Diabetes mellitus (DM) is a major problem of public health around the world and the prevalence of DM has been increasing in the last decades [1]. In 2013, the estimated worldwide number of people with DM was 381.8 million adults in 219 countries, of those with approximately 5.1 million deaths attributed to DM, and a 592 million people (55%, increase) is expected by 2035 [2]. DM is the 5th leading cause of death in South Korea [3]. The prevalence of diabetes was 10.7% in males and 9.1% in females in Korea National Health and Nutrition Examination Survey (KNHANES) 2013 [4]. Although lifestyle and diet patterns play a major role in this increase [5], the pathogenesis of DM is still not fully established as multiple factors appear to be involved.

Iron is critical for metabolic processes, and its dysregulation can lead to oxidative stress, a key contributor to DM development [6]. Serum ferritin concentrations are indicators of iron stores in the body and have been implicated in the pathogenesis of many systemic inflammatory disorders [7]. Amounts of serum ferritin, as an acute phase protein, are increased in inflammatory conditions, which function as the development of DM and cardiovascular disease [8, 9].

There have been many studies, both cross-sectional and prospective, that have examined the association between elevated ferritin levels and DM. Most of these studies have reported the positive association between serum ferritin and DM. Recently, a meta-analysis of 13 studies including 5,117 type 2 DM (T2DM) and 54,451 non-diabetic individuals reported a summary relative risk of 1.91 (1.53∼2.37) [10]. However, there are also several studies indicating no association between serum ferritin and DM [11, 12]. Furthermore, several studies indicated a nonlinear association between serum ferritin levels and DM [13, 14]. Therefore, the aim of this study is to investigate the association between serum ferritin levels and DM in a Korean population. We also evaluated whether nonlinear associations existed between serum ferritin and DM.

1. Subjects

This Study is an ongoing prospective population-based study that was designed to investigate the prevalence, incidence, and risk factors for chronic disease in urban elderly population. Subjects of 9,260 aged 50 years and older were enrolled from 2007 to 2010 in Gwangju Metropolitan City in South Korea. After excluding subjects with missing key variables such as fasting glucose and HbA1c levels, a total of 8,991 (3,588 males and 5,403 females) subjects were included. All participants provided informed consent, and the study was conducted in accordance with the guidelines in The Declaration of Helsinki. The study was approved by the Institutional Review Board of Chonnam National University Hospital (IRB No. I-2008-05-056).

2. Anthropometric Data and Lifestyle Factors

Information on each subject’s medical history and lifestyle characteristics was obtained using the Korean version of the Baecke questionnaire in NWS and the International Physical Activity Questionnaire [15]. Smoking status was categorized as current smokers and nonsmokers (including ex-smokers). Current alcohol intake was categorized as nondrinkers and current-drinkers. Regular exercise was categorized as irregular or regular based on the frequency of recreational activity and exercise during a week. Education was categorized into middle school or less and high school or more. Anthropometric measurements were made in light clothing and without shoes. Body weight and body composition were measured in indoor clothing or light gown without shoes using a calibrated Inbody 520 (Biospace Co.). Height was measured to the nearest 0.1 cm. Waist circumference (WC) was measured to the nearest 0.1 cm at the midpoint between the lower border of the rib cage and the upper hip bone during expiration. Blood pressure was measured in the right upper arm using a mercury sphygmomanometer (Baumanometer; WA Baum Co, Inc) with an appropriately sized cuff after subjects rested at least 5 minutes while seated. Three consecutive measurements of systolic and diastolic blood pressures (DBP) were performed at 1-minute intervals, and the average was used in the analysis.

