- Email: [email protected]
High ferritin and low transferrin saturation are associated with pre-diabetes among a national representative sample of U.S. adults
Clinical Nutrition 32 (2013) 1055e1060
Contents lists available at SciVerse ScienceDirect
Clinical Nutrition journal homepage: http://www.elsevier.com/locate/clnu
High ferritin and low transferrin saturation are associated with pre-diabetes among a national representative sample of U.S. adults Ching-Lung Cheung, Tommy T. Cheung, Karen S.L. Lam, Bernard M.Y. Cheung* Department of Medicine, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Pokfulam, Hong Kong
a r t i c l e i n f o
s u m m a r y
Article history: Received 18 July 2012 Accepted 26 November 2012
Background & aims: Iron overload is known to cause diabetes. However, the underlying mechanism is poorly understood. We therefore studied the association of different markers of iron metabolism, namely ferritin, erythrocyte protoporphyrin and transferrin saturation (TSAT, as defined by a percentage of transferrin that is saturated with iron) with pre-diabetes (preDM) in US adults without chronic kidney disease, anemia, and iron deficiency. Methods: Data on 2575 participants of the National Health and Nutrition Examination Survey (NHANES) 1999e2002 who were free of diabetes, chronic kidney disease, iron deficiency, and anemia were analyzed. Data on 3876 participants of the NHANES III (1988e1994) were used as replication. Homeostasis model assessment of insulin resistance (HOMA-IR), blood glycosylated hemoglobin level (HbA1C), fasting glucose, insulin, and preDM (defined as a fasting plasma glucose 100e125 mg/dl or an HBA1C value 5.7e6.4%) were measured as the outcomes. Results: Logistic regression analyses indicated independent associations of high ferritin (Ptrend ¼ 0.028) and low TSAT (Ptrend ¼ 0.029) with preDM after adjusting for sociodemographics, physical activity (active/sedementary), metabolic and inflammatory markers (triglycerides, total cholesterol, HDL cholesterol, mean arterial pressure, CRP, white cell count, and albumin), and liver enzymes (GGT, Alk phos, AST, and ALT). The NHANES III data showed similar associations. Combining the results showed a more significant association for high ferritin (Pmeta ¼ 0.016) and low TSAT (Pmeta ¼ 0.002). Moreover, TSAT was associated with HbA1C, fasting glucose, insulin, and HOMA-IR (Pmeta 0.001). Conclusions: Higher ferritin and lower TSAT are associated with higher risk of preDM in a general population without confounding diseases. Further research is needed to examine the underlying mechanism of these two indices, especially TSAT, in the pathophysiology of preDM. Ó 2013 Elsevier Ltd and European Society for Clinical Nutrition and Metabolism. All rights reserved.
Keywords: Ferritin Transferrin saturation Pre-diabetes Iron Inflammation
1. Introduction Iron overload induces free radicals formation which have deleterious effects on various cellular structures.1 For instance, iron overload is well known to cause diabetes in patients with hereditary hemochromatosis.2 It is of interest to investigate the effect of iron metabolism on the development of this disease. Although studies
Abbreviations: TSAT, transferrin saturation; NHANES, National Health and Nutrition Examination Survey; HOMA-IR, homeostasis model assessment of insulin resistance; HbA1C, glycosylated hemoglobin; CRP, C-reactive protein; GGT, gammaglutamyl transpeptidase; Alk phos, alkaline phosphatase; AST, aspartate aminotransferase; ALT, alanine transaminase; GFR, glomerular filtration rate; MAP, mean arterial pressure; IFG, impaired fasting glucose. * Corresponding author. Department of Medicine, The University of Hong Kong, Pokfulam, Hong Kong, China. Tel.: þ852 22554347; fax: þ852 28186474. E-mail address: [email protected] (B.M.Y. Cheung).
have shown that high ferritin level (a marker of iron overload) was associated with diabetes, this may not be due to an alteration in iron metabolism.3e6 Elevated serum ferritin concentrations can be caused by increased ferritin release from injured cells or increased synthesis secondary to other medical conditions,7 such as inflammation, chronic kidney or liver diseases.8 Therefore, iron overload is only one of the mechanisms leading to high ferritin levels and the association between ferritin and diabetes may be confounded. Since the regulation of ferritin is complicated and multifactorial, other markers of iron metabolism, such as transferrin saturation (TSAT) and erythrocyte protoporphyrin, are used together with serum ferritin concentration to delineate the true iron status. However, the associations of these markers with preDM and diabetes related traits have not been well evaluated. Using a national representative sample of U.S. adults who were free of co-morbidities, we aimed to examine the association of three different markers of iron metabolism with risk of preDM, and
0261-5614/$ e see front matter Ó 2013 Elsevier Ltd and European Society for Clinical Nutrition and Metabolism. All rights reserved. http://dx.doi.org/10.1016/j.clnu.2012.11.024
1056
C.-L. Cheung et al. / Clinical Nutrition 32 (2013) 1055e1060
to evaluate whether the association was independent of other potential confounders, including serum C-reactive protein (CRP) levels, total white cell count, and albumin levels for systemic inflammation,9 and serum gamma-glutamyl transpeptidase (GGT), alkaline phosphatase (Alk phos), aspartate aminotransferase (AST), and alanine transaminase (ALT) for liver disease or liver dysfunction, that have not been accounted for in the previous studies.
[fasting glucose level (in mmol/l) fasting insulin level (in mU/ml)]/ 22.5]. The specific laboratory methods used to measure levels of iron markers have been described elsewhere.13 Briefly, serum ferritin and free erythrocyte protoporphyrin were measured using immunoradiometric and fluorescence extraction assays, respectively. Serum iron and iron binding capacity were measured using a modified automated AAII-25 colorimetric method.
2. Research design and methods
2.2. Independent variables
The current study utilized data from NHANES 1999e2002.10,11 The NHANES survey included a stratified multistage probability sample which represented the civilian non-institutionalized U.S. population. Selection was based on counties, blocks, households, and individuals within households. It also included non-Hispanic blacks and Mexican-Americans to provide an adequate estimate of these ethic groups. Participants were required to sign a consent form before their participation, and ethical approval was obtained from the Human Subjects Committee of the U.S. Department of Health and Human Services. Our analysis included participants aged 20 or above whose fasting glucose levels were available (n ¼ 4337). We excluded participants with diabetes (n ¼ 539, defined as a plasma glucose level 126 mg/dl [7.0 mmol/L] after fasting for a minimum of 8 h, a glycohemoglobin (HbA1C) value 6.5%, self-reported diabetes or self-reported current use of oral hypoglycemic medication or insulin). Other exclusion criteria included chronic kidney disease (n ¼ 356, defined as a glomerular filtration rate (GFR) 60 ml/min per 1.73 m2 or GFR from 60 to 90 ml/min per 1.73 m2 with microalbuminuria),12 iron deficiency (n ¼ 281, defined as any two of the following: erythrocyte protoporphyrin level >70 mg/dl erythrocytes, ferritin level 15 mg/l, and TSAT <16%),13 anemia (n ¼ 81, defined as hemoglobin <13 g/dl for men and <12 g/dl for women),14 pregnancy (n ¼ 253, self-reported or had a positive result in pregnancy test), or missing data (n ¼ 252). The multivariable model included level of education, smoking status, alcohol use, systolic or diastolic blood pressure, body mass index (BMI), waist circumference, metabolic markers15 (triglycerides, total cholesterol and HDL cholesterol levels), liver enzymes15 (GGT, Alk phos, AST and ALT), inflammatory markers9,15 (total white blood cell count, serum CRP, and serum albumin). With the above exclusion criteria, 2575 participants were included in the analysis (46.9% women), and 990 (32.8%) of those had preDM (defined as a fasting plasma glucose 100e125 mg/dl n ¼ 851, (impaired fasting glucose, IFG) or a HbA1C value 5.7e6.4%; n ¼ 381).16,17 To confirm the findings from NHANES 1999e2002, a replication study was conducted in the third NHANES (NHANES III). NHANES III was conducted by the National Center for Health Statistics (NCHS) from 1988 to 1994 using a stratified multistage probability sample which represented the civilian non-institutionalized U.S. population. Using the same eligibility criteria, 3876 participants were included in the analysis (41.5% women). In NHANES III, only subjects with fasting glucose data (8 he24 h fasting) were included.
