemia in Postmenopausal Women: A Cross-Sectional Study of NHANES 2007-2018

1. Introduction

Hyperuricemia (HUA) is a major public health issue, affecting roughly 21% of adults globally [1]. High serum uric acid (SUA) levels are closely linked to the onset of cardiovascular disease (CVD), chronic kidney disease, gout, and metabolic syndrome [2,3,4]. Postmenopausal women (PMW) face a heightened risk of HUA due to reduced estrogen-mediated uric acid (UA) excretion [5,6], hormonal changes, and age-related declines in renal function [7,8,9].

Vitamin D (VD) is essential in bone metabolism and calcium homeostasis. Nevertheless, its possible extra-skeletal effects have drawn growing attention [10]. Observational studies have reported an inverse relation of serum 25(OH)D to various metabolic disorders: type 2 diabetes [11], hypertension [12], and CVD [13]. Nonetheless, the link between VD and HUA remains debated. Suggested mechanisms connecting VD to UA metabolism include modulation of inflammatory pathways via the vitamin D receptor (VDR) [14], improvement of endothelial function and renal perfusion with enhanced UA excretion [15], and suppression of the renin-angiotensin-aldosterone system [16]. Epidemiological findings, however, are conflicting: some revealed marked inverse associations [17,18], but others detected no meaningful correlation [19].

PMW exhibit distinct metabolic characteristics, and both VD deficiency [7] and elevated UA [20] are more common in this population. Yet few studies have relied on nationally representative data to assess the VD-HUA association specifically in this high-risk group, and even fewer have described its dose-response pattern. We therefore examined the serum 25(OH)D-HUA association among PMW using NHANES 2007-2018, focusing particularly on nonlinear and threshold effects.

2. Materials and Methods

2.1. Study Design and Population

NHANES, a cross-sectional, stratified, multistage probability survey, was carried out by the National Center for Health Statistics (NCHS) under the Centers for Disease Control and Prevention (CDC) [21], offering nationally representative data on the health and nutrition of the U.S. civilian noninstitutionalized population. All protocols gained approval from the NCHS Research Ethics Review Board, with every participant giving informed consent in writing. Six consecutive two-year cycles (2007-2008, 2009-2010, 2011-2012, 2013-2014, 2015-2016, and 2017-2018) were pooled. Eligible participants were women aged ≥20 identified as postmenopausal based on the reproductive health questionnaire; among the 59,842 participants screened across these cycles, 7,888 were classified as postmenopausal; men, pre-/peri-menopausal women, and those with missing menopausal status (n=51,954) were excluded. Participants lacking data on body mass index (BMI), serum 25(OH)D, or SUA (n=685) were excluded thereafter. The final analytic sample consisted of 7,203 PMW. The participant selection flow chart followed the format of a previous NHANES study [22] (Figure 1).

2.2. Exposure Assessment

Serum 25(OH)D (nmol/L) was quantified utilizing the DiaSorin Liaison 25-OH VD total assay at the CDC-NHANES Lipid Satellite Laboratory [21]. 25(OH)D was divided into quartiles according to the weighted distribution of the analytic sample (Q1 <56.5; Q2 56.5-78.0; Q3 78.0-99.1; Q4 ≥99.1 nmol/L).

2.3. Outcome Assessment

HUA was defined as SUA ≥360 µmol/L (≥6.0 mg/dL), the threshold recommended by the American College of Rheumatology for female gout risk [23].

2.4. Covariates

Demographic information (age, education, family poverty-to-income ratio [PIR], marital status, and race/ethnicity), lifestyle factors (drinking, physical activity (PA), and smoking), and medical history (hypertension and diabetes) were collected through standardized questionnaires. BMI was weight (kg) divided by height (m²). Estimated glomerular filtration rate (eGFR) was derived from serum creatinine via the CKD-EPI equation [24].

