L E T T E R S 362 DIABETES CARE, VOLUME 22, NUMBER 2, FEBRUARY 1999 OBSERVATIONS Publication of Abstracts Presented at Diabetes Meetings P resentation of abstracts at scientific meetings is a valid and a popular way to disseminate research findings. However, the ultimate way to have research methodology and results vali- dated is publication in a peer-reviewed journal. No information is available in the diabetes literature regarding the ultimate fate of abstracts presented at national meetings. The aim of this study was to deter- mine the publication rates of abstracts pre- sented at diabetes meetings. Books of abstracts were obtained for three meetings: the 28th Annual Meeting of the European Association for the Study of Diabetes (EASD) in 1992, the 52nd Annual Meeting of the American Diabetes Associa- tion (ADA) in 1992, and the Annual Meet- ing of the Australian Diabetes Society (ADS) and the Australian Diabetes Educators Association in 1990. MEDLINE and Cumulative Index to Nursing and Allied Health Literature (CINAHL) searches were made of a random sample of the abstracts presented at each meeting to determine whether they were published in full. At the ADA meeting, 710 abstracts were presented; of the 71 abstracts included in the search, 38 (53%) were published in full. At the EASD meeting, 840 abstracts were presented; of the 84 abstracts included in the search, 41 (49%) were published in full. At the ADS meet- ing, 126 abstracts were presented; of the 31 included in the search, 8 (26%) were published in full. The publication rates of abstracts pre- sented at diabetes meeting of 26% (Aus- tralia), 49% (Europe), and 53% (Ameri- can) are consistent with reports in the lit- erature for other disciplines (4–9). CRAIG PAYNE, MPH From the School of Human Biosciences, LaTrobe Uni- versity, Bundoora, Australia. E-mail: school.human. [email protected]. Address correspondence to Craig Payne, School of Human Biosciences, LaTrobe University, Bundoora, Victoria 3083, Australia. References 1. Abstracts from the 28th Annual Meeting of the European Association for the Study of Diabetes (EASD), 1992. Diabetologia 35 (Suppl. 1), 1992 2. Abstracts from the 52nd Annual Meeting of the American Diabetes Association (ADA), 1992. Diabetes 41 (Suppl. 1), 1992 3. Proceedings of the Annual Meeting of the Australian Diabetes Society (ADS) and the Australian Diabetes Educators Association, Adelaide, Australia,1990 4. Scherer RW, Dickersin K, Langenberg P: Full publication of results originally pre- sented in abstracts. JAMA 272:158–162, 1994 5. Gavazza JB, Foulkes GD, Meals RA: Publi- cation pattern of papers presented at the American Society for Surgery of the Hand annual meeting. J Hand Surg [Am] 21:742–745, 1996 6. Agustsdottir A, Holcombe J, Wright P, Daf- fin P, Ogletree G: Publication of patient- related oncology nursing research. Oncol Nurs Forum22:827–830, 1995 7. Juzych MS, Shin DH, Coffey JB, Parrow KA, Tsai CS, Briggs KS: Pattern of publication of ophthalmic abstracts in peer-reviewed jour- nals. Ophthalmology 98:553–556, 1991 8. Goldman L, Loscalzo A: Fate of cardiology research originally printed in abstract form. N Engl J Med 303:255–259, 1980 9. McCormick MC: Publication of research presented at the pediatric meetings: change in selection. Am J Dis Child 139:122–126, 1985 Influence of Glycemic Control on Zinc Urinary Excretion in Patients With Type 1 Diabetes S everal studies have found alterations in zinc status of patients with dia- betes. Zinc and insulin interactions and the effects of diabetes on zinc metabo- lism have been particularly discussed (1). Zinc nutriture has been assessed by bio- chemical measurement in tissues and bio- logical fluids. Abnormalities have been associated with duration of the disease, metabolic control, and presence of chronic complications in both types of diabetes (2,3). Hyperzincuria has been shown to be the most consistent finding (4,5). In type 2 diabetes, low serum zinc concentration was found in association with hyperzincu- ria and decreased gut absorption (6). Sev- eral evidences have suggested that zinc deficiency affects the progress of the dis- ease and the development of microvascu- lar complications (1). In a study on the kinetics of zinc in diabetic children, an inverse correlation was detected between zinc clearance, which could be secondary to glomerular hyperfiltration, and growth rate (7). The underlying mechanisms for hyperzincuria in diabetes are obscure, and the importance of glycemic control on zinc excretion is not completely eluci- dated. The aim of this study was to inves- tigate the influence of the improvement in glycemic control on urinary zinc excretion of subjects with type 1 diabetes. Twenty-three subjects aged 9–16 years, with a mean type 1 diabetes duration of 5.3 ± 2.0 years, were studied at an educational camp for diabetic children. Informed con- sent was obtained, and the experimental design was approved by the local ethics committee. No patient suf fered from any disease other than type 1 diabetes, and none took medications other than insulin. Long-term complications were excluded (normal neurological examination, fundus- copy, and serum creatinine; albumin excre- tion rate 20 μg/min). The subjects partic- ipated in a program of exercises and ade- quate diet. In addition to their usual NPH insulin injections, subjects were supple- mented with regular insulin according to capillary glycemia measurements obtained with glucose oxidase strips. Anthropomet- ric measurements were taken on the first and last mornings. Metabolic control was assessed by comparison of each subject’s initial and final fasting glycemia, 24-h uri- nary glucose excretion, and insulin require- ment. The same urine samples were col- lected in metal-free glassware. On the first day, a fasting blood sample was obtained for zinc and glucose determinations. Serum and urine aliquots were stored frozen at 20°C for further determinations. Urinary glucose was assayed by a colorimetric method. Zinc was measured by atomic absorption spectrophotometry. Quality control was tested by Community Bureau of Reference bovine liver, reference material 185. The reference value of the certificate of analysis was 142.0 ± 3.0 μg Zn/g and the mean value found in our lab was 145.0 ± 1.2 μg Zn/g. Food consumption from five randomly selected days was frozen at 20°C and was assessed by duplicate-por- tion technique. Nutrient intake was calcu- lated by chemical analysis performed
Transcript
L E T T E R S
362 DIABETES CARE, VOLUME 22, NUMBER 2, FEBRUARY 1999
O B S E RVAT I O N S
Publication ofAbstracts Pre s e n t e dat Diabetes M e e t i n g s
Presentation of abstracts at scientificmeetings is a valid and a popular wayto disseminate re s e a rch fin d i n g s .
H o w e v e r, the ultimate way to havere s e a rch methodology and results vali-dated is publication in a peer- re v i e w e dj o u rnal. No information is available in thediabetes literature re g a rding the ultimatefate of abstracts presented at nationalm e e t i n g s .
The aim of this study was to deter-mine the publication rates of abstracts pre-sented at diabetes meetings.
Books of abstracts were obtained fort h ree meetings: the 28th Annual Meeting ofthe European Association for the Study ofDiabetes (EASD) in 1992, the 52nd AnnualMeeting of the American Diabetes Associa-tion (ADA) in 1992, and the Annual Meet-ing of the Australian Diabetes Society (ADS)and the Australian Diabetes EducatorsAssociation in 1990. MEDLINE andCumulative Index to Nursing and AlliedHealth Literature (CINAHL) searches weremade of a random sample of the abstractsp resented at each meeting to determ i n ewhether they were published in full.
At the ADA meeting, 710 abstractsw e re presented; of the 71 abstractsincluded in the search, 38 (53%) werepublished in full. At the EASD meeting,840 abstracts were presented; of the 84abstracts included in the search, 41 (49%)w e re published in full. At the ADS meet-ing, 126 abstracts were presented; of the31 included in the search, 8 (26%) werepublished in full.
The publication rates of abstracts pre-sented at diabetes meeting of 26% (Aus-tralia), 49% (Europe), and 53% (Ameri-can) are consistent with re p o rts in the lit-e r a t u re for other disciplines (4–9).
CR A I G PAY N E, M P H
F rom the School of Human Biosciences, LaTrobe Uni-v e r s i t y, Bundoora, Australia. E-mail: school.human.b i o s c i e n c e s @ l a t ro b e . e d u . a u .
A d d ress correspondence to Craig Payne, Schoolof Human Biosciences, LaTrobe University, B u n d o o r a ,Victoria 3083, Australia.
R e f e re n c e s1 . Abstracts from the 28th Annual Meeting of
the European Association for the Study ofDiabetes (EASD), 1992. Diabetologia 3 5(Suppl. 1), 1992
2 . Abstracts from the 52nd Annual Meetingof the American Diabetes Association(ADA), 1992. Diabetes 41 (Suppl. 1), 1992
3 . P roceedings of the Annual Meeting of theAustralian Diabetes Society (ADS) and theAustralian Diabetes Educators Association,Adelaide, Australia,1990
4 . S c h e rer RW, Dickersin K, Langenberg P:Full publication of results originally pre-sented in abstracts. J A M A 2 7 2 : 1 5 8 – 1 6 2 ,1 9 9 4
5 . Gavazza JB, Foulkes GD, Meals RA: Publi-cation pattern of papers presented at theAmerican Society for Surg e ry of the Handannual meeting. J Hand Surg [Am]21:742–745, 1996
6 . Agustsdottir A, Holcombe J, Wright P, Daf-fin P, Ogletree G: Publication of patient-related oncology nursing re s e a rch. O n c o lNurs Foru m22:827–830, 1995
7 . Juzych MS, Shin DH, Coffey JB, Parrow KA,Tsai CS, Briggs KS: Pattern of publication ofophthalmic abstracts in peer- reviewed jour-nals. O p h t h a l m o l o g y98:553–556, 1991
8 . Goldman L, Loscalzo A: Fate of card i o l o g yre s e a rch originally printed in abstractf o rm. N Engl J Med303:255–259, 1980
9 . M c C o rmick MC: Publication of re s e a rc hp resented at the pediatric meetings: changein selection. Am J Dis Child1 3 9 : 1 2 2 – 1 2 6 ,1 9 8 5
I n fluence of GlycemicC o n t rol on Zinc U r i n a ry Excretion inPatients With Type 1D i a b e t e s
Several studies have found alterationsin zinc status of patients with dia-betes. Zinc and insulin interactions
and the effects of diabetes on zinc metabo-lism have been particularly discussed (1).Zinc nutriture has been assessed by bio-chemical measurement in tissues and bio-logical fluids. Abnormalities have beenassociated with duration of the disease,metabolic control, and presence of chro n i ccomplications in both types of diabetes(2,3). Hyperzincuria has been shown to bethe most consistent finding (4,5). In type 2diabetes, low serum zinc concentrationwas found in association with hyperz i n c u-
ria and decreased gut absorption (6). Sev-eral evidences have suggested that zincd e ficiency affects the pro g ress of the dis-ease and the development of micro v a s c u-lar complications (1). In a study on thekinetics of zinc in diabetic children, aninverse correlation was detected betweenzinc clearance, which could be secondaryto glomerular hyperfiltration, and gro w t hrate (7). The underlying mechanisms forh y p e rzincuria in diabetes are obscure, andthe importance of glycemic control onzinc excretion is not completely eluci-dated. The aim of this study was to inves-tigate the influence of the improvement inglycemic control on urinary zinc excre t i o nof subjects with type 1 diabetes.
