DIABETES CARE, VOLUME 23, NUMBER 1, JANUARY 2000 125 O B S E RVAT I O N S Diabetes Scre e n i n g Practices Among I n d i v i d u a l s A g e d 4 5 Ye a r s a nd O l d e r I n 1997, the American Diabetes Associa- tion (ADA) adopted new re c o m m e n d a- tions for screening the general popula- tion aged 45 years for diabetes every 3 years with an emphasis on those at high risk for undiagnosed diabetes (1). Few studies have examined the extent to which this screening has been adopted. This re p o rt describes the results of a telephone s u rvey of Montana residents aged 4 5 years to assess the diabetes screening prac- tices in this population. F rom October to December 1998, the Montana Department of Public Health and Human Services conducted a random household telephone survey of Montana residents aged 45 years living in 18 counties. Respondents indicated whether they ever had been told by a physician that they had diabetes, the number of visits they made to a health care provider during the past year, their family history of dia- betes, whether they had ever been told they had high cholesterol and/or high blood pre s s u re, and their height and weight. Respondents were asked the fol- lowing question to identify whether they had ever been screened for diabetes: “Glu- cose or sugar is a substance found in your blood. Have you ever had your blood glu- cose or sugar checked to see if you have diabetes?” When respondents re s p o n d e d “yes” to this question, they were asked to identify when screening was completed (“When was the last time your blood glu- cose or sugar level was measured by a health care professional?”). The re s p o n s e categories for this question included within the past year, within the past 3 years, 3 years ago, do not know/not sure, and refused to answer. Pearson 2 tests were used to assess associations between dia- betes screening and risk factors for dia- betes. Logistical re g ression analyses were conducted to identify independent vari- ables associated with screening for diabetes during the past year. Odds ratios (95% CIs) were calculated. Of the 1,204 respondents, 92 (7.6%) re p o rted that they had diagnosed diabetes. The remaining 1,112 respondents re p o rt e d that they did not have diagnosed diabetes and are included in the following analyses. Of the respondents, 39% re p o rted a family h i s t o ry of diabetes, 32% re p o rted a BMI 27 kg/m2, 28% re p o rted having hyper- tension, and 28% re p o rted having high c h o l e s t e rol. Excluding age, 34% of re s p o n- dents had one risk factor for diabetes, and 40% had two or more risk factors. Of the 1,112 respondents without diagnosed dia- betes, 39% re p o rted that they had been s c reened for diabetes during the past year, 14% re p o rted screening from 1 to 3 years ago, and 47% re p o rted screening 3 years ago or having never been scre e n e d . Respondents who re p o rted being s c reened for diabetes during the past year w e re more likely to be age 65 years and to have a family history of diabetes, two or m o re visits to a health care provider dur- ing the past year, hypertension, and high c h o l e s t e rol levels (Table 1). We found no association between recent screening and sex (40% men vs. 38% women), Ameri- can Indian ancestry (41% yes vs. 38% no), or BMI (42% 27 kg/m2 vs. 37% 27 kg/m2). Respondents with three or m o re risk factors (e.g., aged 45 years, American Indian ancestry, family history of diabetes, hypertension, high choles- t e rol, or BMI 27 kg/m2) were more likely to be screened for diabetes compared with respondents with only one risk factor (46 vs. 28%, respectively). However, 48% of individuals with two risk factors for dia- betes and 37% of individuals with more than three risk factors had not been s c reened during the past 3 years. Based on logistical re g ression analysis, t h ree factors were associated with scre e n i n g for diabetes during the past year: two or m o re visits to a health care provider during the past year (2.34 [95% CI 1.76–3.11]), high cholesterol level (1.37 [1.03–1.82]), and family history of diabetes (1.45 [1.12–1.89]). Respondents aged 45–54 years were less likely to re p o rt re c e n t s c reening than those aged 65 years (0.62 [ 0 . 4 6 – 0 . 