Independent variables were selected on the basis of the literature on diabetes risk factors.3,9,15 Age (years), gender (male/female), race/ethnicity (Hispanics, non-Hispanic white, non-Hispanic Black, others), smoking status (current/former/non-smokers), alcohol use (drinkers/non-drinkers), level of education (high school), and physical activity (active/sedentary) were assessed using a questionnaire. Individuals who had not smoked more than 100 cigarettes in their lifetime were considered nonsmokers; those who had smoked more than 100 cigarettes in their lifetime were considered former smokers if they answered negatively to the question “Do you smoke now?” and current smokers if they answered affirmatively. Individuals who had 12 alcohol drinks or more per year were considered as drinkers. The summary measure of physical activity of interest was assessed based on whether the participant engaged in moderate or vigorous activity for at least 10 min over the past 30 days. BMI (kg/m2) was calculated as weight in kilograms divided by height in meters squared. Seated systolic and diastolic blood pressures (mm/Hg) were measured using a mercury sphygmomanometer according to the American Heart Association and Seventh Joint National Committee recommendations.20 Up to three measurements were averaged for systolic and diastolic blood pressures. Mean arterial pressure (MAP, mm/Hg) was calculated as 2/3 diastolic blood pressure þ 1/3 systolic blood pressure. All definitions of the covariates in NHANES III were the same as in NHANES 1999e2002 except for physical activity. Participants who engaged in any physical activity with 3e5.9 metabolic equivalent tasks (METs) and 5 or more times per week or any physical activity with 6 or more METs and 3.0 or more times per week were classified as physically active (ideal category). Triglycerides (mg/dL), total cholesterol (mg/ dL), HDL cholesterol (mg/dL), mean arterial pressure (mm/Hg), CRP (mg/dL), total white blood cell count (SI), albumin (g/dL), GGT (U/L), Alk phos (U/L), AST (U/L), and ALT (U/L) were also included as independent variables. Details regarding the measurement of the independent variables are available at the NHANES website.10,11,21
2.1. Main outcome measurements Serum insulin, plasma glucose, and HbA1C were measured at the Diabetes Diagnostic Laboratory, University of Missourie Columbia. Serum insulin and plasma glucose were measured from fasting blood samples (from participants who had fasted for 8 h or more) using RIA and a hexokinase enzymatic method, respectively. HbA1C was measured using Primus CLC330 and Primus CLC 385 analyzers (Primus Corporation, Kansas City, MO).18 Homeostasis model assessment of insulin resistance (HOMA-IR) was used as a measure of insulin resistance (IR)19 and calculated as
2.3. Statistical analyses We categorized the level of ferritin, protoporphyrin, and TSAT into quartiles (ferritin: <87, 87e141, 142e220, >220 for men, <31, 31e54, 55e100, >100 for women; protoporphyrin: <36, 36e41, 42e51, >51 for men, <41, 42e49, 50e58, >58 for women; TSAT: <22.1, 22.1e28.5, 28.6e36.5, >36.5 for men, <19.5, 19.5e24.9, 25e31.3, >31.3 for women). Using the lowest quartile as the reference, the odds ratio (OR) and 95% CI of preDM for each quartile were calculated using multivariable logistic regression models. P-trend analysis was done by treating the quartiles as a continuous variable in the regression analyses. Later, we also used iron variables as continuous variables in the model. We used four models: model 1 was adjusted for age, sex, ethnicity (non-Hispanic whites, non-Hispanic blacks, MexicanAmericans, and others), educational level (high school), smoking (never, former, and current), drinking, physical activity (no moderate/vigorous activity, had moderate/vigorous activity), and BMI; model 2 was further adjusted for four liver enzymes (ALT, AST, GGT, and Alk phos), triglycerides, total
C.-L. Cheung et al. / Clinical Nutrition 32 (2013) 1055e1060
cholesterol, HDL cholesterol, waist circumference, and mean arterial pressure; and model 3 was further adjusted for three inflammatory markers (CRP, total white cell count, and albumin). Previous study suggested a potential interaction between iron markers and sex and race/ethnicity,22 we therefore formally evaluated the associations between different markers of iron metabolism and sex, and race/ ethnicity by including multiplicative interaction terms in the corresponding multivariable models. We also conducted association analyses between different markers and various glycemic traits (fasting glucose, insulin, HbA1C, HOMA-IR) using ANCOVA models in the same adjustment models (models 1e3). Meta-analysis was done by inverse variance method using R. We also tested whether iron supplement would affect our findings by incorporated iron supplement as an independent variable (yes/no), iron supplement was defined as any supplement that contained iron.23 Variables with skewed distributions were log-transformed before analysis. Sample weights that account for the unequal probabilities of selection, oversampling, and non-response were applied for all analyses using complex sampling module in SPSS version 18.0 software (SPSS Inc, Chicago, IL). All values presented are weighted to represent the U.S. civilian population. 3. Results Among these 2575 participants, 990 had preDM, of whom 242 had elevated HbA1C, 609 had IFG, and 139 had both elevated
1057
HbA1C and IFG. Demographic data of the participants are shown in Table 1. Those with preDM were older, more often smokers, had lower education level and physical activity. They also had higher mean BMI, waist circumference, HbA1C, fasting glucose levels, insulin concentrations, HOMA-IR, total cholesterol, triglyceride levels, MAP, ALT, GGT, Alk phos, CRP, white cell count, ferritin, and protoporphyrin levels. Conversely they had lower HDL cholesterol levels, serum albumin, and TSAT. Non-hispanic white or nonhispanic black participants had a lower risk of preDM among the participants. As shown in Table 2, higher ferritin levels and lower TSAT were associated with higher odds of preDM after adjustment for all the confounders, including various metabolic markers, liver enzymes, and inflammatory markers. Models evaluating the trend of this association were also statistically significant (P < 0.05). When ferritin and TSAT were included as continuous variables in the models, the associations in model 3 were attenuated (0.05 < P < 0.1, Table 3). Demographic data of the participants are shown in Supplementary Table 1. As shown in Table 3, similar findings were observed in NHANES III; higher ferritin levels and lower TSAT were associated with higher odds of preDM (P < 0.05). Meta-analysis of these two datasets using the inverse variance fixed-effect method showed OR of preDM being 1.32 (95% CI: 1.05e1.66; P ¼ 0.016) and 0.47 (95% CI: 0.29e0.76; P ¼ 0.002) for each 10 unit increase in ferritin and TSAT respectively. Furthermore, lower TSAT was also
Table 1 Characteristics of the study population (NHANES 1999e2002). Characteristics Female (%) Physical activity (%) Had moderate or vigorous activity Smoking never former current Drinker (%) Ethnicities (%) Mexican American Other Non-hispanic White Non-hispanic Black Education (%)High school Age (yr) Body Mass Index (kg/m2) Waist circumference (cm) Glycohemoglobin (%) Glucose, plasma (mg/dL) Insulin (uU/mL) HOMA-IR Total cholesterol (mg/dL) HDL cholesterol (mg/dL) Triglycerides (mg/dL) Mean arterial pressure (mm/Hg) ALT (U/L) AST (U/L) GGT (U/L) ALP (U/L) C-reactive protein(mg/dL) Total white blood cell count (SI) Albumin (g/dL) Ferritin (ng/mL) Protoporphyrin (ug/dL erythrocyte) Transferrin saturation (%)
Normoglycemia (n ¼ 1585)
With Pre-diabetesa (n ¼ 990)
52.3 1.4
36 1.3
69 1.5
62.1 1.8
P-value <0.001 <0.001 0.016
51.1 22.4 26.5 23.8
2.1 1.8 1.7 2.4
43.5 31.8 24.6 23.9
2.4 2 1.8 2.2
6.2 9.3 75.8 8.8
1 1.6 1.9 1.1
8.4 10.7 73.9 7
1.1 2.2 2.8 1.3
16 25.7 58.3 40.53 26.44 91.02 5.13 90.76 9.46 2.13 198.88 52.83 117.48 88.00 25.25 24.46 26.57 70.00 0.33 6.52 4.42 115.63 46.32 29.16
1.1 2 2.5 0.60 0.15 0.42 0.02 0.19 0.14 0.03 1.53 0.55 2.99 0.37 0.64 0.64 0.89 1.01 0.01 0.06 0.02 2.81 0.37 0.43
24.8 26.7 48.5 50.27 29.05 100.22 5.48 105.06 13.60 3.55 212.02 47.84 169.98 91.05 30.33 26.94 35.32 77.00 0.46 6.92 4.37 164.64 48.24 27.77
1.7 2.3 2.1 0.72 0.27 0.74 0.02 0.35 0.42 0.11 2.02 0.72 7.74 0.47 1.05 0.79 1.54 1.28 0.04 0.06 0.02 6.99 0.65 0.54
0.902 0.037
< 0.001
<0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.031 0.236 0.006 <0.001 0.001 <0.001 0.03 <0.001 0.025 0.020
Data are means SE or %SE (for categorical variables) unless otherwise indicated. All values presented are weighted to represent the U.S. civilian population, 1998e2002. P-value represents differences in means (SE) or proportions, using ANOVA or c2 test. a Among 990 had pre-diabetes, of whom 242 had elevated A1C, 609 had IFG, and 139 had both elevated A1C and IFG.