Tw e n t y - t h ree subjects aged 9–16 years,with a mean type 1 diabetes duration of 5.3± 2.0 years, were studied at an educationalcamp for diabetic children. Informed con-sent was obtained, and the experimentaldesign was approved by the local ethicscommittee. No patient suff e red from anydisease other than type 1 diabetes, andnone took medications other than insulin.L o n g - t e rm complications were excluded( n o rmal neurological examination, fundus-c o p y, and serum creatinine; albumin excre-tion rate 20 µg/min). The subjects part i c-ipated in a program of exercises and ade-quate diet. In addition to their usual NPHinsulin injections, subjects were supple-mented with regular insulin according toc a p i l l a ry glycemia measurements obtainedwith glucose oxidase strips. Anthro p o m e t-ric measurements were taken on the fir s tand last mornings. Metabolic control wasassessed by comparison of each subject’sinitial and final fasting glycemia, 24-h uri-n a ry glucose excretion, and insulin re q u i re-ment. The same urine samples were col-lected in metal-free glassware. On the fir s td a y, a fasting blood sample was obtained forzinc and glucose determinations. Seru mand urine aliquots were stored frozen at
20°C for further determinations. Urinaryglucose was assayed by a colorimetricmethod. Zinc was measured by atomicabsorption spectro p h o t o m e t ry. Qualityc o n t rol was tested by Community Bure a uof Reference bovine liver, re f e rence material185. The re f e rence value of the cert i ficate ofanalysis was 142.0 ± 3.0 µg Zn/g and themean value found in our lab was 145.0 ±1.2 µg Zn/g. Food consumption from fiv erandomly selected days was frozen at
20°C and was assessed by duplicate-por-tion technique. Nutrient intake was calcu-lated by chemical analysis perf o rm e d
DIABETES CARE, VOLUME 22, NUMBER 2, FEBRUARY 1999 363
Letters
a c c o rding to the Association of Official Ana-lytical Chemists (8). Statistical analysisincluded paired Student’s t test, Wi l c o x o n ’stest, and Pearson’s correlation coeff i c i e n t .
Mean glycated protein at entry was 6.3± 2.3%. BMI (19.0 ± 2.0 kg/m2) did notchange during the period. Total energ yintake was adequate (2,115 ± 230kcal/day), and the distribution of caloriesf rom proteins (17.3%), lipids (36.1%), andcarbohydrates (46.5%), and zinc intake(15.8 mg), were within the ranges re c o m-mended by the American Diabetes Associ-ation (9) and given in the RecommendedD i e t a ry Allowances (RDA) of the NationalAcademy of Sciences (10). A mean seru mzinc value of 105.5 ± 14.1 µg/dl wasfound. A negative correlation was detectedbetween serum zinc and duration of type 1diabetes (r = 0.45, P = 0.024) but notbetween serum zinc and the glycated pro-tein levels. The 24-h urine volume wass i g n i ficantly correlated with glucose andzinc excretions. Also, urinary glucose andzinc were shown to be correlated (r = 0.59,P 0.005). Initial urinary zinc loss washigh and decreased (from 954.0 ± 370.3 to667.2 ± 341.5 µg/24 h) at the end of thep rogram, reaching values close to the nor-mal range (300–600 µg/24 h). Such re d u c-tion was accompanied by improvement inglycemic control, which was expressed bya significant decrease in insulin re q u i re-ment (from 31.6 ± 10.1 to 18.7 ± 11.2U / d a y, P 0.001), fasting glycemia (166 ±84 to 128 ± 68 mg/dl, P 0.05), and atendency toward reduced urinary glucose( f rom 16.2 ± 18.0 to 13.2 ± 13.3 g/24 h).
The initial hyperzincuria found in ourstudy is in agreement with the literature(1,4,5). The improvement in metabolicc o n t rol was accompanied by a reduction inzinc excretion, and the positive corre l a t i o nwith urinary glucose suggested a role forglycemic control of zinc excretion. Wespeculate that hyperzincuria may be attrib-uted at least in part to glomerular hyperfil-tration, consequent to chronic hyperg l y-cemia. It would be reasonable to expect thatsuch a chronic exaggerated zinc excre t i o nmight lead to tissue depletion and lowerc i rculating levels. In our study, subjectsreceiving a diet with a normal zinc contenthad serum zinc within the normal range.H o w e v e r, variable results have been shownin diabetic subjects, ranging from high tolow levels (11). Although normal seru mzinc concentrations were found, the inversec o rrelation detected with the duration ofdiabetes may suggest a deterioration of zinc
n u t r i t u re as the disease pro g ressed. How-e v e r, in the course of the disease, persistentpoor glycemic control has been associatedwith glycation of zinc-binding pro t e i n s ,which could result in decreased affinity bythis element (2). Zinc is shown to be impor-tant not only for immune responses in dia-betic patients but also for reductions of theoxidative stress involved in the micro v a s c u-lar and neurological disease process (1,3).T h e re f o re, in addition to the benefits ofglycemic control per se in pre v e n t i n gc h ronic complications, the correction of thedisturbed zinc metabolism could help indelaying their occurrence. In conclusion,our data demonstrated that near norm a l i z a-tion of the exaggerated zinc excretion inunstable diabetes could be achieved by thei m p rovement of glycemic control, both ofwhich are desirable conditions for avoidingl o n g - t e rm diabetic complications.
LÚ C I A F.C. PE D R O S A, PHD
SA N D R A R.G. FE R R E I R A, PHD
PA U L O R. CE S A R I N I, M D
SI LV I A M . F. CO Z Z O L I N O, PHD
F rom the Department of Preventive Medicine( S . R . G . F.) and the Division of Endocrinology (P. R . C . ) ,Federal University of São Paulo; the Department ofFood and Experimental Nutrition (S.M.F.C.), Facultyof Pharmaceutical Sciences, University of São Paulo,São Paulo; and the Department of Public Health andNutrition (L.F. C . P.), Federal University of Rio Grandedo Norte, Natal, Brazil.
A d d ress correspondence to Dr. Sandra R.G. Fer-reira, Departamento de Medicina Preventiva, Univer-sidade Federal de São Paulo, UNIFESP/EPM, RuaBotucatu, 740, CEP 04023-900, São Paulo, SP, Brazil.E-mail: ferre i r a @ m e d p re v. e p m . b r.
R e f e re n c e s1 . Chausmer AB: Zinc, insulin and diabetes. J
Am Coll Nutr 17:109–115, 19982 . C a n field WK, Hambidge KM, Johnson LK:
Zinc nutriture in type I diabetes mellitus:relationship to growth measures and meta-bolic control. J Pediatr Gastro e n t e rol Nutr3:577–584, 1984
3 . F a u re P, Benhamou PY, Perard A, Halimi S,Roussel AM: Lipid peroxidation in insulin-dependent diabetic patients with earlyretina degeneration lesions: effects of anoral zinc supplementation. Eur J Clin Nutr49:282–288, 1995
4 . Cunningham JJ, Fu A, Mearkle PL, Bro w nRG: Hyperzincuria in individuals withinsulin-dependent diabetes mellitus: con-c u rrent zinc status and the effect of high-dose zinc supplementation. M e t a b o l i s m43:1558–1562, 1994
5 . Heise CC, King JC, Costa FM, Kitzmiller JL:H y p e rzincuria in IDDM women: re l a t i o n-ship to measures of glycemic control, re n a l
function, and tissue catabolism. D i a b e t e sC a re 11:780–786, 1988
6 . Kinlaw WB, Levine AS, Morley JE, SilvisSE, McClain CJ: Abnormal zinc metabo-lism in type II diabetes mellitus. Am J Med75:273–277, 1983
7 . Nakamura T, Higashi A, Nishiyama S,Fujimoto S, Matsuda I: Kinetics of zinc sta-tus in children with IDDM. Diabetes Care14:553–557, 1991
8 . Association of Official Analytical Chemists(AOAC): O fficial Methods of Analysis.Arlington, AOAC, 1984
9 . American Diabetes Association: Nutritionrecommendations and principles for peo-ple with diabetes mellitus. Diabetes Care21 (Suppl. 1):S32–S33, 1998
1 0 . National Academy of Sciences: R e c o m m e n d e dD i e t a ry Allowances, 10th ed. Wa s h i n g t o n ,DC, Washington Academy Press, 1989
1 1 . Walter RM, Uriu-Hare JY, Olin KL, OsterMH, Anawalt BD, Critchfield JW, Keen CL:C o p p e r, zinc, manganese, and magnesiumstatus and complications of diabetes melli-tus. Diabetes Care14:1050–1056, 1991
High Levels ofC i rc u l a t i n gP ro i n fla m m a t o ryCytokines andLeptin in Urban, butNot Rural, Indians
A potential explanation fori n c reased risk of diabetes andc o ro n a ry heart disease
The incidence of type 2 diabetes andcoronary heart disease (CHD) isi n c reasing in Indians, particularly in
urban areas and abroad. Its cause isu n c l e a r, but this “epidemic” has beenattributed to changes in diet and exerc i s eleading to central obesity, and to stre s s .South Asian subjects show many feature sof the insulin resistance syndrome, andinsulin resistance as well as its associatedrisk factors predicts the incidence of bothtype 2 diabetes and CHD.
Adipose tissue expresses the pro i n-fla m m a t o ry cytokine tumor necrosis fac-t o r- ( T N F - ), which has actions oninsulin signaling and lipid metabolism andhas been proposed as the link betweenobesity and insulin resistance. We haverecently shown production of a similarcytokine, interleukin-6 (IL-6), by adiposetissue in vivo (1). We have found stro n g
364 DIABETES CARE, VOLUME 22, NUMBER 2, FEBRUARY 1999
Letters
c o rrelations of circulating concentrationsof these cytokines with features of theinsulin resistance syndrome in 107E u ropid subjects, as well as with markersof endothelial damage (2), which plays amajor role in athero t h rombosis (3). Wepostulated that elevated levels of thesecytokines may exist in urban Indians.
We re c ruited 111 subjects from thre epopulations: villages near Pune (n = 43),urban slums in Pune (n = 28), and anurban middle-class residential area inPune (n = 40). Subjects with interc u rre n tillness or clinically evident infections wereexcluded. Anthropometric and bloodp re s s u re measurements (Dinamapp; Criti-con, Tampa, FL) were re c o rded, and fast-ing blood was collected to measure con-centrations of glucose, lipids, TNF- , IL-6(R&D Systems, Minneapolis, MN), andleptin (Linco, St. Charles, MO).
