OTc

YF

M

Tc

FDDN

S2

Rl

RIGINAL ARTICLEShe relationship between 24-hour integrated glucoseoncentrations and % glycohemoglobin

OUSSEF HASSAN, BRADFORD JOHNSON, NICOLE NADER, MARY C. GANNON, andRANK Q. NUTTALL

INNEAPOLIS, MINNESOTA

Objective: Since glycohemoglobin values are widely used clinically as a surrogatefor average glucose concentration over an extended period of time, we decided todetermine the actual relationship between 24-hour integrated glucose values andpercent total glycohemoglobin (%tGHb) in cohorts of people with and withoutdiabetes. Research Design and Methods: In 48 people without known diabetes withknown stability of fasting glucose over a 1-year period of time, the calculated24-hour integrated glucose concentration was compared with their %tGHb. In 15normal young medical students, the glucose area response was determined from46 venous blood samples obtained during a 24-hour period and compared withtheir %tGHb. In 18 people with type 2 diabetes, interstitial glucose concentrationswere monitored using the Continuous Glucose Monitoring System (Medtronic Min-iMed, Inc., Sylmar, Calif) for 3 days at 20-day intervals over 100 days. %tGHb wasperformed at 20-day intervals simultaneously. In 29 people with untreated type 2diabetes, glucose area response was determined from 46 venous blood samplesobtained during a 24-hour period and compared with their %tGHb after being on astandardized diet provided to the subjects for at least 5 weeks. The %tGHb and24-hour profiles were stable. Results: There was an excellent correlation betweenthe mean 24-hour glucose concentration and the %tGHb among subjects withdiabetes. The correlation was poor among subjects without diabetes. The relation-ship was curvilinear when plotted as a single group. Alternatively when data fromsubjects with or without diabetes were plotted separately, the slopes were identicalbut the y-intercepts were different. Conclusion: The relationship between the meanglucose concentration integrated over an extended period of time and the %tGHbis not linear. The reason for this nonlinearity remains to be determined. This non-linearity needs to be considered in the clinical interpretation of %tGHb (and prob-ably HbA1c) in reference to glucose values. (J Lab Clin Med 2006;147:21–26)

Abbreviations: %HbA � percent hemoglobin A ; %tGHb � percent total glycohemoglobin;

1c 1c

RBCs � red blood cells; HPLC � high-performance liquid chromatography

tctp

1n

0

©

d

he %HbA1c or %tGHb adjusted to representHbA1c values is widely used as an index of theaverage (24-hour integrated) blood glucose con-

entration over an extended period of time. That is, it is

rom the Endocrinology, Metabolism and Nutrition Section, theepartment of Veterans Affairs Medical Center, Minneapolis, theepartment of Medicine, and the Department of Food Science andutrition, University of Minnesota, Minneapolis, Minnesota.

ubmitted for publication February 24, 2005; revision submitted May4, 2005; accepted for publication May 25, 2005.

eprint requests: Frank Q. Nuttall, Chief, Endocrinology, Metabo-

ism, and Nutrition Section (111G), Veterans Affairs Medical Center,

acitly assumed that the %HbA1c or %tGHb is an ac-urate surrogate for the average blood glucose concen-ration for some months before the time of blood sam-ling. Indeed, the national criteria for acceptable blood

Veterans Drive, Minneapolis, MN 55417; e-mail:utta001@umn.edu.

022-2143/$ – see front matter

2006 Mosby, Inc. All rights reserved.

oi:10.1016/j.lab.2005.05.009

21

gov

cRgI

ttoonct

w

5%cmTaas

2gsputtsrl

t4tMS2tdTue[

um

caksofhmts

M

ACsdhpoBamuw

midrc

4

t

R

wlt�cdfotcttco0c

J Lab Clin Med22 Hassan et al January 2006

lucose control in people with diabetes now are basedn glycohemoglobin values and not on blood glucosealues.1

It also generally is assumed that there is a linearorrelation between the % of hemoglobin molecules inBCs derivatized by glucose and the average bloodlucose concentration over the life span of the RBCs.2

n this study, we tested this assumption.Specifically, we determined the relationship between

he 24-hour integrated blood glucose concentration (es-imated or measured) and the %tGHb over a wide rangef glucose values. The %tGHb as determined by bor-nate affinity chromatography was used because it doesot measure derivitization of hemoglobin by non-glu-ose adducts and is little affected by hemoglobinopa-hies.