1058
C.-L. Cheung et al. / Clinical Nutrition 32 (2013) 1055e1060
Table 2 Association between markers of iron metabolism (as quartile) and pre-diabetes in NHANES 1999e2002. Iron indices
Model
Quartile 1 (reference)
Quartile 2
Quartile 3
Quartile 4
Ferritin
N 1 2 3 N 1 2 3 N 1 2 3
639 1 1 1 645 1 1 1 649 1 1 1
654 1.35 1.28 1.30 641 0.93 0.91 0.92 636 0.90 0.94 0.97
640 1.76 1.56 1.58 642 1.02 1.04 1.06 647 0.76 0.80 0.81
642 (1.33e2.44) (0.98e1.92) (1e1.97) 647 1.18 (0.79e1.78) 1.12 (0.75e1.67) 1.13 (0.76e1.69) 643 0.60 (0.43e0.84) 0.65 (0.46e0.93) 0.67 (0.46e0.98)
Protoporphyrin
Transferrin Saturation
(0.94e1.93) (0.86e1.91) (0.88e1.91) (0.71e1.22) (0.69e1.2) (0.69e1.21) (0.71e1.16) (0.73e1.22) (0.75e1.26)
(1.29e2.42) (1.12e2.17) (1.13e2.2) (0.71e1.46) (0.73e1.47) (0.75e1.5) (0.56e1.05) (0.58e1.11) (0.58e1.13)
P trend <0.001 0.035 0.028 0.388 0.511 0.445 0.004 0.016 0.029
Model 1: Adjusted for age, sex, ethnicity, educational level, smoking, drinking, physical activity and BMI; Model 2: Further adjusted for GGT, ALP, AST, ALT, triglycerides, total cholesterol, HDL cholesterol, waist circumference and mean arterial pressure; Model 3: Further adjusted for CRP, white cell count and albumin. Data are expressed as OR (95% CI).
significantly associated with higher HbA1C, fasting glucose, insulin levels, and HOMA-IR after the adjustment for various confounders (meta-analysis p-values 0.001; Table 4), but no significant associations were observed for ferritin levels. No significant associations between protoporphyrin levels, preDM, and various glycemic traits were observed in our study. For all iron indices, no significant interaction with sex or race/ethnicity was observed (P > 0.2). 4. Discussion This study examined the association of different markers of iron metabolism with preDM in a population without chronic kidney disease, anemia, or iron deficiency. In a large multi-ethnic, nationally representative sample, we found that high serum ferritin level and low TSAT were associated with preDM. This association was independent of multiple co-variables including liver enzymes, metabolic and inflammatory markers. Several studies have demonstrated a positive association between serum ferritin levels and diabetes in different populations,3e5,22,24 although the association in those studies were confounded by co-morbidities. The serum level of ferritin is associated with that of adipokines, such as adiponectin,25 retinolbinding protein-4,26 and visfatin.27 Although serum ferritin level is generally used to estimate body iron stores, it is also an acutephase reactant, therefore it is not prudent to conclude that an increase in serum ferritin levels in people with diabetes is due to an increase in body iron store. In order to support this statement, two other markers that are commonly used together with serum ferritin to define the iron status13 were also analyzed in our study. We initially observed that
low TSAT was significantly associated with preDM in NHANES 1999e2002, and this observation was successfully replicated using the NHANES III data. This observation is in concordance with the Hemochromatosis and Iron Overload Screening study, in which the participants with self-reported diabetes had higher ferritin levels but lower TSAT.22 However, in a recent study on three Denmark cohorts, elevated TSAT was found to be associated with increased risk of diabetes.28 Notably, this study did not exclude subjects with confounding diseases or include other iron markers in the model. In addition, only sex was adjusted in the model. Therefore, whether the observed associations between elevated TSAT and diabetes are confounded by other factors are unknown. Our robust findings in two NHANES data suggested that perturbed iron metabolism might underlie the development of preDM and diabetes.29 High ferritin (elevated iron stores) and low TSAT are known as functional iron deficiency.30 There are only a few clinical conditions that display a similar profile of serum ferritin and TSAT, such as chronic kidney disease requiring hemodialysis31 and anemia of chronic illness (also known as anemia of inflammation).32 In those scenarios, there is an increase in the uptake and retention of iron in the reticuloendothelial system due to chronic immune activation.32 Although participants with anemia and chronic kidney disease were all excluded in this study, it is possible that the same mechanism also contribute to the development of preDM. There was no overall significant association between ferritin and glycemic traits. Although several studies have shown that high serum ferritin levels were associated with an increased diabetes risk, low ferritin level was also found to be associated with metabolic disease in several molecular studies. Ndisang et al. demonstrated an improvement in insulin sensitivity in hemin-treated
Table 3 Association between markers of iron metabolism (as continuous variables) and pre-diabetes in NHANES 1999e2002 and NHANES III. Iron indices
Ferritin
Protoporphyrin
Transferrin Saturation
Modela
1 2 3 1 2 3 1 2 3
NHANES 1999e2002 (N ¼ 2575)
NHANES III (N ¼ 3876)
ORb
95% CI
P-value
ORb
95% CI
95% CI
Inverse variance meta-analysis (N ¼ 6451) ORb
95% CI
Pfixed-effect
Q-value
PQ
1.77 1.32 1.34 2.12 1.88 1.98 0.39 0.49 0.52
(1.29e2.44) (0.94e1.85) (0.96e1.86) (0.58e7.79) (0.51e6.87) (0.55e7.18) (0.2e0.75) (0.23e1.02) (0.23e1.13)
0.001 0.11 0.083 0.249 0.331 0.286 0.006 0.057 0.095
1.64 1.31 1.31 0.74 0.82 0.83 0.35 0.41 0.45
(1.14e2.36) (0.95e1.81) (0.95e1.79) (0.31e1.78) (0.35e1.90) (0.36e1.91) (0.20e0.62) (0.23e0.73) (0.25e0.82)
0.009 0.095 0.093 0.491 0.632 0.652 0.001 0.004 0.011
1.71 1.32 1.32 1.03 1.05 1.07 0.37 0.44 0.47
(1.35e2.18) (1.04e1.66) (1.05e1.66) (0.50e2.13) (0.52e2.13) (0.53e2.17) (0.24e0.56) (0.28e0.69) (0.29e0.76)
<0.001 0.020 0.016 0.940 0.899 0.841 <0.001 < 0.001 0.002
0.10 0.00 0.01 1.72 1.10 1.24 0.06 0.11 0.06
0.756 0.992 0.930 0.190 0.294 0.265 0.802 0.738 0.799
a Model 1: Adjusted for age, sex, ethnicity, educational level, smoking, drinking, physical activity and body mass index; Model 2: Further adjusted for gamma-glutamyl transpeptidase, alkaline phosphatase, aspartate aminotransferase, alanine transaminase, triglycerides, total cholesterol, HDL cholesterol, waist circumference and mean arterial pressure; Model 3: Further adjusted for C-reactive protein, total white blood cell count and albumin. b OR per 10 unit change.
C.-L. Cheung et al. / Clinical Nutrition 32 (2013) 1055e1060
1059
Table 4 Association between markers of iron metabolism (as continuous variables) and various glycemic traits in NHANES 1999e2002 and NHANES III. Iron indices
Ferritin
Glycemic traits
HbA1C Fasting Glucose Insulin HOMA-IRa Protoporphyrin HbA1C Fasting Glucose Insulin HOMA-IRa Transferrin HbA1C Saturation Fasting Glucose Insulin HOMA-IRa
NHANES 1999e2002 (N ¼ 2575)
NHANES III (N ¼ 3876)
b
95% CI
P-value b
0.016 1.402 0.132 0.003 0.03 0.121 1.811 0.472 0.107 4.162 1.614 0.459
(0.029 to 0.062) (0.065 to 2.738) (1.06 to 0.795) (0.246 to 0.252) (0.106 to 0.167) (4.524 to 4.283) (0.058 to 3.564) (0.034 to 0.909) (0.205 to 0.008) (7.046 to 1.277) (3.267 to 0.038) (0.895 to 0.024)
0.47 0.041 0.772 0.979 0.652 0.956 0.043 0.035 0.034 0.006 0.055 0.039
0.036 0.591 0.273 0.087 0.200 0.368 0.339 0.040 0.150 3.354 1.818 0.477
(0.096 (0.545 (0.357 (0.079 (0.338 (3.177 (2.327 (0.669 (0.244 (5.146 (3.145 (0.791
Inverse variance meta-analysis (N ¼ 6451) P-value b
95% CI to to to to to to to to to to to to
0.025) 0.24 0.003 1.727) 0.298 0.931 0.902) 0.385 0.145 0.252) 0.296 0.061 0.062) 0.006 0.084 2.441) 0.792 0.296 3.005) 0.798 1.367 0.75) 0.908 0.353 0.056) 0.003 0.129 1.563) < 0.001 3.579 0.491) 0.009 1.738 0.164) 0.004 0.471
95% CI
Pfixed-effect Q-value PQ
(0.039 to 0.034) 0.891 (0.065 to 1.797) 0.035 (0.376 to 0.666) 0.586 (0.077 to 0.199) 0.386 (0.181 to 0.014) 0.092 (2.664 to 2.072) 0.806 (0.098 to 2.832) 0.067 (0.019 to 0.725) 0.063 (0.197 to 0.062) < 0.001 (5.101 to 2.057) < 0.001 (2.773 to 0.703) 0.001 (0.726 to 0.217) < 0.001
1.80 0.82 0.50 0.30 5.40 0.01 0.82 1.03 0.39 0.22 0.04 0.00
0.179 0.365 0.479 0.585 0.02 0.926 0.366 0.311 0.530 0.641 0.851 0.948
Results are presented after adjusted for age, sex, ethnicity, educational level, smoking, drinking, physical activity, body mass index, gamma-glutamyl transpeptidase, alkaline phosphatase, aspartate aminotransferase, alanine transaminase, triglycerides, total cholesterol, HDL cholesterol, waist circumference, mean arterial pressure, C-reactive protein, total white blood cell count and albumin. a HOMA-IR was calculated as fasting glucose level (in mmol/l) fasting insulin level (in mU/ml)]/22.5.