In the rural subjects, levels of TNF-and IL-6 were comparable to those foundin an urban Europid population (2). Inboth urban groups, concentrations weres i g n i ficantly elevated (Table 1); TNF- c o n-centrations were similar between theg roups, but IL-6 concentrations werehigher in the subjects from the slums. Lep-tin concentrations were significantly re l a t e dto BMI (r = 0.59, P 0.001) and waist-to-hip ratio (WHR) (r = 0.18, P = 0.05), andw e re higher in women than in men (P0.001), but remained higher in urban Indi-ans after controlling for confounders. Con-centrations of TNF- and IL-6 were alsorelated to BMI (TNF- : r = 0.36, P0.001; IL-6: r = 0.22, P = 0.025) but werenot related to WHR. Levels of both weres i g n i ficantly higher in each urban gro u pwhen data were controlled for age, sex, anda d i p o s i t y, as evidenced by BMI, WHR, andleptin concentrations (TNF- : P 0 . 0 0 1for subjects from the slums, P 0.001 formiddle-class subjects; IL-6: P 0.001 forsubjects from the slums, P = 0.003 for mid-
dle-class subjects). There was no re l a t i o n-ship between the concentrations of any ofthe three molecules and the levels of bloodp re s s u re, glucose, or lipids.
The emergence of C-reactive pro t e i nas a cardiovascular risk factor has led tothe recognition of the potential contribu-tion to athero t h rombosis from infla m m a-t o ry cytokines. In addition, TNF- h a sp o w e rful metabolic effects, many of whicha re shared by IL-6, and may contribute toinsulin resistance and dyslipidemia (4).
The origin of these cytokines re m a i n sspeculative. Adipose tissue expresses allt h ree molecules and is a major contributorto circulating IL-6 (1). Despite anthro p o-metric adjustment for obesity, and for totalfat mass using leptin concentrations (5), itis possible that regional diff e rences in adi-pose tissue distribution or cytokinee x p ression explain our findings. UrbanIndians are more vulnerable to re s p i r a t o ryand gastrointestinal infections, includingHelicobacter pylori infections, because ofpoor sanitation and crowding, especiallyin slum areas. Atmospheric pollutioninduces cytokine expression in alveolarm a c rophages and may contribute to circ u-lating cytokine levels. Psychological stre s scan also elevate concentrations of IL-6.D e t e rmination of whether these mecha-nisms underlie the vulnerability of urbanand migrant Indians to CHD and type 2diabetes will re q u i re prospective studies.
JOHN S. YUDKIN, MD, FRCPCHITTARANJAN S. YAJNIK, MD
VIDYA MOHAMED-ALI, PHDKAREN BULMER, MSC
F rom the Centre for Diabetes and Card i o v a s c u l a rRisk (J.S.Y., V.M.-A., K.B.), Department of Medicine,University College London Medical School, Whit-tington Hospital, London, U.K.; and the DiabetesR e s e a rch Unit (C.S.Y.), King Edward Memorial Hos-pital, Pune, India.
A d d ress correspondence to Professor John S.Yudkin, Centre for Diabetes and Card i o v a s c u l a r
Risk, Department of Medicine, University CollegeLondon Medical School, G Block, Archway Wi n g ,Whittington Hospital, Archway Road, London N193UA, U.K. E-mail: [email protected].
A c k n o w l e d g m e n t s — The Diabetes Researc hC e n t re of the King Edward Memorial Hospitalin Pune is supported by grants from the We l l-come Trust, and the Centre for Diabetes andC a rdiovascular Risk of the University CollegeLondon Medical School is supported by aG roup Support Grant from the British DiabeticAssociation. K.B. is supported by a grant(PG/97150) from the British Heart Foundation.
R e f e re n c e s1 . Mohamed-Ali V, Goodrick S, Rawesh A,
Miles JM, Katz DR, Yudkin JS, CoppackS W: Subcutaneous adipose tissue re l e a s e sinterleukin-6, but not tumor necrosis fac-t o r- in vivo. J Clin Endocrinol Metab82:4196–4200, 1997
2 . Yudkin JS, Stehouwer CDA, Emeis JJ, Cop-pack SW: C-reactive protein in healthy sub-jects—associations with obesity, insulinresistance and endothelial dysfunction: apotential role for cytokines originating fro madipose tissue? A rterioscler Thromb Vasc Biol.In pre s s
3 . Vallance P, Collier J, Bhagat K: Infection,i n flammation, and infarction: does acuteendothelial dysfunction provide a link?L a n c e t349:1391–1392, 1997
4 . Hotamisligil GS, Spiegelman BM: Tu m o rn e c rosis factor : a key component of the obesity-diabetes link. D i a b e t e s4 3 : 1 2 7 1 – 1278, 1994
5 . Dua A, Hennes MI, Hoffmann RG, MaasDL, Krakower GR, Sonnenberg GE, Kisse-bah AH: Leptin: a significant indicator oftotal body fat but not of visceral fat andinsulin insensitivity in African-Americanwomen. D i a b e t e s45:1635–1637, 1996
E n t e roviral RNAand IgM Antibodiesin Early Pre g n a n c yand Risk for C h i l d h o o d - O n s e tIDDM in O ff s p r i n g
The prevalence of diabetes is very highin individuals with the rubella embry-opathy syndrome, and recent studies
have shown increased levels of entero v i r a l ,and specifically coxsackie IgM, antibodiesduring pregnancy among mothers whosec h i l d ren later developed diabetes compare d
Table 1—Levels of IL-6, TNF- , leptin, and obesity in three Indian populations
Data are medians (interq u a rtile range). *P 0.001, †P 0.05 vs. rural subjects when data were contro l l e dfor age and sex; ‡P 0.001 vs. rural subjects when data were controlled for age, sex, BMI, and WHR.
DIABETES CARE, VOLUME 22, NUMBER 2, FEBRUARY 1999 365
Letters
with that found in control subjects (1–3).We have now analyzed the prevalence ofe n t e roviral RNA and IgM antibodies in thefirst trimester of control mothers and ofmothers whose children became diabetic.We also analyzed antibodies to glutamicacid decarboxylase (GAD), a -cell antigenthat has been shown to share an amino acidsequence with one of a coxsackie virus pro-tein (4). Serum samples collected ro u t i n e l yduring the first trimester of pregnancies in85 mothers whose children developed dia-betes before the age of 15 as re c o rded in theSwedish Childhood diabetes register (5)w e re compared with sera from 172 moth-ers whose children had not developed dia-betes. Control subjects were selected at random in strata by county and year ofb i rth. The distribution of month of birt hwas similar for both groups of subjects.E n t e ro v i rus RNA was analyzed using anested polymerase chain reaction (PCR)(6). Entero v i rus IgA, IgM, and IgG titersw e re analyzed using conventional enzyme-linked immunosorbent assay (7) and GADantibodies by radioimmunoassay (8).
T h ree mothers of diabetic patients andone mother of control subjects were RNApositive. In addition, mothers of thre epatients had IgM antibodies to coxsackie Bv i rus, but no control mother was positive.Thus 6 of 85 mothers of diabetic childre nhad signs of an ongoing or re c e n te n t e ro v i rus infection in early pre g n a n c yc o m p a red with 1 of 172 control mothers,equal to an odds ratio of 12.9 (95% CI2.43–69.52). IgG and IgA antibody con-centrations did not differ between theg roups. The distribution of birth monthswas similar among the two groups. Themean age at onset of diabetes was 7.01 ±3.5 years among all patients, and the age atonset of patients with possible intrauterineinfection was not significantly diff e re n t
(6.5 ± 3.0 years). The mean GAD indexwas higher among mothers of diabeticpatients compared with mothers of con-t rol subjects (P = 0.03), but when settingthe cutoff for GAD positivity at 99% ofc o n t rol subjects, there was no diff e rence inpositivity rate between mothers of diabeticpatients and control subjects. None of theP C R mothers were GAD ( Table 1).
This latter finding does not rule out thepossibility of antigen mimicry, but there areother possible mechanisms to explain anassociation between fetal virus infectionand later development of type 1 diabetes.One could be that the perinatal infectionleads to a persistent low-grade infectionthat might slowly hit the -cell dire c t l yand/or lead to peri-insular infection. In thelatter case, the resulting infla m m a t i o ncould then act on the -cell according tothe so-called “innocent bystander” theory.
To our knowledge, for the first time,e n t e ro v i rus RNA has been detected early inp regnancy in mothers of children who laterbecome diabetic in a higher frequency thanthat found in mothers of control subjects.The numbers are low and there f o re needc o n firmation, preferably by large pro s p e c-tive cohort studies, where sera could beanalyzed immediately to ensure that RNA isnot destroyed during long-term storage.Taken together with the higher frequency ofcoxsackie IgM positivity re p o rted in thisstudy and others (2,3), and with the pre v i-ous demonstration of fetal rubella infectionas a strong risk factor for type 1 diabetes,evidence is accumulating that fetal viralinfections might be etiologically associatedwith childhood type 1 diabetes, intro d u c i n gnew possibilities for primary pre v e n t i o n .
GISELA G. DAHLQUIST, MD, PHDJENS E. BOMAN, MDPER JUTO, MD, PHD
F rom the Departments of Pediatrics (G.G.D.) andClinical Vi rology (J.E.B., P.J.), Umeå University Hos-pital, Umeå, Sweden.
A d d ress correspondence to Gisela Dahlquist,D e p a rtment of Pediatrics, Umeå University Hospital, S-901 85 Umeå Sweden. E-mail: gisela.dahlquist@p e d i a t r i . u m u . s e .
A c k n o w l e d g m e n t s — This study was sup-p o rted by grants from the Swedish MedicalR e s e a rch Council (project no. 07351), theSwedish Diabetes Association, and the Countyof Västerbotten.
We sincerely thank Irene Eriksson for skilledtechnical assistance.
R e f e re n c e s1 . Dahlquist GG, Ivarsson S, Lindberg B, Fors-
g ren M: Maternal enteroviral infection dur-ing pregnancy as a risk determinant forchildhood IDDM: a population-based case-c o n t rol study. D i a b e t e s44:408–413, 1995
2 . Dahlquist G, Frisk G, Ivarsson SA, SvanbergL, Forsgren M, Diderholm H: Indicationsthat maternal coxsackie B virus infectionduring pregnancy is a risk factor for child-hood onset IDDM. D i a b e t o l o g i a1 3 7 1 – 1 3 7 3 ,1 9 9 5
3 . Hyöty H, Hiltunen M, Knip M, LaakkonenM, Vähäsalo P, Karjalainen J, Koskela P,Roivninen M, Leiniki P, Hovi T, ÅkerblomH, and the Childhood Diabetes in Finland(DiMe) Study Group: A prospective study ofthe role of coxsackie B and other entero v i ru sinfections in the pathogenesis of IDDM.D i a b e t e s44:652–657, 1995
4 . Kaufman DL, Erlander MG, Clare-Saltzer M,Atkinson MA, MacLaren NK, Tobin AJ:Autoimmunity to two forms of glutamatedecarboxylase in insulin-dependent dia-betes mellitus. J Clin Invest 8 9 : 2 8 3 – 2 9 2 ,1 9 9 2
5 . Dahlquist G, Mustonen L: Childhood onsetdiabetes: time trends and climatological fac-tors. Int J Epidemiol23:1239–1241, 1994
6 . Puig M, Jofre J, Lucena F, Alland A, Wa d e l lG, Girones R: Detection of adenoviruses ande n t e ro v i ruses in polluted waters by nestedPCR amplification. Appl Environm Micro b i o l60:2963–2970, 1994
7 . Boman J, Nilsson B, Juto P: Serum IgA, IgGand IgM responses to diff e rent entero v i ru s e sas measured by a coxsackie B5-based indi-rect ELISA. J Med Vi ro l38:32–35, 1992
8 . F a l o rni A, Örtqvist E, Persson B, Lern m a r kÅ: Radioimmunoassays for glutamic aciddecarboxylase (GAD65) and GAD65autoantibodies using 35S or 3H re c o m b i-nant human ligands. I m m unol Meth186:89–91, 1995
Table 1—Mothers with enterovirus RNA or IgM antibodies: case/control status, GAD antibodies index, and age at onset of proband
Case/control Entero IgM Age at onset ofstatus PCR status antibody titer GAD index IDDM (years)
Cutoff for the entero IgM antibody titer is 0.05 absorbance units. For the GAD index, the 99th centile ofcontrol subjects = 0.144.
