serum ferritin as a component of the insulin - diabetes care

E p i d e m i o I o g y / H e a 11 h S e r v i c e s / P s y e h o s o e i a I R e s e a r c h

N A L A R T I C L E

Serum Ferritin as a Component of theInsulin Resistance Syndrome

JOSE-MANUEL FERNANDEZ-REAL, MDWlFREDO RlCART-ENGEL, MDENRIC ARROYORAFAEL BALANCA

ROSER CASAMITJANA-ABELLA, MDDOLORES CABRERO, MDMIQUEL FERNANDEZ-CASTANER, MDJOAN SOLER, MD

OBJECTIVE — In epidemiological studies, serum ferritin was the second-strongest deter-minant of blood glucose (after BMI) in regression models and the third-strongest determinantof serum insulin (after BMI and age). Its concentration also correlated positively with plasmatriglycerides and apolipoprotein B concentrations, and negatively with HDL2 cholesterol. Wehypothesized that serum ferritin could be a marker of insulin resistance.

RESEARCH DESIGN AND METHODS — Oral glucose tolerance and insulin sensi-tivity (Sh minimal model method) were prospectively evaluated in 36 healthy subjects. Therelationship between serum ferritin and metabolic control (as measured by HbAlc levels) wasalso studied in 76 consecutive NIDDM patients.

RESULTS — In healthy subjects, log-transformed serum ferritin (LOGFER) correlated withbasal serum glucose (r = 0.44, P = 0.007), but not with BMI, age, systolic or diastolic blood pres-sure, total cholesterol, VLDL cholesterol, HDL cholesterol, total triglycerides, VLDL triglyc-erides, serum insulin, or HbAlc (all P = NS). Identical results were obtained when the twolowest quartiles of serum ferritin were evaluated separately. However, in the two highest quar-tiles, LOGFER correlated with BMI (0.50, P = 0.02), diastolic blood pressure (r = 0.8, P <0.0001), serum LDL cholesterol (r = 0.57, P = 0.01), VLDL cholesterol (r = 0.48, P = 0.03), totalcholesterol and HDL2 and HDL3 subfractions of HDL cholesterol (r = -0.68, -0.76, -0.55, P= 0.001, <0.0001, and 0.01, respectively), total triglycerides (r = 0.60, P = 0.006), HDL2/HDL3

quotient (P = -0 .71, P = 0.001), VLDL triglycerides (r = 0.65, P = 0.004), and serum uric acid(r = 0.51, P = 0.03), but not with systolic blood pressure (r = 0.38, P = 0.15). After adjustingfor BMI, only the correlations between LOGFER and diastolic blood pressure (r = 0.7, P =0.002) and HDL2/HDL3 quotient (r = -0.63, P = 0.01) remained significant. Strong correlationsbetween LOGFER and glucose area under the curve during oral glucose tolerance test (Pear-son's r = 0.73, P = 0.001) and Sx (r = -0.68, P = 0.001), which remained significant after con-trolling for BMI, were observed. LOGFER (p = -0.44, P = 0.01) and BMI (p = -0.52, P =0.004) constituted independent predictors of insulin sensitivity in a multivariate analysis (R2

= 0.68). In 76 consecutive NIDDM outpatients, serum glucose (P < 0.00001) and LOGFER(P = 0.03) independently predicted the value of HbAlc (R

2 = 0.40) in a multiple linear regres-sion analysis.

CONCLUSIONS — The correlations among serum ferritin and diastolic blood pressure,HDL quotient, glucose area under the curve, and Si suggest that serum ferritin could be amarker of the insulin resistance syndrome. Serum ferritin may also be an independent deter-minant of poor metabolic control in the diabetic patient.

From the Section of Endocrinology (J.-M.F.-R., WR.-E.) and the Department of Biochemistry (R.B., D.C.),Hospital de Girona, Girona; the Hormonal Laboratory (R.C.), Hospital Clinic, Barcelona; and the Endocrinol-ogy Service (M.E-C., J.S.), Hospital de Bellvitge, UHospitalet de Llobregat, Barcelona, Spain.

Address correspondence and reprint requests to Jose-Manuel Fernandez-Real, MD, Department ofEndocrinology, Hospital de Girona, Ctra. Franga s/n, 17007 Girona, Spain. E-mail: [email protected].

Received for publication 11 July 1997 and accepted in revised form 18 September 1997.Abbreviations: ASF, abdominal skinfold; AUC, area under the curve; BSF, biceps skinfold; FSIGTT, fre-

quently sampled intravenous glucose tolerance test; IRG, insulin response to glucose; LOGFER, log-trans-formed serum ferritin; OGTT, oral glucose tolerance test; SG, glucose effectiveness index; Sh insulin sensitivityindex; SSF, subscapular skinfold; TSF, triceps skinfold, WHR, waist-to-hip ratio.

