Chromium Supplementation Does Not Improve Glucose Tolerance, Insulin Sensitivity, or Lipid Profile:...

OBSERVATIONS

Critical Evaluation ofModels to IdentifyIndividuals WithInsulin Resistance

C linical indexes of insulin resistance(IR) have acquired increasing im-portance with the development of

various drugs that improve endogenousinsulin action (1). Recently, the largestdatabase on insulin clamp studies hasbeen established. This database includes2,321 subjects, of whom 2,138 are non-diabetic (92%), from 19 sites worldwide(2). Using classification trees, three mod-els have been derived. Model 1 is based onhomeostasis model assessment of insulinresistance (HOMA-IR) �4.65, BMI�28.9 kg/m2, or HOMA-IR �3.60 andBMI �27.5 kg/m2. Model 2 is based onBMI �28.7 kg/m2, or BMI �27.0 kg/m2

and a positive family history of diabetes.Model 3 is based on BMI �28.7 kg/m2,BMI �27.0 kg/m2 and a positive diabetesfamily history, or triglycerides �2.44mmol/l and a negative family history ofdiabetes. These three models should allaccurately identify insulin-resistant indi-viduals (2). We have evaluated the prev-alence and characteristics of subjects withIR based on these models using data fromthe KORA Survey 2000, an oral glucosetolerance test (OGTT)-based, population-based survey in Germany (n � 1,352 in-dividuals aged 55–74 years withoutpreviously known diabetes) (3).

In the KORA Survey, proportions(95% CI) with IR were 47.4% (44.7–50.1), 45.8% (43.1– 48.5), and 49.1%(46.4–51.8) for models 1, 2, and 3, re-spectively. Agreement of the models washigh (� coefficients 0.78 – 0.94). Al-though HOMA-IR significantly increasedwith worsening glucose tolerance (geo-metric means [SDF]: normal glucose tol-erance 2.17 [1.83], impaired glucosetolerance [IGT] 3.39 [2.12], and newlydiagnosed diabetes 4.67 [2.16]; all P �0.05), the sensitivities (0.67, 0.60, and0.64 for models 1, 2, and 3, respectively)and specificities (0.60, 0.59, and 0.56) ofthe IR models for detecting IGT or diabe-tes were only moderate. Model 1 included

HOMA-IR as a surrogate measure of IR. Inmultiple age-sex–adjusted logistic regres-sion including all three models, model 1(odds ratio 4.3, 95% CI 2.8–6.8) wasmore closely related to IGT/diabetes (de-pendent variable) than the others (model2: 0.5, 0.2–1.1; model 3: 1.6, 0.7–3.4).

Overall, about one-third of the sub-jects with IGT or previously undiagnoseddiabetes would not have been includedwhen applying the proposed rules foridentifying individuals with IR in our el-derly population. This may indicate a lim-ited diagnostic validity of the IR models,because a large body of evidence showsthat IGT and type 2 diabetes are charac-terized by moderate-to-severe IR (4). Onthe other hand, defective insulin secretionrather than IR may be present in somesubjects with IGT and type 2 diabetes (4).This could partly explain the low sensitiv-ities of the models to detect glucose dis-orders in our population. In conclusion,the recently proposed IR models need tobe further validated, using measures of�-cell function across the whole range ofglucose intolerance, before they shouldbe incorporated into clinical trials andclinical practice (2).

WOLFGANG RATHMANN, MD, MSPH1

BURKHARD HAASTERT, PHD1

GUIDO GIANI, PHD1

ROLF HOLLE, PHD2

WOLFGANG KOENIG, MD3

CHRISTIAN HERDER, PHD4

HANNELORE LOWEL, MD5

From the 1Institute of Biometrics and Epidemiology,German Diabetes Center, Leibniz Institute at theHeinrich-Heine-University, Dusseldorf, Germany;2GSF National Research Center for Environmentand Health, Institute of Health Economics andHealth Care Management, Neuherberg, Germany;3University of Ulm Medical Center, Ulm, Germany;4German Diabetes Clinic, German Diabetes Center,Leibniz Institute at the Heinrich-Heine-University,Dusseldorf, Germany; 5GSF National Research Cen-ter for Environment and Health, Institute of Epide-miology, Neuherberg, Germany.

Address correspondence to Dr. Wolfgang Rath-mann, MSPH (USA), Institute of Biometrics andEpidemiology, German Diabetes Center, Auf�mHennekamp 65 D-40225 Dusseldorf, Germany. E-mail: [email protected].

© 2005 by the American Diabetes Association.

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. McAuley KA, Williams SM, Mann JI,

Walker RJ, Lewis-Barned NJ, Temple LA,Duncan AW: Diagnosing insulin resis-tance in the general population. DiabetesCare 24:460–464, 2001

2. Stern SE, Williams K, Ferrannini E, De-Fronzo RA, Bogardus C, Stern MP: Iden-tification of individuals with insulinresistance using routine clinical measure-ments. Diabetes 54:333–339, 2005

3. Rathmann W, Haastert B, Icks A, LowelH, Meisinger C, Holle R, Giani G: Highprevalence of undiagnosed diabetes mel-litus in southern Germany: target popula-tions for efficient screening: the KORASurvey 2000. Diabetologia 46:182–189,2003

4. Ferrannini E, Gastaldelli A, Miyazaki Y,Matsuda M, Mari A, DeFronzo RA: �-Cellfunction in subjects spanning the rangefrom normal glucose tolerance to overt di-abetes: a new analysis. J Clin EndocrinolMetab 90:493–5 00, 2005

C-Reactive Proteinfor CardiovascularRisk Assessment

M alik et al. (1) report on the useful-ness of C-reactive protein (CRP)in stratifying risk in patients with

the metabolic syndrome and diabetes.Studies on normal volunteers showed

intraindividual variability, which in�50% of individuals was sufficient tochange their CRP-related risk category(2).

Bogaty et al. (3) have also demon-strated what seems to be spontaneousfluctuation of CRP in stable patients withcoronary artery disease (CAD).

Intraindividual biological variationdata for high-sensitivity CRP (hsCRP)needs to be established for each individ-ual before using the level to estimate riskand prognosis.

Reliance on single or even the mean oftwo measurements 2–4 weeks apart isclearly unacceptable, and there is conflictin the published literature regarding thenumber of samples that should be testedand the time span (3).

Many laboratories use conventionalCRP assays, which report levels �5 mg/las normal. These assays are obviously un-suitable for risk assessment, as it nowseems that levels as low as 2 mg/l conferadditional risk (3,4). Laboratories shouldbe able to provide hsCRP assays for thepurpose of cardiovascular risk stratification.

CRP seems to be an important playerin the inflammatory component of ath-erosclerosis and an independent predic-

L E T T E R S

DIABETES CARE, VOLUME 28, NUMBER 7, JULY 2005 1833

tor of adverse CAD outcomes (5,6).Adding it formally to risk stratificationscoring methods would improve our abil-ity to identify high-risk patients in bothprimary and secondary prevention. Be-fore adopting this strategy, it is importantto decide how many measurementsshould be checked, over what interval,and under what conditions.

Also, what weight should be given tothe presence of an elevated CRP? Shouldit influence risk scores qualitatively, likediabetes, or quantitatively, like systolicblood pressure and cholesterol; thehigher the level, the higher the risk? Howshould CRP level be incorporated intostandard scoring systems such as the Fra-mingham risk score?

Until these issues are addressed, CRPmeasurements should perhaps be re-served for problematic or borderlinecases, where the decision to use an inter-vention, whether medical or procedural,is difficult.

MONA A. KHOLEIF, MD, FRCP, FESC

From the Department of Medicine, King Abdul AzizMedical City, Jeddah, Kingdom of Saudi Arabia.

Address correspondence to Dr. Mona A. Kholeif,Consultant Cardiologist, Cardiology Section, De-partment of Medicine, King Abdul Aziz MedicalCity, P.O. Box 9515, Jeddah 21423, Kingdom ofSaudi Arabia. E-mail: [email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Malik S, Wong ND, Franklin S, Pio J, Fair-

child C, Chen R: Cardiovascular diseasein U.S. patients with metabolic syndrome,diabetes, and elevated C-reactive protein.Diabetes Care 28:690–693, 2005

2. de Maat MPM, de Bart ACW, Hennis BC,Meijer P, Havelaar AC, Mulder PG, KluftC: Interindividual and intraindividualvariability in plasma fibrinogen, TPA an-tigen, PAI activity, and CRP in healthy,young volunteers and patients with an-gina pectoris. Arterioscler Thromb Vasc Biol16:1156–1162, 1996

3. Bogaty P, Brophy JM, Boyer L, Simard S,Joseph L, Bertrand F, Dagenais GR: Fluc-tuating inflammatory markers in patientswith stable ischemic heart disease. ArchIntern Med 165:221–226, 2005

4. Ridker PM, Cannon CP, Morrow D, RifaiN, Rose LM, McCabe CH, Pfeffer MA,Braunwald E, for the Pravastatin or Ator-vastatin Evaluation and Infection Therapy–Thrombolysis in Myocardial Infarction22 (PROVE IT –TIMI 22) Investigators:C-Reactive protein levels and outcomesafter statin therapy. N Engl J Med 352:20–

28, 20055. Danesh J, Pepys MB: C-reactive protein in

healthy and in sick populations. EurHeart J 21:1564–1565, 2000

6. Ehrenstein MR, Jury EC, Mauri C: Statinsfor atherosclerosis: as good as it gets?N Engl J Med 352:73–75, 2005

Age and A1C AreImportant ClinicalPredictors ofContinuousSubcutaneous InsulinInfusion Efficacy inType 1 DiabeticPatients

Only a few studies have reported along-term follow-up in a significantnumber of patients using continu-

ous subcutaneous insulin infusion (CSII)(1). Who stands to benefit more from thiscostly insulin therapy is still unclear.