366 DIABETES CARE, VOLUME 22, NUMBER 2, FEBRUARY 1999
Letters
American DiabetesAssociation Criteriafor DiabetesD i a g n o s i s
Another perspective
We read with interest article by Har-ris et al. (1) on diagnosis of dia-betes by 1997 American Diabetes
Association (ADA) criteria (2) compare dwith 1980–1985 World Health Org a n i z a-tion (WHO) diagnostic criteria (3,4).
We agree with the finding, given in thea rticle (1), that the diagnosis of diabetesand impaired fasting glucose (IFG) shouldbe determined by the ADA criterion of fast-ing plasma glucose (FPG) measurement inthe U.S. population (2). The loss of sensi-tivity is only 2%, and more cases of dia-betes could be diagnosed through bettercompliance to evaluation by FPG.
We studied FPG and 2-h post–glucoseload plasma glucose (2-h PG) values in400 apparently healthy subjects and 75patients with angiographically pro v e nc o ro n a ry art e ry disease (CAD) from thea rmed forces (serving and re t i red) whow e re 35–75 years of age.
We observed that diabetes was diag-nosed in 0.5 (n = 2) and 8% (n = 6) ofa p p a rently healthy and CAD patients,re s p e c t i v e l y, by ADA criteria. Diabetes wasdiagnosed in 2.25 (n = 9) and 21.3% (n =16) of apparently healthy individuals andCAD patients, re s p e c t i v e l y, by WHO crite-ria (FPG and 2-h PG). However, based onWHO FPG criteria alone (FPG 1 4 0mg/dl), only 1 of 9 and 6 of 16 diabeticpatients were identified in the healthy andCAD groups, re s p e c t i v e l y.
T h e re is a significant diff e rence in thesensitivity of diagnosing IGT/IFG byWHO (2-h PG) and ADA (FPG) criteria.The incidence of IFG by the ADA criterionis 1.7 (n= 6) and 8% (n = 6) among appar-ently healthy subjects and CAD patients,re s p e c t i v e l y, while by WHO criteria, IGTwas detected in 7.5 (n = 30) and 17.3% (n= 13) of apparently healthy subjects andCAD patients, re s p e c t i v e l y.
We also found that all of the individu-als in the healthy and CAD groups couldbe correctly classified according to WHOcriteria into normal GTT, IGT, and dia-betic, based on 2-h PG value alone.
In the Indian group studied, abnorm a lglucose tolerance (diabetes and IGT/IFG)
by WHO criteria is 9.75% for appare n t l yhealthy individuals and 38.6% for thosein the high-risk CAD group, while by ADAcriteria it is only 2.2 and 16% in thehealthy and CAD groups, re s p e c t i v e l y.
T h e re f o re, 7.5% of the appare n t l yhealthy individuals and 22.6% of those inthe CAD group have been diagnosed withn o rmal glucose tolerance test in the pre s-ence of either IGT or diabetes. This may bedue to the fact that South Asians and Indi-ans are re p o rted to have a higher incidenceof insulin resistance and hyperinsulinemia.The impaired release of insulin in re s p o n s eto glucose load and insulin re s i s t a n c edevelops early in the course of type 2 dia-betes, and even a small decrease in insulins e c re t o ry re s e rve or action impairs abilityto handle a glucose load. There f o re, 2-h PG(WHO) is probably a more sensitive indexthan the FPG (ADA) criterion for the pre s-ence of disease (diabetes and IGT).
We recommend that for the diagnosisof diabetes or IGT, either WHO re c o m m e n-dations (FPG and 2-h PG) be followed, or ifa single value is considered for diagnosis, itshould be the 2-h PG value rather than theFPG value, as recommended by the ADA,in ethnic groups known to have high pre v a-lence of insulin resistance and hyperinsu-linemia. Following 2-h PG, the number ofindividuals known to have abnormal car-bohydrate metabolism would not re m a i nundiagnosed and would not pro g ress toh y p e rglycemic complications, such asi n c reased incidence of CAD. Of CADpatients, 24% have normal FPG, accord i n gto the ADA criterion (FPG 110 mg/dl),but an abnormal 2-h PG value of 140 and
199 mg/dl in 16% (n = 12) and 2 0 0mg/dl in 8% (n = 6). We are also studyingthe incidence of abnormal 2-h PG valuesand normal FPG in other patients in whomh y p e rglycemia is an associated risk factor,such as those who develop cataracts at anearly age.
T h e re f o re, we emphasize that the ADArecommendation for using FPG to diagnosisIFG and diabetes, based on U.S. populationstudies, should be further studied in com-parison to 2-h PG in ethnic groups such asIndians, who are known to have a highp revalence of hyperinsulinemia and insulinresistance, so that no individual who isp resently classified to have an abnormal glu-cose tolerance according to WHO criteria isc l a s s i fied as normal. Detection of individu-als with such an abnormal carbohydratemetabolism may delay or prevent complica-tions in these individuals, and at the same
time would not impart a false sense of secu-rity associated with normal FPG in the pre s-ence of an abnormal 2-h PG value.
BABU LAL SOMANI, PCB (ITALY)SIDDHANA S. BANGAR, MD
RAJVIR BHALWAR, MD
F rom the Departments of Biochemistry (B.L.S.,S.S.B.) and Preventive and Social Medicine (R.B.),A rmed Forces Medical College, Pune, India.
A d d ress correspondence to B.L. Somani, PCB(Italy), Scientist-F, Department of Biochemistry,A rmed Forces Medical College, Pune-411 040,India. E-mail: [email protected] e t . i n .
R e f e re n c e s1 . H a rris MI, Eastman RC, Cowie CC, Flegal
KM, Eberh a rdt MS: Comparison of dia-betes diagnostic categories in the U.S. pop-ulation according to the 1997 AmericanDiabetes Association and 1980–1985World Health Organization diagnostic cri-teria. Diabetes Care20:1859–1862, 1997
2 . The Expert Committee on the Diagnosisand Classification of Diabetes Mellitus:R e p o rt of the Expert Committee on theDiagnosis and Classification of Diabetesmellitus. Diabetes Care 2 0 : 1 1 8 3 – 1 1 9 7 ,1 9 9 7
3 . World Health Organization: Second Reportof the WHO Expert Committee on DiabetesM e l l i t u s. Geneva, World Health Org., 1980( Tech. Rep. Ser., no. 646)
4 . World Health Organization: R e p o rt of theE x p e rt Committee on Diabetes Mellitus.Geneva, World Health Org., 1985 (Te c h .Rep. Ser., no. 727)
P n e u m a t o s i sC y s t o i d e sIntestinalis AfterTreatment With an- G l u c o s i d a s e
I n h i b i t o r
Side effects of -glucosidase inhibitor( -GI) include flatulence andabdominal distention resulting fro m
f e rmentation of unabsorbed carbohydratesby intestinal bacteria; it has rarely causedileus (1,2). We re p o rt here for the fir s ttime pneumatosis cystoides intestinalis(PCI) as a side effect of - G I .
A 64-year-old woman was admitted toour hospital in late June 1998 because ofabdominal distention. She denied consti-pation, nausea, and vomiting. For 20years, she had been receiving insulin ther-
DIABETES CARE, VOLUME 22, NUMBER 2, FEBRUARY 1999 367
Letters
apy for type 2 diabetes: 22 U Penfil 30Rs.c. daily at the time of admission. Exceptfor an antilipemic agent, she received noother drugs until April 1998, when theH b A1 c level was found to be 8.0%, and an
-GI, voglibose (0.6 mg/day), was pre-scribed in addition to the insulin at thebeginning of May. Excessive flatulence wasnoted 1 month later. By the middle ofJune, abdominal distention had appeare dand gradually pro g ressed. The patient wasc o n s i d e red to have moderate diabetic neu-ro p a t h y, without such symptoms asnumbness or orthostatic hypotention, andmild diabetic retinopathy; however, thesecomplications were stable. She had no dia-betic nephro p a t h y. The patient’s past surg i-cal history included a Cesarean section.She was 149 cm in height and 53 kg inweight. Bowel sounds were normal. Fast-ing plasma glucose concentration was 8.3mmol/l, and the HbA1 c level was 7.0%. Anabdominal radiograph showed distentionof the ascending and proximal transversecolon with cystic radiolucencies, indicat-ing intramural gas. Abdominal computedtomography indicated subserosal cystica reas of gas and distention of the involvedsegments. A barium enema study showedtranslucent areas of gas clustered along thed i s t o rted contours of the ascending andtransverse colon. A diagnosis of PCI wasmade, voglibose was discontinued, andfasting was imposed. Flatulence andabdominal distention then improved, andcystic gas-filled loculations disappeare df rom radiographs on the 4th day.
PCI is a condition characterized by thep resence of gas-filled cysts in the submu-cosa or subserosa of the gastro i n t e s t i n a ltract (3). Since diseases previously associ-ated with PCI were ruled out and PCIi m p roved when the -GI was discontin-ued, we had little doubt that in thisinstance PCI was related to the - G I .Although diabetes is rarely associated withPCI, we have previously written that wors-ening of diabetic neuropathy could be con-nected with PCI (4). Neuro p a t h y, however,was not worsening in this case. Becauselactulose administration or jejunoilealbypass surg e ry may result in PCI from bac-terial fermentation (5,6), -GI tre a t m e n tmay also be a cause of PCI. In fact, tre a t-ment with -GI re p o rtedly has causedileus in diabetic patients, particularly inthose with previous abdominal surg e ryand relatively advanced age (1,2), at leastin part from excessive gas formation. Sincethe symptoms of PCI and the side effects of
-GI both involve nonspecific gastro i n-testinal complaints, we should consider thepossibility of PCI when an -GI is given todiabetic patients, especially in the contextsof advancing age, past abdominal surg e ry,and possible autonomic neuro p a t h y.
TETSUO HAYAKAWA, MDMANABU YONESHIMA, MD
TOSHIO ABE, MDGAKUJI NOMURA, MD
F rom the Department of Internal Medicine,Kanazawa Municipal Hospital, Kanazawa, Japan.
A d d ress correspondence to Tetsuo Hayakawa,MD, Department of Internal Medicine, KanazawaMunicipal Hospital, 3-7-3 Heiwa-machi, Kanazawa,Ishikawa, 921-8105, Japan.