2.5. Statistical Analysis

The intricate survey design was accounted for in all analyses by using the provided sample weights along with the stratification and clustering information of NHANES [25]. Since a small proportion of participants had missing covariates, multiple imputation by chained equations with random-forest imputation models was employed [26,27]. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated via survey-weighted multivariable logistic regression (LR), with serum 25(OH)D split into weighted quartiles (Q1 as reference): Model 1 was crude; Model 2 controlled for age, education, marital status, PIR, and race/ethnicity; and Model 3 was additionally adjusted for diabetes, drinking, eGFR, high-density lipoprotein (HDL) cholesterol, hypertension, PA, smoking, and total cholesterol (TC). Because BMI is a strong determinant of both VD status and HUA and may lie on the causal pathway, BMI was not included in the main models. Instead, analyses were stratified by BMI at the cutoff of 25 kg/m², with effect modification assessed through cross-product terms, and a sensitivity model that additionally included BMI was carried out (Supplementary Table S1). Restricted cubic spline (RCS) regression with four knots [28], fitted by weighted maximum likelihood with normalized survey weights, characterized the dose-response relation and tested departure from linearity. A piecewise two-slope model with a grid search was used to further identify a threshold (Supplementary Table S2). Subgroup analyses were conducted within strata of age (<60 vs. ≥60), BMI (<25 vs. ≥25 kg/m²), diabetes, and hypertension, with serum 25(OH)D modeled in weighted quartiles within each stratum (Supplementary Table S3), and interactions were tested using cross-product terms. All analyses were performed with R 4.4.3 (R Foundation for Statistical Computing, Vienna, Austria) [29]. Our tests were two-sided, and P < 0.05 was regarded as statistically significant.

3. Results

3.1. Demographic and Clinical Characteristics

Of the 7,203 included PMW, 1,783 (24.8%) had HUA, which yielded a weighted prevalence of 22.5%. Table 1 presents the baseline characteristics according to HUA status. Women with HUA tended to be older than those with normal SUA (65.4 ± 10.6 vs. 61.8 ± 10.9, P < 0.001), exhibited a greater BMI (32.9 ± 7.8 vs. 28.6 ± 6.6 kg/m², P < 0.001), and showed a higher burden of hypertension (78.2% vs. 51.9%, P < 0.001) and diabetes (33.1% vs. 16.6%, P < 0.001). They also presented with reduced eGFR (68.4 ± 22.4 vs. 83.7 ± 17.7 mL/min/1.73 m², P < 0.001) and lower HDL cholesterol concentrations (1.4 ± 0.4 vs. 1.6 ± 0.5 mmol/L, P < 0.001). No significant difference in serum 25(OH)D was noted across groups (79.0 ± 35.5 vs. 80.4 ± 32.9 nmol/L; P=0.32).

3.2. Nonlinear Dose-Response Analysis

RCS regression showed a notable nonlinear, U-shaped serum 25(OH)D-HUA association (P for overall association=0.005; P for nonlinear=0.002). As presented in Figure 2, the OR was highest at low 25(OH)D concentrations, declined steadily to a nadir at 71.0 nmol/L, and rose again at higher concentrations, exceeding the null above roughly 100 nmol/L, with wide confidence intervals reflecting the small number of participants at high values. A piecewise two-slope analysis yielded consistent findings, placing the turning point at approximately 60 nmol/L; below this level, the odds of HUA decreased as 25(OH)D increased, while above it no association was detected (Supplementary Table S2).

3.3. Association of Serum 25(OH)D with HUA Across Quartiles

Given the nonlinear dose-response relation, serum 25(OH)D was examined categorically in weighted quartiles (Table 2). Compared with Q1 (<56.5 nmol/L), ORs in Model 3 were 0.79 (95% CI: 0.65-0.97) for Q2 (56.5-78.0 nmol/L) and 0.75 (0.59-0.95) for Q3 (78.0-99.1 nmol/L), while the odds in Q4 (≥99.1 nmol/L) were not significantly different (OR=0.83, 95% CI: 0.64-1.08). The lower odds seen in the middle quartiles but not in the highest quartile reflect the U-shaped dose-response curve. Table 2 presents the findings from both the unadjusted model and the model adjusted for demographic characteristics.