3. Laboratory Measurements

Blood samples were drawn from an antecubital vein in the morning after a 12-hour overnight fast. Serum was separated within 30 minutes and stored at –70℃ until analyzed. Serum total cholesterol, high-density lipoproteincholesterol (HDL), triglycerides (TG), and fasting blood glucose levels were measured using enzymatic methods. All samples were analyzed using an automatic analyzer (model 7600 chemical analyzer; Hitachi Ltd). C-reactive protein (CRP) level was determined by means of particle-enhanced immunenephelometry using BN II nephelometer (Dade Behring). HbA1c levels were analyzed by high-performance liquid chromatography using the VARIANT II system (Bio-Rad). Concentrations of serum ferritin were measured by chemiluminescent microparticle immunoassay (ARCHTECT i2000, Abbott). The coefficient of variation for the total analytic precision of this assay was ≤10%. The lower detection limits of this assay were 3.0 ng/mL. DM was defined as a fasting glucose of at least 126 mg/dL HbA1c≥6.5% or use of antidiabetic medications.

4. Statistical Analysis

Differences in the values of continuous variables and in the proportion of categorical variables for the subjects according to quintiles of serum ferritin were compared using analysis of variance (ANOVA) and the Chi-square test, respectively. Since serum ferritin distribution was marked skewed, logarithmic transformation was used before statistical testing. Second order fractional polynomial regression was used to describe the association of serum ferritin levels with fasting glucose and HbA1c levels. Serum ferritin levels were divided into quartiles for comparison. Multiple logistic regression and linear regression analyses were used to evaluate the association of the quintiles of serum ferritin levels with the risk of DM, fasting glucose and HbA1c levels after adjusting for age, sex, body mass index (BMI), WC, smoking, alcohol intake, regular exercise, anti-hypertensive medication, dyslipidemia medication, systolic blood pressure (SBP), DBP, total cholesterol, log-transformed TG, HDL cholesterol and log-transformed high sensitivity CRP (hs-CRP). Statistical analyses were conducted using Stata version 12 (Stata Co.).

General characteristics of the population, stratified by quintiles of serum ferritin, are shown in Table 1. The mean age of study participants was 65.1 (8.2) years and 39.9% were males. Subjects in the highest quintile of serum ferritin were more likely to be older, male, and highly educated, smokers, drinkers and physically active. Furthermore, they had higher BMI, higher WC, higher DBP, higher hs-CRP levels, higher TG and lower HDL-cholesterol. The prevalence of antihypertensive and SBP was lowest in the mid-quintile of ferritin levels.

Table 1

Characteristics of study subjects according to the quintile of serum ferritin levels