R e f e re n c e s1 . Nishi Y, Aizawa T, Hashizume K: Ileus: a
r a re side effect of acarbose (Letter). D i a -betes Care 19:1033, 1996
2 . Odawara M, Bannai C, Saitoh T,Kawakami Y, Yamashita K: Potentiallylethal ileus associated with acarbose tre a t-ment for NIDDM (Letter). Diabetes Care20:1210–1211, 1997
3 . Priest RJ, Goldstein F: Pneumatosis cys-toides intestinalis. In Bockus Gastro e n t e ro l-o g y. 4th ed. Berk JE, Ed. Philadelphia, WBSaunders, 1985, p. 2473–2483
4 . Hayakawa T, Nomura G: A case of non-insulin-dependent diabetes mellitus withpneumatosis cystoides intestinalis. J JpnDiabetes Soc 40:801–806, 1997
5 . Z i m m e rman AL, Gupta JK, Ingegno AP:Pneumatosis coli following treatment withlactulose. NY State J Med 7 9 : 1 8 9 6 – 1 8 9 9 ,1 9 7 9
6 . S i c a rd GA, Vaughan R, Wise L: Pneumato-sis cystoides intestinalis: an unusual com-plication of jejunoileal bypass. S u rg e ry79:480–484, 1976
C O M M E N T S A N DR E S P O N S E S
Do the NewS c reening and Diagnostic Criteria of DiabetesP roposed by theAmerican DiabetesAssociation ReallyM a t c h ?
The new criteria proposed by theAmerican Diabetes Association (ADA)for the diagnosis of diabetes (1) have
recently been applied to a large sample ofobese subjects by Mannucci et al. (2) andhave resulted in an increased prevalence ofboth diabetes and impaired fasting glucose(IFG). When these authors used fastingplasma glucose (FPG) on its own to esti-mate the prevalence of diabetes, their diag-nosis, when compared with the oral glu-cose tolerance test (OGTT) criteriaadvanced by the ADA, failed for 57% of thepopulation under study. There f o re, weplanned to verify the effective corre l a t i o nbetween FPG and 120-min glycemia afteradministration of a glucose oral load (75 g)for both the screening and the diagnosis ofdiabetes in a consecutive series of 414 out-patients (327 female and 87 male sub-jects), ranging from normal in weight toobese and attending the program forweight control at the Catholic University ofR o m e ’s section of metabolic disease. Theyhad no previous history of diabetes, andthey had the following characteristics: BMI,34.65 ± 8 kg/m2; age, 38.1 ± 14 years; cho-l e s t e rol, 185.4 ± 43.3 mg/dl; and triglyc-erides, 138 ± 67.9 mg/dl. They were testedfor FPG and 120-min glycemia after anoral glucose load (75 g). According to thenew ADA diagnostic criteria, using theOGTT as the diagnostic pro c e d u re, 261patients (63.7%) were found to be norm a l .The prevalence of diabetes was 8.5% (35patients), that of impaired glucose toler-ance (IGT) was 27.8% (115 patients), andthe diagnosis of IFG was established foronly 3 obese patients (0.7%). Conversely,when FPG was used alone for diabetess c reening, diabetes was diagnosed in onlyfour obese patients (1%). Because bothBMI and plasma glucose levels (basal val-ues and those measured 120 min after theoral glucose load) had normal distribu-tions, Pearson’s correlation was calculated.A positive significant correlation was foundbetween BMI and basal glycemia (R = 0.34;P 0.0001) and between BMI and 120-min glycemia (R = 0.21; P 0 . 0 0 0 1 ) .T h e re f o re, because Mannucci et al. (2) stud-ied only obese subjects, they overe s t i m a t e dthe prevalence of diabetes, IGT, and IFG.
In conclusion, using the 120-minOGTT screening pro c e d u re, the pre v a l e n c eof type 2 diabetes was similar to that cur-rently re f e rred by literature (10–15%),while FPG clearly underestimated thep revalence of diabetes mellitus. The ExpertCommittee on the Diagnosis and Classifi-
368 DIABETES CARE, VOLUME 22, NUMBER 2, FEBRUARY 1999
Letters
cation of Diabetes Mellitus of the ADA re c-ommends FPG as a screening pro c e d u re ,and the test of 120-min glycemia after glu-cose load as a diagnostic pro c e d u re (1), butthese criteria have resulted in diff e rent esti-mates of diabetes prevalence in both obeseand nonobese subjects.
DONATELLA GNIULI, MDGELTRUDE MINGRONE, MD, PHD
FELICE STROLLO, MDGIOVANNA STROLLO, MD
ALDO V. GRECO, MD
F rom the Section of Metabolic Disease (D.G., G.M.,A . V.G.), Department of Internal Medicine, CatholicUniversity of Rome, Rome; and Istituto NazionaleRiposo e Cura Anziani (I.N.R.C.A.) (F.S., G.S.),Rome, Italy.
A d d ress correspondence to Donatella Gniuli,MD, Istituto di Medicina Interna e Geriatria, Sectionof Metabolic Disease, Catholic University, Larg oGemelli 8, 00168 Rome, Italy. E-mail:a g re c o @ rm . u n i c a t t . i t .
R e f e re n c e s1 . The Expert Committee on the Diagnosis
and Classification of Diabetes Mellitus:R e p o rt of the Expert Committee on theDiagnosis and Classification of DiabetesMellitus. Diabetes Care 2 0 : 1 1 8 3 – 1 1 9 7 ,1 9 9 7
2 . Mannucci E, Bardini G, Ognibene A,Rotella CM: Screening for diabetes inobese patients using the new diagnosticcriteria (Letter). Diabetes Care 2 1 : 4 6 8 ,1 9 9 8
Response to Gniuliet al.
The principal aim of our study (1) wasto compare the estimates of diabetesp revalence among obese patients
based on either the World Health Org a n i z a-t i o n ’s (2) or the American Diabetes Associa-t i o n ’s (ADA) (3) criteria. The sample studiedwas composed of obese outpatients seekingt reatment for weight loss, and it was notintended to re p resent the general popula-tion. In fact, the prevalence of diabetes isknown to be significantly higher in obesepatients when compared with norm a l -weight subjects; this diff e rence is confirm e dby the positive correlation between BMI andplasma glucose described by Gniuli et al. intheir sample. There f o re, we think that thissample was adequate for the purpose of thes t u d y. In this re g a rd, it should be observ e dthat the sample selected by Gniuli et al.,
which was composed of subjects seekingweight loss and having a mean BMI of 34.6k g / m2, is equally unre p resentative of thegeneral population.
We re p o rted that the adoption of fastingplasma glucose (FPG) as the only scre e n i n gmethod for diabetes while maintaining 2-hpostload plasma glucose 200 mg/dlamong the diagnostic criteria, as suggestedby the ADA’s Expert Committee on theDiagnosis and the Classification of DiabetesMellitus (3), could lead to a re m a r k a b l eu n d e restimate of diabetes prevalence inobese patients (1). This result was con-firmed by Gniuli et al. in a diff e rent, moreh e t e rogeneous sample. They enrolled intheir study all patients seeking treatment forweight loss, irrespective of their actualweight; there f o re, their sample also includedn o rmal-weight and underweight individu-als, many of whom were probably aff e c t e dby eating disorders and could be scarc e l yre p resentative of normal-weight subjects inthe general population. Considering that themean BMI was well above 30 kg/m2, even ifthe number of normal-weight and under-weight patients is not specified, the vastmajority of the sample was probably com-posed of obese individuals. There f o re, theconclusions of Gniuli et al. about the inef-fectiveness of FPG in the screening of dia-betes can be applied to overweight subjects,but not to normal-weight subjects. Furt h e rstudies are needed to assess the sensitivityand specificity of FPG in the screening ofdiabetes in the general population.
EDOARDO MANNUCCI, MDGIANLUCA BARDINI, MD
AGOSTINO OGNIBENE, MDCARLO M. ROTELLA, MD
F rom the Section of Metabolic Diseases and Diabetol-ogy (E.M., G.B., C.M.R.), Division of Endocrinology,University of Florence; and the Laboratory ofEndocrinology (A.O.), Careggi Hospital, Florence, Italy.
A d d ress correspondence to Carlo M. Rotella,MD, Division of Endocrinology, Section of Diabetol-o g y, Viale Pieraccini 6-50134 Florence, Italy. E-mail:c . ro t e l l a @ m e d . n u c 2 . d f c . u n i fi. i t
R e f e re n c e s1. Mannucci E, Bardini G, Ognibene A,
Rotella CM: Screening for diabetes inobese patients using the new diagnosticcriteria (Letter). Diabetes Care 2 1 : 4 6 8 ,1 9 9 8
2. World Health Organization: Diabetes Melli -tus: Report of a WHO Study Gro u p. Geneva,World Health Org., 1985 (Tech. Rep. Ser. ,n o . 7 2 7 )
3. The Expert Committee on the Diagnosisand the Classification of Diabetes Mellitus:R e p o rt of the Expert Committee on theDiagnosis and the Classification of DiabetesMellitus. Diabetes Care 2 0 : 1 1 8 3 – 1 1 9 7 ,1 9 9 7
H b A1 c D e t e rm i n a t i o nin Patients Wi t hH e m o g l o b i n o p a t h i e s
Th e re is an increasing number ofhemoglobin (Hb) variants, causingfalse HbA1c results in determination of
H b A1 c in diabetic patients of all ethnicg roups. We read with interest the study ofR o b e rts et al. (1) on the effects of sickle celltrait and HbC trait in a black study popula-tion on determination of HbA1 c by animmunoassay method (DCA 2000, BayerDiagnostics, Elkhart, IN). In patients withHbC trait, the immunoassay methodshowed significant positive bias compare dwith that found with high perf o rmance liq-uid chromatography (HPLC) (Diamat, Bio-Rad Clinical Labs, Hercules, CA) (1). Usinga cation-exchange chro m a t o g r a p h y, hemo-globinopathies C and S are alreadyre p o rted to falsely lower glycohemoglobinvalues in diabetic patients (2). Althoughr a re in Caucasians, HbC and HbS are, witha prevalence of up to 3 and 9%, a commono c c u rrence in African-Americans (1).
We re p o rt an estimated 0.6‰ pre v a-lence of silent hemoglobinopathiesdetected in 15,000 HbA1 c d e t e rm i n a t i o n sduring a time period of 6.5 years. Thismeans that there were nine diabeticpatients of Caucasian origin with silent Hbvariants (3,4). There are several re p o rts onhemoglobinopathies and wrong HbA1 c
results in diabetic patients of Asian origin(5). Despite advances in the standard i z a-tion of methods for glycohemoglobins,several hemoglobinopathies cause falseresults in HbA1 c d e t e rminations (3). HbA1 c
is highly sensitive to blood glucose eleva-tions, and a deviation of 1% in HbA1 c
results re flects a change of 1.4–1.9 mmol/lin average blood glucose. If the clinicali m p ression and HbA1 c test results do notmatch, in case of inappropriate HbA1 c
results or additional peaks in HPLC chro-matograms, it is suggested to determ i n ethe HbA1 c values with a second methodbased on a diff e rent principle (4). Sinceblood glucose measurements and fru c-tosamine results are not always well corre-
DIABETES CARE, VOLUME 22, NUMBER 2, FEBRUARY 1999 369
Letters
lated (6), correct HbA1 c values and meanblood glucose are the most appro p r i a t emethods, but only if the contribution ofHb variants is properly accounted for. Wi t hrespect to local appearances of Hb variants,and concerning the ethnic origin of a pop-ulation, every individual laboratory stillhas to establish and secure its own assaymethod. Calculation of mean blood glu-cose and/or measurement of fru c t o s a m i n e(7) may be alternative methods for long-t e rm control in diabetic patients withh e m o g l o b i n o p a t h i e s .