Subjects studied. Four different groups of subjectsere studied.Group One. In 48 normal subjects with a mean age of

0 years (range 31–82 years), the stability of thetGHb, fasting glucose, Hb, RBC count, and reticulo-

yte count were determined monthly over a 12–26-onth period. All laboratory values were very stable.3

he fasting glucose data obtained in that study weredjusted upward by 9% to convert the fasting values tonticipated 24-hour integrated values4 and used in thistudy.

Group Two. In 15 young medical students (mean age3 years, range of 20–26 years), the 24-hour serumlucose concentrations and %tGHb were obtained on aingle occasion. The diet consumed during the 24-houreriod of study was designed to mimic the students’sual eating habits as determined by the research die-ician. Blood was obtained at 15-minute intervals forhe first hour after meals, at one-half hour during theecond hour after each meal, and hourly throughout theemainder of the 24-hour period of the study. Reticu-ocyte counts were only obtained in 4 subjects.

Group Three. In 18 subjects with type 2 diabetes,reated only with oral agents (mean age 61 years, range5–76 years), the interstitial fluid glucose concentra-ions were monitored using the Continuous Glucose

onitoring System (CGMS; Medtronic MiniMed, Inc.,ylmar, Calif) for an average of 3 days (2–4 days) at0-day intervals for 100 days. The %tGHb and %re-iculocytes were determined at 20-day intervals for 100ays as well and were very stable over this time period.he interstitial glucose concentrations were adjustedpward by 5 mg/dL to account for the reported differ-nce between interstitial and serum glucose valuesCGMS, Instructions for Use (IFU), 2000].

Group Four. This group consisted of 29 subjects withntreated type 2 diabetes, enrolled in studies to deter-

ine the effect of diet composition on blood glucose t

ontrol.5,6 The diet was specifically defined, constant,nd typical of the American diet. The %tGHb wasnown to be stable for 5 weeks or greater, and 24-hourerum glucose profiles were determined after 5 weeksn the diet. Blood was obtained at 15-minute intervalsor the first hour, at one-half hour during the secondour after each meal, and hourly throughout the re-ainder of the 24-hour period of the study. The %re-

iculocytes also was determined. The mean age of theubjects was 61 years (range 39–79 years).

ETHODS

The study was approved by the Department of Veteransffairs Medical Center and the University of Minnesotaommittees on Human Subjects and written informed con-

ent was obtained from all participants. The %tGHb wasetermined by a HPLC boronate affinity method, which isighly specific for glycation of globin in hemoglobin and is arecise system (coefficient of variation � 2% of the valuebtained) (Primus Corp., Kansas City, Mo, and Variant II,io-Rad, Hercules, Calif). Glucose was determined by anutomated standard enzymic method (Vitros glucose instru-ent; Johnson & Johnson, Rochester, NY). Hemoglobin val-

es and reticulocyte counts and percentages were determinedith a Coulter GEN-S apparatus (Miami, Fla).The mean blood glucose in group 1 was determined byultiplying the fasting glucose concentration by 1.09, as

ndicated. In groups 2 and 4, the mean blood glucose wasetermined by dividing the 24 hour integrated glucose areaesponse by 24 hours. In group 3, the mean blood glucose wasalculated automatically from the CBGM.

Statistical analysis was performed with GraphPad Prism.0 (San Diego, Calif) for the Macintosh computer.The research was carried out according to the principles of

he Declaration of Helsinki.

ESULTS

Initially, the data were analyzed separately for peopleith (n � 47) and without (n � 63) diabetes, using

inear regression. The slope for the 2 groups was iden-ical (0.027). However, the intercepts were different (y

3.20 and y � 1.82, respectively). The correlationoefficient was excellent for the data from people withiabetes (R2 � 0.866), but relatively poor for the datarom people without diabetes (R2 � 0.263) as noted bythers.7–10 Since a priori there is no reason to expecthe glycation rate not to be continuous, the data wereombined. As expected, when the data were combined,he relationship could be described as more curvilinearhan linear (Figure 1, A and B). The correlation coeffi-ient for a curvilinear relationship was best for a sec-nd-degree polynomial (quadratic) relationship (R2 �.92), and the y-intercept was �1.1 � 0.4, the 95%onfidence interval was �1.9 to 0.3 and the line crossed