3.4. Modification by BMI

To examine modification by adiposity, analyses were stratified by BMI at the cutoff of 25 kg/m² (Table 3). The inverse 25(OH)D-HUA association was evident among women with BMI ≥25 kg/m² (Q2: OR: 0.76, 95% CI: 0.62-0.93; Q3: OR: 0.76, 0.59-0.98 vs. Q1) but was absent among women with BMI <25 kg/m², although the formal interaction was statistically non-significant (P=0.70) and the stratum-specific estimates in normal-weight women were imprecise. In the sensitivity model that additionally adjusted for BMI, the inverse associations for Q2 and Q3 were attenuated toward the null (Supplementary Table S1), supporting the decision to reserve BMI for stratification rather than covariate adjustment. The remaining subgroup analyses, with 25(OH)D modeled in quartiles, are provided in Supplementary Table S3.

4. Discussion

In this nationally representative, cross-sectional study on 7,203 American PMW, serum 25(OH)D exhibited a significant nonlinear, U-shaped association with HUA, with the lowest odds of HUA at approximately 71 nmol/L and elevated odds at both lower and higher concentrations. This pattern remained robust across quartile and spline models and was most evident among overweight or obese women, in whom the inverse association reached statistical significance. To our knowledge, this study is the first to characterize the dose-response relation of the VD-HUA association specifically in PMW using a nationally representative sample, and it suggests that the relevant question may not be whether VD is linearly protective, but where along the range of VD status the balance of benefit and harm lies.

Our findings differ from and extend those of earlier studies importantly. Previous cross-sectional studies in general populations, including Chinese [30,31] and Korean [32,33] adults, generally modeled a linear association between VD and UA and reported inverse relationships, whereas meta-analyses have highlighted substantial heterogeneity across studies [34]. Our study focused on PMW, a high-risk group in whom the estrogen-related enhancement of UA excretion is lost, and both VD deficiency and HUA are common [7,20,35]. It also benefits from the nationally representative, survey-weighted NHANES design and a large analytic sample. More importantly, when the dose-response relationship was examined rather than assumed to be linear, the overall linear association was null, whereas the inverse association was confined to the mid-range of 25(OH)D concentrations and attenuated at both extremes. This nonlinear pattern may help reconcile some of the inconsistent linear estimates reported previously.

Why might both low and high 25(OH)D correlate with higher HUA risk? Serum 25(OH)D is the most widely accepted biomarker of vitamin D status. With a half-life of around 2-3 weeks, it is more stable than the active hormone and reflects VD obtained from both cutaneous synthesis and dietary intake; accordingly, it is a principal VD indicator in nutritional epidemiology [36]. Several interrelated mechanisms may plausibly underlie a urate-lowering role of VD within the insufficient-to-adequate range. First, the active VD hormone acts via VDR expressed in vascular endothelium, renal tubular epithelial cells, and immune cells. VDR signaling suppresses NF-κB-driven transcription of pro-inflammatory cytokines and may attenuate the chronic low-grade inflammatory state associated with HUA and impaired renal urate handling [14,37,38]. Second, 1,25-dihydroxyvitamin D inhibits renin expression and may thereby attenuate renin-angiotensin-aldosterone system activity. Reduced angiotensin II signaling may improve renal perfusion and decrease sodium-dependent urate reabsorption in the proximal tubule, consistent with the mild uricosuric effects observed with pharmacological blockade of this system [15,37]. Third, experimental studies suggest that VDR activation may regulate the expression of renal urate transporters: ABCG2, GLUT9, and URAT1, which contribute to the reabsorption and secretion of urate. The modulation of these transporters could therefore affect urinary urate excretion [14,38]. These mechanisms suggest that renal urate handling may be particularly responsive as VD status improves from insufficient toward adequate levels, whereas the incremental benefit of further increases may diminish once adequate status is reached. The upturn observed at higher concentrations is considered below.