We conclude that in treatment of dia-betic patients the knowledge of hemoglo-binopathies influencing HbA1 c d e t e rm i n a-tion methods is essential (8) and needs tobe investigated thoroughly because Hbvariants contribute to mismanagement ofdiabetic patients due to false HbA1c re s u l t s .
WOLFGANG J. SCHNEDL, MDMARTIN TRINKER, MD
RAINER W. LIPP, MD
F rom the Department of Internal Medicine, Karl-Franzens University, Graz, Austria.
A d d ress correspondence to Wolfgang J. Schnedl,MD, Department of Internal Medicine, Karl-FranzensU n i v e r s i t y, Auenbruggerplatz 15, 8036 Graz, Austria.E-mail: [email protected].
R e f e re n c e s1 . R o b e rts WL, McCraw M, Cook CB: Eff e c t s
of sickle cell trait and hemoglobin C traiton determinations of HbA1 c by animmunoassay method. Diabetes Care21:983–986, 1998
2 . Aleyassine H: Low pro p o rtions of glycosy-lated hemoglobin associated with hemo-globin S and hemoglobin C. Clin Chem25:1484–1486, 1979
3 . Halwachs-Baumann G, Katzensteiner S,Schnedl W, Pürstner P, Pieber T, Wi l d e r s -Truschnig M: Comparative evaluation oft h ree assay systems for the automatedd e t e rmination of hemoglobin A1c. C l i nC h e m43:511–517, 1997
4 . Schnedl WJ, Reisinger EC, KatzensteinerS, Lipp RW, Schreiber F, Hopmeier P, Kre j sGJ: Haemoglobin O Padova and falselylow haemoglobin A1c in a patient withtype I diabetes. J Clin Pathol5 0 : 4 3 4 – 4 3 5 ,1 9 9 7
5 . Nakanishi T, Miyazaki A, Kishikawa M,Shimizu A, Aoki Y, Kikuchi M: Hb Peter-b o rough [beta 111(G13)Va l→Phe] inJapan: rapid identification by ESI/MSusing proteolytic digests of oxidized glo-bin. H e m o g l o b i n22:23–35, 1998
6 . Aebi-Ochsner CH, Grey VL: Fru c t o s a m i n em e a s u rements in the adolescent with typeI diabetes (Letter). Diabetes Care1 8 : 1 6 1 9 –1620, 1995
7 . Onyemelukwe GC, Isah HS, Mba EC,Aw u n n o r-Renner C, Mohammed I: Glyco-sylated haemoglobin (HbA1) for diabeticc o n t rol in Africans: pre l i m i n a ry fin d i n g swith the microcolumn technique. Tro pGeogr Med35:347–351, 1983
8 . Schnedl WJ, Lipp RW, Tr i n k e r M, Hop-m e i e r P : Hemoglobin D [ 1 2 1 ( G H 4 )Glu>Gln] causing falsely low and highH b A1 c-values in high perf o rmance liquidc h romatography (Letter). Clin Chem44:1999–2000, 1998
Response to Schnedlet al.
We appreciate the authors’ interest inour findings (1,2). We agree thathemoglobin variants may pose a
potential problem in ascertaining glycemicc o n t rol, particularly when working withpopulations that contain a high pre v a l e n c eof them. The effect of a particular hemoglo-bin variant on glycohemoglobin measure-ments is highly method-dependent (3).C o rrelation of glycohemoglobin values withother measures of glycemic control, such asf ructosamine or mean blood glucosesobtained from patient self-monitoring, mayhelp identify those patients with spuriousresults due to hemoglobin variants. Furt h e rinvestigation into the impact of their pre s-ence on decision-making in the manage-ment of diabetes is needed. The data fro mour paper and this letter underscore theneed for additional studies on the effects ofvariant hemoglobins on the numerous gly-cohemoglobin methods used in clinical lab-oratories and physician off i c e s .
WILLIAM L. ROBERTS, MD, PHDCURTISS B. COOK, MD
F rom the Department of Pathology (W.L.R.), Uni-versity of Utah, Salt Lake City, Utah; and theD e p a rtment of Internal Medicine (C.B.C.), EmoryU n i v e r s i t y, Atlanta, Georg i a .
A d d ress correspondence to William L. Robert s ,MD, PhD, ARUP Laboratories, Inc., 500 ChipetaWa y, Salt Lake City, UT 84108.
R e f e re n c e s1 . R o b e rts WL, MacCraw M, Cook CB:
E ffects of sickle cell trait and hemoglobinC trait on determinations of HbA1 c by animmunoassay method. Diabetes Care21:983–986, 1998
2 . Schnedl WJ, Trinker M, Lipp RW: HbA1 c
d e t e rmination in patients with hemoglo-binopathies (Letter). Diabetes Care 2 2 : 3 6 3 –364, 1999
3 . Weykamp CW, Penders TJ, Muskiet FA J ,van der Slik W: Influence of hemoglobinvariants and derivatives on glycohemoglo-bin determinations, as investigated by 102laboratories using 16 methods. Clin Chem39:1717–1723, 1993
Fasting Glucose andP ro g ression ofI m p a i red Glucose Tolerance
Iwrite in response to the article by Chouet al. (1) published in the July issue ofDiabetes Care. The authors of this art i c l e
examine pro g ression to diabetes in ac o h o rt of subjects according to their base-line glucose tolerance, which is subdi-vided according to normal glucose toler-ance (NGT), impaired glucose tolerance(IGT), and persistent fasting hyperg l y-cemia (PFH) (defined as fasting plasmaglucose [FPG] 5.6–7.8 mmol/l). They con-clude that PFH may be a “transitional con-dition that precedes” IGT and is interm e-diate between NGT and IGT
I am concerned, however, that theselection bias inherent in the design of thestudy does not allow this conclusion to bed r a w n .
Selection bias was introduced by theuse of an initial FPG cutoff value of 5.6mmol/l to decide who re q u i red an oralglucose tolerance test (OGTT). In ourexperience among Hong Kong Chinese,the mean FPG value of subjects with IGTis only 5.3 mmol/l—well below the cutoffvalue used in this study. Thus many, quitepossible the majority, of subjects with IGTbased on World Health Organization crite-ria may have been excluded from the out-set, and the IGT group would have beenp reselected for a higher FPG. Thesepatients would there f o re constitute a high-risk subgroup of IGT, not necessarilyre flecting the behavior of IGT as a whole.
Also, it is not clear how those subjectswith both PFH (FPG 5.6–7.8) and dia-betes (2-h plasma glucose 11.1) werehandled. I estimate that these patientsaccount for 20% of the PFH gro u p .
Putting it another way, all of the sub-jects by definition have PFH, since sub-jects with FPG 5.6 mmol/l wereexcluded as normal and subjects with FPG
7.8 mmol/l were diagnosed as havingdiabetes. Thus, what the authors have
370 DIABETES CARE, VOLUME 22, NUMBER 2, FEBRUARY 1999
Letters
found is that subjects with PFH are morelikely to pro g ress to type 2 diabetes if theyalso have a 2-h glucose in the IGT range.While interesting, this tells us nothingabout the pro g ression of IGT as a whole asc o m p a red with PFH. To answer this point,d i rect comparison would need to be madebetween an IGT group unselected by FPGand a PFH group selected by OGTT for 2-h glucose 7.8 mmol/l.
Thus I would disagree with the word-ing of the conclusions by the authors,since the IGT group is biased toward“ m o re severity,” while the PFH group maybe biased in the other dire c t i o n .
CLIVE S. COCKRAM, BS, FRCP, MD
F rom the Department of Medicine and Therapeu-tics, Chinese University of Hong Kong, Prince ofWales Hospital, Hong Kong.
A d d ress correspondence to Clive S. Cockram,D e p a rtment of Medicine and Therapeutics, ChineseUniversity of Hong Kong, Room 114028 9/F ClinicScience Building, Prince of Wales Hospital, Shatin,N T, Hong Kong.
R e f e re n c e s1 . Chou P, Li C-L, Wu G-S, Tsai S-T: Pro g re s-
sion to type 2 diabetes among high-riskg roups in Kin-Chen, Kinmen: exploringthe natural history of type 2 diabetes. D i a -betes Care47:1183–1187, 1998
Response to Cockram
We would like to thank Dr. Cockram(1) for his response to our study(2). The major concern of his letter
is that the method of our study design mayhave led to bias in the selection of studysubjects. However, in the title of our study,we clarified that our study subjects weref rom a high-risk group rather than fro mthe general population. The main objectiveof our study was to explore the conversionrate of people among a high-risk group totype 2 diabetes. Thus, there should be noselection bias introduced by the use of aninitial fasting plasma glucose (FPG) valueof 5.6–7.8 mmol/l to decide who re q u i re dan oral glucose tolerance test (OGTT).What we were more concerned with waspossible selection bias introduced byinevitable losses between the two-stages c reening steps. As mentioned by Ham-man (3), this was one of the disadvantagesof the two-stage screening approach toidentifying people with diabetes.
In our study, glucose tolerance classifi-cation was based on currently used Wo r l dHealth Organization (WHO) criteria with a75-g OGTT (4). Those with a history ofdiabetes were excluded first. Impaired glu-cose tolerance (IGT) was defined as FPG
7.8 mmol/l and a 2-h plasma glucose(PG) 7.8 and 11.1 mmol/l. Type 2 dia-betes was defined as either FPG 7 . 8mmol/l or a 2-h PG 11.1 mmol/l. Persis-tent fasting hyperglycemia (PFH), whichpatients used to be classified as having nor-mal glucose tolerance (NGT) according toWHO criteria, was defined as FPG of5.6–7.8 and a 2-h PG 7.8 mmol/l. NGT,PFH, IGT, and diabetes were all mutuallyexclusive. There f o re, there is no overlap-ping among these groups. In fact, we didmake direct comparison between an IGTg roup and a PFH group based on OGTT.
What we showed in this study wasthat, among the high-risk groups in Kin-Chen, Kinmen, those with PFH are morelikely to develop diabetes than those withNGT but less likely than those with IGT,suggesting that PFH may precede IGT inthe pro g ression toward type 2 diabetes.We agree with Hamman (3), who men-tions that it is difficult to determine from asingle follow-up OGTT the temporalsequence of fasting versus postchallengeglucose elevations. Additional pro s p e c t i v estudies of sequential tests over longer timeperiods are certainly needed.