Serum ferritin has been proposed as acardiovascular risk factor. In a study bySalonen et al. (1), serum ferritin con-

centration had a significant positive correla-tion with blood glucose, serum triglycerides,and serum apolipoprotein B concentrations,and an inverse correlation with serum HDL2

cholesterol, all components of what has beentermed the insulin resistance syndrome (2).In a more recent study by this group, involv-ing 1,013 men, serum ferritin was the sec-ond-strongest determinant of blood glucose(after BMI) in regression models and thethird strongest determinant of serum insulin(after BMI and age) (3). In another epidemi-ological study, men with the higher intake ofheme iron had increased serum ferritin andan increased risk of coronary heart disease.The relative risk of coronary heart disease forthe highest versus the lowest quintile ofheme iron intake was 7.3 among diabeticpatients and 1.36 among nondiabetic sub-jects (4). In fact, the changes of stored iron asa function of sex and age are quantitativelyquite similar to changes in the incidence ofcoronary heart disease as a function of ageand sex, increasing after menopause (5).

On the other hand, several lines of evi-dence suggest that iron stores may influ-ence metabolic control in the diabeticpatient. Oxidant stress is increased in dia-betes (6) because of generation of oxygenfree radicals during protein glycation andglucose autooxidation (7). This reaction iscatalyzed by transition metals as iron (8).Furthermore, extracellular ferritin may bethe source of iron for oxidative damage:when ferritin is becoming glycated, itreleases free metal ions (9,10).

Serum ferritin concentration is consid-ered a good measure of body iron stores inhealthy people (11). In order to investigatethe relationship between serum ferritin andthe components of the insulin resistancesyndrome across different degrees of storediron, we evaluated abdominal obesity, sys-tolic and diastolic blood pressure, serumHDL cholesterol, serum triglycerides, andserum ferritin in 36 healthy subjects in rela-tion to insulin sensitivity As in the recentstudy by Tuomainen et al. (3), in whichboth blood glucose and serum insulin con-centrations were only elevated at mildlyincreased serum ferritin concentrations

62 DIABETES CARE, VOLUME 21, NUMBER 1, JANUARY 1998

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/21/1/62/585122/21-1-62.pdf by guest on 20 February 2022

Fernandez-Real and Associates

Table 1—Anthropometric and clinical variables of healthy (group 1) subjects

Variables

nAge (years)BMI (kg/m2)WHRTSF (%)SSF (%)ASF (%)Systolic blood pressure (mmHg)Diastolic blood pressure (mmHg)

Men

1536.3 ± 228.2 ±1.21.01 ±0.01

148.6 ±16.3163.7 ± 17.7135.5 ±10.7130.3 ±3

80 ±2.3

Women

2133.6 ±1.4527.9 ± 1.40.94 ±0.01

140.1 ±9.1177 ±18.6

140.6 ±11.9120 ±2.5

71.2 ±2.7

(> 150 ug/1, the two upper quintiles in theirstudy), the healthy subjects were dividedinto quartiles of serum ferritin. To testwhether higher serum ferritin is associatedwith insulin resistance in diabetic patients,we also studied the relationship betweenmetabolic control (as measured by HbAlc

levels) and serum ferritin in 76 consecutivepatients with NIDDM.

RESEARCH DESIGN ANDMETHODS

SubjectsGroup 1. Thirty-six healthy volunteers,studied as outpatients at the Clinical Inves-tigation Unit, were recruited. Some of thesesubjects were studied as recently reported(12). Acute or chronic liver disease, alcoholdrinking over 80 g a day in men or 40 g aday in women, malignant processes, uremia,inflammatory diseases, hyperthyroidism,hemolytic anemia, pregnancy, recent blooddonation or transfusion, and recent intake ofiron therapy were excluded in all subjects byclinical history and biochemical work-up.None of the individuals was taking any med-ications. Blood pressure was measured in thesupine position on the right arm after a 10-min rest; a standard sphygmomanometer ofappropriate cuff size was used and the firstand fifth phases were recorded. Values usedin the analysis are the average of three read-ings taken at 5-min intervals.Anthropometric measurements. All sub-jects were evaluated, in addition to BMI(calculated as weight in kilograms dividedby height in meters squared) and the waist-to-hip ratio (WHR) (13), by the followinganthropometric parameters: triceps skin-fold thickness (TSF), biceps skinfold thick-ness (BSF), subscapular skinfold thickness(SSF), and abdominal skinfold thickness.TSF, BSF, SSF, and ASF were measured with

a skinfold caliper (Holtain, Cambridge,U.K.) as described in the World HealthOrganization Monographic Series (13). Val-ues for each variable were expressed as apercentage of the 50th percentile adjustedby sex and age as obtained from a largesample of healthy population living in thesame area covered by our hospital (14)(Table 1).

The subjects were instructed to con-sume at least 250 g of carbohydrates for the3 days preceding each of the following tests.

The oral glucose tolerance test (OGTT)was performed according to the recom-mendations of the National Diabetes DataGroup (15). Glucose was ingested in a doseof 75 g. and blood samples were obtainedat 0, 30, 60, 90, and 120 min for measure-ment of serum glucose and insulin. Theglucose and insulin total areas under thecurve (AUCs) during the OGTT weredetermined by the trapezoidal method.