The aim of our observational, retro-spective study was to evaluate the possiblepredictors of the degree of improvementof metabolic control with CSII in 82 con-secutive type 1 diabetic patients (age37.9 � 13.4 years, 42 men and 40women, duration of diabetes 19.7 � 9.9years) who started CSII in the DiabetesUnit of Bergamo Hospital between June1999 and March 2004.

The patients had been treated withmultiple daily injection (MDI) therapy(regular [n � 22] or rapid-acting analoginsulin [n � 60] before meals plus NPH[n � 72] or glargine [n � 10] as basalinsulin) for at least 1 year. During CSIItreatment, lispro or aspart analogs wereused.

All patients were evaluated every 3months both before and during CSII. Themean duration of CSII treatment was31.9 � 14.5 months (range 4–55).

Only three patients discontinued CSII.Every patient performed self-monitoringof blood glucose (SMBG) (four to sevendaily determinations).

Data are expressed as means � SD.The means of parametric data of the pe-riod of MDI treatment were comparedwith those of the CSII period using theStudent’s t test for paired data. The differ-ences between groups were compared us-ing the Student’s t test for unpaired data.

Compared with MDI therapy, HbA1c(A1C) significantly decreased with 3months of CSII therapy (CSII vs. MDI:8.35 � 1.06 vs. 9.39 � 1.35%, P �0.001). The significant decrease of A1Cwas maintained over the whole CSII treat-ment with a mean change in A1C of1.15 � 0.84% (P � 0.001).

During CSII treatment, as comparedwith MDI treatment, there was a signifi-cant decrease of severe hypoglycemic ep-isodes (0.35 � 0.07 per patient/yearduring MDIs vs. 0.10 � 0.02 during CSII,P � 0.001) and insulin requirement(52.1 � 17.5 units/day vs. 38.8 � 12.3,P � 0.001). CSII was not associated withany significant increase in BMI. Incidenceof ketoacidosis was negligible during bothMDI and CSII treatment.

To evaluate the possible predictors ofCSII effect on A1C changes, multiple lin-ear regression analysis performed on allpatients revealed that age (� � 0.16, P �0.05) and baseline A1C (� � 0.21, P �0.008) were independently associatedwith A1C improvement after 3 months ofCSII (F � 5.41, adjusted R2 � 0.28). BMI,diabetes duration, insulin requirement,and frequency of SMBG were unrelated toA1C changes. Age and baseline A1C wereeven better predictors of the mean A1Cchanges during the whole follow-up pe-riod (F � 11.87, adjusted R2 � 0.48).

This observational, clinic-based studyof a significant number of type 1 diabeticpatients who represent all pump-treatedpatients in the province of Bergamo, Italy,confirmed that CSII significantly de-creased A1C levels with respect to MDIs,as reported in a recent meta-analysis of 52CSII studies (2). The reduction of severehypoglycemic episodes and the concom-itant negligible frequency of ketaoacidosisconfirmed the safety of CSII (3–4).

Although most of our patients duringMDI treatment used rapid-acting analogs,only a few used glargine as basal insulin. Itis possible that the extensive use of thislong-term analog during MDI treatmentwould reduce the difference in metaboliccontrol, as suggested (5).

There are discordant data about thepersistence of initial lowering of A1Cachieved with CSII (6). Our data showedthat improved glucose control persistedduring the whole long-term follow-up pe-riod, which was longer (mean duration2.6 years) than most CSII studies.

Most important, our study demon-strated that CSII was particularly advan-

Letters

1834 DIABETES CARE, VOLUME 28, NUMBER 7, JULY 2005

tageous in patients with the poorestmetabolic control, as suggested by a ran-domized controlled trial that comparedCSII and MDI treatments using a rapid-acting analog (7). In our population,those with a baseline A1C �10% (n � 25)had an average decline in A1C of 1.5 �0.6%, significantly greater (P � 0.001)than that observed in those with a base-line value �8% (n � 16; average decline0.6 � 0.5%).

CSII was also more advantageous inpatients older than 50 years (n � 14; av-erage A1C decline 1.45 � 0.7%) than inthose younger than 20 years (n � 11; av-erage A1C decline 0.5 � 0.8%; P � 0.01).A similar observation suggested that CSIIis useful and safe in older adults with type1 diabetes (8).

In conclusion, we suggest that in type 1diabetic patients who have sufficient abilityto master CSII therapy, a poor metaboliccontrol, despite MDI therapy, and older ageare better predictors of CSII efficacy.

GIUSEPPE LEPORE, MD

ALESSANDRO R. DODESINI, MD

ITALO NOSARI, MD

ROBERTO TREVISAN, MD, PHD

From the Diabetes Unit, Ospedali Riuniti Bergamo,Bergamo, Italy.

Address correspondence to Dr. Giuseppe Lepore,U.O. Diabetologia, Ospedali Riuniti di Bergamo,Largo Barozzi, 1, 24128 Bergamo, Italy. E-mail:[email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Linkeshova R, Raoul M, Bott U, Berger M,

Spraul M: Less severe hypoglycaemia, bet-ter metabolic control, and improved qual-ity of life in type 1 diabetes mellitus withcontinuous subcutaneous insulin infu-sion (CSII) therapy: an observationalstudy of 100 consecutive patients fol-lowed for a mean of 2 years. Diabet Med19:746–751, 2002

2. Pickup J, Keen H: Continuous subcutane-ous insulin infusion at 25 years. DiabetesCare 25:593–598, 2002

3. Bode BW, Steed RD, Davidson PC: Reduc-tion in severe hypoglycaemia with long-term continuous subcutaneous insulininfusion in type 1 diabetes. Diabetes Care19:324–327, 1996

4. Boland EA, Grey M, Osterle A, Fredrick-son L, Tamborlane W: Continuous sub-cutaneous insulin infusion: a new way tolower risk of severe hypoglycemia, im-prove metabolic control, and enhancecoping in adolescents with type 1 diabe-

tes. Diabetes Care 22:1779–1784, 19995. Lepore G, Dodesini AR, Nosari I, Trevisan

R: Both continuous insulin infusion and amultiple daily insulin injection regimenwith glargine as basal insulin are equallybetter than traditional multiple daily in-sulin injection treatment (Letter). DiabetesCare 26:1321–1322, 2003

6. Plotnick LP, Clark LM, Brancati FL, Er-linger T: Safety and effectiveness of insu-lin pump therapy in children andadolescents with type 1 diabetes. DiabetesCare 26:1142–1146, 2003

7. DeVries JH, Snoek FJ, Kostense PJ, Ma-surel N, Heine RJ: A randomized trial ofcontinuous subcutaneous insulin infu-sion and intensive injection therapy intype 1 diabetes for patients with long-standing poor glycemic control. DiabetesCare 25:2074–2080, 2002

8. Siegel-Czarkowski L, Herold KC, GolandRS: Continuous subcutaneous insulin infu-sion in older patients with type 1 diabetes(Letter). Diabetes Care 27:3022–3023,2004

ContinuousSubcutaneous InsulinInfusion VersusMultiple DailyInjections

Modeling predicted benefits inrelationship to baseline A1c

W ith either continuous subcutane-ous insulin infusion (CSII) ormultiple daily insulin injection

(MDII) therapy, the optimal meal insulinis a rapid-acting analog (lispro or aspart).To date, the efficacy of CSII versus MDII

therapy has been evaluated in a limitednumber of randomized controlled trialsin which rapid-acting analogs were usedfor both regimens. In this context, we re-cently conducted a pooled analysis (1) us-ing raw trial data from three such studiesundertaken in adults with type 1 diabetes(2–4). This analysis suggested that CSII isassociated with better glycemic control,particularly in those patients with poorinitial control. Indeed, the relative benefitof CSII over MDII was found to increasewith higher baseline A1c. To provide di-rect clinical context, we have now re-analyzed this data to evaluate the impactof CSII and MDII in relation to specificbaseline A1c categories using the pooleddataset (139 patients representing 529patient-months on MDII and 596 patient-months on CSII).

Treatment effect on A1c was studiedusing a mixed linear modeling approach(MIXED procedure in SAS 9.1.3), with anisotropic exponential spatial covariancestructure used to model intrasubject cor-relation of the repeated measurementsand random effects used to model patient,patient*treatment, study, and study*month effects. All fixed and random ef-fects were initially allowed to differbetween studies, with the Akaike Infor-mation Criterion (AIC) approach used toreduce model complexity. Fixed effects inthe final model included the following: 1)baseline A1c, 2) treatment modality, and3) the interaction between baseline A1cand treatment effect.

These models predicted that, withboth MDII and CSII, the reduction in A1cwill progressively increase as baseline A1crises (Fig. 1). Importantly, however, CSII

Figure 1—The predicted relative benefit of CSII over MDII in lowering A1c increases as baselineA1c rises. �, MDII; f, CSII.