he x-axis at �18-mg/dL glucose at 0%tGHb. The

cstdcHqassw

tbm%Tl

diswn

D

mioAfgbbataicck

%(poTccttociottb2lur

p2ombsit

Fcetcagm

J Lab Clin MedVolume 147, Number 1 Hassan et al 23

orrelation coefficient obtained when a linear regres-ion was applied to the data was R2 � 0.89. However,he y-intercept was 1.34 � 0.2 %tGHb (95% confi-ence interval � 1.0 to 1.7). The correlation coeffi-ients for the slopes were not different statistically.owever, the correlation coefficient was best for auadratic relationship. Visual inspection of the datalso indicated that a curvilinear relationship best de-cribed the results. Both intercepts were statisticallyignificantly different from zero, but the differencesere in opposite directions.We previously had reported that RBC survival, de-

ermined both by a carbon monoxide method as well asy determination of the proportion (%) of the red cellass attributable to reticulocytes, was reduced as thetGHb increased in people with type 2 diabetes.herefore, we plotted the %tGHb against the %reticu-

ig 1. Relationship between 24-hour integrated mean plasma glucoseoncentration and %tGHb in 4 groups of subjects. (A) There is anxcellent correlation between integrated plasma glucose concentra-ion over 24 hours and %tGHb (R2 � 0.89). (B) The relationship isurvilinear; thus, a second-degree polynomial (quadratic equation) isbetter description of the data (R2 � 0.92), particularly at the lowerlucose values. However, the glucose intercept at 0 %tGHb was �18g/dL.

ocyte count in this study, which also included non-

iabetic subjects. The reticulocyte count was obtainedn all subjects with the exception of 11 normal medicaltudents. The %reticulocyte count correlated poorlyith the %tGHb (y � 4.4 � 1.5x) (R2 � 0.065) (dataot shown).

ISCUSSION

The %tGHb as determined by boronate affinity chro-atography is specific for the ketoamine adduct result-

ng from glucose attachment to primary amino groupsn amino acids in the globin molecules in hemoglobin.ttachment of non-glucose-derived adducts could af-

ect the binding of glucose to specific sites on thelobin protein peptide chains but would not be detectedy this method. Thus, differences in the relationshipetween the %tGHb and mean glucose concentrationre due to differences in either the ketoamine attachedo the amino groups in the globin protein at the time ofnalysis or the differences in the life span of the RBCs,e, the duration of time they are exposed to a givenirculating glucose concentration. Alternatively, theyould be due to a variation in the rate of removal of theetamine adduct from the globin.11–13

Previously we have determined the stability of thetGHb, fasting glucose concentration, Hb, RBC count

and presumably RBC mass), and reticulocyte count ineople without known diabetes over an extended periodf time. All parameters were extraordinarily stable.3

his allowed us to use these data to determine theorrelation between the %tGHb and fasting glucoseoncentration over a period of time greater than theurnover time of the RBC mass. The contribution of theransient postprandial glucose elevations in people withr without mild diabetes to the 24-hour integrated glu-ose concentration is only �3–9% over that of thenitial fasting concentration as was shown by us andthers.4,14–16 It is similar to the relationship betweenhe fasting values and the 7-point glucose profiles de-ermined in the DCCT.2 Thus, the fasting glucose cane adjusted upward modestly to obtain an approximate4-hour integrated glucose concentration. In this pub-ication, we increased the fasting value by 9%. If 3% issed, it makes little difference in the slope of theelationship, but it does affect the intercept modestly.

Knowledge regarding the stability of the %tGHb ineople without diabetes also allowed us to use a single4-hour integrated glucose concentration as a surrogatef mean glucose concentration over the range of esti-ated RBC survivals and to determine the relationship

etween it and the %tGHb in young students. Theseubjects were likely to have low values for both anntegrated glucose concentration and %tGHb17 and ishe reason for inclusion in the study.