The frequently sampled intravenousglucose tolerance test (FSIGTT) was per-formed 7-10 days after the OGTT in menand on days 3-8 of the menstrual cycle,27-33 days after the OGTT, in women. Foreach subject, the experimental protocolstarted between 8:00 and 8:30 A.M., after anovernight fast. A butterfly needle wasinserted into an antecubital vein, andpatency was maintained with a slow salinedrip. After a 20-min rest period, basalblood samples were drawn at —15 and —5min, after which glucose (0.3 g/kg bodyweight) was injected over 1 min starting attime 0. Additional samples were obtainedfrom a contralateral antecubital vein at 1,2,3, 4, 5, 6, 7, 8, 10, 12, 14, 16, 19, 20, 22,23, 24, 25, 27, 30, 40, 50, 60, 70, 80, 90,100,120,140,160, and 180 min. Sampleswere rapidly collected via a three-way stop-cock connected to the butterfly needle.

Informed consent was obtained fromall subjects. The protocol was approved bythe local ethics committee.Group 2. Seventy-six consecutive NIDDMoutpatients were prospectively studied(Table 4). Exclusion criteria were the sameas explained above, including acute hyper-glycemia.

Analytical methodsThe serum glucose level during the FSIGTTwas measured in duplicate by the glucoseoxidase method. The coefficient of varia-tion was 2.9%. The serum insulin levelduring the FSIGTT was measured in dupli-cate by monoclonal immunoradiometricassay (Medgenix Diagnostics, Fleunes, Bel-

Table 2—Biochemical characteristics of healthy (group 1) subjects

Characteristics

nFasting glucose (mmol/1)Fasting insulin (mU/1)Total cholesterol (mmol/1)HDL cholesterol (mmol/1)HDL2 cholesterol (mmol/1)HDL3 cholesterol (mmol/1)VLDL cholesterol (mmol/1)Total serum triglycerides (mmol/1)VLDL serum triglycerides (mmol/1)Glucose AUC (mmol/1)Insulin AUC (mU/1)IRG (mU/mmol)S, (min-1 • raU"1 • I"1)Uric acid (umol/1)

Men

155.45 ±0.2110.7 ±1.45.42 ±0.271.11 ±0.070.25 ±0.030.86 ±0.050.39 ±0.061.42 ±0.210.82 ±0.168.42 ±0.5888.6 ± 16.620.8 ±3.92.02 ±0.35312 ±18

Women

215.18 ±0.2110.5 ± 1.14.91 ±0.191.23 ±0.070.31 ±0.040.92 ± 0.040.36 ± 0.081.21 ±0.160.70 ±0.128.14 ±0.75

67.07 ± 10.225.8 ±5.52.94 ± 0.41

222.6 ± 16.2

DIABETES CARE, VOLUME 21, NUMBER 1, JANUARY 1998 63

Dow


Serum ferritin and the insulin resistance syndrome

^ 6 . 5

o

E,"IT 5.5woo3O)

1.5

r=0.44p=0.007

0 0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 3

LOGFER

Figure 1—Relationship between serum glucose and LOGFER in healthy subjects.

gium). The lowest limit of detection was4.0 mU/1. The intra-assay coefficient of vari-ation was 5.2% at a concentration of 10mU/1 and 3.4% at 130 mU/1. The inter-assay coefficients of variation were 6.9%and 4.5% at 14 and 89 mU/1, respectively.

Serum ferritin was determined byMicroparticle Enzyme ImmunoAssay(AXSYMTM, Abbot Laboratories, AbbottPark, IL), with a coefficient of variationbelow 6% for intra- and interassay. SerumHbAj and HbAlc concentrations were ana-lyzed by high-pressure liquid chromatogra-phy (Merck, Rahway NJ), with coefficientsof variation below 4%. The range for HbAlc

established in our laboratory in normal-glucose-tolerant subjects was 3.8-5.43%.Serum transferrin and uric acid (Beckman,Fullerton, CA), iron (Hitachi 917), andwhole blood hemoglobin level and hemat-ocrit (EDTA sample, Coulter Electronics,Hialeah, FL) were determined by routinelaboratory tests.

Total serum cholesterol was measuredthrough the reaction of cholesterolesterase/cholesterol oxidase/peroxidase (16)using a BM/Hitachi 747. VLDL cholesterolwas measured after ultracentrifugation at45000g. HDL cholesterol was quantifiedafter precipitation with polyethylene glycol atroom temperature (17). Total serum triglyc-erides were measured through the reaction ofglycerol-phosphate-oxidase and peroxidase(18). VLDL triglycerides were measured afterultracentrifugation at 45000g.

Data analysisData from the FSIGTT were input into

computer programs that calculate the char-acteristic metabolic parameters by fittingglucose and insulin to the minimal modelthat describes the time course of glucoseand insulin concentrations. The glucosedisappearance model, by accounting forthe effect of insulin and glucose on glucosedisappearance, provides the parameterinsulin sensitivity index (Si; 10~4 perminute per milliunit per liter, a measure ofthe effect of insulin concentrations abovethe basal level to enhance glucose disap-pearance), and glucose effectiveness index(SG; per min), defined as the effect of glu-cose itself, at basal insulin, to promote itsown disposal through uptake by massaction into the tissues and through sup-pression of endogenous glucose produc-tion. The estimation of model parameterswas performed according to the MINMODcomputer program (19). The 5! parameteris equivalent to that obtained with theeuglycemic glucose clamp (19,20).Expressed in the same units (dl/min perpU/ml), the Sl from the FSIGTT was 44 ±4% lower than the Si from the clamp in arecent study (20).