Our study has provided useful data forexploring the natural history of type 2 dia-betes from Chinese patients with fastingh y p e rglycemia. Researchers should con-tinue to seek the best methods and cutofflevels to permit early detection and appro-priate treatment for high-risk populations.
PESUS CHOU, DRPHCHIA-LIN LI, MS
SHIH-TZER TSAI, MD
F rom the Community Medicine Research Centerand Institute of Public Health (P.C., C.-L.L.),National Yang-Ming University; the Department ofMedicine (S.-T. T.), Veterans General Hospital-Taipei, Taipei; and the Department of Health CareManagement (C.-L.L.), Chang Gung University,Ta o - Yuan, Taiwan, Republic of China.
A d d ress correspondence to Pesus Chou, DrPH,National Yang-Ming University, Institute of PublicHealth, Shih-Pai, Taipei, 11221, Taiwan, R.O.C. E-mail: [email protected].
R e f e re n c e s1 . Cockram CS: Fasting glucose and pro g re s-
sion of impaired glucose tolerance. D i a b e t e sC a re 2 2 :369–370, 1999
2 . Chou P, Li C-L, Wu G-S, Tsai S-T: Pro g re s-sion to type 2 diabetes among high-riskg roups in Kin-Chen, Kinmen: exploring thenatural history of type 2 diabetes. D i ab e t e sC a re 21:1183–1187, 1998
3 . Hamman RF: Natural history of type 2diabetes in Chinese: lessons from theislands of Taiwan (Editorial). Diabetes Care21:1035–1036, 1998
4 . World Health Organization: Diabetes Melli -tus: Report of a WHO Study Group. G e n e v a ,World Health Org., 1985 (Tech. Rep. Ser. ,no. 727)
M e n i n g e a lS y n d rome inD i a b e t i cK e t o a c i d o s i s
Consider cerebral edema
We read with great interest the casere p o rted by Albareda et al. (1).They described a case of diabetic
ketoacidosis (DKA) occurring in a 32-year-old woman who presented with headacheand nuchal rigidity, symptoms ofmeningeal irritation. Cere b rospinal flu i d(CSF) analysis revealed a raised pro t e i nwith no evidence of infective cause.A l b a reda et al. subsequently concluded thatDKA and meningeal syndrome are causallyrelated, but they cannot specify any meta-bolic abnormality to account for this (1).
In this case, we believe that the diagno-sis of cerebral edema should have been con-s i d e red. Cerebral edema is an uncommoncomplication of DKA that occurs mainly inc h i l d ren (2), although it has also beenre p o rted in the adult population (3). In fact,a significant number of patients with DKAhad subclinical cerebral edema confirm e dby computed tomography (CT) (4). A CTbrain scan in this patient before the lumbarp u n c t u re (LP) would have been very valu-able to detect brain swelling, as well as toexclude other causes of raised intracranialp re s s u re, which would make LP hazard o u s .CSF pre s s u re, which would have givensome indication of whether cerebral edemawas present, was not documented. In DKA-associated cerebral edema, the presence ofwhite cells in CSF (absent in this case) canoccur in the absence of infection, whichmay create a diagnostic problem. It wouldalso be valuable for the authors to includevalues of plasma electrolytes, part i c u l a r l yplasma sodium in this patient, given that
DIABETES CARE, VOLUME 22, NUMBER 2, FEBRUARY 1999 371
Letters
she had been on diuretic therapy (spiro n o-lactone). It has been shown that hypona-t remia is a predicting factor in the develop-ment of this potentially serious complica-tion of DKA (6). Symptoms of meningeali rritation often settle after the restoration offluid and electrolyte balance; however, itmay also deteriorate during early tre a t m e n t(7). Several mechanisms have been pro-posed for the development of cere b r a ledema in DKA, although the precise causeremains speculative (8).
C e rebral edema should always be con-s i d e red in patients with DKA who havemeningeal symptoms not only at pre s e n t a-tion but also during early rehydration. CTof the brain is mandatory to establish earlydiagnosis because this complication has av e ry high mortality (2).
NORMAN N. CHAN, MRCP, DCHDANIEL DARKO, MRCP
DONAL O’SHEA, MD, MRCPI
F rom the Department of Endocrinology, ImperialCollege School of Medicine, Charing Cross Hospi-tal, London, U.K.
A d d ress correspondence to Dr. N.N. Chan,M R C P, DCH, Department of Endocrinology, Imper-ial College School of Medicine, Charing Cross Hos-pital, London W6 8RF, U.K.
R e f e re n c e s1 . A l b a reda MM, Wägner A, Puig M, Corc o y
R: Is diabetic ketoacidosis a cause ofmeningeal syndrome? Case re p o rt (Letter).Diabetes Care 21:1572–1573, 1998
2 . Rosenbloom AL: Intracerebral crises dur-ing treatment of diabetic ketoacidosis. D i a -betes Care 13:22–33, 1990
4 . Krane EJ, Rockoff MA, Wallman JK, Wo l t s-d o rf JI: Subclinical brain swelling in chil-d ren during treatment of diabetic ketoaci-dosis. N Engl J Med 312:1147–1151, 1985
5 . Young E, Bradley RF: Cerebral edema withi rreversible coma in severe diabetic ketoaci-dosis. N Engl J Med 276:665–669, 1967
6 . Hale PM, Rezvani I, Braunstein AW, Lip-man TH, Martinez N, Garibaldi L: Factorsp redicting cerebral oedema in young chil-d ren with diabetic ketoacidosis and newonset type 1 diabetes. Acta Paediatr86:626–631, 1997
8 . Van der Meulen J, Klip A, Grinstein S: Possible mechanism for cerebral oedemain diabetic ketoacidosis. Lancet i i(8554):306–308, 1987
Response to Chanet al.
Is diabetic ketoacidosis a cause ofmeningeal syndro m e ?
We have carefully read the com-ments of Chan and colleagues (1)on our re p o rt of diabetic ketoaci-
dosis (DKA) as a potential cause ofmeningeal syndrome (2). We agree thatc e rebral edema (a well-known complica-tion of DKA [3]) can be responsible forsymptoms of meningeal irritation (4).H o w e v e r, this was not the case in thepatient we re p o rted. Before lumbar punc-t u re was attempted, an ophthalmoscopydocumented the absence of papilledema,and a cranial computed tomography scandid not disclose any signs of cere b r a ledema. The cerebral fluid pre s s u re wasnot measured. So, we can aff i rm that in there p o rted patient, meningeal syndrome wasnot due to cerebral edema (not even sub-clinical) and, thus, must have been due tosomething else.
We would also like to add that theclinical course of the re p o rted patient didnot suggest at all the presence of cere b r a ledema. This complication is infre q u e n t ,mainly occurs in children (3), and isexceptional before the initiation of therapy(5). Even more important, the patient’s sta-tus improved steadily after the initiation oft reatment, which is the opposite of whatone would expect in the case of cere b r a le d e m a .
In conclusion, we agree that cere b r a ledema has to be considered as a potentialcause of meningeal syndrome in DKA.H o w e v e r, DKA per se seems to be suff i-cient to produce it.
MAMERCÈ ALBAREDA, MDANA WÄGNER, MD
MIREIA PUIG, MDROSA CORCOY, MD, PHD
F rom the Departments of Endocrinology (M.A.,A . W., R.C.) and Internal Medicine (M.P.), Hospitalde Sant Pau, Barcelona, Spain.
A d d ress correspondence to Dra. Rosa Corc o y,D e p a rtment of Endocrinology, Hospital de SantPau, Sant Antoni M’ Claret, 167, 08025 Barc e l o n a ,Spain. E-mail: rc o rc o y @ s a n t p a u . e s .
R e f e re n c e s1 . Chan NN, Darko D, O’Shea D: Meningeal
2 . A l b a reda MM, Wägner A, Puig M, Corc o yR: Is diabetic ketoacidosis a cause ofmeningeal syndrome? Case re p o rt (Letter).D i a b e t e sC a re 21:1572–1573, 1998
3 . Rosenbloom AL: Intracerebral crises dur-ing treatment of diabetic ketoacidosis. D i a -betes Care13:22–33, 1990
4 . Chadwick D, Cartlidge N, Bates D: Raisedintracranial pre s s u re and cerebral tumors.In Medical Neuro l o g y. Singapore, Churc h i l lLivingstone, 1990, p. 324–330
Analytic Accuracyof Calculated L o w - D e n s i t yL i p o p rotein C h o l e s t e ro lC o n c e n t r a t i o n s
We would like to discuss the con-clusions of Branchi et al. (1)re g a rding the reason for inaccu-
rate LDL cholesterol values in diabeticsubjects calculated with Friedewald’s for-mula. Although there was a good corre l a-tion between the calculated and deter-mined LDL cholesterol concentrations (r =0.95 and 0.97 in samples with triglyceride[TG] concentrations 4.52 mmol/l), thebias was frequently large, especially in thediabetic subjects and in subjects whoseplasma TG concentrations were 2 . 2 6mmol/l. Consequently, 25% of the sub-jects were misclassified for coro n a ry heartdisease risk using cutoff points of theNational Cholesterol Education Pro g r a m .
The re p o rt of Branchi et al. (1) adds ton u m e rous other studies published on thissubject. All of these re p o rts suggest that thedeviating chemical composition of theVLDL and LDL fractions must be the majorreason for the observed deviations. How-e v e r, it can be reasoned that this arg u m e n tis of limited value for clinical decisionmaking. Pre v i o u s l y, we found that the cho-l e s t e rol content in atherogenic remnants, asp resent in familial dysbetalipopro t e i n e m i a(FD), is included in the calculated LDLc h o l e s t e rol value, thereby more accuratelyre flecting the clinical risk than the deter-mined LDL cholesterol value (2). Thisfinding can be explained by the fact that in
372 DIABETES CARE, VOLUME 22, NUMBER 2, FEBRUARY 1999
Letters
these patients, the plasma TG concentra-tion is relatively low. Consequently, the cal-culated VLDL cholesterol concentration isonly a fraction of the actual cholestero lcontent in the density 1.006 g/ml con-taining mainly remnants. In subjects withremnant accumulation not typical for FD,i.e., in those with the apoprotein (apo) Ephenotypes E2/E3 or E2/E4, a similar clin-ical risk correction can be expected. Inaddition, we found in two independentstudies that the ratio of VLDL cholestero lto plasma TG was relatively independent ofthe plasma TG up to a TG concentration of8 mmol/l (usually, 400 mg/dl [4.52mmol/l] is considered the limit).