In 29 subjects with untreated type 2 diabetes, the diet

wkkLtticvsooe(�0

cg%dgg

btcavtsuccra

Rmat%sRb

crma

itc

tp%ldio

ocRrtmccs

teasbctp

taAfgfdOtHcct

dtjrirct

sci

J Lab Clin Med24 Hassan et al January 2006

as known, with all food supplied by our metabolicitchen over a 5-week period of time. The %tGHb wasnown to be stable over this same 5 weeks or longer.5,6

astly, in a subgroup of subjects with type 2 diabetesreated with oral agents, the interstitial glucose concen-ration determined continuously over 3 days at 20-dayntervals for 100 days was determined directly andompared with the individual %tGHb values. Thus, aariety of subjects both with and without diabetes wastudied. Interestingly, the data obtained with continu-us glucose monitoring and those obtained in the groupn a defined diet yielded a line of identity that wasssentially identical, indicating the accuracy of the datadata not shown). Defined diet slope � 0.26, intercept

3.4, and R2 � 0.87. Continuous monitoring slope �.25, intercept � 3.1, and R2 � 0.82.When all of the data in this study are pooled, a

urvilinear relationship between the 24-hour integratedlucose values (determined or calculated) and thetGHb was observed (Fig 1, B). Others also have

etermined the relationship to be curvilinear in a smallroup of subjects using ion-exchange methodolo-ies18,19 and an isoelectric focusing technique.11

If it is accepted that there is a curvilinear relationshipetween the integrated 24-hour plasma glucose concen-rations and %tGHb, the reason remains unknown. Re-ently we have reported that an increase in %tGHb isssociated with a progressive shortening of RBC sur-ival in people with type 2 diabetes.20 This suggestedhat the mean blood glucose concentration affects RBCurvival. It also could result in the %tGHb (or %HbA1c)nderestimating the actual average blood glucose con-entration particularly at chronically high glucose con-entrations.20–22 This then could produce a curvilinearelationship between the mean glucose concentrationnd the %tGHb or %HbA1c.To better define the relationship between %GHb andBC survival, we obtained reticulocyte data from nor-al subjects with a spectrum of glucose concentrations

nd %tGHbs, as well as in subjects with type 2 diabe-es. In this study, there was no correlation between the

tGHb and %reticulocyte count (R2 � 0.065, data nothown). Therefore, it is unlikely that an alteration inBC survival contributed to the nonlinear relationshipetween the %tGHb and mean glucose concentration.The extrapolated curve in this study crossed the x-

oordinate at �18-mg/dL glucose concentration; ie, theesidual plasma glucose concentration would be �18g/dL at a 0 %tGHb (Fig 1, B). This suggested an

dditional factor was present that affected the results.In overnight fasted, non-diabetic people, the RBC

ntracellular glucose concentration has been determinedo be �15–30 mg/dL lower than the plasma glucose

oncentration.23 Thus, the discrepancy between the ex- (

rapolated plasma glucose of 18 mg/dL and an antici-ated plasma glucose concentration of 0 when thetGHb was 0 could be explained by a lower intracel-

ular glucose concentration than that in the plasma. Thisifference is in the range expected with the reportedntracellular glucose concentrations23,24 in the vicinityf the globin molecules.The intracellular versus extracellular concentration

f glucose depends on the rate of glucose utilizationompared with the rate of glucose transport across theBC membrane. Presumably, the glucose utilization

ate will be the same at all plasma glucose concentra-ions. However, the transmembrane glucose transportust increase as the plasma glucose concentration in-

reases to maintain osmotic equilibrium. Thus, the con-entration gradient across the membrane should be theame at different plasma water concentrations.

This may not be entirely true. The membrane facul-ative bidirectional glucose transporter (Glut 1) report-dly has a low Km (�1.5 mM) for glucose.25 Thus, atglucose concentration of �8–10 mM, it is essentially

aturated. The glucose flux rate above this level shoulde unchanged. Also, a chronically elevated plasma glu-ose concentration could result in structural changeshat affect the rate of transmembrane glucose trans-ort.26

In addition, the curvilinear relationship could be dueo the rate of formation of the ketoamine adduct having

zero-order (saturation) type of reaction kinetics.19

lternatively, it could be due to an ordering in affinityor glucose binding at the various potential sites on thelobin molecules where glucose can attach and thusorm a ketoamine product.27 It also could be due to aifference in ketoamine formation or removal rate.11,26

riginally the ketoamine derivitization of globin pro-eins was considered to be essentially irreversible.26

owever, recent data indicate that this may not be thease. Thus, differences in the ketoamine off rate alsoould contribute.11,28 It also may be due to a combina-ion of these potential mechanisms.

In addition, the ketoamine formation rate may beifferent in people with and without diabetes, such thathe glucose to %tGHb relationship is different in sub-ects with and without known diabetes; ie, a two-phaseelationship is present, as described. However, the datan the group without diabetes is too limited in the lowerange of %tGHb and the variance in mean serum glu-ose too small for a reliable equation to be generated inhat group.