animals that showed an increase in ferritin levels.33 Moreover, reduction in hyperglycemia was associated with increased serum ferritin levels in a non-obese type 2 diabetes rat model.34 Using a proteomics approach, ferritin heavy chain protein expression was found to be highly elevated (6.3-fold) after incubating murine primary skeletal muscle cells with adiponectin.35 Therefore the role of ferritin in diabetes is complex and multifactorial; and further studies are required to examine its role in diabetes under different conditions. Compared to ferritin, the role of TSAT in diabetes, preDM, or glycemic traits has received little attention. In the current study, lower TSAT was robustly associated with multiple glycemic traits in two NHANES data. The estimates of the effect of TSAT on glycemic traits in NHANES III and NHANES 1999e2002 were homogenous. TSAT is a measure of the amount of iron molecules available in the serum that are bound to transferrin, therefore our findings suggested that low iron saturation is associated with poor glycemic outcome. Since low iron saturation is affected by multiple factors and is not specific for iron deficiency, the underlying mechanism leading to low TSAT in preDM subjects remains to be elucidated. Our study has several strengths. We used data from a reliable nationally representative database. All participants had no confounding co-morbidities such as iron deficiency, anemia, chronic kidney disease, and pregnancy. Therefore, it provided an unique opportunity to study the association between the markers of iron metabolism and preDM in the general population. Moreover, the association of the markers, in particular, serum ferritin levels and transferrin saturations, with preDM remained robust even after adjustment of multiple confounding factors such as blood pressure, liver enzymes, and various metabolic and inflammatory markers. However, our study had limitations. This cross-sectional study cannot imply a causal relationship between the markers of iron metabolism and preDM. Moreover, both serum ferritin level and transferrin saturation can be affected by perturbed iron metabolism as well as inflammation. Since our study did not include other markers less affected by inflammation, such as soluble transferrin receptor, a cautious interpretation is required. In summary, we have observed that high ferritin and low TSAT were associated with preDM. Our findings underscore the public health importance of monitoring iron level, and also suggested that iron metabolism and inflammation may underlie the etiology of preDM. If confirmed in future prospective studies, management of iron level may have a role in the prevention of diabetes.
Disclosure summary The authors have nothing to disclose. Statement of authorship CLC and BMC contributed to conception and design of study; CLC conducted data analysis and interpretation; CLC, TTC, KSL, BMC contributed to drafting or revision of the manuscript. Conflict of Interest The authors reported no conflict of interest. Acknowledgments BMYC received support from the Faculty Research Fund, Li Ka Shing Faculty of Medicine, University of Hong Kong. CL Cheung is supported by the Faculty Development Fund, Li Ka Shing Faculty of Medicine, University of Hong Kong. We thank Dr. KL Ong of the Heart Research Institute, Sydney, Australia, for his advice. Appendix A. Supplementary data Supplementary data related to this article can be found online at http://dx.doi.org/10.1016/j.clnu.2012.11.024. References 1. Lieu PT, Heiskala M, Peterson PA, Yang Y. The roles of iron in health and disease. Mol Aspects Med FebeApr 2001;22(1e2):1e87. 2. Utzschneider KM, Kowdley KV. Hereditary hemochromatosis and diabetes mellitus: implications for clinical practice. Nat Rev Endocrinol Jan 2010;6(1): 26e33. 3. Luan de C, Li H, Li SJ, Zhao Z, Li X, Liu ZM. Body iron stores and dietary iron intake in relation to diabetes in adults in North China. Diabetes Care Feb 2008;31(2):285e6. 4. Ford ES, Cogswell ME. Diabetes and serum ferritin concentration among U.S. adults. Diabetes Care Dec 1999;22(12):1978e83. 5. Shi Z, Hu X, Yuan B, Pan X, Meyer HE, Holmboe-Ottesen G. Association between serum ferritin, hemoglobin, iron intake, and diabetes in adults in Jiangsu, China. Diabetes Care Aug 2006;29(8):1878e83. 6. Forouhi NG, Harding AH, Allison M, Sandhu MS, Welch A, Luben R, et al. Elevated serum ferritin levels predict new-onset type 2 diabetes: results from the EPIC-Norfolk prospective study. Diabetologia May 2007;50(5):949e56. 7. Adams PC, Barton JC. A diagnostic approach to hyperferritinemia with a nonelevated transferrin saturation. J Hepatol Aug 2011;55(2):453e8.
1060
C.-L. Cheung et al. / Clinical Nutrition 32 (2013) 1055e1060
8. Wang W, Knovich MA, Coffman LG, Torti FM, Torti SV. Serum ferritin: past, present and future. Biochim Biophys Acta Aug 2010;1800(8):760e9. 9. Gupta AK, Johnson WD. Prediabetes and prehypertension in disease free obese adults correlate with an exacerbated systemic proinflammatory milieu. J Inflamm (Lond) 2010;7:36. 10. Centers for Disease Control and Prevention (CDC). National Center for Health Statistics (NCHS). National Health and Nutrition Examination survey data. Hyattsville, MD: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention [accessed 01.03.12], http://www.cdc.gov/nchs/ nhanes/nhanes1999-2000/nhanes99_00.htm; 1999. 11. Centers for Disease Control and Prevention (CDC). National Center for health Statistics (NCHS). National Health and Nutrition Examination survey data. Hyattsville, MD: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention [accessed 01.03.12], http://www.cdc.gov/nchs/ nhanes/nhanes2001-2002/nhanes01_02.htm; 2001. 12. Coresh J, Selvin E, Stevens LA, Manzi J, Kusek JW, Eggers P, et al. Prevalence of chronic kidney disease in the United States. JAMA Nov 7 2007;298(17):2038e47. 13. Looker AC, Dallman PR, Carroll MD, Gunter EW, Johnson CL. Prevalence of iron deficiency in the United States. JAMA Mar 26 1997;277(12):973e6. 14. Blanc B, Finch CA, Hallberg L, Lawkowicz W, Layrisse M, Mollin DL. Nutritional anaemias. Report of a WHO Scientific Group. WHO Tech Rep Ser 1968;405:1e40. 15. Kumar PJ, Clark M, editors. Textbook of clinical medicine. 8th ed. London: Saunders; 2012. 16. Diagnosis and classification of diabetes mellitus. Diabetes Care Jan 2008;31(Suppl. 1):S55e60. 17. Standards of medical care in diabetese2010. Diabetes Care Jan 2010;33(Suppl. 1):S11e61. 18. Gould BJ, Hall PM, Cook JG. A sensitive method for the measurement of glycosylated plasma proteins using affinity chromatography. Ann Clin Biochem Jan 1984;21(Pt 1):16e21. 19. Wallace TM, Levy JC, Matthews DR. Use and abuse of HOMA modeling. Diabetes Care Jun 2004;27(6):1487e95. 20. Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, Izzo Jr JL, et al. Seventh report of the Joint National Committee on prevention, detection, evaluation, and treatment of high blood pressure. Hypertension Dec 2003;42(6):1206e52. 21. Centers for Disease Control and Prevention (CDC). National Center for Health STATISTICS (NCHS). National health and Nutrition Examination survey data. Hyattsville, MD: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention [accessed 01.03.12], http://www.cdc.gov/nchs/ nhanes/nh3data.htm; 1988.
22. Acton RT, Barton JC, Passmore LV, Adams PC, Speechley MR, Dawkins FW, et al. Relationships of serum ferritin, transferrin saturation, and HFE mutations and self-reported diabetes in the Hemochromatosis and Iron Overload Screening (HEIRS) study. Diabetes Care Sep 2006;29(9):2084e9. 23. Blanck HM, Cogswell ME, Gillespie C, Reyes M. Iron supplement use and iron status among US adults: results from the third National Health and Nutrition Examination Survey. Am J Clin Nutr Nov 2005;82(5):1024e31. 24. Sun L, Franco OH, Hu FB, Cai L, Yu Z, Li H, et al. Ferritin concentrations, metabolic syndrome, and type 2 diabetes in middle-aged and elderly chinese. J Clin Endocrinol Metab Dec 2008;93(12):4690e6. 25. Ku BJ, Kim SY, Lee TY, Park KS. Serum ferritin is inversely correlated with serum adiponectin level: population-based cross-sectional study. Dis Markers 2009;27(6):303e10. 26. Fernandez-Real JM, Moreno JM, Ricart W. Circulating retinol-binding protein-4 concentration might reflect insulin resistance-associated iron overload. Diabetes Jul 2008;57(7):1918e25. 27. Fernandez-Real JM, Moreno JM, Chico B, Lopez-Bermejo A, Ricart W. Circulating visfatin is associated with parameters of iron metabolism in subjects with altered glucose tolerance. Diabetes Care Mar 2007;30(3):616e21. 28. Ellervik C, Mandrup-Poulsen T, Andersen HU, Tybjaerg-Hansen A, Frandsen M, Birgens H, et al. Elevated transferrin saturation and risk of diabetes: three population-based studies. Diabetes Care Oct 2011;34(10):2256e8. 29. Besarab A. Evaluating iron sufficiency: a clearer view. Kidney Int Dec 2001;60(6):2412e4. 30. Macdougall IC, Hutton RD, Cavill I, Coles GA, Williams JD. Poor response to treatment of renal anaemia with erythropoietin corrected by iron given intravenously. BMJ Jul 15 1989;299(6692):157e8. 31. Rambod M, Kovesdy CP, Kalantar-Zadeh K. Combined high serum ferritin and low iron saturation in hemodialysis patients: the role of inflammation. Clin J Am Soc Nephrol Nov 2008;3(6):1691e701. 32. Weiss G, Goodnough LT. Anemia of chronic disease. N Engl J Med Mar 10 2005;352(10):1011e23. 33. Ndisang JF, Jadhav A. The heme oxygenase system attenuates pancreatic lesions and improves insulin sensitivity and glucose metabolism in deoxycorticosterone acetate hypertension. Am J Physiol Regul Integr Comp Physiol Jan 2010;298(1):R211e23. 34. Ndisang JF, Lane N, Jadhav A. Upregulation of the heme oxygenase system ameliorates postprandial and fasting hyperglycemia in type 2 diabetes. Am J Physiol Endocrinol Metab May 2009;296(5):E1029e41. 35. Ikegami Y, Inukai K, Imai K, Sakamoto Y, Katagiri H, Kurihara S, et al. Adiponectin upregulates ferritin heavy chain in skeletal muscle cells. Diabetes Jan 2009;58(1):61e70.
