This study utilised data from the 2011–2018 National Health and Nutrition Examination Survey (NHANES), which employed a stratified, multistage, probability sampling design to obtain representative samples of the U.S. population. Detailed survey operations and procedures have been described online (https://www.cdc.gov/nchs/nhanes). Briefly, NHANES surveys included an in-home interview and a subsequent health examination at a mobile examination centre (MEC). Information on demographics, lifestyle behaviours, and health conditions was collected through interviews. Physical examinations, body measurements, and laboratory tests were performed at the MEC. Dietary related information was obtained through two follow-up telephone interviews to obtain a 24-hour dietary recall.

We included people who participated in NHANES from 2011 to 2018, and the exclusion criteria were as follows: we excluded patients who were only interviewed, considering the correlation between 25 (OH)D concentration and sun exposure. We included a population aged 20–59 years who participated in the sun exposure and sun protective behaviour interviews. Moreover, dietary intake and supplements can also affect 25 (OH)D levels; hence, individuals who did not participate in dietary follow-up were excluded from the analysis. To ensure the accuracy of the dual-energy X-ray absorptiometry (DXA) scan data, only complete and valid whole-body scans were included in the study. Furthermore, participants were excluded if they wore clothes or carried medical equipment during weight measurements. Participants who did not maintain an upright posture during height measurement were also excluded. Finally, any individuals with missing data on obesity-related indicators was also excluded from the analysis. The detailed screening process is shown in Fig. 1.

Fig. 1.

Flowchart of the study population. Abbreviations: 25(OH)D, 25-hydroxyvitamin D; DXA, dual-energy X-ray absorptiometry; WC, waist circumference; BMC, bone mineral content; VD, vitamin D.

Serum concentrations of 25(OH)D2, 25(OH)D3, and C3-epimer-25(OH)D3 were measured using a fully validated, standardised, high-performance liquid chromatography-tandem mass spectrometry method (Thermo Vantage mass spectrometer and Thermo Accela ultra-high-performance liquid chromatography system, ThermoElectron Corp). The total 25(OH)D concentration was defined as the sum of 25(OH)D2 and 25(OH)D3, excluding C3-epimer-25(OH)D3. C3-epimer-25(OH)D3 is an epimer of 25(OH)D3 and was excluded to avoid overestimating VD levels. The detection limits were constant for all analytes. It should be noted that 25(OH)D2 detection results may be lower than the limit of detection (LOD). For analytes with results below the LOD, an imputed value of 1.45 nmol/L (the LOD divided by the square root of 2) was placed in the analyte results field.

BMI is an internationally recognised classification indicator for overweight and obesity, calculated by dividing weight by the square of height [24]. WC is frequently used as an indicator for evaluating abdominal obesity. Therefore, in this study, we used height, weight, and WC, along with the distribution of body fat, to evaluate the correlation between obesity and 25 (OH)D serum concentration. To ensure the accuracy of our results, we excluded individuals from the NHANES data who were wearing clothes during weight measurement, carrying medical equipment, or those who failed to maintain an upright posture during height measurement, as these factors could potentially affect the results. For the measurement of human body fat, DXA is the most widely accepted method for measuring body composition due in part to its speed, ease of use, and low radiation exposure. Only data obtained from scans covering the entire body and subsequently validated were included in our analysis.

Demographic variables included age, sex, ethnicity, education level, family income-to-poverty ratio, and examination period. Lifestyle factors comprised smoking status, alcohol consumption, time spent outdoors on weekdays and weekends, and physical activity level. Dietary intake of VD, calcium, phosphorus, and magnesium from foods and supplements were also considered as potential confounders. Additionally, biomarkers included serum calcium and phosphorus, alkaline phosphatase (ALP), alanine aminotransferase (ALT), and creatinine. Given the classic role of VD in regulating calcium and phosphorus metabolism and promoting bone development [25, 26], we included dietary intake and supplementation of calcium and phosphorus, along with serum calcium and phosphorus levels, as covariates. Since ALP indicates bone formation and is involved in generating phosphorus ions and bone salt crystals [27], it was also included in the analysis as a covariate. Given that VD is metabolised in the liver and kidneys [28], serum creatinine reflects renal function [29], and ALT reflects liver function, creatinine and ALT were included as covariates.