	Quintile 1		Quintile 2		Quintile 3		Quintile 4		Quintile 5	P-value
	<44.3		44.3∼68.2		68.3∼95.1		95.2∼140.4		>140.4
Number	1,796		1,798		1,800		1,799		1,798
Age (yr)	64.7±8.6		64.7±7.9		65.1±8		65.4±8		65.9±8.3	<0.001
Males	438 (24.3)		420 (23.3)		595 (33.0)		833 (46.3)		1,308 (72.6)	<0.001
Education	926 (51.4)		980 (54.4)		1,020 (56.6)		1,090 (60.6)		1,186 (65.9)	<0.001
Body mass index (kg/m2)	24.1±3		24.3±2.9		24.2±2.9		24.5±2.9		24.6±2.9	<0.001
Waist circumference (cm)	86.9±9.0		87.8±8.8		87.7±8.5		88.4±8.4		89.1±8.0	<0.001
Current smoking	119 (6.6)		125 (6.9)		179 (9.9)		195 (10.8)		365 (20.3)	0.053
Current alcohol intake	642 (35.7)		677 (37.6)		803 (44.6)		882 (49.0)		1,150 (63.9)	0.039
Regular exercise	530 (29.4)		554 (30.8)		603 (33.5)		617 (34.3)		597 (33.1)	0.009
Anti-hypertensive drug use	662 (36.8)		641 (35.6)		607 (33.7)		620 (34.4)		675 (37.5)	0.099
Anti-diabetes drug use	229 (12.7)		219 (12.2)		202 (11.2)		236 (13.1)		274 (15.2)	0.007
Dyslipidemia mediation	167 (9.3)		157 (8.7)		146 (8.1)		152 (8.4)		147 (8.2)	0.714
SBP (mmHg)	123.9±17.7		122.8±17.4		122.2±16.1		123.3±16.7		124.4±16.4	<0.001
DBP (mmHg)	73.9±10.1		73.7±10.3		73.9±9.9		74.5±10.3		75.5±10.5	<0.001
Total cholesterol (mg/dL)	200.5±39.5		204.2±39.3		203.5±38.8		200.9±39.6		197.2±42.2	<0.001
Triglycerides (mg/dL)	111.0 (80.0∼162.8)		113.0 (83.0∼163.0)		115.0 (84.0∼169.0)		119.0 (84.0∼166.8)		133.0 (91.5∼197.0)	<0.001
HDL cholesterol (mg/dL)	52.6±12.1		52.6±11.8		52.3±12		50.5±11.3		49.7±12.1	<0.001
Glucose (mg/dL)	106.4±22.9		106.8±23.4		107.5±23.1		110.6±23.9		116.3±29.5	<0.001
HbA1c (%)	5.85±0.96		5.78±0.68		5.78±0.89		5.84±0.95		5.93±1.02	<0.001
hs-CRP (mg/dL)	0.05 (0.02∼0.09)		0.05 (0.03∼0.11)		0.06 (0.03∼0.12)		0.07 (0.03∼0.15)		0.08 (0.04∼0.18)	<0.001
Ferritin (ng/mL)	31.0 (21.0∼38.5)		56.2 (50.0∼62.0)		80.8 (74.0∼87.4)		113.6 (103.6∼125.5)		192.5 (161.7∼253.7)	<0.001
Diabetes mellitus	352 (19.6)		317 (17.6)		310 (17.2)		408 (22.7)		525 (29.2)	<0.001

Data are means±standard deviation, medians (interquartile range) or number (percentage).

P-value was obtained using Student’s t test for continuous variables and χ² test for categorical variables, respectively.

Abbreviations: Education, middle school or less; HDL, high density lipoprotein; SBP, systolic blood pressures; DBP, diastolic blood pressures; hs-CRP, high sensitivity C-reactive protein.

Figure 1 shows the prevalence of DM by ferritin quintiles. Ferritin levels showed a J-shaped relationship with the risk of DM with higher prevalence at both high and low levels of ferritin. The age and sex adjusted prevalence of DM for each quintile was 20.4%, 18.5% and 17.5%, 22.3% and 27.2%, respectively. Table 2 shows logistic regression models of the association between the quintiles of ferritin and DM. In the age and sex adjusted model, there was a J-shape association between ferritin quintiles and DM. Subjects with upper levels (4th and 5th quintiles) and lower levels (1st quintiles) showed higher odds ratios (OR) of DM compared to those with the 3rd quintile of ferritin. Compared with subjects with mid-quintile of ferritin levels, those with 1st, 4th and 5th quintiles of ferritin levels had a high risk of DM (OR, 1.21; 95% confidence interval [CI], 1.02∼1.43; OR, 1.36; 95% CI, 1.15∼1.16; OR, 1.77; 95% CI, 1.50∼2.09; respectively). After adjusting for cardiovascular risk factors, this association was slightly attenuated but remained significant.

Fig. 1. Age and sex adjusted prevalence of diabetes mellitus according to the quintile of serum ferritin levels.

Table 2

Odds ratio with 95% confidence interval of diabetes according to the quintile of serum ferritin levels

	Serum ferritin range, ng/mL	Model 1	Model 2
Quintile 1	<44.3	1.21 (1.02∼1.43)	1.18 (0.99∼1.41)
Quintile 2	44.3∼68.2	1.07 (0.90∼1.27)	1.04 (0.87∼1.24)
Quintile 3	68.3∼95.1	1.00 (reference)	1.00 (reference)
Quintile 4	95.2∼140.4	1.36 (1.15∼1.60)	1.30 (1.09∼1.54)
Quintile 5	>140.4	1.77 (1.50∼2.09)	1.54 (1.30∼1.83)

Model 1 was adjusted for age and sex. Model 2 was adjusted for age, sex, body mass index, waist circumference, smoking, alcohol intake, regular exercise, anti-hypertensive medication, dyslipidemia medication, systolic blood pressure, diastolic blood pressure, total cholesterol, log-transformed triglyceride, HDL cholesterol and log-transformed hs-CRP.