Letters


is predicted to achieve a greater reductionin A1c at each level of initial glycemia.Moreover, the relative benefit of CSII overMDII in lowering A1c (i.e., difference intreatment effect between the two modali-ties) increases as baseline A1c rises. In-deed, for a patient with an A1c of 12%,CSII would be expected to reduce A1c bya full 1% more than MDII.

In summary, this analysis extends ourearlier work by providing a quantitativeestimate of the anticipated reduction inA1c associated with CSII and MDII ther-apy, respectively, in relation to baselineA1c. Importantly, the relative benefit ofCSII over MDII in lowering A1c increasesas baseline A1c rises. This analysis has sig-nificant clinical implications in that 1) itprovides an anticipated response to inten-sive insulin regimens in adults with type 1diabetes, and 2) it highlights the impor-tance of baseline glycemic control as a fac-tor to consider when choosing betweenCSII and MDII therapy.

It should be noted that NPH insulinprovided basal replacement in MDII ther-apy in the current analysis. Given the su-perior basal pharmacokinetics of long-acting analogs, future clinical trialcomparison between CSII and MDII us-ing both rapid- and long-acting analogswill be of great interest. Based on the cur-rent findings, comparison of these treat-ment modalities across a broad range ofbaseline glycemic control would be par-ticularly relevant.

RAVI RETNAKARAN, MD1,2

J. HANS DEVRIES, MD3

HELENE HANAIRE-BROUTIN, MD4

ROBERT J. HEINE, MD, PHD5

VINCENT MELKI, MD4

BERNARD ZINMAN, MD1,2

From the 1Division of Endocrinology, University ofToronto, Toronto, Ontario, Canada; 2Leadership Si-nai Centre for Diabetes, Mount Sinai Hospital, To-ronto, Ontario, Canada; the 3Department of InternalMedicine, Academic Medical Center, Amsterdam,the Netherlands; 4Service de Diabetologie, Hospitalde Rangueil, CHU de Toulouse, Toulouse, France;and the 5Department of Endocrinology, DiabetesCenter, VU University Medical Center, Amsterdam,the Netherlands.

Address correspondence to Dr. Bernard Zinman,Leadership Sinai Centre for Diabetes, Mount SinaiHospital, Lebovic Building, 5th Floor, Room L5-024, 600 University Ave., Toronto, ON, Canada,M5G 1X5. E-mail: [email protected].

R.J.H. is a member of an advisory board for NovoNordisk, Denmark. B.Z. has received consulting feesfrom Smiths Medical Canada.


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Retnakaran R, Hochman J, DeVries JH,

Hanaire-Broutin H, Heine RJ, Melki V,Zinman B: Continuous subcutaneous in-sulin infusion versus multiple daily injec-tions: the impact of baseline A1c. DiabetesCare 27:2590–2596, 2004

2. DeVries JH, Snoek FJ, Kostense PJ, Ma-surel N, Heine RJ: A randomized trial ofcontinuous subcutaneous insulin infu-sion and intensive injection therapy intype 1 diabetes for patients with long-standing poor glycemic control. DiabetesCare 25:2074–2080, 2002

3. Tsui E, Barnie A, Ross S, Parkes R, ZinmanB: Intensive insulin therapy with insulinlispro: a randomized trial of continuoussubcutaneous insulin infusion versusmultiple daily insulin injection. DiabetesCare 24:1722–1727, 2001

4. Hanaire-Broutin H, Melki V, Bessieres-La-combe S, Tauber JP: Comparison of contin-uous subcutaneous insulin infusion andmultiple daily injection regimens using in-sulin lispro in type 1 diabetic patients onintensified treatment: a randomized study.Diabetes Care 23:1232–1235, 2000

The Effect of BloodSample Volume on11 GlucoseMonitoring Systems

The effect of variable blood samplevolume on the accuracy of 11 glu-cose meters was studied to verify the

reliability of self-monitoring of blood glu-

cose. A total of 11 meters were assessed:OneTouch FastTake, OneTouch Basic,OneTouch Profile, and SureStep (Life-Scan Canada, Burnaby, B.C., Canada);AccuSoft Advantage and AccuSoft Man-ager (Roche, Hoffman-LaRoche, Laval,P.Q., Canada); Precision Pen and Preci-sion QID (MediSense Canada, Missis-sauga, ON, Canada); and GlucometerElite, Glucometer Elite XL, and Glucom-eter DEX (Bayer, Toronto, ON, Canada).Venous blood collected from 16 fastingpatients with diabetes was used to testeach meter brand in triplicate. Samplevolumes tested were of 1, 2, 3, 4, 5, 10,and 20 l. Each patient contributed to the5-l sample plus two other sample vol-umes. The 5-l volume, which is theusual volume required by the manufac-turer of most meters, was considered thereference for comparison with other vol-umes tested, thus excluding the con-founding effects of hematocrit, humidity,hypotension, and hypoxia. Several repli-cates of each volume size were tested. Thenumber of times meters gave no result oran error message was recorded. Resultswere then calculated as percentages of thereference value and considered accurate ifwithin 20% of the reference, as recom-mended by the Food and Drug Adminis-tration/National Committee for ClinicalLaboratory Standards (1). It has been rec-ognized that most current meters do notcomply with the 5% accuracy recom-mended by the American Diabetes Asso-ciation (2–4).

All meters gave mostly nonmisleading

Table 1—Meter brands ranked by performance as estimated by Somers’ d statistic

Meter brand

Percentage ofmisleading results

at 2 l* (n/N)†

Percentage ofmisleading results

at 1 l (n/N)†Somers’

d statistic‡

Precision QID 0 (0/14) 0 (0/12) 0.01Precision Pen 7 (1/14) 0 (0/10) 0.02AccuSoft Advantage 13 (2/16) 0 (0/16) 0.03AccuSoft Manager 13 (2/16) 6 (1/16) 0.05Glucometer Elite 0 (0/16) 31 (5/16) 0.13SureStep 11 (2/18) 44 (8/18) 0.14Elite XL 19 (3/16) 25 (4/16) 0.14OneTouch Profile 75 (12/16) 6 (1/16) 0.17Glucometer DEX 69 (11/16) 19 (3/16) 0.20OneTouch Basic 81 (13/16) 19 (3/16) 0.22OneTouch FastTake 22 (4/18) 44 (8/18) 0.22

*Results �20% from the reference value. †Number of misleading results divided by the number of samplestested. ‡Measures the degree to which more blood volume decreases the chance of a misleading result.Calculation is based on all volumes tested (1, 2, 3, 4, 5, 10, and 20 l).

Letters


results (accurate result or error message) atthe 3-l volume or above. However, belowthe 3-l volume, most meters gave mislead-ing results (Table 1). At the 1-l volume,the mean calculated as a percentage fromthe 5-l reference volume varied between40 and 68%. OneTouch Basic and FastTakewere the most likely to give a misleadingresult for a smaller blood volume as shownby Somers’ d statistic. Only four meters werereliable at all volumes showing no associa-tion with the Jonckheere tests: PrecisionQID, Precision Pen, AccuSoft Advantage,and AccuSoft Manager (data not shown).

Our results confirm that insufficientblood sample volume remains potentiallya major source of user error with many ofthe current meters, including those withbuilt-in detection of insufficient bloodsample. Inaccurate results underesti-mated the real glucose value. Patientsshould be aware of this phenomenonsince falsely low readings may result inunnecessary treatment of hypoglycemiaand weight gain.

ZEINA YARED, MD1,2

KHALED ALJABERI, MD1,2

NANCY RENOUF1,2

JEAN-FRANCOIS YALE, MD1,2

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Royal Victoria Hospital Metabolic DayCentre, McGill University, Montreal, Quebec, Can-ada; and the 2McGill Nutrition Centre, Royal Victo-ria Hospital, MUHC, McGill University, Montreal,Quebec, Canada.

Address correspondence and reprint requests toDr. Jean-Francois Yale, MD, McGill Nutrition Cen-tre, Royal Victoria Hospital, 687 Pine Ave. West,Montreal, Quebec, Canada, H3A 1A1. E-mail: [email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Chen ET, Nichols JH, Show-Hong D,

Hortin G: Performance evaluation ofblood glucose monitoring devices. Diabe-tes Technol Ther 5:749–768, 2003

2. Bohme P, Floriot M, Sirveaux MA, DurainD, Ziegler O, Drouin P, Guerci B: Evolu-tion of analytical performance in portableglucose meters in the last decade. DiabetesCare 26:1170–1175, 2003

3. Trajanoski Z, Brunner GA, Gfrerer RJ,Wach P, Pieber TR: Accuracy of homeblood glucose meters during hypoglyce-mia. Diabetes Care 19:1412–1415, 1996

4. American Diabetes Association: Stan-dards of medical care in diabetes. DiabetesCare 27 (Suppl. 1):S15–S35, 2004

Anticraving Effects ofTopiramate in aDiabetic Patient

Topiramate is an effective antiepilep-tic medication. It holds promise inthe care of diabetic patients by virtue

of its effect on weight loss (1). It is alsoreported efficacious as an adjunct in thetreatment of alcohol dependence (2) andin the management of binge-eating disor-der (3). We report a case illustrating thepotential anticraving effects of topiramateagainst chocolate, leading to significantlyimproved glycemic control in an epilepticpatient with concurrent diabetes.

