Association of serum ferritin with metabolic syndrome in eight cities in China

Abstract Objective This study aims to evaluate the cross‐sectional association of serum ferritin (SF) and the risk of metabolic syndrome (MetS) and its components among adults in eight cities in China. Methods Subjects were recruited using a combination of systematic cluster random sampling and purposive sampling in eight cities in China. The sociodemographic characteristics, data of lifestyle factors, self‐reported disease history, and 24‐hr dietary intake were obtained using a validated questionnaire. Anthropometry was performed, and fasting blood was collected to test the SF, fasting blood glucose (FBG), insulin, high‐sensitivity C‐reactive protein (hs‐CRP), triglycerides (TG), and cholesterols. Logistic and linear regression analyses were conducted to investigate the associations, adjusting for age, city level, smoking, drinking, weekly moderate‐to‐vigorous activity, dietary factors, hs‐CRP, and BMI. Results Serum ferritin level is positively correlated with total cholesterol, TG, FBG, HOMA‐IR, and hs‐CRP after adjusting for age and BMI. The odds ratio (OR) for MetS in the highest quartile of SF was 2.23 (1.32, 3.77) after adjusting for men, compared with the lowest quartile. An elevated ferritin concentration was significantly related to hypertriglyceridemia (p < .001) and elevated glucose (p = .013) among men, but not among women. Furthermore, compared with Q1, the OR for insulin resistance in the ferritin Q4 group was 3.08 (1.50, 6.32) among men and 1.96 (1.19, 3.24) among women. Conclusion A positive association between elevated SF and MetS and its components including hypertriglyceridemia and elevated glucose was found in multivariate analyses among men, and SF levels are independently associated with IR.


| INTRODUC TI ON
Iron is required in the human body as a functional component of many proteins which participate in a number of vital biochemical functions, including oxygen transport, energy production, and cytochrome synthesis (Ganz & Nemeth, 2015). Ferritin is known as the form of iron storage, and serum ferritin (SF) levels are widely applied as an indicator of iron status (Ganz & Nemeth, 2015). SF is identified as an important marker of inflammatory disease. Several studies have revealed that elevated SF may be a risk factor for type 2 diabetes mellitus (T2DM; Aregbesola et al., 2018;Chen et al., 2018;Gao et al., 2017), atherosclerosis (Ma et al., 2015;Seo et al., 2015), insulin resistance (IR; Park et al., 2015), and dyslipidemia .
Nevertheless, a significant association was found between SF and MetS in Korean adults (Shim et al., 2017). In China, some studies have concentrated on exploring the association between SF and

MetS. The Fangchenggang Area Males Health and Examination
Survey (FAMHES) conducted in Guangxi discussed this association in the Chinese male population (Tang et al., 2015). Another study reported that SF levels are independently associated with MetS and IR in the Beijing area (Chen et al., 2017). A study in the Yi ethnic group showed that the risk of MetS was significantly higher in female subjects who had elevated ferritin levels (Wei et al., 2015). One study from the China Health and Nutrition Survey (CHNS) found the association among men after adjustment (Han et al., 2014), while another reported the associations among men and women (Li, Wang, Luo, Li, & Xiao, 2013). Some previous studies were conducted in one area of China, and their results were still inconsistent. Importantly, most of these studies did not consider the potential effects of dietary factors, physical activities, and inflammatory status.
In the present study, we collected clinical parameters and field investigation data from eight cities in China, and the confounding factors were controlled very well. We aim to evaluate the independent association between SF and the risk of MetS and its components among adults in eight cities in China, and to explore whether there are gender differences.

| Subjects
The data used in the present study were extracted from the Chinese Urban Adults Diet and Health Study, a cross-sectional survey conducted from March to July 2016. The sampling methods were described elsewhere . Briefly, subjects were recruited using a combination of systematic cluster random sampling and a purposive sampling method. Firstly, we selected two communities in each first-tier city (Beijing and Guangzhou) and one community in each non-first-tier city (Chengdu, Chenzhou, Jilin, Lanzhou, Wuhu and Xuchang). Secondly, a random sampling method was used to recruit participants aged 18-75 years based on resident registration. Lastly, at least 60, 60 and 50 residents were included for three age groups (18-44, 45-

| Data collection and variable definitions
The sociodemographic characteristics (e.g., age and gender), data of lifestyle factors (smoking status, alcohol intake, and physical activity), self-reported disease history (hypertension, diabetes mellitus, and hyperlipidemia), and 24-hr dietary intake before the investigation day were obtained by trained interviewers through a validated questionnaire. In the analysis, cigarette smoking status was grouped as never or current/former smoker, and if participants self-reported alcohol intake in the past month, they were defined as an alcohol consumer. Weekly moderate-to-vigorous physical activity levels were grouped as <0.5, 0.5-3.9 or >4.0 hr/week, and cities were grouped as first-tier or non-first-tier cities according to their economic status. The 24-hr dietary intake data were used to calculate energy and macronutrient intake according to the Chinese Food Composition Table (Yang, 2004(Yang, , 2009. BP, height, weight, and waist circumference (WC) were collected by trained interviewers or professional nurses in the field. BP was measured twice using an electronic BP monitor (Omron HEM-7124). BMI was calculated as weight (kg) divided by height squared (m 2 ).