Abbreviations: HDL, high-density lipoproteincholesterol; hs-CRP, high sensitivity C-reactive protein.

Figures 2 and 3 plot the fractional polynomial regression line of serum ferritin levels with fasting glucose and HbA1c after adjustment for age and sex. Fasting glucose showed a positive quadratic relationship with log-transformed ferritin levels, with a plateau or slight incline at the higher levels, while HbA1c levels showed a significant J-shape association with log-transformed ferritin level. Table 3 illustrates the adjusted means of fasting glucose by ferritin quintile, showing significantly higher glucose levels in the 4th and 5th quintiles. Subjects with lowest quintile had a lowest levels of fasting glucose. Compared to the lowest quintile of ferritin, those with 4th and 5th quintiles had a significant higher glucose levels. Table 4 shows the adjusted mean of HbA1c according to the quintile of serum ferritin. Similar to the result of the fractional polynomial regression, we found a J-shaped association between ferritin levels and HbA1c levels, with those patients with the mid-quintile of ferritin levels having the lowest levels of HbA1c.

Fig. 2. Fractional polynomial regression line between log-transformed ferritin and fasting glucose levels adjusted for age and sex.

Fig. 3. Fractional polynomial regression line between HbA1c (%) and fasting glucose levels adjusted for age and sex.

Table 3

Adjusted mean of fasting glucose levels (mg/dL) according to the quintile of serum ferritin levels

	Model 1	P-value	Model 2	P-value
Quintile 1	107.0 (105.8∼108.1)		107.6 (106.5∼108.7)
Quintile 2	107.3 (106.2∼108.5)	0.697	107.8 (106.7∼108.9)	0.821
Quintile 3	107.7 (106.6∼108.9)	0.354	108.2 (107.1∼109.4)	0.425
Quintile 4	110.4 (109.2∼111.5)	<0.001	110.3 (109.1∼111.4)	0.001
Quintile 5	115.2 (114.0∼116.4)	<0.001	113.7 (112.5∼114.9)	<0.001

Model 1 was adjusted for age and sex. Model 2 was adjusted for age, sex, body mass index, waist circumference, smoking, alcohol intake, regular exercise, anti-hypertensive medication, dyslipidemia medication, systolic blood pressure, diastolic blood pressure, total cholesterol, log-transformed triglyceride, HDL cholesterol and log-transformed hs-CRP.

Abbreviations: HDL, high-density lipoproteincholesterol; hs-CRP, high sensitivity C-reactive protein.

Table 4

Adjusted mean of HbA1c levels (%) according to the quintile of serum ferritin levels

	Model 1	P-value	Model 2	P-value
Quintile 1	5.84 (5.80∼5.89)	0.019	5.87 (5.83∼5.91)	0.013
Quintile 2	5.78 (5.74∼5.82)	0.900	5.80 (5.75∼5.84)	0.982
Quintile 3	5.78 (5.73∼5.82)		5.80 (5.75∼5.84)
Quintile 4	5.84 (5.79∼5.88)	0.046	5.83 (5.79∼5.88)	0.199
Quintile 5	5.94 (5.89∼5.98)	<0.001	5.88 (5.84∼5.92)	0.008

Model 1 was adjusted for age and sex. Model 2 was adjusted for age, sex, body mass index, waist circumference, smoking, alcohol intake, regular exercise, anti-hypertensive medication, dyslipidemia medication, systolic blood pressure, diastolic blood pressure, total cholesterol, log-transformed triglyceride, HDL cholesterol and log-transformed hs-CRP.