A 67-year-old woman presented tothe epilepsy clinic in September 2003 forevaluation of “possible seizures” and wassubsequently treated with topiramate.She also had an 11-year history of poorlycontrolled diabetes and was a recoveredalcoholic for 30 years. Medications in-cluded levothyroxine, repaglinide, acetyl-salicylic acid, and atorvastatin. Generalexamination was remarkable for an obesewoman with clinical features of hypothy-roidism. She weighed 194 pounds, andneurological examination was significantfor symmetrical peripheral neuropathy.Subsequent follow-ups were remarkablefor a moderate improvement in seizurecontrol, a total of 34 pounds weight loss,and significant amelioration of her diabe-tes. She ascribed the improved glycemiccontrol to her recent aversion to choco-lates and sweets, which she claimed hadoccurred since starting the topiramate.

Initial blood work was remarkable fora fasting blood glucose of 14.3 mmol/l(normal 3.3– 6.0 mmol/l) and HbA1c(A1C) of 8.9% (normal 4.3–6.1%). Eightmonths after starting the topiramate, herfasting blood glucose was 6.1 mmol/l, andA1C dropped from 8.9 to 6.1%. Wechecked fasting blood glucose and A1Cevery 3–4 months after initiating the topi-ramate. Average fasting blood glucose andA1C before introduction of topiramatewere 10.2 mmol/l and 8.4%, and 8months later were 8.7 mmol/l and 6.7%,respectively. The dosage of repaglinidewas unchanged over the last 2 years. Thepatient reported a significant aversion tochocolate after starting the topiramate.

Topiramate is one of the preferredagents in obese epileptic patients, and thiswas one of the reasons for prescribing this

agent to our patient. The 34-poundweight loss was most likely related to theuse of topiramate. Our patient reported avery strong craving for chocolate in pre-vious years, which together with her obe-sity was apparently responsible for herpoor glycemic control. Topiramate isknown to cause weight loss in 1520% ofpatients, but the precise mechanism is un-known. There is a possibility that the an-ticraving effect may in part be responsiblefor the weight loss.

Topiramate is known to be efficaciousas an adjunct treatment for alcohol de-pendence and in the treatment of binge-eating disorder. The proposed mechanismis facilitation of �-amino-butyric acid ac-tivity and inhibition of glutamate functionin the mesocorticolimbic dopamine path-ways (2). Similar mechanisms may be re-sponsible for the anticraving effects notedin our patient. Dietary noncompliancecan adversely affect glycemic control. Al-though some patients are aware of thisfact, they are unable to avert this cravingwithout pharmacological support, thusleading to failure of oral hypoglycemicagents. By virtue of its potential to causeweight loss, topiramate deserves consid-eration when treating diabetic patientswith epilepsy. Our case illustrates the pos-sibility of another potential mechanism, itsanticraving effect, which would supporttopiramate as a useful adjuvant in the treat-ment of diabetes.

SYED NIZAMUDDIN AHMED, MD, FRCPC1

YASMIN RASHID, MD2

From the 1Division of Neurology, University of Al-berta Hospital, Edmonton, Alberta, Canada; and the2All Well Primary Care Centre, Edmonton, Alberta,Canada.

Address correspondence to S. Nizam Ahmed,MD, FRCPC, 2E3.12. University of Alberta Hospital,Edmonton, AB T6G 2B7, Canada. E-mail: [email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Astrup A, Toubro S: Topiramate: a new

potential pharmacological treatment forobesity (Review). Obes Res 12:167S–173S, 2004

2. Johnson BA, Ait-Daoud N, Bowden CL,DiClemente CC, Roache JD, Lawson K,Javors MA, Ma JZ: Oral topiramate fortreatment of alcohol dependence: a ran-domized controlled trial. Lancet 361:1677–1685, 2003

3. Shapira NA, Goldsmith TD, McElroy SL:Treatment of binge-eating disorder with

Letters


topiramate: a clinical case series. J ClinPsychiatry 61:368–372, 2000

COMMENTS ANDRESPONSES

Diabetes Is the MainFactor Accounting forthe High FerritinLevels Detected inChronic Hepatitis CVirus Infection

Response to Lecube et al.

W e read with interest the article byLecube et al. (1) regarding thestrong association they observed

between diabetes and serum ferritin inchronic hepatitis C. We sought to analyzethe contribution of biochemical, meta-bolic, and histological parameters to highferritin levels detected in hepatitis C.

We investigated a large consecutiveseries of 177 patients with chronic hepa-titis C who underwent a diagnostic liverbiopsy. Serum ferritin was tested in a uni-variate analysis against demographics,biochemical parameters, and histologicalfeatures. Patients with cirrhosis or withany alcohol intake were excluded. Themedian age of the patients was 48.4 years(range 19–71) and 97 (54.8%) were men.Using the same cutoff values of Lecube etal., serum ferritin was raised in 92 cases(52.0%). The prevalence of impaired glu-cose tolerance or diabetes was 9.6% (17cases) in our series. Overall, 66 patients(37.3%) had mild fibrosis (F0–F1) and111 patients (62.7%) had moderate to se-vere fibrosis (F2–F3) according to META-VIR. Hepatic iron deposits were found in68 patients (38.4%). Hepatic steatosiswas detected in 132 patients (74.6%). Se-rum ferritin correlated by univariate anal-ysis with male sex (P � 0.05), BMI (P �0.0001), aspartate aminotransferase andalanine aminotransferase levels (P �0.003 and P � 0.0009, respectively),�-glutamyl transferase levels (P �0.00001), hepatic iron (P � 0.00001),and hepatic steatosis (P � 0.01). No cor-relation between serum ferritin and fast-ing glucose could be observed. Moreover,

no significant difference in serum ferritinwas observed in patients with impairedglucose tolerance or diabetes in compari-son with other patients.

The following considerations arisefrom the comparison of our data withthose reported by Lecube et al. First,mean serum ferritin in our study wasmuch higher than that observed byLecube et al. even if we excluded cirrhosisand alcohol consumption. Second, preva-lence of diabetes was much higher (21.7%)in the Spanish series compared with ourown cohort, probably because Lecube etal.’s study was conducted in a tertiary refer-ence center for both diabetes and hepatitisC. Third, we did not observe any associa-tion between raised serum ferritin and dia-betes. In chronic hepatitis C there are manyconditions that could elevate serum ferritinsuch as necroinflammation, steatosis, andhepatic iron deposition (2). In our series, wefound an association with metabolic factorsand with markers of inflammation. On theother hand, there was also a strong associa-tion with proper hepatic iron deposition.

We can therefore conclude that the in-crease of serum ferritin in chronic hepatitisC could be linked to diabetes, as Lecube etal. have clearly suggested, but the pathogen-esis is multifactorial. The weight of differentdeterminant factors on the elevation of fer-ritin depends on their prevalence in the an-alyzed series of patients.

GIADA SEBASTIANI, MD

ALESSANDRO VARIO, MD

ALFREDO ALBERTI, MD

From the Department of Clinical and ExperimentalMedicine, University of Padova, Padova, Italy.

Address correspondence to Prof. Alfredo Alberti,MD, Department of Clinical and Experimental Med-icine, University of Padova, Via Giustiniani 2, 35100Padova, Italy. E-mail: [email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Lecube A, Hernandez C, Genesca J, Es-

teban JI, Jardı R, Garcıa L, Simo R: Diabe-tes is the main factor accounting for thehigh ferritin levels detected in chronichepatitis C virus infection. Diabetes Care27:2669–2675, 2004

2. Metwally MA, Zein CO, Zein NN: Clinicalsignificance of hepatic iron depositionand serum iron values in patients withchronic hepatitis C infection. Am J Gastro-enterol 99:286–291, 2004

Diabetes Is the MainFactor Accounting forthe High FerritinLevels Detected inChronic Hepatitis CVirus Infection

Response to Sebastiani et al.

A fter reading the comments by Se-bastiani et al. (1) on our article (2)regarding the association between

diabetes and ferritin in chronic hepatitis Cvirus (HCV) infection, we would like tomake the following comments. In con-trast to our results, the authors did notfind any relationship between serum fer-ritin levels and glucose abnormalities inHCV-infected patients. However, itshould be noted that the number of pa-tients included in our study was muchlarger and, thereby, a statistical multivar-iate analysis considering sex (a major con-founding factor) could be performed. Inaddition, because in our study a group ofdiabetic patients without HCV infectionand a group of anti–HCV-negative nondi-abetic control subjects were analyzed, wewere able to conclude that the increase inferritin levels detected in HCV patientswas closely related to the presence of di-abetes (2). Another concern of Sebastianiet al. is the high prevalence of diabetes inour population (21.7%). Although someinfluence could be attributed to the ter-tiary reference center setting of our study,it is more important to note that the HCV-infected patients included in our studywere �10 years older that those reportedby Sebastiani et al. In addition, our resultsagree with a previous study by our groupthat specifically addressed this issue (3).Concerning the higher serum ferritin lev-els detected in Sebastiani et al.’s popula-tion, it should be emphasized that most ofthe patients included in their study ap-pear to have been in more advancedstages of chronic hepatitis than those inour study. Moreover, we are unawarewhether they had ruled out hemochroma-tosis. Finally, we did not deny that thereare other factors apart from diabetes ac-counting for the high serum ferritin levelsdetected in HCV-infected patients. Infact, the relationship between alanineaminotransferase and serum ferritin levels

Letters


observed by Sebastiani et al. in univariateanalysis was also observed by us in a mul-tiple regression analysis (2). However, inlight of our results, it would appear thatdiabetes is not only associated with higherserum ferritin levels but also is a signifi-cant factor accounting for the higher fer-ritin levels detected in HCV-infectedpatients.