The location of the nadir at approximately 71 nmol/L also has a plausible biological and clinical basis. The Endocrine Society clinical practice guideline classifies 50-75 nmol/L as VD insufficiency and generally considers 75 nmol/L the lower limit of sufficiency [7]. The nadir observed in our study therefore lies just below the conventional sufficiency threshold. Below this level, VD-dependent physiological processes may remain relatively under-supplied, whereas as concentrations approach and enter the sufficient range, VDR-mediated anti-inflammatory and uricosuric pathways may become progressively less responsive to further increases. This pattern of diminishing marginal benefit is consistent with the nonlinear relationship observed in our data and may help explain two apparently paradoxical findings: quartile-specific estimates showed lower odds of HUA in the middle quartiles but not in the highest quartile, whereas conventional linear models fitted across the entire exposure range yielded null results. The upturn of the spline curve at higher concentrations, which is consistent with the J-shaped VD-UA association reported in previous studies [13], should nevertheless be interpreted cautiously because relatively few participants had 25(OH)D concentrations in this range. Residual confounding by lifestyle factors associated with VD supplement use also cannot be excluded.

The apparent modification by adiposity deserves comment, although the formal interaction was statistically non-significant (P=0.70). These analyses should be considered exploratory. The inverse relation was noted only within females with BMI ≥25 kg/m² and absent among normal-weight women, among whom only 226 HUA events occurred, limiting statistical power. After menopause, loss of estrogen may promote visceral fat accumulation. Therefore, obesity is both a driver of HUA and a determinant of low VD availability, making adiposity a plausible modifier of the VD-HUA association in this population [37,38]. Mechanistically, first, VD is lipophilic and may be sequestered in adipose tissue, reducing the bioavailability of circulating 25(OH)D [39]. Second, obesity-related chronic low-grade inflammation may attenuate the anti-inflammatory effects of VD [40]. Third, obesity-related insulin resistance may impair renal UA excretion and potentially offset any urate-lowering effect associated with VD status [18,41]. Because further adjustment for BMI in Model 3 attenuated the quartile associations toward the null (Supplementary Table S1), BMI may capture part of the shared metabolic pathway underlying these associations. The stratified analyses therefore provide additional context for interpreting the observed pattern while avoiding reliance on BMI adjustment alone. We also observed a nominally significant interaction with diabetes (P=0.047, Supplementary Table S3), with the inverse association observed only among women without diabetes. Given the number of interaction tests performed, this finding should nevertheless be deemed exploratory and requires replication before any firm inference can be drawn.

There are several limitations. First, because of the cross-sectional design, causality cannot be established. Second, 25(OH)D and UA were measured at a single time point, and dietary purine intake and other unmeasured confounders could not be fully accounted for. Third, subgroup and interaction analyses were exploratory, and although covariate imputation is a standard approach, it may not fully address missingness that is not at random. Fourth, the standardized 25(OH)D assay changed across survey cycles, which may have introduced measurement heterogeneity. Future studies should include longitudinal cohorts of PMW with repeated measurements and randomized controlled trials of VD supplementation among women with confirmed insufficiency [42], with UA and incident HUA as outcomes.

5. Conclusions

In U.S. PMW, serum 25(OH)D showed a U-shaped association with HUA, with the lowest odds of HUA observed at approximately 71 nmol/L and higher odds at both lower and higher concentrations. The inverse association was observed only among overweight or obese women. These findings suggest that maintaining 25(OH)D within a mid-range of approximately 50-75 nmol/L, rather than maximizing concentrations, may be relevant to HUA prevention in this population. However, the cross-sectional nature of the study does not permit identification of an optimal 25(OH)D target range, which requires confirmation in prospective studies. Randomized controlled trials are needed to establish causality.