Abbreviations: HDL, high-density lipoproteincholesterol; hs-CRP, high sensitivity C-reactive protein.

In the present population-based cross-sectional study, ferritin levels displays a J-shape association with the risk of DM and a U-shape with HbA1c, and a quadratic association with fasting glucose. These association persisted after adjustment for potential confounders. In addition, this association was no significant difference between males and females. To our knowledge, this is the biggest population-base study to examine this association in Korean population.

In this study, the serum ferritin levels showed a J-shaped relationship with the risk of DM. There were six cross-sectional and four prospective studies that had more than 1,000 participants. Table 5 shows the OR according to levels of ferritin for the prospective studies that used at least three categories of ferritin levels and thus provided more information of the ferritin–T2DM association [11, 16-18]. Furthermore, two of four studies showed similar distribution with our finding J-shaped. Our findings align with those of middle-aged and elderly Chinese populations [19, 20] and the Korean general population [21, 22]. Kim et al [21] found a significantly graded relationship between glucose level and quartile group of serum ferritin in Korean general population. In a representative US population, Jehn et al [13] demonstrated that DM was more common in the highest compared with the lowest quartile of serum ferritin in females and in males. All reported higher risk for developing DM for individuals with the highest serum ferritin quintile group compared with those with the lowest. However, the above studies have produced mixed findings about difference in lowest quartile of serum ferritin level. We divided the subjects into five groups, and more specifically analyzed. We did observe that not only highest ferritin level but only a lowest ferritin level was significantly associated with increased risk of DM (OR 1.21, 95% CI 1.02∼1.43). We compared our result and four prospective studies [11, 16-18] that evaluated the ferritin-DM association in subjects of 1,000 population over. The OR for incident T2DM in a comparison of individuals in the highest ferritin levels for the nine studies in fully adjusted analysis was significant.

Table 5

Odds ratio with 95% confidence interval of diabetes according to categories of ferritin levels for prospective studies that provided results for quartiles or quintiles of ferritin levels

Lead author, year	Study	Quintile 1	Quintile 2	Quintile 3	Quintile 4	Quintile 5	P-value
Rajpathak, 2009 [16]	DPP	1.00	0.65 (0.37∼1.14)	1.05 (0.60∼1.84)	1.65 (0.90∼3.02)		0.020
Jiang, 2004 [17]	NHS	1.00	1.19 (0.81∼1.75)	1.53 (1.06∼2.23)	1.38 (0.94∼2.03)	3.20 (2.22∼4.62)	<0.001
Le, 2009 [11]	ACLS	1.00	1.26 (0.80∼1.99)	1.39 (0.88∼2.18)	1.67 (1.05∼2.66)		0.027
Forouhi, 2007 [18]	EPIC-Norfolk	1.0	0.8 (0.5∼1.2)	0.8 (0.6∼1.3)	1.3 (0.9∼1.9)	7.5 (3.7∼15.0)	<0.001

Abbreviations: DPP, Diabetes Prevention Program; NHS, Nurses’ Health Study; ACLS, Aerobics Center Longitudinal Study; EPIC-Norfolk, European Prospective Investigation of Cancer Norfolk.

The explanation for the observed J-shaped association between serum ferritin levels and elevated risk of DM is unknown. In a NHANES study, Jehn et al [13] reported the J-shape association between serum ferritin and insulin resistance in postmenopausal females. They suggested that iron deficiency may contribute to insulin resistance and several studies have shown similar results [23, 24].

Although the pathogenesis linking serum ferritin and DM development is not clearly understood, several potential mechanisms have been suggested to explain this association [25]. First, iron is a prooxidant capable of producing oxidative stress which potentially lead to insulin resistance and diabetes. Second, increased iron levels in specific tissues such as liver, pancreas, muscle and fat may increase type 2 diabetes risk. Iron accumulation in the liver may impair insulin extraction and glucose regulation, contributing to insulin resistance [26]. Increased muscle iron stores may enhance free fatty acid oxidation and thereby could interfere with glucose disposal. Excess body iron may also cause iron deposition in the pancreatic β-cells resulting in impaired insulin secretion [24, 27].