Contents lists available at SciVerse ScienceDirect
Clinical Nutrition journal homepage: http://www.elsevier.com/locate/clnu
High ferritin and low transferrin saturation are associated with pre-diabetes among a national representative sample of U.S. adults Ching-Lung Cheung, Tommy T. Cheung, Karen S.L. Lam, Bernard M.Y. Cheung* Department of Medicine, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Pokfulam, Hong Kong
a r t i c l e i n f o
s u m m a r y
Article history: Received 18 July 2012 Accepted 26 November 2012
Background & aims: Iron overload is known to cause diabetes. However, the underlying mechanism is poorly understood. We therefore studied the association of different markers of iron metabolism, namely ferritin, erythrocyte protoporphyrin and transferrin saturation (TSAT, as defined by a percentage of transferrin that is saturated with iron) with pre-diabetes (preDM) in US adults without chronic kidney disease, anemia, and iron deficiency. Methods: Data on 2575 participants of the National Health and Nutrition Examination Survey (NHANES) 1999e2002 who were free of diabetes, chronic kidney disease, iron deficiency, and anemia were analyzed. Data on 3876 participants of the NHANES III (1988e1994) were used as replication. Homeostasis model assessment of insulin resistance (HOMA-IR), blood glycosylated hemoglobin level (HbA1C), fasting glucose, insulin, and preDM (defined as a fasting plasma glucose 100e125 mg/dl or an HBA1C value 5.7e6.4%) were measured as the outcomes. Results: Logistic regression analyses indicated independent associations of high ferritin (Ptrend ¼ 0.028) and low TSAT (Ptrend ¼ 0.029) with preDM after adjusting for sociodemographics, physical activity (active/sedementary), metabolic and inflammatory markers (triglycerides, total cholesterol, HDL cholesterol, mean arterial pressure, CRP, white cell count, and albumin), and liver enzymes (GGT, Alk phos, AST, and ALT). The NHANES III data showed similar associations. Combining the results showed a more significant association for high ferritin (Pmeta ¼ 0.016) and low TSAT (Pmeta ¼ 0.002). Moreover, TSAT was associated with HbA1C, fasting glucose, insulin, and HOMA-IR (Pmeta 0.001). Conclusions: Higher ferritin and lower TSAT are associated with higher risk of preDM in a general population without confounding diseases. Further research is needed to examine the underlying mechanism of these two indices, especially TSAT, in the pathophysiology of preDM. Ó 2013 Elsevier Ltd and European Society for Clinical Nutrition and Metabolism. All rights reserved.
Keywords: Ferritin Transferrin saturation Pre-diabetes Iron Inflammation
1. Introduction Iron overload induces free radicals formation which have deleterious effects on various cellular structures.1 For instance, iron overload is well known to cause diabetes in patients with hereditary hemochromatosis.2 It is of interest to investigate the effect of iron metabolism on the development of this disease. Although studies
Abbreviations: TSAT, transferrin saturation; NHANES, National Health and Nutrition Examination Survey; HOMA-IR, homeostasis model assessment of insulin resistance; HbA1C, glycosylated hemoglobin; CRP, C-reactive protein; GGT, gammaglutamyl transpeptidase; Alk phos, alkaline phosphatase; AST, aspartate aminotransferase; ALT, alanine transaminase; GFR, glomerular filtration rate; MAP, mean arterial pressure; IFG, impaired fasting glucose. * Corresponding author. Department of Medicine, The University of Hong Kong, Pokfulam, Hong Kong, China. Tel.: þ852 22554347; fax: þ852 28186474. E-mail address: [email protected] (B.M.Y. Cheung).
have shown that high ferritin level (a marker of iron overload) was associated with diabetes, this may not be due to an alteration in iron metabolism.3e6 Elevated serum ferritin concentrations can be caused by increased ferritin release from injured cells or increased synthesis secondary to other medical conditions,7 such as inflammation, chronic kidney or liver diseases.8 Therefore, iron overload is only one of the mechanisms leading to high ferritin levels and the association between ferritin and diabetes may be confounded. Since the regulation of ferritin is complicated and multifactorial, other markers of iron metabolism, such as transferrin saturation (TSAT) and erythrocyte protoporphyrin, are used together with serum ferritin concentration to delineate the true iron status. However, the associations of these markers with preDM and diabetes related traits have not been well evaluated. Using a national representative sample of U.S. adults who were free of co-morbidities, we aimed to examine the association of three different markers of iron metabolism with risk of preDM, and
0261-5614/$ e see front matter Ó 2013 Elsevier Ltd and European Society for Clinical Nutrition and Metabolism. All rights reserved. http://dx.doi.org/10.1016/j.clnu.2012.11.024
1056
C.-L. Cheung et al. / Clinical Nutrition 32 (2013) 1055e1060
to evaluate whether the association was independent of other potential confounders, including serum C-reactive protein (CRP) levels, total white cell count, and albumin levels for systemic inflammation,9 and serum gamma-glutamyl transpeptidase (GGT), alkaline phosphatase (Alk phos), aspartate aminotransferase (AST), and alanine transaminase (ALT) for liver disease or liver dysfunction, that have not been accounted for in the previous studies.
[fasting glucose level (in mmol/l) fasting insulin level (in mU/ml)]/ 22.5]. The specific laboratory methods used to measure levels of iron markers have been described elsewhere.13 Briefly, serum ferritin and free erythrocyte protoporphyrin were measured using immunoradiometric and fluorescence extraction assays, respectively. Serum iron and iron binding capacity were measured using a modified automated AAII-25 colorimetric method.
2. Research design and methods
2.2. Independent variables
The current study utilized data from NHANES 1999e2002.10,11 The NHANES survey included a stratified multistage probability sample which represented the civilian non-institutionalized U.S. population. Selection was based on counties, blocks, households, and individuals within households. It also included non-Hispanic blacks and Mexican-Americans to provide an adequate estimate of these ethic groups. Participants were required to sign a consent form before their participation, and ethical approval was obtained from the Human Subjects Committee of the U.S. Department of Health and Human Services. Our analysis included participants aged 20 or above whose fasting glucose levels were available (n ¼ 4337). We excluded participants with diabetes (n ¼ 539, defined as a plasma glucose level 126 mg/dl [7.0 mmol/L] after fasting for a minimum of 8 h, a glycohemoglobin (HbA1C) value 6.5%, self-reported diabetes or self-reported current use of oral hypoglycemic medication or insulin). Other exclusion criteria included chronic kidney disease (n ¼ 356, defined as a glomerular filtration rate (GFR) 60 ml/min per 1.73 m2 or GFR from 60 to 90 ml/min per 1.73 m2 with microalbuminuria),12 iron deficiency (n ¼ 281, defined as any two of the following: erythrocyte protoporphyrin level >70 mg/dl erythrocytes, ferritin level 15 mg/l, and TSAT <16%),13 anemia (n ¼ 81, defined as hemoglobin <13 g/dl for men and <12 g/dl for women),14 pregnancy (n ¼ 253, self-reported or had a positive result in pregnancy test), or missing data (n ¼ 252). The multivariable model included level of education, smoking status, alcohol use, systolic or diastolic blood pressure, body mass index (BMI), waist circumference, metabolic markers15 (triglycerides, total cholesterol and HDL cholesterol levels), liver enzymes15 (GGT, Alk phos, AST and ALT), inflammatory markers9,15 (total white blood cell count, serum CRP, and serum albumin). With the above exclusion criteria, 2575 participants were included in the analysis (46.9% women), and 990 (32.8%) of those had preDM (defined as a fasting plasma glucose 100e125 mg/dl n ¼ 851, (impaired fasting glucose, IFG) or a HbA1C value 5.7e6.4%; n ¼ 381).16,17 To confirm the findings from NHANES 1999e2002, a replication study was conducted in the third NHANES (NHANES III). NHANES III was conducted by the National Center for Health Statistics (NCHS) from 1988 to 1994 using a stratified multistage probability sample which represented the civilian non-institutionalized U.S. population. Using the same eligibility criteria, 3876 participants were included in the analysis (41.5% women). In NHANES III, only subjects with fasting glucose data (8 he24 h fasting) were included.
