Association of Serum Ferritin with Total and Regional Fat Measured by Dual-Energy X-Ray Absorptiometry in a Nationally Representative Survey

Obesity is a prevalent, complex, and severe disease that imposes significant public health and economic burdens worldwide [1]. In the USA, the age-adjusted prevalence of general obesity increased from 35.4% in 2011–12 to 43.4% in 2017–18 [2]. Obesity adversely affects human health, increasing the risk of cardiometabolic diseases, mental illnesses, and mortality [3, 4].

The health risks associated with obesity are related not only to fat content itself, but more strongly to fat distribution [5]. Abdominal obesity, characterized by an “apple-shape,” emerges as a stronger predictor for cardiometabolic morbidity and mortality than body mass index (BMI) or body fat mass [6]. Conversely, a “pear-shape” with most fat accumulating in the lower body is found to have cardiometabolic protective effects [7]. Hence, it is crucial to take into account the location of body fat when assessing the risk of cardiometabolic diseases and mortality.

Iron, an essential element, is fundamental to several vital biological processes, including nutrient energy transport and storage [8]. Serum ferritin is widely used as a reliable marker for body iron stores, providing a more precise stores’ measurement than direct quantification of iron in serum [9, 10]. Recent evidence has increasingly linked elevated iron stores, indicated by serum ferritin levels, to metabolic syndrome [11], malignancy [12], fatty liver disease [13], neurodegenerative disorders [14], and autoimmune diseases [15]. Additionally, the effects of iron metabolism on obesity have garnered growing interest. Several observational studies have identified an association between iron stores and obesity [16‒18]. However, these studies typically measure obesity using BMI and waist circumference (WC). Although these measures are convenient and useful, they do not directly quantify body fat and reflect its distribution [19]. Accumulating evidence has indicated that individuals with similar body weights could have substantially different comorbidities and health risks [20]. Research on the association between ferritin levels and fat distribution in specific body regions remains scarce. To fill this gap, our study aimed to investigate the association between ferritin levels and body fat distribution using dual-energy X-ray absorptiometry (DXA) in the general population of US adults, from the 2017–2018 National Health and Nutrition Examination Survey (NHANES).

This study analyzed data from the 2017–2018 NHANES cycle conducted by the US National Center for Health Statistics (NCHS). NHANES employs a cross-sectional survey design with stratified, multistage, clustered probability sampling to represent the general, non-institutionalized population. The study received ethical approval by the NCHS Institutional Review Board, and all participants provided written informed consent.

A total of 3,419 adults aged 20–59 years were included in the 2017–2018 NHANES cycle, with DXA performed on 2,764 participants. Those with missing ferritin values (n = 118) were excluded, resulting in a final sample of 2,646 US adults for this study.

Participants underwent thorough assessments through household interviews and physical examinations conducted in a mobile examination center. Data collection included standardized questionnaires capturing information on age, gender, race, education level, family income-to-poverty ratio (PIR), lifestyle-related risk factors, and medical history. Race and ethnicity were self-reported and categorized into Mexican American, other-Hispanic White, other-Hispanic Black, and other races. Education levels were classified as less than high school, high school graduate, and beyond high school. PIR was determined by the ratio of the midpoint of the observed family income category to the official poverty threshold and was categorized into three groups (≤1.30, 1.31–3.49, and ≥3.50). Smoking status was classified as current (having smoked at least 100 cigarettes in one’s lifetime and currently smoking), former (having smoked at least 100 cigarettes in one’s lifetime but not smoking currently), and never (never smoked or smoked less than 100 cigarettes) [21]. Alcohol consumption was categorized as never or rarely (drinking less than once a month), sometimes (drinking once a month or more but less than once a week), and often (drinking once a week or more) [22]. Physical activity was defined as engaging in moderate-intensity sports, fitness, or recreational activities (such as brisk walking, bicycling, swimming, or volleyball) that cause a slight increase in breathing or heart rate for at least 10 min [22]. BMI was calculated as the weight in kilograms divided by the height in square meters.

Participants aged 8–59 years underwent whole-body DXA scans performed by professional radiology technicians using Hologic Discovery Model A densitometers (Hologic, Inc., Bedford, MA, USA). The scanned images were analyzed using Hologic APEX software (version 4.0) using the NHANES body composition analysis option. Subject to quality control and analysis, these scans provided measurements of soft tissues and bones in various body parts. The trunk and leg regions, excluding the head and arms, were delineated by angled lines defining the pelvic triangle. Fat mass index (FMI) was calculated by dividing fat mass (kg) by height squared (m²). The leg-to-trunk fat ratio (LTR) was defined as the ratio of absolute leg fat mass to trunk fat mass to assess fat accumulation in the legs relative to the trunk.