In this study, fasting glucose showed a positive quadratic relationship with ferritin levels while HbA1c levels showed a J-shape association. In most of the previous studies, fasting glucose and HbA1c did not show the similar association with ferritin levels [17, 28, 29]. This contradiction is most likely due to differences in age distribution, sex, menopausal status, and ethnicity of the study participants.

Despite our data have much larger sample size, our study also had several limitations. First, it was cross-sectional study, which prevents us from making inferences about the directionality of the relationships. Therefore, it is needed to conduct longitudinal studies evaluating these relationships among Korean population. Second, we did not measure insulin resistance. Many previous studies have shown that there are a close relation between insulin resistance and ferritin levels. Third, subjects with hemochromatosis were not checked in this study and were not eliminated in data because hemochromatosis is very rare in Korean population. Fourth, recently the role of serum ferritin on adipose tissue has attracted a lot of attention [30]. On the contrary, we did not measure the adiposity-related factors such as leptin and adipokines. Finally, Hepcidin, a key iron-regulatory hormone, was not measured in this study, limiting our understanding of iron metabolism’s role in DM.

In conclusion, our data show that high ferritin levels are associated with higher risk of DM in Korean adults. The associations may be of considerable public health importance considering that modest levels of iron storage may occur in otherwise healthy individuals probably attributable to changes in dietary habits along with iron supplementation. Considering the growing prevalence of DM and changes in dietary habits, strategies for monitoring and managing iron storage may play a crucial role in public health interventions.

본 연구는 도시에 거주하는 50세 이상의 한국인을 대상으로 혈중 페리틴 수준이 당뇨에 미치는 영향을 알아보고자 하였다. 지역사회 기반 코호트 연구에 참여한 50세 이상 한국인 8,991명(남자: 3,588, 여자: 5,403)을 대상으로 하였다. 공복 시 혈당이 126 mg/dL 이상, 당화혈색소의 농도가 6.5% 이상, 또는 당뇨약 복용 여부를 기준으로 한 가지 이상을 만족하면 당뇨로 정의하였다. 혈청 페리틴 농도는 화학발광-미세입자-면역분석법을 이용하여 측정하였다. 인구학적 요인, 신체계측 요인, 대사 요인, 생활습관 요인을 보정한 후 선형회귀분석과 로지스틱 회귀분석을 이용하여 혈중 페리틴 농도와 공복 시 혈당, 당화혈색소의 농도 및 당뇨 유병률과의 연관성을 분석하였다. 페리틴 농도와 당뇨와의 관계는 J자형 그래프를 보여주었다. 당뇨의 위험은 혈중 페리틴 중간 분위에 비해 1분위와 4,5분위가 높게 나타났다(각각 odds ratio [OR], 1.21; 95% confidence interval [CI], 1.02∼1.43; OR, 1.36; 95% CI, 1.15∼1.16; OR, 1.77; 95% CI, 1.50∼2.09). 공복 시 혈당은 로그변환 한 5분위 페리틴의 농도와 상관관계가 있었고 당화혈색소의 농도는 유의한 J자형 그래프 관계를 나타냈다. 대규모 표본집단 연구를 통하여 5분위로 구분한 혈중 페리틴 농도는 당뇨 위험도, 당화혈색소의 농도와 J자형 관계가 있었고. 공복 시 혈당 농도와는 양의 상관관계를 보였다. 이 연구결과는 체내 저장된 철이 당뇨에서 중요한 역할을 한다는 것을 보여준다.

None

This paper was supported by Wonkwang Health Science University in 2022.

None

Lee JH, Professor.

The article is prepared by a single author.

All procedures were performed in accordance with protocols approved by the Institutional Review Board of Chonnam National University Hospital (IRB No. I-2008-05-056).