The initial pancreatic insulin responseto glucose (IRG) was estimated, accordingto the method validated by Cederholm andWibell (21), as the ratio of the incrementalinsulin AUC above the fasting level (I30—Io;milliunits per minute per liter) to the incre-mental glucose AUC above the fasting level(G3o~GO) millimoles per minute per liter)during the first 30 min of the OGTT. It wasexpressed as IRG (mU/mmol) = (I30—IoV(G30—Go).

Statistical analysisDescriptive results of continuous variablesare expressed as means ± SE. Before statis-tical analysis, normal distribution andhomogeneity of the variables were tested.Parameters that did not fulfill these tests(serum ferritin) were log-transformed. Weused the x2 test for comparisons of pro-portions, and unpaired t tests or ANOVAwith Sheffe analysis for comparisons ofquantitative variables. Pearson's correlationcoefficient or the Spearman's test were useddepending on the normal distribution ofthe variables.

Multiple regression analysis was per-formed in a step wise manner to predict Sb

glucose AUC, and HbAlc levels. Levels ofstatistical significance (two-tailed) were setat P < 0.05. Statistical analyses were per-formed with the BMDP statistical package(BMDP Statistical Software, Cork, Ireland).

RESULTS

Group 1 (healthy subjects)Thirty-six subjects (15 men and 21 pre-menopausal women), mean age 34.7 ± 1.2(range 19-46) years, mean BMI 28 ± 0.97(range 16.8-37.9), were evaluated (Table1). Note that WHRs are greater in our pop-ulation than in other Caucasian popula-tions, as recently reported (12). Biochemicalcharacteristics are shown on Table 2. Log-transformed serum ferritin (LOGFER) cor-related with basal serum glucose (r = 0.44,P = 0.007, Fig. 1), blood hemoglobin (r =0.72, P < 0.0001), and hematocrit (r =0.71, P < 0.0001) but not with BMI, age,systolic or diastolic blood pressure, serumtotal cholesterol, VLDL cholesterol, HDLcholesterol, total triglycerides, VLDL triglyc-erides, serum insulin, or HbAk (all P = NS).There were no relationships betweenLOGFER and glucose AUC after OGTT,insulin AUC, acute IRG, or S{.

Identical results were obtained whenthe two lowest quartiles of serum ferritinwere evaluated separately. However, in thetwo highest quartiles, LOGFER correlatedwith BMI (0.50, P = 0.02), diastolic bloodpressure (r = 0.8, P < 0.0001), serum LDLcholesterol (r = 0.57, P = 0.01), VLDL cho-lesterol (r = 0.48, P = 0.03), total and HDL2

and HDL3 subfractions of HDL cholesterol(r = -0.68, -0.76, -0.55, P = 0.001,<0.0001 and 0.01, respectively), totaltriglycerides (r = 0.60, P = 0.006),HDL2/HDL3 quotient (P = -0 .71 , P =0.001),VLDL triglycerides (r = 0.65, P =0.004), and serum uric acid (r = 0.51, P =


Dow



3 4

DE 3

e 2

-1

r=-0.68p=0.001

1.5 1.8 2.1 2.4

LOGFER

Figure 2—Relationship between insulin sensitivity and LOGFER in healthy subjects in the 3rd and 4thquartiles of serum ferritin.

0.03), but not with systolic blood pressure(r = 0.38, P = 0.15), WHR, ASF, or SSEOnly the correlations between LOGFERand diastolic blood pressure (r = 0.7, P =0.002) and HDL2/HDL3 quotient (r =—0.63,P = 0.01) remained significant afteradjusting for BM1. Strong correlationsbetween LOGFER and glucose AUC (Pear-son's r = 0.73, P = 0.001) and Si (r = -0.68,P = 0.001, Fig. 2), which remained signifi-cant after controlling for BMI, wereobserved. LOGFER ((3 = -0.44, P = 0.01)and BMI (|3 = -0.52, P = 0.004) wereindependent predictors of insulin sensitiv-ity in a multivariate analysis (R2 = 0.68). Inanother model, with glucose AUC as thedependent variable and LOGFER and BMIas independent variables, only LOGFERcontributed to the variance of glucose AUC(P = 0.49, P = 0.03).

The subjects in the two highest quartileswere comparable in BMI, age, WHR, meanbasal serum glucose and insulin levels, alco-hol, and smoking (Table 3). However,greater glucose AUC, insulin AUC, 2-hOGTT serum glucose, and HbAlc and lowerinsulin sensitivity were observed in thosesubjects in the highest quartile (Table 3).

Group 2 (NIDDM patients)Clinical and demographic variables areshown on Table 4. Mean duration ofNIDDM was 7.3 ± 1 years. The medianserum ferritin was 174.6 ± 18.2 pg/1. Serumferritin levels below 200 pg/1 were found in70% of patients, between 200 and 300 pg/1in 16%, and above 300 pg/1 in 14%. Tenpatients were taking antihypertensives, and

14 were taking anticholesteremic agents.Serum ferritin levels were comparable inpatients with (8 patients) or without coro-nary heart disease, previous stroke (4patients), claudication (6 patients), diabeticretinopathy (14 patients), microalbuminuria(22 patients), peripheral neuropathy(defined as bilateral absence of vibrationsense and/or bilateral absence of Achillestendon reflexes, 16 patients), smoking (22patients) or hypertension (10 patients).Serum ferritin was also comparable betweenpatients taking antihypertensives or anti-cholesteremic agents and patients who didnot take these drugs. When patients taking

antihypertensives or anticholesteremicagents were excluded from the analysis,those with serum ferritin above the medianwere more likely to exhibit arterial hyper-tension (P = 0.04) and hypertriglyceridemia(P = 0.03).