ALBERT LECUBE, MD1

CRISTINA HERNANDEZ, MD1

JOAN GENESCA, MD2

RAFAEL SIMO, MD1

From the 1Diabetes Research Unit, EndocrinologyDivision, Hospital Universitari Vall d’Hebron, Bar-celona, Spain; and the 2Liver Unit, Hospital Univer-sitari Vall d’Hebron, Barcelona, Spain.

Address correspondence to Dr. Rafael Simo, Di-abetes Research Unit, Endocrinology Division, Hos-pital General Vall d’Hebron, Pg. Vall d’Hebron 119-129, 08035 Barcelona, Spain. E-mail: [email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Sebastiani G, Vario A, Alberti A: Diabetes

is the main factor accounting for the highferritin levels detected in chronic hepatitisC virus infection (Letter). Diabetes Care28:1838, 2005

2. Lecube A, Hernandez C, Genesca J, Es-teban JI, Jardı R, Garcıa L, Simo R: Diabe-tes is the main factor accounting for thehigh ferritin levels detected in chronichepatitis C virus infection. Diabetes Care27:2669–2675, 2004

3. Lecube A, Hernandez C, Genesca J, Es-teban JI, Jardı R, Simo R: High prevalenceof glucose abnormalities in patients withhepatitis C virus infection: a multivariateanalysis considering the liver injury. Dia-betes Care 27:1171–1175, 2004

Glycemic Index,Glycemic Load, andGlycemic ResponseAre Not the Same

The paper by Hodge et al. (1) pub-lished in the November 2004 issueof Diabetes Care aptly contrasts the

potential benefits of moderately high-carbohydrate diets with a low glycemicindex (GI) versus diets that have a lowerglycemic load (GL) by virtue of a low car-bohydrate content. In their prospectiveanalysis of a cohort of �36,000 adults fol-lowed for 4 years, Hodge et. al found that

higher-carbohydrate diets were associ-ated with a lower risk of development oftype 2 diabetes. However, the type of car-bohydrate was equally important: low-GIcarbohydrates reduced the risk, whilehigh-GI carbohydrates increased the risk.Thus, low GI and low GL are not equiva-lent and produce different clinical out-comes.

Because this issue may be confusingto some readers, it is important to clarifythe difference between GI and GL. Boththe quality and quantity of carbohydratedetermines an individual’s glycemic re-sponse to a food or meal (2). By defini-tion, the GI compares equal quantities ofavailable carbohydrate in foods and pro-vides a measure of carbohydrate quality.Available carbohydrate can be calculatedby summing the quantity of available sug-ars, starch, oligosaccharides, and malto-dextrins. As defined (3), the GL is theproduct of a food’s GI and its total avail-able carbohydrate content: glycemicload � [GI � carbohydrate (g)]/100.

Therefore, the GL provides a sum-mary measure of the relative glycemic im-pact of a “typical” serving of the food.Foods with a GL �10 have been classifiedas low GL, and those with a value �20 ashigh GL (4). In healthy individuals, step-wise increases in GL have been shown topredict stepwise elevations in postpran-dial blood glucose and/or insulin levels(5).

It can be seen from the equation thateither a low-GI/high-carbohydrate foodor a high-GI/low-carbohydrate food canhave the same GL. However, while theeffects on postprandial glycemia may besimilar, there is evidence that the two ap-proaches will have very different meta-bolic effects, including differences in�-cell function (6), triglyceride concen-trations (7), free fatty acid levels (7), andeffects on satiety (8). Hence, the distinc-tion has important implications for theprevention and management of diabetesand cardiovascular disease. Our concernis that the use of the GL or “glycemicresponse” in isolation may lead to thehabi tua l consumpt ion of lower-carbohydrate diets.

The simplest way to consume a mod-erately high-carbohydrate, but low-GIdiet is to follow the new 2005 DietaryGuidelines for Americans (9) and to in-corporate the recommendations of theWorld Health Organization/Food andAgriculture Organization (10); that is, the

GI should be used to compare foods ofsimilar composition within food groups.By choosing the lower-GI options withina food category (breads, breakfast cereals,etc.), an individual automatically choosesthose with a lower GL. Because most fruitand vegetables, other than potatoes, arenot major contributors to carbohydrateintake, their GI should not be the basis forrestriction.

The important message for clinicians,nutritionists, and food industry profes-sionals is that the evidence, as it stands,suggests that for preventing type 2 diabe-tes, we ought to encourage low-GI carbo-hydrate foods but not those that simplyhave low “net carbs,” low GL, or producea low glycemic response.

ALAN W. BARCLAY, BSC, GRADDIPDIET,1

JENNIE C. BRAND-MILLER, PHD1

THOMAS M.S. WOLEVER, MD, PHD2

From the 1School of Molecular and Microbial Bio-sciences, University of Sydney, Sydney, Australia;and the 2Department of Nutritional Sciences, Uni-versity of Toronto, Toronto, Ontario, Canada.

Address correspondence to Alan Barclay, Diabe-tes Australia, GPO Box 9824, Sydney, NSW, 2001,Australia. E-mail: [email protected].

T.M.S.W. is president of the Board of Directorsof, holds stock in, and has received grant/researchsupport from G.I. Testing. T.M.S.W. also holdsstock in G.I. Laboratories.


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Hodge AM, English DR, O’Dea K, Giles

GG: Glycemic index and dietary fiber andthe risk of type 2 diabetes. Diabetes Care27:2701–2706, 2004

2. Sheard NF, Clark NG, Brand-Miller JC,Franz MJ, Pi-Sunyer FX, Mayer-Davis E,Kulkarni K, Geil P: Dietary carbohydrate(amount and type) in the prevention andmanagement of diabetes: a statement bythe American Diabetes Association. Dia-betes Care 27:2266–2271, 2004

3. Salmeron J, Manson JAE, Stampfer MJ,Colditz GA, Wing AL, Jenkins DJ, WingAL, Willett WC: Dietary fiber, glycemicload, and risk of non-insulin-dependentdiabetes mellitus in women. JAMA 277:472–477, 1997

4. Brand-Miller JC, Holt SHA, Petocz P:Reply to R. Mendosa. Am J Clin Nutr 77:994–995, 2003

5. Brand-Miller JC, Thomas M, Swan V, Ah-mad ZI, Petocz P, Colagiuri S: Physiolog-ical validation of the concept of glycemicload in lean young adults. J Nutr133:2728–2732, 2003

6. Wolever TMS, Mehling C: High-carbohy-drate/low-glycaemic index dietary advice

L E T T E R S


improves glucose disposition index insubjects with impaired glucose tolerance.Br J Nutr 87:477–487, 2002

7. Wolever TMS, Mehling C: Long-term ef-fect of varying the source or amount ofdietary carbohydrate on postprandialplasma glucose, insulin, triacylglycerol,and free fatty acid concentrations in sub-jects with impaired glucose tolerance.Am J Clin Nutr 76:5–56, 2002

8. Ball SD, Keller KR, Moyer-Mileur LJ, DingYW, Donaldson D, Jackson WD: Prolon-gation of satiety after low versus moder-ately high glycemic index meals in obeseadolescents. Pediatrics 111:488–494,2003

9. Dietary Guidelines for Americans 2005[article online]. Department of Healthand Human Services and the U.S. De-partment of Agriculture. Available fromwww.healthierus.gov/dietaryguidelines.Accessed 16 January 2005

10. Food and Agriculture Organization/World Health Organization: Carbohy-drates in human nutrition: report of aJoint FAO/WHO expert consultation.FAO Food and Nutrition Paper 66:1–140,1998

�-GlucosidaseInhibitors forPatients WithType 2 Diabetes

Response to van de Laar et al.

The authors of the Cochrane system-atic review carefully analyzed allavailable studies that fulfilled the

criteria of randomized clinical trials of atleast 12 weeks’ duration (1). With the ex-ception of one study (2), all registeredmortality and morbidity as secondary ob-jectives. Glycemic control was the pri-mary objective in 40 of 41 of these trials.Thus, the major legitimate conclusion ofthis careful analysis was that “AGIs[ -glucosidase inhibitors] have clear ben-eficial effects on glycemic control” mainlythrough their dose-dependent effect onpostprandial hyperglycemia.

However, the authors also state as oneof their main conclusions that they “foundno evidence for an effect on mortality ormorbidity.” Although this statement maybe mathematically correct, it is misleadingas it purports to be based on a solid anal-ysis of the data from the 41 studies. This isnot the case in their meta-analysis. Most

of the selected trials had a treatment pe-riod of �24 weeks; many were of3-month duration only and were there-fore not designed and powered to inves-tigate hard clinical end points such asmorbidity or mortality. This is well re-flected by the fact that, as reported by theauthors, information on morbidity ormortality could only be retrieved in 3 ofthe 41 trials. While one study showed asignificant treatment effect regarding car-diovascular events, the others presentedonly general statements without provid-ing any detail. It is well known that sam-ple sizes of individual clinical trials areoften too small to detect clinically impor-tant effects reliably and that this is one ofthe reasons why meta-analysis is em-ployed (3,4). However, hard end pointssuch as cardiovascular mortality are goingto be very rare in short-term duration stud-ies unless compensated for by a hugepopulation sample. Therefore, includingshort-term duration studies in their meta-analysis dilutes the cases of cardiovascularmortality. That biases the interpretation ofthe data analyzed.