Serum ferritin levels were measured using the Roche Tina-quant immunoturbidimetric assay on the Cobas e601 clinical analyzer (Roche Diagnostics, Basel, Switzerland). Glycated hemoglobin (HbA1c) levels were measured in whole blood samples using high-performance liquid chromatography. Total cholesterol (TC) levels were measured colorimetrically using the peroxidase endpoint method. C-reactive protein (CRP), a marker of chronic inflammation, was detected using Beckman UniCel analyzers with the lowest limits of 0.015 mg/dL. Values below this limit were replaced by the detection limit divided by the square root of 2.

Weighted estimates of the population parameters were calculated following the NHANES Analytic and Reporting Guidelines to account for the complex sampling design. All statistical analyses were conducted using IBM SPSS Statistics, Version 26 (IBM Corporation, Armonk, NY, USA), and R (version 4.4.1; R Foundation). A two-sided p value of <0.05 was considered statistically significant.

The baseline characteristics of participants were presented by quartiles of serum ferritin concentrations. Continuous variables were reported as mean (standard error), while categorical variables were expressed as proportions. Linear regression models for continuous variables were used to test linear trends across serum ferritin quartiles, considering serum ferritin quartiles as continuous one. The Rao-Scott chi-square test was employed to compare categorical variables across ferritin quartiles. Multiple linear regression was applied to examine the association of serum ferritin with total and regional FMI. The serum ferritin concentrations were used as both continuous covariates and categorical covariates (four quartiles) in the linear regression models. Data were expressed as beta (β) coefficients with 95% confidence intervals (CI). Model 1 was unadjusted; Model 2 was adjusted for age, race, educational level, PIR, smoking and drinking status, physical activity, HbA1c and TC. Given the significant negative association between LTR and ferritin, trunk and leg fat were hypothesized to confound each other. Consequently, Model 3 was developed, which included all confounding factors from Model 2 and mutually adjusted leg FMI and trunk FMI.

Mediation analysis was also conducted to investigate whether the inflammatory marker CRP mediated the association between the exposure variable (serum ferritin) and the outcome (regional FMI). A thousand bootstraps were used in our analysis. The size of the indirect pathway effect, proportion of the mediating effect, and p value for the mediating effect were all evaluated.

The general characteristics of the participants are summarized in Table 1. The study included 1,282 men and 1,364 women. Median (interquartile range [IQR]) of serum ferritin levels were 0.168 μg/mL (0.104–0.269) for men and 0.053 μg/mL (0.026–0.102) for women. The mean total, trunk, and leg FMI were 8.26 ± 0.14, 4.21 ± 0.09, and 2.68 ± 0.05 kg/m² for men and 11.58 ± 0.33, 5.45 ± 0.19, and 4.34 ± 0.10 kg/m² for women, respectively. Moreover, the mean LTR was 0.69 ± 0.01 for men and 0.87 ± 0.01 for women. For men, compared with the participants in the lowest quartile, those in the highest quartile were older, more likely to consume alcohol, and had worse metabolic profiles, including higher BMI and TC. For women, BMI, HbA1c, and TC increased with increasing serum ferritin quartiles. Women in the highest serum ferritin quartile group were older and less likely to be current smokers.

As is shown in Figure 1, men and women exhibited a significant decreasing trend in LTR across serum ferritin levels (all p for trend <0.01). In men, the LTR decreased from 0.77 ± 0.02 in the lowest serum ferritin quartile to 0.62 ± 0.02 in the highest quartile. In women, the LTR decreased from 0.90 ± 0.02 in the lowest quartile to 0.80 ± 0.03 in the highest quartile.

Table 2 presents the results of the linear regression analysis examining the associations of serum ferritin levels with total and regional FMI. In men, the unadjusted model (Model 1) showed that higher serum ferritin levels were significantly associated with lower LTR (β = −0.264), higher total FMI (β = 2.653), trunk FMI (β = 2.204), and leg FMI (β = 0.526) (all p < 0.05). Adjusting for age, race, educational level, PIR, smoking, drinking, physical activity, HbA1c, and TC did not weaken the observed associations (Table 2, Model 2).