HbAj and HbAlc levels (mean 9.8 ± 0.3and 8.66 ± 0.2%, respectively) correlatedwith LOGFER (r = 0.33, P = 0.018 and r =0.25, P = 0.03, respectively; Fig. 3). HbAk

also correlated with serum glucose (r =0.57, P < 0.0001) and weakly with age (r= 0.19, P = 0.1) but not with BMI (r = 0.17,P = NS) or serum iron. The correlationbetween HbAlc levels and serum ferritinremained significant after controlling forage (r = 0.25, P = 0.03). Serum ferritin neg-atively correlated with serum transferrin (r= -0.28, P = 0.03). We built a multiple lin-ear regression in a stepwise manner to pre-dict serum HbAlc. Variables tested in thismodel were those that correlated in theunivariant analysis and included glucose,age, and LOGFER. Serum glucose (P <0.00001) and LOGFER (P = 0.03) inde-pendently predicted the value of HbAlc inNIDDM patients (R2 = 0.40). In this regres-sion model, the same results were obtainedif HbA! levels were used as the dependentvariable (R2 = 0.42).

CONCLUSIONS— The main findingin this study was the relationship betweenserum ferritin and several components ofthe insulin resistance syndrome in the twohighest quartiles of serum ferritin. Interest-ingly, in the recent study by Toumainen et

Table 3—Comparison between the 3rd and 4th quartiles of serum ferritin

nSex (M/F)Age (years)BMI (kg/m2)WHRAlcohol (g/day)Fasting glucose (mmol/1)Fasting insulin (mU/l)2-h OGTT glucose (mmol/1)2-h OGTT insulin (mU/l)HbAlc (%)Glucose AUC (mmol/1)Insulin AUC (mU/l)IRG (mU/mmol)S^mirr1 • mU"1 -I"1)ScOrun-1)

3rd quartile(32-125 ug/1)

97/2

39.6 ± 1.729.2 ±1.41.01 ±0.0220.2 ±105.1 ±0.18

9.51 ±0.65.5 ±0.47

48.4 ±6.44.67 ± 0.0912.8 ±0.84

124.6 ±23.727.8 ±7.63.1 ±0.3

0.0176 ± 0.003

4th quartile(> 125 ug/1)

99/0

33.1 ±2.526 ±1.8

0.97 ±0.0225.6 ±155.56 ±0.1211.1 ±2.48.98 ±1.23

156.8 ±425.17 ±0.1419.1 ±1.45

221.2 ±33.614.4 ± 1.71.47 ±0.4

0.0165 ±0.002

P value

NSNSNSNSNSNSNS

0.020.040.010.0030.030.130.01NS


Dow


Serum ferritin and the insulin resistance syndrome

Table 4—Clinical characteristics ofNIDDM(group 2) patients

Characteristic

nAge (years)Sex (M/F)BMI (kg/m2)WHR

MenWomen

DietHypoglycemic agentsInsulin

Value

765 6 ± 1 434/42

30.5 ±1.5

1.11 ±0.120.92 ±0.1134 (45%)24 (31%)28 (37%)

al. (3), the elevation in serum insulin con-centration, a marker of insulin resistance(22), was more apparent in their two high-est quintiles

In our study, serum ferritin was propor-tional to serum glucose concentration, as hasbeen found in an epidemiological basis(1,3), and it was directly correlated withglucose AUC after OGTT and several meas-urements of insulin resistance. Furthermore,serum ferritin was positively correlated withdiastolic blood pressure after adjusting forBMI. Oshaug et al. (23) have recently foundthat diastolic blood pressure was a signifi-cant predictor for serum ferritin levels. Inthis sense, the therapeutic efficiency of phle-botomy has been demonstrated in resistanthypertension (24) and posttransplant hyper-tension (25), when the body iron stores havebeen depleted. Increased serum ferritin hasalso been found to be a predictor of highermorbidity and mortality after acute ischemicstroke (26). Finally, serum ferritin wasinversely correlated with HDL2/HDL3 quo-tient. In the report by Oshaug et al. (23),HDL cholesterol was also a significant pre-dictor for serum ferritin levels. An interac-tion between iron stores and classic riskfactors for atherosclerosis has been proposed(27). The previous reports and our findingsconcerning the independent relationshipbetween diastolic blood pressure, HDL quo-tient, and serum ferritin concentration, sup-port the idea of serum ferritin as acomponent of the insulin resistance syn-drome. Serum ferritin also correlated withuric acid, a simple marker of insulin resis-tance and an inherent component of theinsulin resistance syndrome (28). This is thefirst report in which insulin resistance hasbeen found to be independently related toserum ferritin levels in individuals without

14

13

12

11

£ 1 0

2 9<•° ft

7

6

5

4

r=0.25p=0.03

0 0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 3 3.3 3.6 3.9

LOGFER

Figure 3—Relationship between LOGFER and HbA}c in group 2 (NIDDM) patients.

evidence of iron overload. The cross-sec-tional design, however, does not enable us toinfer causality between serum ferritin andinsulin resistance.