Independent variables were selected on the basis of the literature on diabetes risk factors.3,9,15 Age (years), gender (male/female), race/ethnicity (Hispanics, non-Hispanic white, non-Hispanic Black, others), smoking status (current/former/non-smokers), alcohol use (drinkers/non-drinkers), level of education (
2.1. Main outcome measurements Serum insulin, plasma glucose, and HbA1C were measured at the Diabetes Diagnostic Laboratory, University of Missourie Columbia. Serum insulin and plasma glucose were measured from fasting blood samples (from participants who had fasted for 8 h or more) using RIA and a hexokinase enzymatic method, respectively. HbA1C was measured using Primus CLC330 and Primus CLC 385 analyzers (Primus Corporation, Kansas City, MO).18 Homeostasis model assessment of insulin resistance (HOMA-IR) was used as a measure of insulin resistance (IR)19 and calculated as
2.3. Statistical analyses We categorized the level of ferritin, protoporphyrin, and TSAT into quartiles (ferritin: <87, 87e141, 142e220, >220 for men, <31, 31e54, 55e100, >100 for women; protoporphyrin: <36, 36e41, 42e51, >51 for men, <41, 42e49, 50e58, >58 for women; TSAT: <22.1, 22.1e28.5, 28.6e36.5, >36.5 for men, <19.5, 19.5e24.9, 25e31.3, >31.3 for women). Using the lowest quartile as the reference, the odds ratio (OR) and 95% CI of preDM for each quartile were calculated using multivariable logistic regression models. P-trend analysis was done by treating the quartiles as a continuous variable in the regression analyses. Later, we also used iron variables as continuous variables in the model. We used four models: model 1 was adjusted for age, sex, ethnicity (non-Hispanic whites, non-Hispanic blacks, MexicanAmericans, and others), educational level (
C.-L. Cheung et al. / Clinical Nutrition 32 (2013) 1055e1060
cholesterol, HDL cholesterol, waist circumference, and mean arterial pressure; and model 3 was further adjusted for three inflammatory markers (CRP, total white cell count, and albumin). Previous study suggested a potential interaction between iron markers and sex and race/ethnicity,22 we therefore formally evaluated the associations between different markers of iron metabolism and sex, and race/ ethnicity by including multiplicative interaction terms in the corresponding multivariable models. We also conducted association analyses between different markers and various glycemic traits (fasting glucose, insulin, HbA1C, HOMA-IR) using ANCOVA models in the same adjustment models (models 1e3). Meta-analysis was done by inverse variance method using R. We also tested whether iron supplement would affect our findings by incorporated iron supplement as an independent variable (yes/no), iron supplement was defined as any supplement that contained iron.23 Variables with skewed distributions were log-transformed before analysis. Sample weights that account for the unequal probabilities of selection, oversampling, and non-response were applied for all analyses using complex sampling module in SPSS version 18.0 software (SPSS Inc, Chicago, IL). All values presented are weighted to represent the U.S. civilian population. 3. Results Among these 2575 participants, 990 had preDM, of whom 242 had elevated HbA1C, 609 had IFG, and 139 had both elevated
1057
HbA1C and IFG. Demographic data of the participants are shown in Table 1. Those with preDM were older, more often smokers, had lower education level and physical activity. They also had higher mean BMI, waist circumference, HbA1C, fasting glucose levels, insulin concentrations, HOMA-IR, total cholesterol, triglyceride levels, MAP, ALT, GGT, Alk phos, CRP, white cell count, ferritin, and protoporphyrin levels. Conversely they had lower HDL cholesterol levels, serum albumin, and TSAT. Non-hispanic white or nonhispanic black participants had a lower risk of preDM among the participants. As shown in Table 2, higher ferritin levels and lower TSAT were associated with higher odds of preDM after adjustment for all the confounders, including various metabolic markers, liver enzymes, and inflammatory markers. Models evaluating the trend of this association were also statistically significant (P < 0.05). When ferritin and TSAT were included as continuous variables in the models, the associations in model 3 were attenuated (0.05 < P < 0.1, Table 3). Demographic data of the participants are shown in Supplementary Table 1. As shown in Table 3, similar findings were observed in NHANES III; higher ferritin levels and lower TSAT were associated with higher odds of preDM (P < 0.05). Meta-analysis of these two datasets using the inverse variance fixed-effect method showed OR of preDM being 1.32 (95% CI: 1.05e1.66; P ¼ 0.016) and 0.47 (95% CI: 0.29e0.76; P ¼ 0.002) for each 10 unit increase in ferritin and TSAT respectively. Furthermore, lower TSAT was also
Table 1 Characteristics of the study population (NHANES 1999e2002). Characteristics Female (%) Physical activity (%) Had moderate or vigorous activity Smoking never former current Drinker (%) Ethnicities (%) Mexican American Other Non-hispanic White Non-hispanic Black Education (%)
Normoglycemia (n ¼ 1585)
With Pre-diabetesa (n ¼ 990)
52.3 1.4
36 1.3
69 1.5
62.1 1.8
P-value <0.001 <0.001 0.016
51.1 22.4 26.5 23.8
2.1 1.8 1.7 2.4
43.5 31.8 24.6 23.9
2.4 2 1.8 2.2
6.2 9.3 75.8 8.8
1 1.6 1.9 1.1
8.4 10.7 73.9 7
1.1 2.2 2.8 1.3
16 25.7 58.3 40.53 26.44 91.02 5.13 90.76 9.46 2.13 198.88 52.83 117.48 88.00 25.25 24.46 26.57 70.00 0.33 6.52 4.42 115.63 46.32 29.16
1.1 2 2.5 0.60 0.15 0.42 0.02 0.19 0.14 0.03 1.53 0.55 2.99 0.37 0.64 0.64 0.89 1.01 0.01 0.06 0.02 2.81 0.37 0.43
24.8 26.7 48.5 50.27 29.05 100.22 5.48 105.06 13.60 3.55 212.02 47.84 169.98 91.05 30.33 26.94 35.32 77.00 0.46 6.92 4.37 164.64 48.24 27.77
1.7 2.3 2.1 0.72 0.27 0.74 0.02 0.35 0.42 0.11 2.02 0.72 7.74 0.47 1.05 0.79 1.54 1.28 0.04 0.06 0.02 6.99 0.65 0.54
0.902 0.037
< 0.001
<0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.031 0.236 0.006 <0.001 0.001 <0.001 0.03 <0.001 0.025 0.020
Data are means SE or %SE (for categorical variables) unless otherwise indicated. All values presented are weighted to represent the U.S. civilian population, 1998e2002. P-value represents differences in means (SE) or proportions, using ANOVA or c2 test. a Among 990 had pre-diabetes, of whom 242 had elevated A1C, 609 had IFG, and 139 had both elevated A1C and IFG.
1058
C.-L. Cheung et al. / Clinical Nutrition 32 (2013) 1055e1060
Table 2 Association between markers of iron metabolism (as quartile) and pre-diabetes in NHANES 1999e2002. Iron indices
Model
Quartile 1 (reference)
Quartile 2
Quartile 3
Quartile 4
Ferritin
N 1 2 3 N 1 2 3 N 1 2 3
639 1 1 1 645 1 1 1 649 1 1 1
654 1.35 1.28 1.30 641 0.93 0.91 0.92 636 0.90 0.94 0.97
640 1.76 1.56 1.58 642 1.02 1.04 1.06 647 0.76 0.80 0.81
642 (1.33e2.44) (0.98e1.92) (1e1.97) 647 1.18 (0.79e1.78) 1.12 (0.75e1.67) 1.13 (0.76e1.69) 643 0.60 (0.43e0.84) 0.65 (0.46e0.93) 0.67 (0.46e0.98)
Protoporphyrin
Transferrin Saturation
(0.94e1.93) (0.86e1.91) (0.88e1.91) (0.71e1.22) (0.69e1.2) (0.69e1.21) (0.71e1.16) (0.73e1.22) (0.75e1.26)
(1.29e2.42) (1.12e2.17) (1.13e2.2) (0.71e1.46) (0.73e1.47) (0.75e1.5) (0.56e1.05) (0.58e1.11) (0.58e1.13)
P trend <0.001 0.035 0.028 0.388 0.511 0.445 0.004 0.016 0.029
Model 1: Adjusted for age, sex, ethnicity, educational level, smoking, drinking, physical activity and BMI; Model 2: Further adjusted for GGT, ALP, AST, ALT, triglycerides, total cholesterol, HDL cholesterol, waist circumference and mean arterial pressure; Model 3: Further adjusted for CRP, white cell count and albumin. Data are expressed as OR (95% CI).