In women, the unadjusted model (Model 1) indicated that higher serum ferritin levels were significantly associated with lower LTR (β = −0.456), higher total FMI (β = 6.387), and trunk FMI (β = 4.929) (all p < 0.01). These associations persisted after adjusting for age, race, educational level, PIR, smoking, drinking, physical activity, HbA1c, and TC (Table 2, Model 2).

When trunk FMI and leg FMI were mutually adjusted, serum ferritin remained positively associated with trunk FMI (β = 0.983) and negatively with leg FMI (β = −0.324) in men. In women, serum ferritin showed a significant positive association with trunk FMI (β = 2.085).

Using the lowest quartile of serum ferritin levels as a reference (assuming zero change in total and regional FMI), significant positive trends were observed for serum ferritin levels with total FMI and trunk FMI in men and women in the fully adjusted model (all p for trend<0.01) (Fig. 2). Moreover, a significant negative association was observed between serum ferritin levels and leg FMI in men (p for trend = 0.019). In the fully adjusted model, men in the highest quartile of serum ferritin levels had a 10.2% lower LTR (95% CI: 5.0%–15.5%) than those in the lowest quartile (Fig. 2). In women, marginally significant negative trends were observed between serum ferritin levels and leg FMI (p for trend = 0.068) and LTR (p for trend = 0.102).

A mediation analysis was conducted between serum ferritin and regional FMI stratified by sex. The results of this analysis are shown in Table 3. In men, the path model showed that serum ferritin had a positive effect on CRP, and CRP was positively associated with trunk FMI (total effect: 2.925 [0.957, 3.068]; mediation effect: 0.478 [0.151, 0.933]; and proportion mediated: 16.4%). Additionally, CRP significantly mediated the association between serum ferritin and trunk FMI in women (total effect: 4.579 [1.644, 6.081]; mediation effect: 1.036 [0.126, 2.810]; and proportion mediated: 22.6%). On the contrary, CRP did not significantly mediate the relationship between serum ferritin and leg FMI in men and women.

This study utilized data from the NHANES 2017–2018 cycle, encompassing a large, nationally representative sample of US adults. It investigated the potential association of iron stores, measured as serum ferritin levels, with total and regional FMI using DXA. The findings revealed that higher ferritin concentrations were associated with total FMI, with a substantial association with increased trunk fat and decreased leg fat. These associations remained significant even after adjusting for various sociodemographic, lifestyle-related, and metabolic risk factors.

Elevated iron stores, indicated by serum ferritin concentration, have been associated with obesity [16‒18]. However, most of these studies assessed adiposity using BMI or WC. For instance, data from the China Health and Nutrition Survey indicated that elevated ferritin levels were significantly associated with an increased risk of overweight and obesity (defined by BMI) [16]. Similarly, Marta Ledesma et al. [9] found a strong positive association between serum ferritin and central obesity (diagnosed on WC). However, BMI is frequently criticized for its inability to differentiate between fat and fat-free mass and distinguish fat distribution [19]. WC is influenced by respiratory phases and stomach fullness, and is subject to significant measurement variability and reproducibility [23], which highlights the importance of precisely quantifying body fat distribution using imaging techniques as regional fat depots have been causally linked to cardiometabolic risk. Recently, a Korean study on peritoneal dialysis patients found a positive association between serum ferritin and body fat percentage [24]. Nevertheless, this study had a relatively small sample size (n = 350), and body composition was assessed using a portable whole-body bioimpedance spectroscope [24]. Similarly, Abdelhamid Kerkadi et al. [25] reported an increase in the android-to-gynoid fat ratio with higher ferritin concentrations in 1,000 healthy Qatari adults; however, the study did not perform region-specific analysis. Consequently, whether elevated iron stores are associated with fat mass in specific body regions remains unclear. Using DXA, an imaging technique that provides accurate assessment of body fat distribution, the current study extends previous findings by demonstrating that higher ferritin levels are associated with increased trunk fat and decreased leg fat in a large population of general adults.