Moirand et al. (29) have recentlydescribed a new syndrome of liver ironoverload with normal transferrin satura-tion. In all the patients (with high-normalferritin levels), hemochromatosis wasexcluded by HLA-A3 antigen analysis.Most of the patients (95%) had one ormore of the following conditions: obesity,hyperlipidemia, abnormal glucose metabo-lism, and hypertension. However, insulinresistance was not measured. Again, thesefindings fit well with the hypothesis ofserum ferritin as a component of theinsulin resistance syndrome.

In fact, serum ferritin level may consti-tute a marker of insulin resistance in sub-jects over a threshold of iron stores (>125pg/1 in our study). In the recent report byTuomainen et al. (3), both blood glucoseand serum insulin were elevated at mildlyincreased (150 ug/1) serum ferritin concen-trations, a level very similar to that obtainedin our study. Below this threshold of serumferritin, the interaction between serum fer-ritin and insulin resistance is possiblydiluted, and the influence of other factors(for example, obesity) predominates. Therelationship between serum ferritin andhistochemical assessment of stainable tissueiron contributes to defining threshold val-ues for serum ferritin that indicateexhausted, small, normal, ample, andincreased iron stores (30). However, thebarrier between "normal" and "small" or

"ample" iron stores is not well defined andremains controversial. Only 20% of menand 8% of women heterozygous forhemochromatosis had serum ferritin con-centrations that exceeded the 95th per-centile value for the age-matched malecontrol subjects (31). Recently, it has beenclaimed that normal iron stores are nulliron stores (32).

Elevated iron stores could enhance oxi-dation of lipids, especially of free fatty acids,through accelerated production of free rad-icals, as ferrous iron is a potent catalyst. Thedegree of iron overload necessary to induceinsulin resistance is unknown. It appearsthat the initial and most common defect inpatients with an earlier stage of damageinduced by iron overload is one of liver-mediated insulin resistance (33-35). Areduced hepatic insulin extraction, resultingin hyperinsulinemia, correlates with thedegree of iron overload (36). Iron deposi-tion in the liver may also cause insulin resis-tance by interfering with the ability ofinsulin to suppress hepatic glucose produc-tion. Furthermore, increased iron contenthas also been found in the peripheral mus-cle tissue (37), the principal site of overallglucose disposal (38). In patients withhemochromatosis and hemosiderosis,insulin resistance—evaluated by the eugly-cemic glucose clamp (39) or minimal modelmethod (40,41)—correlated with iron over-load, even in the presence of normal glucosetolerance (40,41). Dmochowski et al. (40)found that serum ferritin level inversely cor-related with S\ (r = —0.58). However, nodata about the relation of S] with serum fer-


Dow



ritin was investigated in control subjects.Cavallo-Perin et al. (41) found that S\ wasreduced by 40% in thalassemic subjects,and again it was inversely correlated withiron overload (r = —0.70). Although controlsubjects were also evaluated, the relation-ship between iron stores and Si was notexamined in this group. The tissue respon-sible for insulin resistance in situations ofhigh-normal iron stores can not be inferredfrom these studies, although model-derivedSi is dominated by extrahepatic insulineffects (19).

If serum ferritin is important in thedevelopment of insulin resistance, thehigher the ferritin levels, the higher theincidence of NIDDM. We are unaware ofstudies aimed to test this hypothesisdirectly, but given the usually strong rela-tionships between iron stores and thehematocrit level, the report by Wan-namethee et al. (42) is interesting. In thisstudy, an independent linear associationbetween hematocrit level and the risk ofNIDDM was found in a general populationsample of middle-aged men after 12 yearsof follow-up (42). In another recent study,the glucose clearance rate, as calculated bythe hyperinsulinemic isoglycemic clamp,was inversely related to the hematocritlevel. This relationship was independent ofsex, BMI, and age (43). This association didnot result from acute hemodynamicchanges on insulin sensitivity, and "maytherefore reflect the presence of a commondeterminant" (43). High hemoglobin levelshad also predicted subsequent glucoseintolerance (44) and diabetes (45).

Serum ferritin concentrations aresignificantly increased in patients withpoorly controlled diabetes (46,47), andshort-term improvement in glucose controlis associated with a marked decrease inserum ferritin concentration (46). Aderangement in other serum parameters ofiron metabolism is frequently found inpatients with poorly controlled diabetes(46-48). In NIDDM, about 10% of patientswith high ferritin levels had transferrin sat-urations greater than 40% (46). A correla-tion between serum ferritin and HbAlc hasalready been reported (46-48), but to ourknowledge, serum ferritin had not beendescribed to be an independent predictorof HbAlc levels. However, although serumferritin may be involved in the develop-ment of insulin resistance in diabeticpatients, it is also possible that the presenceof insulin resistance could affect serum fer-ritin levels. Only a prospective study design

could address these issues. More research isneeded to evaluate whether high serumferritin levels reflects increased body ironstores in NIDDM—through liver magneticresonance studies—and whether modifica-tions of iron stores (iron chelators or phle-botomy) in these subjects affects glucosemetabolism. Finally, serum ferritin shouldbe cautiously evaluated in NIDDMpatients, because it may indicate falselynormal iron stores.