significantly associated with higher HbA1C, fasting glucose, insulin levels, and HOMA-IR after the adjustment for various confounders (meta-analysis p-values 0.001; Table 4), but no significant associations were observed for ferritin levels. No significant associations between protoporphyrin levels, preDM, and various glycemic traits were observed in our study. For all iron indices, no significant interaction with sex or race/ethnicity was observed (P > 0.2). 4. Discussion This study examined the association of different markers of iron metabolism with preDM in a population without chronic kidney disease, anemia, or iron deficiency. In a large multi-ethnic, nationally representative sample, we found that high serum ferritin level and low TSAT were associated with preDM. This association was independent of multiple co-variables including liver enzymes, metabolic and inflammatory markers. Several studies have demonstrated a positive association between serum ferritin levels and diabetes in different populations,3e5,22,24 although the association in those studies were confounded by co-morbidities. The serum level of ferritin is associated with that of adipokines, such as adiponectin,25 retinolbinding protein-4,26 and visfatin.27 Although serum ferritin level is generally used to estimate body iron stores, it is also an acutephase reactant, therefore it is not prudent to conclude that an increase in serum ferritin levels in people with diabetes is due to an increase in body iron store. In order to support this statement, two other markers that are commonly used together with serum ferritin to define the iron status13 were also analyzed in our study. We initially observed that
low TSAT was significantly associated with preDM in NHANES 1999e2002, and this observation was successfully replicated using the NHANES III data. This observation is in concordance with the Hemochromatosis and Iron Overload Screening study, in which the participants with self-reported diabetes had higher ferritin levels but lower TSAT.22 However, in a recent study on three Denmark cohorts, elevated TSAT was found to be associated with increased risk of diabetes.28 Notably, this study did not exclude subjects with confounding diseases or include other iron markers in the model. In addition, only sex was adjusted in the model. Therefore, whether the observed associations between elevated TSAT and diabetes are confounded by other factors are unknown. Our robust findings in two NHANES data suggested that perturbed iron metabolism might underlie the development of preDM and diabetes.29 High ferritin (elevated iron stores) and low TSAT are known as functional iron deficiency.30 There are only a few clinical conditions that display a similar profile of serum ferritin and TSAT, such as chronic kidney disease requiring hemodialysis31 and anemia of chronic illness (also known as anemia of inflammation).32 In those scenarios, there is an increase in the uptake and retention of iron in the reticuloendothelial system due to chronic immune activation.32 Although participants with anemia and chronic kidney disease were all excluded in this study, it is possible that the same mechanism also contribute to the development of preDM. There was no overall significant association between ferritin and glycemic traits. Although several studies have shown that high serum ferritin levels were associated with an increased diabetes risk, low ferritin level was also found to be associated with metabolic disease in several molecular studies. Ndisang et al. demonstrated an improvement in insulin sensitivity in hemin-treated
Table 3 Association between markers of iron metabolism (as continuous variables) and pre-diabetes in NHANES 1999e2002 and NHANES III. Iron indices
Ferritin
Protoporphyrin
Transferrin Saturation
Modela
1 2 3 1 2 3 1 2 3
NHANES 1999e2002 (N ¼ 2575)
NHANES III (N ¼ 3876)
ORb
95% CI
P-value
ORb
95% CI
95% CI
Inverse variance meta-analysis (N ¼ 6451) ORb
95% CI
Pfixed-effect
Q-value
PQ
1.77 1.32 1.34 2.12 1.88 1.98 0.39 0.49 0.52
(1.29e2.44) (0.94e1.85) (0.96e1.86) (0.58e7.79) (0.51e6.87) (0.55e7.18) (0.2e0.75) (0.23e1.02) (0.23e1.13)
0.001 0.11 0.083 0.249 0.331 0.286 0.006 0.057 0.095
1.64 1.31 1.31 0.74 0.82 0.83 0.35 0.41 0.45
(1.14e2.36) (0.95e1.81) (0.95e1.79) (0.31e1.78) (0.35e1.90) (0.36e1.91) (0.20e0.62) (0.23e0.73) (0.25e0.82)
0.009 0.095 0.093 0.491 0.632 0.652 0.001 0.004 0.011
1.71 1.32 1.32 1.03 1.05 1.07 0.37 0.44 0.47
(1.35e2.18) (1.04e1.66) (1.05e1.66) (0.50e2.13) (0.52e2.13) (0.53e2.17) (0.24e0.56) (0.28e0.69) (0.29e0.76)
<0.001 0.020 0.016 0.940 0.899 0.841 <0.001 < 0.001 0.002
0.10 0.00 0.01 1.72 1.10 1.24 0.06 0.11 0.06
0.756 0.992 0.930 0.190 0.294 0.265 0.802 0.738 0.799
a Model 1: Adjusted for age, sex, ethnicity, educational level, smoking, drinking, physical activity and body mass index; Model 2: Further adjusted for gamma-glutamyl transpeptidase, alkaline phosphatase, aspartate aminotransferase, alanine transaminase, triglycerides, total cholesterol, HDL cholesterol, waist circumference and mean arterial pressure; Model 3: Further adjusted for C-reactive protein, total white blood cell count and albumin. b OR per 10 unit change.
C.-L. Cheung et al. / Clinical Nutrition 32 (2013) 1055e1060
1059
Table 4 Association between markers of iron metabolism (as continuous variables) and various glycemic traits in NHANES 1999e2002 and NHANES III. Iron indices
Ferritin
Glycemic traits
HbA1C Fasting Glucose Insulin HOMA-IRa Protoporphyrin HbA1C Fasting Glucose Insulin HOMA-IRa Transferrin HbA1C Saturation Fasting Glucose Insulin HOMA-IRa
NHANES 1999e2002 (N ¼ 2575)
NHANES III (N ¼ 3876)
b
95% CI
P-value b
0.016 1.402 0.132 0.003 0.03 0.121 1.811 0.472 0.107 4.162 1.614 0.459
(0.029 to 0.062) (0.065 to 2.738) (1.06 to 0.795) (0.246 to 0.252) (0.106 to 0.167) (4.524 to 4.283) (0.058 to 3.564) (0.034 to 0.909) (0.205 to 0.008) (7.046 to 1.277) (3.267 to 0.038) (0.895 to 0.024)
0.47 0.041 0.772 0.979 0.652 0.956 0.043 0.035 0.034 0.006 0.055 0.039
0.036 0.591 0.273 0.087 0.200 0.368 0.339 0.040 0.150 3.354 1.818 0.477
(0.096 (0.545 (0.357 (0.079 (0.338 (3.177 (2.327 (0.669 (0.244 (5.146 (3.145 (0.791
Inverse variance meta-analysis (N ¼ 6451) P-value b
95% CI to to to to to to to to to to to to
0.025) 0.24 0.003 1.727) 0.298 0.931 0.902) 0.385 0.145 0.252) 0.296 0.061 0.062) 0.006 0.084 2.441) 0.792 0.296 3.005) 0.798 1.367 0.75) 0.908 0.353 0.056) 0.003 0.129 1.563) < 0.001 3.579 0.491) 0.009 1.738 0.164) 0.004 0.471
95% CI
Pfixed-effect Q-value PQ
(0.039 to 0.034) 0.891 (0.065 to 1.797) 0.035 (0.376 to 0.666) 0.586 (0.077 to 0.199) 0.386 (0.181 to 0.014) 0.092 (2.664 to 2.072) 0.806 (0.098 to 2.832) 0.067 (0.019 to 0.725) 0.063 (0.197 to 0.062) < 0.001 (5.101 to 2.057) < 0.001 (2.773 to 0.703) 0.001 (0.726 to 0.217) < 0.001
1.80 0.82 0.50 0.30 5.40 0.01 0.82 1.03 0.39 0.22 0.04 0.00
0.179 0.365 0.479 0.585 0.02 0.926 0.366 0.311 0.530 0.641 0.851 0.948
Results are presented after adjusted for age, sex, ethnicity, educational level, smoking, drinking, physical activity, body mass index, gamma-glutamyl transpeptidase, alkaline phosphatase, aspartate aminotransferase, alanine transaminase, triglycerides, total cholesterol, HDL cholesterol, waist circumference, mean arterial pressure, C-reactive protein, total white blood cell count and albumin. a HOMA-IR was calculated as fasting glucose level (in mmol/l) fasting insulin level (in mU/ml)]/22.5.
animals that showed an increase in ferritin levels.33 Moreover, reduction in hyperglycemia was associated with increased serum ferritin levels in a non-obese type 2 diabetes rat model.34 Using a proteomics approach, ferritin heavy chain protein expression was found to be highly elevated (6.3-fold) after incubating murine primary skeletal muscle cells with adiponectin.35 Therefore the role of ferritin in diabetes is complex and multifactorial; and further studies are required to examine its role in diabetes under different conditions. Compared to ferritin, the role of TSAT in diabetes, preDM, or glycemic traits has received little attention. In the current study, lower TSAT was robustly associated with multiple glycemic traits in two NHANES data. The estimates of the effect of TSAT on glycemic traits in NHANES III and NHANES 1999e2002 were homogenous. TSAT is a measure of the amount of iron molecules available in the serum that are bound to transferrin, therefore our findings suggested that low iron saturation is associated with poor glycemic outcome. Since low iron saturation is affected by multiple factors and is not specific for iron deficiency, the underlying mechanism leading to low TSAT in preDM subjects remains to be elucidated. Our study has several strengths. We used data from a reliable nationally representative database. All participants had no confounding co-morbidities such as iron deficiency, anemia, chronic kidney disease, and pregnancy. Therefore, it provided an unique opportunity to study the association between the markers of iron metabolism and preDM in the general population. Moreover, the association of the markers, in particular, serum ferritin levels and transferrin saturations, with preDM remained robust even after adjustment of multiple confounding factors such as blood pressure, liver enzymes, and various metabolic and inflammatory markers. However, our study had limitations. This cross-sectional study cannot imply a causal relationship between the markers of iron metabolism and preDM. Moreover, both serum ferritin level and transferrin saturation can be affected by perturbed iron metabolism as well as inflammation. Since our study did not include other markers less affected by inflammation, such as soluble transferrin receptor, a cautious interpretation is required. In summary, we have observed that high ferritin and low TSAT were associated with preDM. Our findings underscore the public health importance of monitoring iron level, and also suggested that iron metabolism and inflammation may underlie the etiology of preDM. If confirmed in future prospective studies, management of iron level may have a role in the prevention of diabetes.