Serum 25(OH)D concentrations were categorised into quartiles (49.0, 49.0–64.2, 64.3–80.3, $\ge$80.4 nmol/L). Demographic, lifestyle, dietary, anthropometric, and clinical characteristics across 25(OH)D quartiles were compared using Kruskal-Wallis tests for continuous variables and chi-square tests for categorical variables.

We used multiple imputation to fill in missing data. Predictive mean matching was used to impute numeric features, logistic regression to impute binary variables, and Bayesian polytomous regression to impute factor features. Simultaneously, we performed statistical analysis on both imputed and non-imputed data to evaluate the stability of the model.

We incorporated all previously mentioned covariates into the multivariable linear regression models. The models were generated to examine the associations of weight, WC, head fat, arm fat, leg fat, subcutaneous fat, and visceral adipose tissue with serum 25(OH)D levels separately. Due to the inconsistency of evaluation index units, we standardised the data. Model 1 was adjusted for demographic characteristics, including age, sex, height, ethnicity, education level, examination period, and family income. Model 2 was further adjusted for lifestyle factors, including smoking history, alcohol consumption, sunlight exposure, and physical activity. Model 3 incorporated dietary intake and supplements (VD, calcium, magnesium, and phosphorus), bone mineral content (BMC), and biomarkers, including serum calcium and phosphorus, ALP, ALT, and creatinine. Simultaneously, we used bidirectional stepwise regression to explore which component of fat plays a key role in serum 25(OH)D levels. Stratified analysis based on sex was also performed.

All statistical analyses accounted for the complex survey design and sample weights in NHANES. Two-sided p 0.05 was considered statistically significant. All analyses were performed using R version 4.2.2 (R Foundation for Statistical Computing, Vienna, Austria), with the aid of the “survey” and “mice” packages.

Using imputed data from 9168 participants categorised by serum 25(OH)D levels, we stratified them into quartiles: quartile 1 (49.0 nmol/L), quartile 2 (49.0–64.2 nmol/L), quartile 3 (64.3–80.3 nmol/L), and quartile 4 ($\ge$80.4 nmol/L). Key demographic factors—age, sex, ethnicity, education level, and family income-to-poverty ratio—were correlated with serum 25(OH)D concentrations (all p 0.001). As participants aged and as education and family income rose, 25(OH)D concentrations increased. Notably, women and non-Hispanic Whites tended to have higher 25(OH)D levels. Meanwhile, seasonal variations were conspicuous, with individuals tested in summer showing generally elevated concentrations (p 0.001). Lifestyle choices influenced 25(OH)D levels: alcohol consumers had higher concentrations (p = 0.009), and reduced outdoor exposure corresponded with diminished 25(OH)D levels (p 0.001). Anthropometric indicators, including weight, WC, and fat, sequentially decreased from quartile 1 to quartile 4 (all p 0.001). Dietary patterns played a role: augmented intakes of VD, calcium, phosphorus, and magnesium correlated with heightened 25(OH)D concentrations (all p 0.001). Biochemical parameters, including ALP, creatinine, and serum calcium and phosphorus, underscored pronounced differences across quartiles (all p 0.05). Detailed insights from the imputed data are presented in Table 1. Preliminary findings from unimputed data largely aligned with those from imputed data (Table 2). However, while the imputed data highlighted statistical differences in ALT and physical activity across concentrations (p 0.05), the unimputed dataset did not (p $\ge$ 0.05).