In summary, the correlations betweenserum ferritin and diastolic blood pressure,HDL cholesterol, glucose AUC, and Si sug-gest that the former may be a simplemarker of the insulin resistance syndrome.Serum ferritin may also be a marker ofpoor metabolic control in the diabeticpatient.

Acknowledgments — This work was partiallysupported by grant 93/1315 from the Fondo deInvestigaciones Sanitarias, National Health Insti-tute.

References1. Salonen JT, Nyyssonen K, Korpela H,

Tuomilehto J, Seppanen R, Salonen R: Highstored iron levels are associated with excessrisk of myocardial infarction in EasternFinnish men. Circulation 86:803-811,1992

2. Reaven GM: Role of insulin resistance inhuman disease. Diabetes 37:1595-1607,1988

3. Tuomainen T-P, Nyysonen K, Salonen R,Tervahauta A, Korpela H, Lakka T, KaplanGA, Salonen JT: Body iron stores are asso-ciated with serum insulin and blood glu-cose concentrations. Diabetes Care 20:426-428,1997

4. Ascherio A, Willet WC, Rimm EB, Gio-vanucci EL, Stampfer MJ: Dietary ironintake and risk of coronary heart diseaseamong men. Circulation 89:969-974,1994

5. Sullivan JL: Stored iron and ischemic heartdisease: empirical support for a new para-digm. Circulation 86:1036-1037, 1992

6. Lyons TJ: Oxidized low-density lipopro-teins: a role in the pathogenesis of athero-sclerosis in diabetes. Diabet Med 8:411-419,1991

7. Huntjy Smith CC, WolffSP: Autooxidativeglycosylation and possible involvement ofperoxides and free radicals in LDL modifi-cation by glucose. Diabetes 39:1420-1424,1990

8. Wolff SP, Dean RT: Glucose autooxidationand protein oxidation: the role of autox-idative glycosylation in diabetes mellitusand ageing. Biochem] 245:243-250, 1987

9. Reif DW: Ferritin as a source of iron for

oxidative damage. Free Rad Biol Med12:417-427,1992

10. Fujimoto S, Kawakami N, Ohara A:Nonenzymatic glycation of transferrin:decrease of iron-binding capacity andincrease of oxygen radical production. Bio!Pharm Bull 6:396-400, 1995

11. CookJD, Finch CA, Smith NJ: Evaluationof the iron stores of a population. Blood3:449-455, 1976

12. Fernandez-Real JM, Gutierrez C, Ricart W,Casamitjana R, Fernandez-Castaner M,Vendrell J, Richart C, Soler J: The TNF-otgene Nco I polymorphism influences therelationship among insulin resistance, per-cent body fat, and increased serum leptinlevels. Diabetes 46:1468-1472,1997

13. Jeliffe DB: I7ie Assessment of Nutritional Sta-tus of the Community. Geneva, World HealthOrg., 1966 (WHO Monographs Ser, no.53)

14. Ricart W, Gonzalez-Huix F, Conde V, Grupper l'Evaluacio de la Composicio Gorporalde la Poblacio de Catalunya: The evaluationof the state of nutrition through the deter-mination of anthropometric parameters:new tables in the working population inCatalonia (Spain). Med Clin (Bare) 100:681-691,1993

15. National Diabetes Data Group: Classifica-tion and diagnosis of diabetes mellitus andother categories of glucose intolerance. Dia-betes 28:1039-1057,1979

16. SiedelJ, Hagele EO, Ziegenhorn J, Wahle-feld AW: Reagent for the enzymatic deter-mination of serum total cholesterol withimproved lipolytic efficiency. Clin Q\em 29:1075-1080,1983

17. Kostner GM, Molinari E, Pichler P: Evalu-ation of a new HDL2/HDL3 quantitationmethod based on precipitation with poly-ethylene glycol. Gin Chim Ada 148:139-147, 1985

18. Wahlefeld AW In Methoden der Enzymatis-chen Analyse. 3rd ed. Vol II. Bergmeyer HU,Ed. Weinheim, Germany, Verlag Chemie:187-188, 1974

19. Bergman RN, Prager R, Volund A, OlefskyJM: Equivalence of the insulin sensitivityindex in man derived by the minimalmodel method and euglycaemic glucoseclamp. J din Invest 79:790-800, 1987

20. Saad MF, Steil GM, Kades WW, Ayad MF,Elsewafy WA, Boyadjian R, Jinagouda SD,Bergman RN: Differences between thetolbutamide-boosted and the insulin-mod-ified minimal model protocols. Diabetes46:1167-1171, 1997

21. Cederholm J, Wibell L: Insulin release andperipheral sensitivity at the oral glucosetolerance test. Diabetes Res Clin Pract10:167-175,1990

22. Laakso M: How good a marker is insulinlevel for insulin resistance? Am] Epidemiol137:959-965, 1993

23. Oshaug A, Bugge KH, Bjonnes CH, Boreh-


Dow


Serum jerritin and the insulin resistance syndrome

lohnsen B, Neslein I-L: Associationsbetween serum ferritin and cardiovascularrisk factors in healthy young men: a crosssectional study. EurJ Qin Nutr 49:430-438,1995