The association between body iron stores and ectopic fat accumulation is biologically plausible. First, iron and its regulation are linked to inflammatory responses [26]. Iron potentiates the inflammatory phenotype, and inflammatory cells secrete proinflammatory cytokines such as interleukin-6 and tumor necrosis factor-alpha, which are implicated in excessive trunk fat accumulation [27, 28]. Actually, in the present analysis, the inflammatory marker CRP mediated 16.4% and 22.6% of the potential effects of serum ferritin on trunk FMI in men and women, respectively. Therefore, the inflammation hypothesis may be proposed as an important potential mechanism linking iron stores and trunk fat deposition. Second, elevated serum ferritin levels may contribute to ectopic fat deposition by reducing insulin sensitivity. Evidence increasingly supports a positive association between serum ferritin levels and insulin resistance [29]. Impaired insulin signaling can enhance free fatty acids through de novo lipogenesis and unsuppressed lipolysis [30]. Elevated free fatty acids in the bloodstream can overflow into the upper-body visceral adipose tissue, contributing to unhealthy fat distribution. Third, ferroptosis, a recently identified form of cell death, is distinct from other types in morphology and biochemistry. It is primarily defined by iron-dependent accumulation of lipid reactive oxygen species [31]. Although the role of ferroptosis in ectopic fat deposition remains unclear, preclinical studies offer crucial insights. For instance, adipocytes are particularly susceptible to oxidative stress due to their weak antioxidant defenses, which can be mitigated by overexpressing the ferroptosis regulatory gene glutathione peroxidase 4 (GPX4) [32, 33]. These results suggest that ferroptosis may play a role in reactive oxygen species-induced ectopic fat deposition, which deserves further investigation.

Evaluating the location of body fat is essential for risk assessment and treatment decisions, and our study has significant clinical and public health implications. Ectopic fat distribution, characterized by increased trunk fat, strongly predicts cardiometabolic risk, independent of overall obesity. In contrast, lower-body subcutaneous fat may offer metabolic protection by storing excess energy [34]. Understanding the association between iron stores and specific fat depots is crucial as it may help clarify the mechanisms linking iron metabolism and obesity. Our findings emphasize the importance of assessing ferritin concentrations in clinical practice as it may be a marker for unfavorable fat distribution. Notably, body iron stores can be influenced by low-cost interventions such as dietary interventions [35]. Therefore, our study suggests that modifying iron status could help prevent obesity and improve body fat distribution, meriting further investigation.

A significant strength of this analysis lies in its data source. NHANES employs population-based cluster random sampling, allowing us to derive a nationally representative sample reflective of the entire US population. Moreover, the extensive covariate data available in NHANES enables comprehensive multivariate adjustments to control for potential confounders. Our study also benefits from using DXA instead of anthropometric measurements to assess body fat distribution. Additionally, we conducted a mediation analysis to explore the pathway linking iron stores to excess trunk fat accumulation via inflammation, providing a deeper understanding of the underlying mechanisms. However, several limitations must be considered. First, the cross-sectional design of the study precludes causal inferences. Second, DXA screening was only conducted on young and middle-aged adults, limiting the generalizability of the findings to the older population.

Elevated serum ferritin levels were significantly associated with the total fat depot. More importantly, elevated ferritin levels correlated with increased trunk fat and decreased leg fat. Further longitudinal and experimental studies are needed to confirm our findings and elucidate the underlying mechanisms.

This study protocol was reviewed and approved by the National Center for Health Statistics Ethics Review Committee, Approval No. #2011-17. All participants completed written informed consent forms before participation.

The authors declare that there is no duality of interest associated with this manuscript.

This study was supported by the National Natural Science Foundation of China (82100846; 82474398), Shanghai Municipal Health Commission (20234Y0161), the Shanghai “Rising Stars of Medical Talent” Youth Development Program (SHWSRS (2024)_070), and the Shanghai Key Laboratory of Chinese Clinical Medicine (20DZ2272200). The funders played no role in the study design, data collection, management, analysis, interpretation, manuscript preparation, review, or approval.

Dr. Chi Chen, Dr. Junfei Xu, and Prof. Hao Lu are the guarantors of this work. They had full access to all the study data and were responsible for the integrity and accuracy of the data analysis. C.C., Y.C., and C.W. analyzed the data and prepared the manuscript. C.C., Y.X., J.X., and H.L. reviewed and edited the manuscript. T.H., X.Y., and G.M. contributed to the discussion.

Yuan Chen, Chao Wang, and Yanyan Xiao contributed equally to this work.

All data analyzed during this study are publicly available on the NHANES website (https://www.cdc.gov/nchs/nhanes/index.htm).