Fig. 2 delineates the outcomes of multiple linear regression analyses in the general population, and the male and female subsets, which explored the correlation between distinct metrics—namely weight, WC, and fat in different parts of the body—and serum 25(OH)D levels. Adjustments were made and all indicators of the adjusted model showed statistical significance (all p 0.05). In the comprehensive model using interpolated data, the relatively superior evaluation indicators include weight, WC, and subcutaneous fat [weight: $\beta$ 95% CI: –5.12 (–6.00, –4.24), pseudo R²: 0.408, AIC: 75,985.36; WC: $\beta$ 95% CI: –4.67 (–5.45, –3.90), pseudo R²: 0.408, AIC: 75,980.16; subcutaneous fat: $\beta$ 95% CI: –4.76 (–5.54, –3.99), pseudo R²: 0.407, AIC: 75,990.86, respectively]. Among men, WC was the superior index [$\beta$ 95% CI: –4.42 (–5.44, –3.41), pseudo R²: 0.400, AIC: 37,487.01], while weight was a stronger predictor in women [$\beta$ 95% CI: –4.92 (–6.19, –3.65), pseudo R²: 0.425, AIC: 38,323.89]. Regarding fat distribution, visceral adipose tissue volume in men showed a stronger correlation with serum 25(OH)D levels [$\beta$ 95% CI: –4.67 (–5.70, –3.64), pseudo R²: 0.403, AIC: 37,466.70], whereas subcutaneous fat volume was more closely associated with serum 25(OH)D levels in women [$\beta$ 95% CI: –3.95 (–5.06, –2.58), pseudo R²: 0.422, AIC: 38,347.76]. Consistency was observed between results derived from the imputed and non-imputed data.

Fig. 2.

Association between obesity indicators and serum 25(OH)D in total population (a), and male (b), and female subpopulations (c): Forest plot from multiple linear regression. Model 1 adjusted for demographic characteristics, including age, sex, height, ethnicity, education level, examination period and family income. Model 2 further adjusted for lifestyle factors, including smoking history, alcohol consumption, sunlight exposure, and physical activity. Model 3 incorporated dietary intake and supplements (VD, calcium, magnesium, and phosphorus), bone mineral content, and biomarkers including serum calcium and phosphorus, alkaline phosphatase, alanine aminotransferase, and creatinine.

To comprehensively consider the effects of body fat composition on serum 25(OH)D levels, we conducted a rigorous bidirectional stepwise regression analysis on the entire population, and the male and female subpopulations. Head fat, visceral adipose tissue, and subcutaneous fat were independent determinants in the entire population [head fat: $\beta$ 95% CI: –9.24 (–16.46, –2.03), p = 0.012; visceral adipose tissue: $\beta$ 95% CI: –6.74 (–12.66, –0.81], p = 0.026; subcutaneous fat: $\beta$ 95% CI: –3.40 (–4.93, –1.88), p 0.001, respectively). In men, a strong inverse correlation was observed between serum 25(OH)D concentrations and both visceral adipose tissue and subcutaneous fat, with visceral adipose tissue having a more pronounced impact [visceral adipose tissue: $\beta$ 95% CI: –13.04 (–18.10, –7.99), p 0.001; subcutaneous fat: $\beta$ 95% CI: –4.69 (–8.76, –0.16), p = 0.024]. In contrast, no significant correlation was observed between fat distribution and serum 25(OH)D levels in women, although weight was identified as an important influencing factor [$\beta$ 95% CI: –0.20 (–0.28, –0.12), p 0.001]. Additionally, demographic factors like age, sex, ethnicity, examination period, and family income; lifestyle aspects including outdoor duration; dietary elements encompassing dietary intake of VD, phosphorus, and magnesium; and supplementary intake of VD and calcium are intricately linked with 25(OH)D. Notably, biomarkers such as ALP, creatinine (µmol/L), and serum calcium (mmol/L) also share a close association with 25(OH)D. Please refer to Table 3 for a detailed overview.