The MERIA (MEta-analysis of RiskImprovement under Acarbose) analysis ofseven placebo-controlled, long-term, ran-domized studies examining the effect ofacarbose on cardiovascular-related mor-tality and morbidity showed a reductionof cardiovascular events in patients withtype 2 diabetes (5). This analysis is basedon all available acarbose studies with aminimum treatment duration of 52 weeksfrom a database including individual pa-tient data. Because of this, publication andselection bias were already ruled out, asdiscussed in the response (6) to the criti-cism raised by van de Laar and Lucassen.Unfortunately, the same criticism voicedpreviously is repeated in their meta-analysis without taking the detailed re-sponse into consideration. In summary,the MERIA analysis showed a beneficialeffect on cardiovascular complications inpatients with established type 2 diabetes,a finding which is in accordance with theresults from the STOP-NIDDM trial in sub-jects with impaired glucose tolerance (7).

We fully agree with the authors’ state-ment that prospective trials with theprimary objective of investigating cardio-vascular events and mortality are requiredto confirm the beneficial effect of acarboseon cardiovascular events in these high-risk populations. However, the combineddata from the STOP-NIDDM trial and the

MERIA meta-analysis are highly sugges-tive of the preventive effects of acarboseon cardiovascular complications in sub-jects with glucose intolerance.

MARKOLF HANEFELD, MD, PHD1

ROBERT G. JOSSE, MBBS2

JEAN-LOUIS CHIASSON, MD3

From the 1Centre for Clinical Studies, Science andTechnology Transfer–Technical UniversityDresden, Dresden, Germany; the 2Department ofMedicine, St. Michael’s Hospital, University of To-ronto, Toronto, Canada; and the 3Centre Hospitalierde l’Universite de Montreal-Hotel-Dieu and Depart-ment of Medicine, Montreal University, Montreal,Canada.

Address correspondence to Markolf Hanefeld,MD, PhD, Science and Technology Transfer–GWTTU Dresden, Centre for Clinical Studies, Fiedler Str.34, Dresden 01307, Germany. E-mail: [email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. van de Laar FA, Lucassen PL, Akkermans

RP, von de Lisdonk EH, Rutten GE, vanWeel C: -Glucosidase inhibitors for pa-tients with type 2 diabetes: results from aCochrane systematic review and meta-analysis. Diabetes Care 28:154–163, 2005

2. Holman RR, Cull CA, Turner RC: A ran-domized double-blind trial of acarbose intype 2 diabetes shows improved glycemiccontrol over 3 years (U.K. Prospective Di-abetes Study 44). Diabetes Care 22:960–964, 1999 (erratum in Diabetes Care 22:1922, 1999)

3. Thompson SG: Biostatistics in Clinical Tri-als. Redmond CK, Colton T, Eds. Chi-chester, NY, Wiley, 2001

4. Committee for Propietary MedicinalProducts: Points to Consider on ApplicationWith 1: Meta-analyses, 2: One Pivotal Study.London, European Agency for the Evalu-ation of Medicinal Products, 2001

5. Hanefeld M, Cagatay M, Petrowitsch T,Neuser D, Petzinna D, Rupp M: Acarbosereduces the risk for myocardial infarctionin type 2 diabetic patients: meta-analysisof seven long-term studies. Eur Heart J 25:10–16, 2004

6. Hanefeld MG: Meta-analysis of long-termstudies to assess the effect of acarbose oncardiovascular risk reduction scientifi-cally credible: reply (Letter). Eur Heart J25:1179–1180, 2004

7. Chiasson JL, Josse RG, Gomis R, HanefeldM, Karasik A, Laakso M; STOP-NIDDMTrial Research Group: Acarbose treatmentand the risk of cardiovascular disease andhypertension in patients with impairedglucose tolerance: the STOP-NIDDM trial.JAMA 290:486–494, 2003

Letters


�-GlucosidaseInhibitors forPatients WithType 2 Diabetes

Response to Hanefeld et al.

H anefeld et al. (1) assert that the con-clusion (“no evidence for an effecton mortality or morbidity”) from

our systematic review on the effects of -glucosidase inhibitors for patients withtype 2 diabetes was biased. Furthermore,they claim to have found evidence forsuch an effect based on their own meta-analysis. We disagree with both statements.

First, we would like to underline thatthe solid basis of our results is a systematicreview and that meta-analyses were onlyapplied when this was methodologicallysound. The extensive search for all possi-ble trials investigating -glucosidase in-hibitor monotherapy yielded only onestudy with prospectively collected data onmorbidity or mortality (2), so a meta-analysis could not be done with these endpoints; therefore, we concluded that noevidence for an effect on mortality andmorbidity could be found (which is es-sentially different from “evidence for noeffect”). In the above-mentioned study, itwas reported that for the entire treatmentgroup ( -glucosidase inhibitors givenboth as monotherapy and as additionaltherapy), no effects of acarbose on cardio-vascular end points were found.

This makes it quite remarkable thatthis particular study (2) was not includedin the MERIA (MEta-analysis of Risk Im-provement under Acarbose) study (3).Hanefeld et al. assert that this meta-analysis shows a beneficial effect of acar-bose on the occurrence of myocardialinfarctions. If it had been included in theMERIA study, it would have been thestudy with the second longest duration, itwould have nearby doubled the numberof patients, and it would have been theonly study with a sound method of col-lecting end points. This points to the factthat the sole use of a manufacturer’s data-base is not a reliable method for the selec-tion of studies for a meta-analysis and thatan extensive systematic review is neces-sary to reduce the risk of selection bias.

Other differences between the con-clusions of MERIA and our Cochrane re-

view can be explained by differences ininclusion and methodological robustness.Three of the seven studies in MERIA werealso included in our Cochrane review, butno reliable data on cardiovascular out-comes could be obtained. The four otherpublications were excluded from our re-view, mainly because no data on -gluco-sidase inhibitor monotherapy wereavailable or accessible. Moreover, it shouldbe noted that there was no quality assess-ment of the studies included in MERIA.Other serious concerns about the MERIAstudy were expressed in our previous let-ter and remain largely unresolved (4).

In conclusion, there is currently noevidence for an effect on cardiovascularmorbidity and mortality of monotherapywith -glucosidase inhibitors in patientswith type 2 diabetes. In the near future,indirect evidence may be derived fromanother Cochrane review on the effects of -glucosidase inhibitors for patientswith glucose intolerance (5). We arepleased that the authors of the main stud-ies in this field already have assured theircooperation.

FLORIS A. VAN DE LAAR, MD

PETER L. LUCASSEN, MD, PHD

From the Department of General Practice, RadboudUniversity Nijmegen Medical Centre, Nijmegen, theNetherlands.

Address correspondence to Floris van de Laar,Radboud University Nijmegen Medical Centre, De-partment of General Practice, 229 HAG, P.O. Box9101, 6500 HB Nijmegen, Netherlands. E-mail:[email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Hanefeld M, Josse RG, Chiasson J-L:

-Glucosidase inhibitors for patients withtype 2 diabetes (Letter). Diabetes Care 28:1840, 2005

2. Holman RR, Cull CA, Turner RC: A ran-domized double-blind trial of acarbose intype 2 diabetes shows improved glycemiccontrol over 3 years (U.K. Prospective Di-abetes Study 44). Diabetes Care 22:960–964, 1999 (erratum in Diabetes Care 22:1922, 1999)

3. Hanefeld M, Cagatay M, Petrowitsch T,Neuser D, Petzinna D, Rupp M: Acarbosereduces the risk for myocardial infarctionin type 2 diabetic patients: meta-analysisof seven long-term studies. Eur Heart J 25:10–16, 2004

4. van de Laar FA, Lucassen PL: No evidencefor a reduction of myocardial infarctionsby acarbose (Letter). Eur Heart J 25:1179,

2004 (reply 1179–1180)5. van de Laar FA, Lucassen PLBJ, Akker-

mans RP, Van de Lisdonk EH, De GrauwWJC: Alpha-glucosidase inhibitors forpeople with impaired glucose toleranceor impaired fasting blood glucose (Proto-col). Cochrane Database Syst Rev 2004. Is-sue 4. Art. no. CD005061. DOI: 10.1002/14651858

ChromiumSupplementationDoes Not ImproveGlucose Tolerance,Insulin Sensitivity, orLipid Profile: ARandomized,Placebo-Controlled,Double-Blind Trial ofSupplementation inSubjects WithImpaired GlucoseTolerance

Response to Gunton et al.

W e read the recent article by Gun-ton et al. (1) with great interestand feel that it warrants com-

ment. In this study, the authors statedthat they “found no beneficial effect ofchromium supplementation in the treat-ment of people with IGT [impaired glu-cose tolerance].” The results are inconflict with other clinical studies thatshowed chromium picolinate can en-hance or normalize impaired glucose me-tabolism, as described in a recent review(2). The lack of effect described by theauthors may be explained by the apparentlow dose of elemental chromium used inthe study.