In any regard, the clinical consequence of the ob-erved curvilinear relationship between integrated glu-ose concentration and %tGHb (and probably %HbA1c)s that there is a considerable loss in sensitivity of tGHb

or %HbA1c) as an index of integrated glucose concen-

tr

tmfcp

lpcmstcdlbhucaaatmh

itSea

ct

ttMA

N

R

1

1

1

1

1

Fgc

J Lab Clin MedVolume 147, Number 1 Hassan et al 25

rations at higher average glucose concentrations but aeduction in variance.

For example, an increase in mean glucose concentra-ion from 100 to 125 mg/dL would be represented by aean increase of 1.2 %tGHb. Whereas, an increase

rom 275 to 300 mg/dL would represent only an in-rease of 0.4% tGHb (Fig 2). However, it should beointed out that the variance is large.Based on the current data, it is interesting to specu-

ate that the curvilinear relationship of %tGHb (andresumably, %HbA1c) to mean glucose concentrationould explain the curvilinear relationship between theean %HbA1c and the rate of development, or progres-

ion, of retinopathy noted in the DCCT29; ie, the rela-ionship between the 24-hour integrated blood glucoseoncentration over an extended period of time and theevelopment or progression of retinopathy could beinearly related. If we assume that the relationshipetween %tGHb and glucose is curvilinear, and that atigh glucose concentrations, the %tGHb progressivelynderestimates the glucose concentration, then oneould speculate that the exponential increase in retinop-thy in relation to GHb noted in the DCCT is actuallylinear increase in retinopathy when compared with

ctual mean glucose concentrations in vivo. The rela-ionship only seems curvilinear because %tGHb is noteasuring the true mean glucose concentration at

igher GHb values.In summary, several factors potentially can affect the

nterpretation of integrated blood glucose concentra-ions in the context of either %tGHb or %HbA1c values.ome of these have been addressed in this study. How-ver, the variables need to be more fully characterized

ig 2. Effect of an increase of 25-mg/dL integrated mean 24-hourlucose concentration on %tGHb at a low vs. a high glucose con-entration.

nd explained if %tGHb or %HbA1c is to be used

orrectly as an index of integrated glucose concentra-ions over a prolonged period of time.

The authors would like to thank the volunteers for participating inhe studies, Ms. Heidi Hoover, RD, MS, the staffs of the SDTU andhe Clinical Chemistry Laboratory for excellent technical support, Dr.

ichael Kuskowski for advice on statistical analysis of the data, andnn Emery for clerical support.

Dr. Hassan was a Fellow in Endocrinology and Drs. Johnson andader were fourth year medical students at the time of this study.

EFERENCES

1. American Diabetes Association. Clinical Practice Recommenda-tions. Nutrition principles and recommendations in diabetes DiabCare 2004;27(Suppl 1):S36–46.

2. Rohlfing CL, Wiedmeyer HM, Little RR, England JD, Tennill A,Goldstein DE. Defining the relationship between plasma glucoseand HbA(1c): analysis of glucose profiles and HbA(1c) in theDiabetes Control and Complications Trial. Diab Care 2002;25:275–8.

3. Nuttall FQ, Gannon MC, Swaim W, Adams MJ. Stability overtime of glycohemoglobin, glucose and red blood cell survival inhematologically stable people without diabetes. Metabolism2004;53:1399–1404.

4. Nuttall FQ, Gannon MC, Wald JL, Ahmed M. Plasma glucoseand insulin profiles in normal subjects ingesting diets of varyingcarbohydrate, fat and protein content. J Am Coll Nutr 1985;4:437–50.

5. Gannon MC, Nuttall FQ, Saeed A, Jordan K, Hoover H. Anincrease in dietary protein improves the blood glucose responsein persons with type 2 diabetes. Am J Clin Nutr 2003;78:734–41.

6. Gannon MC, Nuttall FQ. Effect of a high-protein, low-carbohy-drate diet on blood glucose control in people with type 2 diabe-tes. Diabetes 2004;53:2375–82.

7. Meigs JB, Nathan DM, Wilson PWF, Cupples A, Singer DE.Metabolic risk factors worsen continuously across the spectrumof non-diabetic glucose tolerance. The Framingham OffspringStudy. Ann Intern Med 1998;128:524–33.

8. Kilpatrick ES, Maylor PW, Keevil BG. Biological variation ofglycated hemoglobin. Implications for diabetes screening andmonitoring. Diab Care 1998;21:261–4.