24. Zidek W, Tenschert W, Karoff CH, Vetter H:Treatment of resistant hypertension byphlebotomy. Klin Wochenschr 63:762-764,1985

25. Barenbrock M, Spieker C, Rahn KH, ZidekW: Therapeutic efficiency of phlebotomy inposttransplant hypertension associatedwith erythrocytosis. CM Nephwl 40:241-243,1993

26. Davalos A, Fernandez-Real JM, Ricart W,Soler S, Molins A, Planas E, Genis D: Ironrelated damage in acute ischemic stroke.Stroke 25:1543-1546, 1994

27. Fernandez-Real JM: Iron as a cardiovascularrisk factor: its interaction with classical riskfactors. MedClin (Bare) 105:617-619,1995

28. Vuorinen-Markkola H, Yki-Jarvinen H:Hyperuricemia and insulin resistance. J QinEndoerinol Mctab 78:25-29, 1994

29. Moirand R, Mortaji AM, Loreal O, PaillardF, Brissot P, Deugnier Y: A new syndrome ofliver iron overload with normal transferrinsaturation. Lancet 349:95-97, 1997

30. Milman N: Serum ferritin in Danes: studiesof iron status from infancy to old. Int JHematol 63:103-135, 1996

31. Bulaj ZB, Griffen LM, Jorde LB, EdwardsCQ, Kushner JP: Clinical and biochemicalabnormalities in people heterozygous forhemochromatosis. N Engl] Med 335:1799-1805,1996

32. Sullivan JL: Iron versus cholesterol: per-spectives on the iron and heart disease

debate. ] Qin Epidemiol 49:1345-1352,1997

33. Felber J-P, Ferrannini E, Golay A, Meyer A,Theibaud D, Curchod B, Maeder E, JequierE, DeFronzo RA: Role of lipid oxidation inpathogenesis of insulin resistance of obesityand type II diabetes. Diabetes 36:1341-1350,1987

34. DeFronzo RA: The triumvirate: (3-cell,muscle, liver. Diabetes 37:667-687, 1988

35. Dandona P, Hussain MAM, Varghese Z,Politis D, Flynn DM, Hoffbrand AV: Insulinresistance and iron overload. Ann QinBiochem 20:77-79, 1983

36. Niederau C, Berger M, Stremmel W, StarkeA, Strohmeyer G, Ebert R, Siegel E,Creutzfeldt W: Hyperinsulinemia in non-cirrhotic haemochromatosis: impairedhepatic insulin degradation? Diabetologia26:441-444, 1984

37. Shafer AI, Cheron RG, Dluhy R, Cooper B,Gleason RE, Soeldner JS, Bunn HF: Clinicalconsequences of acquired transfusionaliron overload in adults. N Engl J Med304:319-324, 1981

38. Katz LD, Glickman MG, Rapoport S, Fer-rannini E, DeFronzo RA: Splanchnic andperipheral glucose disposal of oral glucosein man. Diabetes 32:675-679, 1983

39. Merkel PA, Simonson DC, Amiel SA, PleweG, Sherwin RS, Pearson riA, TamborlaneWV: Insulin resistance and hyperinsuline-mia in patients with thalassemia majortreated by hypertransfusion. N Engl J Med318:809-814,1988

40. Dmochowski K, Finegood DT, FrancombeW, Tyler B, Zinman B: Factors determiningglucose tolerance in patients with tha-

lassemia major. J Qin Endoerinol Metab77:478-483, 1993

41. Cavallo-Perin P, Pacini G, Cerutti F, BessoneA, Condo C, Sacchetti L, Piga A, Pagano G:Insulin resistance and hyperinsulinemia inhomozygous 3-thalassemia. Metabolism44:281-286,1995

42. Wannamethee SG, Perry IJ, Shaper AG:Hematocrit and risk of NIDDM. Diabetes45:576-579, 1996

43. Catalano C, Muscelli E, Quinones A, Baldi S,Ciociaro D, Seghieri G, Ferrannini E: Recip-rocal association between insulin sensitivityand the hematocrit in man (Abstract). Dia-betes 45 (Suppl. 2):323A, 1996

44. Wilson PW, McGee DL, Kannel WB: Obe-sity, very low density lipoproteins and glu-cose intolerance over fourteen years: theFramingham Study. Am J Epidemiol114:697-704, 1981

45. Medalie JH, Papier CM, Goldbourt U, Her-man JB: Major factors in the developmentof diabetes mellitus in 10,000 men. ArchIntern Med 135:811-817, 1975

46. Kaye TB, Guay AT, Simonson DC: Non-insulin-dependent diabetes mellitus andelevated serum ferritin level. J Diabet Com-plications 7:246-249, 1993

47. Gallou G, Guilhem I, Poirier J-Y, RuellandA, Legras B, Cloarec L: Increased serum fer-ritin in insulin-dependent diabetes melli-tus: relation to glycemic control. Qin Chem40:947-948, 1994

48. Mauricio D, Perez A, Riera A, Arroyo JA, deLeiva A, Gimferrer E: Serum parameters ofiron metabolism in type I (insulin depen-dent) diabetes mellitus. Diabetes Nutr Metab8:315-316, 1993


Dow


serum ferritin as a component of the insulin - diabetes care

Documents