The authors stated that the chromiumpicolinate “dose (at 800 g/day) was atthe higher end of the ranges used in pre-vious studies” (1). However, chromiumpicolinate administered at 800 g per dayyields a daily dose of 100 g per day ofelemental chromium (i.e., chromium pi-colinate contains 12.4% elemental chro-mium). An elemental chromium dose of100 g a day is half of the suggested min-imum amount (200 g) of elementalchromium previously shown to exhibit

Letters


https://www.researchgate.net/publication/13193145_Microalbuminuria_prevalence_varies_with_age_sex_and_puberty_in_children_with_type_1_diabetes_followed_from_diagnosis_in_a_longitudinal_study_Oxford_Regional_Prospective_Study_Group?el=1_x_8&enrichId=rgreq-01613bfbc1e5ab47321c1f42b0c61838-XXX&enrichSource=Y292ZXJQYWdlOzc3NjEyMzE7QVM6MjU3NDYyMDYyMjg0ODAwQDE0MzgzOTUxNDcyNTQ=


efficacy in glucose and lipid metabolism(2). A daily dose of 200–1,000 g of ele-mental chromium, as chromium picoli-nate, is the efficacious dosage range usedin previous studies.

Bullivants Natural Health Products,the supplier of the study products used bythe authors, stated that 400 g of thechromium picolinate product they pro-duce yields 50 g of elemental chro-mium. The study was conducted inAustralia, and the 50-g elemental chro-mium dose is also the maximum dailydose allowed by the Australian Therapeu-tic Goods Administration (3).

It was also interesting to note that al-though the serum chromium levels signif-icantly rose in the active group, the serumchromium levels were not significantlyhigher in the active group than in the pla-cebo group after 3 months of supplemen-tation (active group 5.2 � 8.9 nmol/l,placebo group 4.4 � 4.0 nmol/l). Forthese reasons, we believe study subjects inthe active group may have been adminis-tered daily doses of 50 g elemental chro-mium, twice daily.

We recommend future studies beconducted in people with impaired glu-cose tolerance (following criteria definedby the American Diabetes Association)using daily doses of chromium picolinateproviding �200–1,000 g of elementalchromium for at least 90 days. We alsorecommend evaluating efficacy using the75-g oral glucose tolerance test with cal-culation of the area under the curve usingthe trapezoidal method.

JAMES KOMOROWSKI, MS

VIJAYA JUTURU, PHD

From the Technical Services & Scientific Affairs De-partment, Nutrition 21, Inc., Purchase, New York.

Address correspondence to Vijaya Juturu, PhD,Nutrition 21, Inc., 4 Manhattanville Rd., Purchase,NY 10577. E-mail: [email protected].

The authors are employees of Nutrition 21, Inc.,which manufactures products containing chro-mium.


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Gunton JE, Cheung NW, Hitchman R,

Hams G, O’Sullivan C, Foster-Powell K,McElduff A: Chromium supplementationdoes not improve glucose tolerance, insu-lin sensitivity, or lipid profile: a random-ized, placebo-controlled, double-blindtrial of supplementation in subjects withimpaired glucose tolerance. Diabetes Care

28:712–713, 20052. Cefalu WT, Hu FB: Role of chromium in

human health and in diabetes. DiabetesCare 27:2741–2751, 2004

3. Complementary Medicines EvaluationCommittee (CMEC), meeting 41, 1 Au-gust 2003, public recommendation sum-mary [summary online]. Available athttp://www.tga.gov.au/docs/html/cmec/cmecdr41.htm. Accessed 1 August 2003

ChromiumSupplementationDoes Not ImproveGlucose Tolerance,Insulin Sensitivity, orLipid Profile: ARandomized,Placebo-Controlled,Double-Blind Trial ofSupplementation inSubjects WithImpaired GlucoseTolerance

Response to Komorowski andJuturu

W e thank Komorowski and Juturu(1) for their interest in our study,and we agree that it is possible

that higher doses of chromium may showsome effects. The subjects in this studyreceived 100 g of elemental chromiumdaily, administered as 800 g of chro-mium picolinate, for 3 months. The smallincrease in serum chromium in the activegroup is probably appropriate for thisdose. It is worth noting that significantuncertainty remains regarding how bestto measure whole-body chromium status,as there are no well-defined outcomes toallow determination of a therapeuticrange. This unfortunately includes assay-ing of serum.

The Australian Therapeutic GoodsAdministration daily dose recommenda-tion was adopted in 2004 and is based onconcerns about the safety of higher dosesas raised by the Complementary Medi-cines Evaluation Committee (2), includ-ing two reports of renal failure (3,4). Ourstudy commenced before 2003 and was

scientifically reviewed by our local ethicscommittee.

We do not agree that the results of thestudy conflict with the literature. Studiesof chromium supplementation in nondi-abetic subjects and people with normalglucose tolerance, insulin resistance,and/or impaired glucose tolerance (IGT)have not produced any consistent bene-fits, as reviewed by Cefalu et al. (5), Yeh etal. (6), Althuis et al. (7), and Gunton et al.(8). In contrast, some studies in subjectswith diabetes have shown significant ben-efits (5,8), and further studies in thisgroup will be of great interest.

Our study was conducted in peoplediagnosed with impaired glucose toler-ance according to the American DiabetesAssociation criteria. Efficacy was alsoevaluated using area under the curve dur-ing glucose tolerance testing, but this wasalso nonsignificant (data not shown). Wefeel that at this dose, chromium picolinatesupplementation in subjects with im-paired glucose tolerance is ineffective.

We note that Komorowski and Juturuare affiliated with Nutrition 21, Inc.,which markets chromium picolinate.

JENNY E. GUNTON, MBBS, FRACP, PHD1,2

N. WAH CHEUNG, MBBS, FRACP3

ROSEMARY HITCHMAN, RNCM1

GRAHAM HAMS, MAPPSC4

CHRISTINE SULLIVAN, RNDEC3

KAYE FOSTER-POWELL, BSC, MNUTR, DIETAPD3

AIDAN MCELDUFF, MBBS, FRACP1

From the 1Department of Endocrinology, RoyalNorth Shore Hospital, St. Leonards, Sydney, Austra-lia; 2Cellular and Molecular Physiology, Joslin Dia-betes Center, Boston, Massachusetts ; the3Department of Diabetes and Endocrinology, West-ern Sydney Area Health Service, Westmead and Ne-pean Hospitals, Sydney, Australia; and 4PacificLaboratory Medical Services, Royal North ShoreHospital, Sydney, Australia.

Address correspondence to Jenny Gunton, c/oKahn Laboratory, Level 6, Joslin Diabetes Center,One Joslin Place, Boston, MA 02215. E-mail:[email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Komorowski J, Juturu V: Chromium sup-

plementation does not improve glucosetolerance, insulin sensitivity, or lipid pro-file: a randomized, placebo-controlled,double-blind trial of supplementation insubjects with impaired glucose tolerance:response to Gunton et al. (Letter). Diabe-tes Care 28:1841–1842, 2005

2. Complementary Medicines Evaluation

Letters


https://www.researchgate.net/publication/8212250_Role_of_Chromium_in_Human_Health_and_in_Diabetes?el=1_x_8&enrichId=rgreq-01613bfbc1e5ab47321c1f42b0c61838-XXX&enrichSource=Y292ZXJQYWdlOzc3NjEyMzE7QVM6MjU3NDYyMDYyMjg0ODAwQDE0MzgzOTUxNDcyNTQ=



Committee (CMEC), meeting 41, 1 Au-gust 2003, public recommendation sum-mary [summary online]. Available athttp://www.tga.gov.au/docs/html/cmec/cmecdr41.htm. Accessed April 2005

3. Wasser WG, Feldman NS: Chronic renalfailure after ingestion of over-the-counterchromium picolinate. Ann Intern Med 126:410, 1997

4. Cerulli DW, Grabe I, Gauthier M, Maloneand McGoldrick MD: Chromium picoli-nate toxicity. Ann Pharmacother 32:428–431, 1998

5. Cefalu WT, Hu FB: Role of chromium inhuman health and in diabetes. DiabetesCare 27:2741–2751, 2004

6. Yeh GY, Eisenberg DM, Kaptchuk TJ,Phillips RS: Systematic review of herbsand dietary supplements for glycemiccontrol in diabetes (Review Article). Dia-betes Care 26:1277–1294, 2003

7. Althuis MD, Jordan NE, Ludington EA,Wittes JT: Glucose and insulin responsesto dietary chromium supplements: ameta-analysis. Am J Clin Nutr 76:148–155, 2002

8. Gunton JE, Hams G, Hitchman R,McElduff A: Serum chromium does notpredict glucose tolerance in late preg-nancy. Am J Clin Nutr 73:99–104, 2001

The Case for BiennialRetinopathyScreening in Childrenand Adolescents

Response to Maguire et al.