9. Yudkin JS, Forrest RD, Jackson CA, Ryle AJ, Davie S, GouldBJ. Unexplained variability of glycated hemoglobin in non-diabetic subjects not related to glycemia. Diabetologia 1990;33:208–15.

0. Modan M, Meytes D, Rozeman P, Yosef SB, Sehayek E, YosefNB, et al. Significance of high HbA1 levels in normal glucosetolerance. Diab Care 1988;11:422–8.

1. Mortensen HB. Glycated hemoglobin. Reaction and biokineticstudies. Clinical application of hemoglobin A1c in the assess-ment of metabolic control in children with diabetes mellitus. DanMed Bull 1985;32:309–28.

2. Delpierre G, Rider MH, Collard F, Stroobant V, Vanstapel F,Santos H, et al. Identification, cloning and heterologous expres-sion of a mammalian fructosamine-3-kinase. Diabetes 2000;49:1627–34.

3. Szwergold BS, Howell S, Beisswenger PJ. Human fructosamine-3-kinase: purification, sequencing, substrate specificity, and ev-idence of activity in vivo. Diabetes 2001;50:2139–47.

4. Holman RR, Turner RC. The basal plasma glucose: a simplerelevant index of maturity-onset diabetes. Clin Endocrinol 1980;

14:279–86.

1

1

1

1

1

2

2

2

2

2

2

2

2

2

2

J Lab Clin Med26 Hassan et al January 2006

5. Paisey RB, Bradshaw P, Hartog M. Home blood glucose con-centrations in maturity onset diabetes. Br Med J 1980;280:596–8.

6. Hansen AP, Johansen K. Diurnal patterns of blood glucose,serum free fatty acids, insulin, glucagon and growth hormone innormals and juvenile diabetics. Diabetologia 1970;6:27–33.

7. Nuttall FQ. Effect of age on the percentage HbA1c and thepercentage of total glycohemoglobin in non-diabetic people.J Lab Clin Med 1999;134:451–3.

8. Svendsen PA, Lauritzen T, Soegaard U, Nerup J. Glycosylatedhaemoglobin and steady-state mean blood glucose concentration intype 1 (insulin-dependent) diabetes. Diabetologia 1982;23:403–5.

9. Graf RJ, Halter JB, Porte D. Glycosylated hemoglobin in normalsubjects and subjects with maturity-onset diabetes. Evidence fora saturable system in man. Diabetes 1978;27:834–9.

0. Virtue MA, Furne JK, Nuttall FQ, Levitt MD. Relationship betweenGHb concentration and erythrocyte survival determined from breathcarbon monoxide concentration. Diab Care 2004;27:931–5.

1. Comi RJ, Hamilton H. Reduction of red cell glucose transporter in-trinsic activity in diabetes running. Horm Metab Res 1994;26:26–32.

2. Peterson CM, Jones RL, Koenig RJ, Melvin ET, Lehrman ML.Reversible hematologic sequelae of diabetes mellitus. Ann Intern

Med 1977;86:425–9.

3. Somogyi M. The distribution of sugar in normal human blood.J Biol Chem 1928;78:117.

4. Higgins PJ, Garlick RL, Bunn HF. Glycosylated hemoglobin inhuman and animal red cells. Role of glucose permeability. Dia-betes 1982;31:743–8.

5. Nelson DL, Cox MM. Lehninger Principles of Biochemistry. 3rded. New York: W.H. Freeman; 2000: 413.

6. Bunn HF, McDonald MJ, Shapiro R. Non-enzymatic glycosyla-tion of hemoglobin and proteins of the erythrocyte membrane.New York: Elsevier North Holland; 1980:455–62.

7. Bunn HF, Shapiro R, McManus M, Garrick L, McDonald M,Gallop PM, et al. Structural heterogeneity of human hemoglobinA due to non-enzymatic glycosylation. J Biol Chem 1979;254:3892–8.

8. Delpierre G, Collard F, Fortpied J, Schaftingen V. Fructosamine3-kinase is involved in an intracellular deglycation pathway inhuman erythrocytes. Biochem J 2002;265:801–8.

9. DCCT, The Diabetes Control and Complications Trial ResearchGroup. The effect of intensive treatment of diabetes on thedevelopment and progression of long-term complications in in-sulin-dependent diabetes mellitus. N Engl J Med 1993;329:977–

86.