| Laboratory measurements
A total of 10 ml of fasting venous blood was drawn from participants and centrifuged and separated into plasma, serum, and red blood cells, then transported to the local hospital and stored in a freezer. A total of 1,696 blood samples were collected. Finally, samples were transported to Beijing and stored at −80°C in freezers. All samples were analyzed by the Lawke Health Laboratory with strict quality control. SF was measured by the chemiluminescence method, and FBG was determined by the glucose oxidase method. Insulin level was analyzed by enzyme-linked immunosorbent assay (ELISA), and serum triglycerides, cholesterol, and lipoprotein cholesterol were measured by using the glycerol oxidase 4-chloro acid method, cholesterol oxidase aminoantipyrine phenol method, and direct method, respectively. In addition, we also used a turbidimetric immunoassay to detect serum hypersensitive C-reactive protein (hs-CRP).

| Statistical analysis
The analyses were stratified by sex. Categorical variables were described by percentages and proportion and compared using the chi-squared test. Normally and non-normally distributed continuous variables were reported by geometric means and standard deviations, and medians and interquartile ranges (25th and 75th percentages), respectively. Student's t test or the Mann-Whitney test were used when continuous variables were compared. Partial correlation coefficients (adjusted for age and BMI) between SF and risk factors of MetS components, fasting insulin, HOMA-IR, and hs-CRP were calculated in the overall studied population. Skewed variables (SF, FBG, fasting insulin, HOMA-IR, and hs-CRP) were all normalized by logarithmic transformation (ln) and were treated as continuous variables in partial correlation.
Logistic regression models were applied to analyze the relationship between SF and MetS as well as its components. Participants were categorized into quartiles based on the distribution of SF and we took the lowest quartile as the reference in both genders. The crude model was the unadjusted model. Model 1 was adjusted for age, city level, smoking, drinking, activity, dietary factors (total energy intake, fat intake, and protein intake), and hs-CRP. All data management and statistical analyses were performed with SPSS version 20.0 (International Business Machines Corporation). Statistical tests were performed two-tailed, and a p-value <.05 was considered statistically significant.

| Characteristics of studied population
A total of 1,659 participants (561 men and 1,098 women) were included, and their demographic characteristics and lifestyle factors were stratified by gender as shown in Table 1. The rates of cigarette smoking and alcohol consumption were significantly higher in men.
BMI, WC, BP, dietary energy, macronutrient intake, FBG, HbA1c, SF, total cholesterol, TG concentration, and HOMA-IR were all higher in men, whereas men had lower HDL-C, fasting insulin, and hs-CRP concentrations than women. There was no significant difference in the duration of weekly moderate-to-vigorous activity and LDL-C concentration.

| Partial correlation coefficients between SF and risk factors of MetS
The partial correlation coefficients (adjusted for age) between SF and BMI were 0.141 (p = .001) in men and 0.027 (p = .363) in women.
The partial correlation coefficients adjusted for age and BMI between SF and the indicators of MetS are shown in Table 2. Overall, SF was significantly positively associated with total cholesterol, TG, FBG, HOMA-IR, and hs-CRP; these associations could also be observed for different genders. Furthermore, SF was significantly correlated with WC, systolic BP, and HDL-C in the overall study sample but not specified for different genders. The significant correlation between SF and fasting insulin was only found in women. There was no statistically significant correlation between SF and other indicators, including diastolic BP and LDL-C.

| Multivariate analysis for MetS and its components
Logistic regression models were applied to evaluate the association between SF and MetS and its components according to sex-specific quartiles of SF in