I have read the article by Maguire et al.(1) with interest. In a large, longitudi-nal cohort of type 1 diabetic children

and adolescents, the study describes thenatural history and prevalence of retinop-athy. Although not the main focus of thearticle, their data also highlight two impor-tant points. First, Maguire et al. highlightthe relationship between puberty and mi-crovascular complications, as evidenced bythe significantly increased incidence of ret-inopathy after 2 years’ follow-up in subjectsaged �11 years and after 5 years’ follow-upin subjects aged �11 years. These findingswere independent of glycemic control. Sec-ond, the study reveals that in 136 subjectswith evidence of retinopathy at outset, 64(47%) regressed, after a median of 3.1 years,to either lower-grade retinopathy or to nor-mal at the end of follow-up, although the

median age at which this occurs is notgiven.

These data are comparable to the nat-ural history of microalbuminuria as de-scribed in longitudinal studies of childrenand adolescents, including the OxfordRegional Prospective Study in the U.K.(2). In this study, puberty (using age 11years as a surrogate marker for onset ofpuberty) conferred a threefold increasedrisk of microalbuminuria, independent ofpoor glycemic control, and these datahave been in part confirmed by previousstudies from Couper et al. (3). This mayrelate in part to pubertal hormonal vari-ables, as recent evidence suggests that mi-croalbuminuria risk in this age-group isassociated with growth hormone hyper-secretion and insulin resistance, particu-larly in females (4). Furthermore, in�60% of subjects, microalbuminuriasubsequently regressed, and in these sub-jects compared with those with persistentmicroalbuminuria, mean HbA1c levelswere similar before onset of microalbu-minuria (median age 15.8 years) but werelower thereafter. Thus, adolescent sub-jects with regression may be individualsin whom microalbuminuria initially man-ifests during the poor glycemic controland insulin resistant state associated withpuberty but demonstrate regression whenglycemic control improves in the postpu-bertal period. One may hypothesize thatmicroalbuminuria may subsequently re-appear in these “at risk” subjects in laterlife; however, this remains unproven.

These same mechanisms may applyto the pathogenesis and natural history ofretinopathy during adolescence. We hopeMaguire et al. will further detail the de-mographic details and risk factors for pro-gression and regression of retinopathy, asadequate longitudinal data in this age-group are currently lacking.

RAKESH AMIN, MBCHB, MRCP

From the Department of Paediatric Endocrinologyand Diabetes, Royal Manchester Children’s Hospi-tal, Manchester, U.K.

Address correspondence to Rakesh Amin,MBCHB, MRCP, Royal Manchester Childrens Hos-pital, Department of Paediatric Endocrinology,Hospital Road, Manchester, M27 4HA U.K. E-mail:[email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Maguire A, Chan A, Cusumano J, Hing S,

Craig M, Silink M, Howard N, Donaghue

K: The case for biennial retinopathyscreening in children and adolescents. Di-abetes Care 28:509–513, 2005

2. Schultz CJ, Konopelska-Bahu T, DaltonRN, Carroll TA, Stratton I, Gale EA, NeilA, Dunger DB: Microalbuminuria preva-lence varies with age, sex, and puberty inchildren with type 1 diabetes followedfrom diagnosis in a longitudinal study:Oxford Regional Prospective StudyGroup. Diabetes Care 22:495–502, 1999

3. Couper JJ, Clarke CF, Byrne GC, JonesTW, Donaghue KC, Nairn J, Boyce D,Russell M, Stephens M, Raymond J, BatesDJ, McCaul K: Progression of borderlineincreases in albuminuria in adolescentswith insulin-dependent diabetes mellitus.Diabet Med 14:766–771, 1997

4. Amin R, Williams RM, Frystyk J,Umpleby M, Matthews D, Orskov H, Dal-ton RN, Dunger DB: Increasing urine al-bumin excretion is associated withgrowth hormone hypersecretion and re-duced clearance of insulin in adolescentsand young adults with type 1 diabetes: theOxford Regional Prospective Study. ClinEndocrinol (Oxf) 62:137–144, 2005

The Case for BiennialRetinopathyScreening in Childrenand Adolescents

Response to Amin

In response to the letter from Amin (1),we provide further details on the riskfactors for progression and regression

of retinopathy in our natural history study(2). Of 136 patients with retinopathy atbaseline, 72 progressed or persisted com-pared with 50 patients (37%) who re-gressed to no retinopathy and 14 whoregressed to a lower grade of retinopathyafter a median follow-up of 3.1 years inboth groups. Those who progressed orpersisted had longer duration (7.8 vs. 5.9years, P � 0.0086) and higher HbA1c (9.1vs. 8.5%, P � 0.034) at baseline and wereolder at follow-up (18.1 vs. 17.4 years,P � 0.037). In multivariate logistic re-gression analysis of baseline factors, gly-cemic control (P � 0.0057) and durationof diabetes (P � 0.017) were significantpredictors of progression/persistence ver-sus regression of retinopathy to normal;these data are consistent with the DiabetesControl and Complications Trial. Neitherhigh baseline BMI nor blood pressure per-

Letters










https://www.researchgate.net/publication/8061990_Increasing_urine_albumin_excretion_is_associated_with_growth_hormone_hypersecretion_and_reduced_clearance_of_insulin_in_adolescents_and_young_adults_with_type_1_diabetes_The_Oxford_Regional_Prospectiv?el=1_x_8&enrichId=rgreq-01613bfbc1e5ab47321c1f42b0c61838-XXX&enrichSource=Y292ZXJQYWdlOzc3NjEyMzE7QVM6MjU3NDYyMDYyMjg0ODAwQDE0MzgzOTUxNDcyNTQ=









https://www.researchgate.net/publication/227756879_Progression_of_borderline_increases_in_albuminuria_in_adolescents_with_insulin-dependent_diabetes_mellitus?el=1_x_8&enrichId=rgreq-01613bfbc1e5ab47321c1f42b0c61838-XXX&enrichSource=Y292ZXJQYWdlOzc3NjEyMzE7QVM6MjU3NDYyMDYyMjg0ODAwQDE0MzgzOTUxNDcyNTQ=







centiles were significant contributors tothe incidence or progression/regression ofretinopathy.

We agree that our data support a re-lationship between puberty and micro-vascular complications (using age �11years as a surrogate marker for puberty).In this cohort, however, we did not find arelationship between pubertal stagingand progression/regression of retinopa-thy, but the small number (n � 13) ofprepubertal patients with retinopathyprevented us from answering this ques-tion. In addition, the permissive effect ofpuberty, growth hormone, and IGF-1may be more pronounced in the patho-genesis of microalbuminuria than reti-nopathy. When we studied an incidentcohort after 6 years’ duration, we foundthat higher pubertal stage (Tanner stage4–5 vs. 1–3) had a larger effect on eleva-

tion of albumin excretion than retinopa-thy (odd ratios 5.2 vs. 3.4). The effect onalbumin excretion rate was independentof HbA1c (3).

ANN M. MAGUIRE, MB, BAO, BCH1

ALBERT K.F. CHAN, MAPPSTAT1

JANINE M. CUSUMANO1

STEPHEN J. HING, MBBS1

MARIA E. CRAIG, PHD1,2,3

MARTIN SILINK, MD1,2

NEVILLE J. HOWARD, MBBS1

KIM C. DONAGHUE, PHD1,2

From the 1Institute of Endocrinology and Diabetes,The Children’s Hospital at Westmead, Sydney, Aus-tralia; the 2Department of Paediatrics and ChildHealth, University of Sydney, Sydney, Australia; andthe 3School of Women’s and Children’s Health, Uni-versity of New South Wales, Sydney, Australia.

Address correspondence to Dr. Ann Maguire, In-stitute of Endocrinology and Diabetes, The Chil-dren’s Hospital at Westmead, Locked Bag 4001,

Sydney, NSW 2145, Australia. E-mail: [email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Amin R: The case for biennial retinopathy

screening in children and adolescents (Let-ter). Diabetes Care 28:1843, 2005

2. Maguire A, Chan A, Cusumano J, Hing S,Craig M, Silink M, Howard N, DonaghueK: The case for biennial retinopathyscreening in children and adolescents. Di-abetes Care 28:509–513, 2005

3. Donaghue KC, Craig ME, Chan AKF,Fairchild JM, Cusumano JM, Verge CF,Crock PA, Hing SJ, Howard NJ, Silink M:Prevalence of diabetes complications sixyears after diagnosis in an incident cohortof childhood diabetes. Diabet Med. Inpress

Letters


https://www.researchgate.net/publication/7999052_The_Case_for_Biennial_Retinopathy_Screening_in_Children_and_Adolescents?el=1_x_8&enrichId=rgreq-01613bfbc1e5ab47321c1f42b0c61838-XXX&enrichSource=Y292ZXJQYWdlOzc3NjEyMzE7QVM6MjU3NDYyMDYyMjg0ODAwQDE0MzgzOTUxNDcyNTQ=





https://www.researchgate.net/publication/7830294_Prevalence_of_diabetes_complications_6_years_after_diagnosis_in_an_incident_cohort_of_childhood_diabetes?el=1_x_8&enrichId=rgreq-01613bfbc1e5ab47321c1f42b0c61838-XXX&enrichSource=Y292ZXJQYWdlOzc3NjEyMzE7QVM6MjU3NDYyMDYyMjg0ODAwQDE0MzgzOTUxNDcyNTQ=







Chromium Supplementation Does Not Improve Glucose Tolerance, Insulin Sensitivity, or Lipid Profile:...

Documents

Transcript of Chromium Supplementation Does Not Improve Glucose Tolerance, Insulin Sensitivity, or Lipid Profile:...