Disclosure summary The authors have nothing to disclose. Statement of authorship CLC and BMC contributed to conception and design of study; CLC conducted data analysis and interpretation; CLC, TTC, KSL, BMC contributed to drafting or revision of the manuscript. Conflict of Interest The authors reported no conflict of interest. Acknowledgments BMYC received support from the Faculty Research Fund, Li Ka Shing Faculty of Medicine, University of Hong Kong. CL Cheung is supported by the Faculty Development Fund, Li Ka Shing Faculty of Medicine, University of Hong Kong. We thank Dr. KL Ong of the Heart Research Institute, Sydney, Australia, for his advice. Appendix A. Supplementary data Supplementary data related to this article can be found online at http://dx.doi.org/10.1016/j.clnu.2012.11.024. References 1. Lieu PT, Heiskala M, Peterson PA, Yang Y. The roles of iron in health and disease. Mol Aspects Med FebeApr 2001;22(1e2):1e87. 2. Utzschneider KM, Kowdley KV. Hereditary hemochromatosis and diabetes mellitus: implications for clinical practice. Nat Rev Endocrinol Jan 2010;6(1): 26e33. 3. Luan de C, Li H, Li SJ, Zhao Z, Li X, Liu ZM. Body iron stores and dietary iron intake in relation to diabetes in adults in North China. Diabetes Care Feb 2008;31(2):285e6. 4. Ford ES, Cogswell ME. Diabetes and serum ferritin concentration among U.S. adults. Diabetes Care Dec 1999;22(12):1978e83. 5. Shi Z, Hu X, Yuan B, Pan X, Meyer HE, Holmboe-Ottesen G. Association between serum ferritin, hemoglobin, iron intake, and diabetes in adults in Jiangsu, China. Diabetes Care Aug 2006;29(8):1878e83. 6. Forouhi NG, Harding AH, Allison M, Sandhu MS, Welch A, Luben R, et al. Elevated serum ferritin levels predict new-onset type 2 diabetes: results from the EPIC-Norfolk prospective study. Diabetologia May 2007;50(5):949e56. 7. Adams PC, Barton JC. A diagnostic approach to hyperferritinemia with a nonelevated transferrin saturation. J Hepatol Aug 2011;55(2):453e8.
1060
C.-L. Cheung et al. / Clinical Nutrition 32 (2013) 1055e1060
8. Wang W, Knovich MA, Coffman LG, Torti FM, Torti SV. Serum ferritin: past, present and future. Biochim Biophys Acta Aug 2010;1800(8):760e9. 9. Gupta AK, Johnson WD. Prediabetes and prehypertension in disease free obese adults correlate with an exacerbated systemic proinflammatory milieu. J Inflamm (Lond) 2010;7:36. 10. Centers for Disease Control and Prevention (CDC). National Center for Health Statistics (NCHS). National Health and Nutrition Examination survey data. Hyattsville, MD: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention [accessed 01.03.12], http://www.cdc.gov/nchs/ nhanes/nhanes1999-2000/nhanes99_00.htm; 1999. 11. Centers for Disease Control and Prevention (CDC). National Center for health Statistics (NCHS). National Health and Nutrition Examination survey data. Hyattsville, MD: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention [accessed 01.03.12], http://www.cdc.gov/nchs/ nhanes/nhanes2001-2002/nhanes01_02.htm; 2001. 12. Coresh J, Selvin E, Stevens LA, Manzi J, Kusek JW, Eggers P, et al. Prevalence of chronic kidney disease in the United States. JAMA Nov 7 2007;298(17):2038e47. 13. Looker AC, Dallman PR, Carroll MD, Gunter EW, Johnson CL. Prevalence of iron deficiency in the United States. JAMA Mar 26 1997;277(12):973e6. 14. Blanc B, Finch CA, Hallberg L, Lawkowicz W, Layrisse M, Mollin DL. Nutritional anaemias. Report of a WHO Scientific Group. WHO Tech Rep Ser 1968;405:1e40. 15. Kumar PJ, Clark M, editors. Textbook of clinical medicine. 8th ed. London: Saunders; 2012. 16. Diagnosis and classification of diabetes mellitus. Diabetes Care Jan 2008;31(Suppl. 1):S55e60. 17. Standards of medical care in diabetese2010. Diabetes Care Jan 2010;33(Suppl. 1):S11e61. 18. Gould BJ, Hall PM, Cook JG. A sensitive method for the measurement of glycosylated plasma proteins using affinity chromatography. Ann Clin Biochem Jan 1984;21(Pt 1):16e21. 19. Wallace TM, Levy JC, Matthews DR. Use and abuse of HOMA modeling. Diabetes Care Jun 2004;27(6):1487e95. 20. Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, Izzo Jr JL, et al. Seventh report of the Joint National Committee on prevention, detection, evaluation, and treatment of high blood pressure. Hypertension Dec 2003;42(6):1206e52. 21. Centers for Disease Control and Prevention (CDC). National Center for Health STATISTICS (NCHS). National health and Nutrition Examination survey data. Hyattsville, MD: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention [accessed 01.03.12], http://www.cdc.gov/nchs/ nhanes/nh3data.htm; 1988.
22. Acton RT, Barton JC, Passmore LV, Adams PC, Speechley MR, Dawkins FW, et al. Relationships of serum ferritin, transferrin saturation, and HFE mutations and self-reported diabetes in the Hemochromatosis and Iron Overload Screening (HEIRS) study. Diabetes Care Sep 2006;29(9):2084e9. 23. Blanck HM, Cogswell ME, Gillespie C, Reyes M. Iron supplement use and iron status among US adults: results from the third National Health and Nutrition Examination Survey. Am J Clin Nutr Nov 2005;82(5):1024e31. 24. Sun L, Franco OH, Hu FB, Cai L, Yu Z, Li H, et al. Ferritin concentrations, metabolic syndrome, and type 2 diabetes in middle-aged and elderly chinese. J Clin Endocrinol Metab Dec 2008;93(12):4690e6. 25. Ku BJ, Kim SY, Lee TY, Park KS. Serum ferritin is inversely correlated with serum adiponectin level: population-based cross-sectional study. Dis Markers 2009;27(6):303e10. 26. Fernandez-Real JM, Moreno JM, Ricart W. Circulating retinol-binding protein-4 concentration might reflect insulin resistance-associated iron overload. Diabetes Jul 2008;57(7):1918e25. 27. Fernandez-Real JM, Moreno JM, Chico B, Lopez-Bermejo A, Ricart W. Circulating visfatin is associated with parameters of iron metabolism in subjects with altered glucose tolerance. Diabetes Care Mar 2007;30(3):616e21. 28. Ellervik C, Mandrup-Poulsen T, Andersen HU, Tybjaerg-Hansen A, Frandsen M, Birgens H, et al. Elevated transferrin saturation and risk of diabetes: three population-based studies. Diabetes Care Oct 2011;34(10):2256e8. 29. Besarab A. Evaluating iron sufficiency: a clearer view. Kidney Int Dec 2001;60(6):2412e4. 30. Macdougall IC, Hutton RD, Cavill I, Coles GA, Williams JD. Poor response to treatment of renal anaemia with erythropoietin corrected by iron given intravenously. BMJ Jul 15 1989;299(6692):157e8. 31. Rambod M, Kovesdy CP, Kalantar-Zadeh K. Combined high serum ferritin and low iron saturation in hemodialysis patients: the role of inflammation. Clin J Am Soc Nephrol Nov 2008;3(6):1691e701. 32. Weiss G, Goodnough LT. Anemia of chronic disease. N Engl J Med Mar 10 2005;352(10):1011e23. 33. Ndisang JF, Jadhav A. The heme oxygenase system attenuates pancreatic lesions and improves insulin sensitivity and glucose metabolism in deoxycorticosterone acetate hypertension. Am J Physiol Regul Integr Comp Physiol Jan 2010;298(1):R211e23. 34. Ndisang JF, Lane N, Jadhav A. Upregulation of the heme oxygenase system ameliorates postprandial and fasting hyperglycemia in type 2 diabetes. Am J Physiol Endocrinol Metab May 2009;296(5):E1029e41. 35. Ikegami Y, Inukai K, Imai K, Sakamoto Y, Katagiri H, Kurihara S, et al. Adiponectin upregulates ferritin heavy chain in skeletal muscle cells. Diabetes Jan 2009;58(1):61e70.