| D ISCUSS I ON
The results of the present cross-sectional study suggest that SF level is independently associated with IR in men as well as in women.
Furthermore, we found a positive association between elevated SF and MetS and its components, including elevated glucose and hypertriglyceridemia, after adjustment for possible confounders in covariance analyses among men rather than women. However, the association of MetS was not significant after adjustment for BMI.
Several previous studies analyzed the association between SF and metabolic disease stratified by gender. In the Korean National Health and Nutrition Examination Survey (KNHANES; Shim et al., 2017), the highest SF quartile exhibited a 1.62-fold (95% CI: 1.28-2.12) increased risk of MetS in males and a 1.36-fold (95% CI: 1.09-1.69) increased risk in females compared with the lowest quartile after adjustment. In CHNS (Han et al., 2014), elevated SF levels were significantly related with a higher risk of MetS (OR 5.46, 95% CI: 3.17, 9.39) among men after adjusting for age, region, smoking, drinking, and dietary factors, but not among women. A strong gender difference in association between SF and MetS was also found in the present study. The influence of sex may be related to the characteristics of the population sampled and differences in iron depletion (Han et al., 2014).
With respect to the association between SF and the components of MetS, the findings in different studies were inconsistent and inconclusive. The Aragon Workers' Health Study (AWHS; Ledesma et al., 2015) suggested that SF is significantly associated  (Kang, Linton, & Shim, 2012) also found an association of SF with high TG and glucose concentrations in men and women, respectively. CHNS (Han et al., 2014) found that an elevated concentration of ferritins was significantly related to a higher risk of the five components of MetS among men, but not among women. A recent meta-analysis (Suarez-Ortegon et al., Body mass index is considered to be an anthropometric predictor of cardiovascular disease (Reis et al., 2015), and it was correlated with SF in males. Obesity, identified by BMI, is also related to elevated SF (Han et al., 2014;Park et al., 2014). Thus, adjusting for BMI allows investigation of whether any association exists independently of obesity. In the present study, SF level was positively correlated with total cholesterol, TG, FBG, HOMA-IR, and hs-CRP after adjusting for age and BMI, but the association of MetS did not reach the level of significance after adjustment for BMI. A meta-analysis (Suarez-Ortegon et al., 2018) also found that the meta-regression for ferritin-MetS association identified weaker associations when the studies adjusted for BMI, but the role of adjustment for BMI in evaluating confounding factors should be considered. FAMHES (Tang et al., 2015) has not yet found an association between SF and  (Kell & Pretorius, 2018). Iron metabolism in the body, consisting of iron conservation and recycling, is controlled by hepcidin. Meanwhile, hepcidin is homeostatically regulated by iron and erythropoietic activity (Ambachew & Biadgo, 2017).
Therefore, iron in a healthy person is in a state of metabolic balance.
A multitude of factors that release free iron into the bloodstream leading to raised SF can initiate iron dysregulation, such as nutritional stress (Schaffer, 2016) and oxidative stress (Nanba et al., 2016). In high iron conditions, hepcidin alone is not sufficient to regulate iron homeostasis (Parmar, Davis, Shevchuk, & Mendes, 2017). Iron dysregulation and elevated SF might result in cell death and microbial reactivation, then inflammagens such as lipopolysaccharide (LPS) and lipoteichoic acid (LTA) will be produced. These inflammagens can damage pancreatic β-cells and cause IR and metabolic disorders (Kell & Pretorius, 2018). Metabolic disorders may also influence iron dysregulation. Consequently, SF levels are associated with metabolic diseases, but it is hard to be certain about the exact sequence of causality. Prospective studies should be conducted to confirm whether elevated SF predicts MetS and its components or is just a biomarker of metabolic disturbance.
The present study was conducted using strict procedures, including the collection of questionnaires, body measurements, and blood samples in eight cities in China. The association between MetS and its components was analyzed after adjusting for possible confounders, including physical activity, dietary factors, inflammation markers, and BMI. Meanwhile, some limitations must be acknowledged. We did not identify a genetic cause or hepcidin levels underlying SF. In addition, as this is a cross-sectional study, some recall biases might exist and causation cannot be inferred, and some confounding information such as family history of disease was not collected.

| CON CLUS IONS
In conclusion, we found a positive association between elevated SF and MetS and its components including elevated glucose and hypertriglyceridemia after adjustment for possible confounders in multivariate analysis among men rather than women. SF levels were independently associated with IR in men and women, respectively. Further studies should be conducted to investigate the pathophysiologic mechanisms and to confirm whether elevated SF predicts metabolic disease.

ACKNOWLEDGMENTS
The authors thank every participant joining the survey. We also sincerely appreciate the invaluable assistance of our colleagues from

TA B L E 3 (Continued)
eight cities. The Inner Mongolia Yili Industrial Group Co., Ltd is acknowledged for supporting this project.

CO N FLI C T O F I NTE R E S T
All the authors declare no conflict of interest.

E TH I C A L S TATEM ENT
The study conforms to the Declaration of Helsinki, US and does not embrace any human or animal testing. Written informed consent from every participant had been obtained and documented, and the study's protocols and procedures were ethically reviewed and approved by the Medical Ethics Research Board of Peking University (No. IRB00001052-15059).