In our recent study (3), we suggestthat the use of imprecise and inaccuratemethods for measurement of plasma lipidsand HDL cholesterol may be the main re a-son for the re p o rted limited value ofF r i e d e w a l d ’s approach in other laborato-ries. This possibility is supported by thehigher correlation coefficients (r = 0.97 to
0.98) we obtained comparing calculatedand determined LDL cholesterol concen-trations in subjects with plasma TG con-centrations up to 8 mmol/l (2,3). Besides,t h e re are other methodological concern sin the study of Branchi et al. (1): qualityc o n t rol data of the methods used are lack-ing, and the re f e rence HDL cholestero lmethod is based on ultracentrifugation,which is inaccurate because the densityfraction 1.063 g/ml contains most of thel i p o p rotein(a) present in plasma (4).Other lessons of our two independentstudies on the reliability of Friedewald’sequation are as follows: 1) In each evaluat-ing study, the relatively small VLDL cho-l e s t e rol value needed to calculate the re f e r-ence LDL cholesterol concentration mustbe determined directly instead of indi-rectly as the diff e rence of two large ana-lytes (total cholesterol minus cholesterol inthe density fraction 1.006 g/ml). The useof a tube-slicing technique instead ofd i rect aspiration is thus pro b l e m a t i c .D i rect aspiration and analysis improve thep recision (coefficient of variation [CV]) ofthe re f e rence VLDL cholesterol value. 2)An HDL cholesterol method with anuncomplicated precipitation of the apoB-containing lipoproteins in sera with a TGconcentration 4.5 mmol/l should beused, e.g., the PEG-6000, the phospho-t u n g s t a t e / M g2 , or a direct HDL choles-t e rol method (5). Surprisingly, the re f e r-ence for the HDL cholesterol method usedby Branchi et al. (1) is not indicated. The
use of an inadequate HDL cholestero lmethod contributes to a relatively highi m p recision, not only of the re f e rence butalso of the calculated LDL cholestero lvalue. 3) Methods to measure TG and cho-l e s t e rol (the latter also at low levels, asp resent in the HDL fraction) should bemaximally precise and accurate. In thisrespect, the linearity of the TG determ i n a-tion should be taken into account: mostmethods commonly used re q u i re a dilu-tion at concentrations 4.0 mmol/l.When a dilution is needed, the CV of theanalyte increases severalfold.
F rom the above discussion, it is obvi-ous that at TG concentrations 4 . 0mmol/l, several factors increase the impre-cision of both the measured and the calcu-lated LDL cholesterol concentrations. Inour opinion, this problem is a major causeof the presumed imprecision of Friede-w a l d ’s formula (3). This view is support e dby the similar under- and overe s t i m a t i o np e rcentages of 16–19% in the study ofBranchi et al. (1), in both the patient andthe control subject groups. Note thatu n d e r- or overestimation was most fre-quent at low concentrations of LDL cho-l e s t e rol ( 3.9 mmol/l), which were char-acteristic for sera with increased plasmaTG concentrations.
PIERRE N.M. DEMACKER, PHDANTON H. STALENHOEF, MD
F rom the Department of Medicine, Division of Gen-eral Internal Medicine, University HospitalNijmegen, Nijmegen, the Netherlands.
A d d ress correspondence to Pierre N.M.D e m a c k e r, PhD, Department of Medicine, Divisionof General Internal Medicine, University HospitalNijmegen, 6500 HB Nijmegen, the Netherlands. E-mail: p.demacker@[email protected].
R e f e re n c e s1 . Branchi A, Rovellini A, To rri A, Sommariva
D: Accuracy of calculated serum low-den-sity lipoprotein cholesterol for the assess-ment of coro n a ry heart disease risk inNIDDM patients. Diabetes Care21:1397–1402, 1998
2 . Demacker PN, Hijmans AG, Bre n n i n k -meijer BJ, Jansen AP, van ‘t Laar A: Fivemethods for determining low-densityl i p o p rotein cholesterol compared. C l i nChem 30:1797–1800, 1984
3 . Demacker PN, Toenhake-Dijkstra H, deRijke YB, Stalenhoef AF, Stuyt PM,Willems HL: On the presumed inaccuracyof the Friedewald formula in hypert r i g l y c-eridemic plasma: a role for impre c i s eanalysis? Clin Chem42:1491–1494, 1996
4 . Demacker PN, Vos-Janssen HE, Jansen AP,
van ‘t Laar A: Evaluation of the dual-pre-cipitation method by comparison with theultracentrifugation method for measure-ment of lipoproteins in seru m. Clin Chem23:1238–1244, 1977
5 . Demacker PN, Hessels M, To e n h a k e - D i j k-stra H, Baadenhuijsen H: Pre c i p i t a t i o nmethods for high-density lipoprotein cho-l e s t e rol measurement compared, and fin a levaluation under routine operating condi-tions of a method with a low sample-to-reagent ratio. Clin Chem 4 3 : 6 6 3 – 6 6 8 ,1 9 9 7
Response toDemacker andS t a l e n h o e f
In their letter, Demacker and Stalenhoefsuggest that the imprecision of LDLc h o l e s t e rol calculation using Friede-
w a l d ’s formula is due mainly to the inaccu-racy of methods for measurement ofplasma lipids and HDL cholesterol ratherthan to variations in VLDL composition.Their opinion is based on two observ a-tions: 1) the high correlation coeff i c i e n t sthey found between measured and calcu-lated LDL cholesterol and 2) the re l a t i v econstancy of the ratio of VLDL cholestero lto plasma triglycerides (TGs) up to aplasma TG level of 1,240 mg/dl (14mmol/l) (1). In our series, too, and inother studies (2), measured and calculatedLDL cholesterol values were highly corre-lated (r = 0.95 in 151 diabetic patients andr = 0.97 in 405 nondiabetic patients);h o w e v e r, a good correlation may mask thep revalence of clinically significant erro r s .Despite the high correlation between thetwo methods, we found a diff e re n c e
10% between measured and calculatedLDL cholesterol values in 34% of samplesf rom diabetic subjects and in 26% of sam-ples from nondiabetic subjects, in accordwith previous re p o rts (2).
VLDL composition has a major impacton LDL cholesterol calculation usingF r i e d e w a l d ’s formula, which is based onthe assumption of a fixed re l a t i o n s h i pbetween VLDL cholesterol and serum TGs.Demacker and Stalenhoef claim that theratio of VLDL cholesterol to plasma TGs isrelatively constant and independent of theplasma TG level, although in the fig u reaccompanying their article (1), the ratio ofVLDL cholesterol to plasma TG is scattere df rom 0.1 to 0.7, for a TG level up to 5
DIABETES CARE, VOLUME 22, NUMBER 2, FEBRUARY 1999 373
Letters
mmol/l. In our series, too, the VLDL cho-l e s t e rol–to–plasma TG ratio showed a widev a r i a b i l i t y, ranging from 0.05 to 0.9. Theg reat variability of the ratio is expectedbecause the serum TG level results notonly from VLDL concentration but fro mthe sum of the TG content of all thel i p o p roteins, and the TG content of all thelipoprotein families, including VLDL,changes when TG metabolism is altere d(3). On the other hand, it is widelyaccepted that variations in serum TG con-centration are the main source of error inF r i e d e w a l d ’s empirical equation (2).
We agree that the method used ind e t e rmining HDL cholesterol is criticallyi m p o rtant in the estimation of LDL choles-t e rol. As stated in our article, we deter-mined the cholesterol content of the den-sity fraction 1.063, which contains someof the lipoprotein(a) present in the plasma,as correctly pointed out by Demacker andStalenhoef. HDL cholesterol determined insuch a way should be overestimated; how-e v e r, this is of minor importance in thefinal results of our study, because the samevalue of HDL cholesterol was used whenre f e rence ultracentrifugation method wasused as when Friedewald’s equation wasused. There f o re, the diff e rence betweenm e a s u red and calculated LDL cholestero lwas entirely due to a diff e rence in the esti-mation of VLDL cholestero l .
Demacker and Stalenhoef observe thatour article lacks data on the re p ro d u c i b i l-ity of determinations. We apologize forour omission; the coefficient of variationwas 1.76% for cholesterol, 3.54% for TGs,7.01% for VLDL cholesterol, 4.57% forLDL cholesterol, and 4.19% for HDL cho-
l e s t e rol. VLDL cholesterol was calculatedas the diff e rence between total cholestero land cholesterol of density 1 . 0 0 6 ,a c c o rding to the method of Bro n z e rt andB rewer (4). Demacker and Stalenhoefo b s e rve that this method of determ i n i n gVLDL cholesterol is less accurate than thed i rect measurement of cholesterol in thel i p o p rotein fraction. In fact, in their study(1), the resulting coefficient of variation ofVLDL cholesterol was low. In our experi-ence, the Bro n z e rt and Brewer method isnot less suitable than the direct determ i n a-tion of VLDL cholesterol in the aspiratedfraction, which often has a very low cho-l e s t e rol concentration (there f o re incre a s-ing imprecision and inaccuracy) andneeds to be corrected for re c o v e ry.
O b v i o u s l y, we agree that methodologyof lipid determination plays a key role inboth the re f e rence method and Friede-w a l d ’s calculation, but we believe thati m p recision of Friedewald’s equation insome conditions, such as hypert r i g l y c-eridemia, is unlikely to be due to onlyi m p recision and inaccuracy of methodsused in lipid determination. A number ofstudies using diff e rent methodologies arein accord with our conclusions.
The main purpose of our study was toevaluate the reliability of Friedewald’sequation in type 2 diabetic patients withrespect to nondiabetic subjects. Asexpected, we found a greater error in LDLc h o l e s t e rol estimation in diabetic patientsthan in nondiabetic subjects, because ofthe more frequent abnormalities in TGmetabolism in the former group. The limi-tations of the estimation of LDL cholestero lwith Friedewald’s formula are well known.
H o w e v e r, in our experience, the methodp roved useful in the correct assignment ofrisk classes of coro n a ry disease in 75% ofnondiabetic and diabetic patients.
ADRIANA BRANCHI, MDANGELO ROVELLINI, MD
ADRIANA TORRI, MDDOMENICO SOMMARIVA, MD
F rom the Department of Internal Medicine (A.B.,A.R.), University of Milan, Maggiore HospitalIRCCS, Milan; and the Department of Internal Med-icine (A.T., D.S.), G. Salvini Hospital, GarbagnateMilanese, Italy.
A d d ress correspondence to Adriana Branchi,MD, Dipartimento di Medicina Interna, PadiglioneGranelli, Via Francesco Sforza 35, 20100 Milano,I t a l y.
R e f e re n c e s1 . Demacker PN, Toenhake-Dijkstra H, de
Rijke YB, Stalenhoef AF, Stuyt PM,Willems HL: On the presumed inaccuracyof the Friedewald formula in hypert r i g l y c-eridemic plasma: a role for impre c i s eanalysis? Clin Chem42:1491–1494, 1996
2 . Branchi A, Rovellini A, To rri A, SommarivaD: Accuracy of calculated serum low-densityl i p o p rotein cholesterol for the assessment ofc o ro n a ry heart disease risk in NIDDMpatients. Diabetes Care21:1397–1402, 1998
3 . Sommariva D, Branchi A, Ti rrito M, Bon-figlioli D, Pini C, Scandiani L, OttomanoC, Fasoli A: Relationship of serum triglyc-eride concentration to lipoprotein compo-sition and concentration in norm o l i p i-demic and hyperlipidemic subjects. R e sClin Lab18:281–290, 1988
4 . B ro n z e rt TJ, Brewer HB Jr: Newm i c romethod for measuring cholesterol inplasma lipoprotein fractions. Clin Chem23:2089–2091, 1977