8 4 ] ) . A limitation of this assessment is that these data are self-re p o rted. Pre v i o u s studies, however, have found that self- re p o rts of conditions such as diabetes and h y p e rtension are reliable (2,3). In addi- tion, the survey was conducted by tele- phone and does not re flect the experience of individuals in Montana homes without t e l e p h o n e s . The findings suggest that diabetes s c reening is being adopted by physicians for individuals aged 45 years at risk for diabetes and that the ADA re c o m m e n d a- tions are being implemented in the general c o m m u n i t y. However, these data also indi- cate a need to develop strategies to encour- L E T T E R S Table 1—Characteristics of respondents aged 45 years reporting screening for diabetes in Montana in 1998 S c reening for diabetes Past year 1–3 years 3 years or never n 4 3 0 1 6 0 5 2 0 Age (years) 4 5 – 5 4 155 (32) 80 (17) 251 (52) 5 5 – 6 4 95 (39) 41 (17) 108 (44) 6 5 180 (47)* 39 (10) 163 (43) Family history of diabetes Ye s 192 (45)* 74 (17) 162 (38) N o 235 (35) 85 (13) 352 (52) Visits to a health care provider during the past year 2 317 (47)* 84 (12) 275 (41) 2 100 (24) 74 (18) 237 (58) H y p e rt e n s i o n Ye s 149 (48)* 41 (13) 122 (39) N o 281 (35) 119 (15) 400 (50) High cholestero l Ye s 148 (47)* 43 (14) 123 (39) N o 280 (35) 116 (15) 396 (50) Data are n (%). *P 0.001. 126 DIABETES CARE, VOLUME 23, NUMBER 1, JANUARY 2000 Letters age screening among all individuals at high risk for diabetes. TODD S. HARWELL, MPH JANE G. SMILIE, BA JANET M. MCDOWALL, RN, BSN STEVEN D. HELGERSON, MD, MPH DOROTHY GOHDES, MD F rom the Montana Diabetes Project, Montana D e p a rtment of Public Health and Human Serv i c e s , Helena, Montana. A d d ress correspondence to Todd S. Harw e l l , MPH, Montana Department of Public Health and Human Services, Cogswell Building, C-317, P. O . Box 202951, Helena, MT 59620-2951. E-mail: t h a rw e l l @ s t a t e . m t . u s . A c k n o w l e d g m e n t s — This project was sup- p o rted through a cooperative agreement (U-32/ CCU-815663-02) with the Centers for Disease C o n t rol and Prevention, Division of Diabetes Translation, Atlanta, Georg i a . We thank Linda Priest and the staff mem- bers at Northwest Resource Consultants for their work on the telephone surv e y. The contents of this letter are solely the responsibility of the authors and do not neces- sarily re p resent the official views of the Centers for Disease Control and Pre v e n t i o n . R e f e re n c e s 1 . American Diabetes Association: Screening for type 2 diabetes. Diabetes Care 22 (Suppl. 1): S20–S23, 1999 2 . Kehoe R, Wu SY, Leske MC, Chylack LT Jr: Comparing self-re p o rted and physician- re p o rted medical history. Am J Epidemiol 139:813–818, 1994 3 . Jackson C, Jatulis DE, Fortmann SP: The Behavioral Risk Factor Survey and the Stan- f o rd Five-City Project Survey: a comparison of cardiovascular risk behavior estimates. Am J Public Health 82:412–416, 1992 H b A1 c Is Not Recommended as a S c reening Test for Diabetes in Cystic Fibro s i s I n the June 1999 issue of Diabetes Care, H u n k e rt et al. (1) recommend the use of H b A1 c for early detection of cystic fib ro- s i s – related diabetes (CFRD). Their re c o m- mendation is based on the finding that mean HbA1 c is slightly higher in cystic fib rosis (CF) patients requiring insulin t h e r a p y, compared with CF patients with i m p a i red or normal glucose tolerance. H o w e v e r, no data on the validity of this a p p roach for the diagnosis of asympto- matic diabetes in patients with cystic fib ro- sis are pre s e n t e d . Our group has a long-standing intere s t in the early diagnosis of CFRD, comparing fasting blood glucose levels with oral glu- cose tolerance test results (2). In our series, we have now 13 CF patients with newly diagnosed diabetes based on a 2-h venous plasma glucose value 200 mg/dl (11.1 mmol/l), and simultaneous determ i n a t i o n of HbA1 c ( h i g h - p e rf o rmance liquid chro- matography method [Pharmacia, Erlan- gen, Germany], normal range 3.5–5.7%). Only 4 out of 13 CF patients (31%) diag- nosed as diabetic according to the Ameri- can Diabetes Association and World Health O rganization criteria (3) had an HbA1 c value above the normal range (individual values 5.9, 6.0, 6.3, and 6.6%). In nine diabetic CF patients with normal HbA1 c, values between 4.6 and 5.7% were e n c o u n t e red (mean ± SD, 5.1 ± 0.4%). The mean 2-h blood glucose value after inges- tion of oral glucose was not signific a n t l y d i ff e rent between diabetic CF patients with n o rmal HbA1 c (263 ± 36 mg/dl, mean ± SD) and diabetic CF patients with elevated H b A1 c (298 ± 34 mg/dl, Student’s t t e s t ) . These data clearly demonstrate that the d e t e rmination of HbA1 c is not able to sub- stitute for the oral glucose tolerance test in the early diagnosis of CFRD. Our fin d i n g s a re in agreement with several re p o rts in the l i t e r a t u re (4–6) as well as the 1998 consen- sus conference on CFRD (7). In addition to its low sensitivity when used as a diagnos- tic tool for the detection of CFRD, the mea- s u rement of HbA1 c has the disadvantage of considerable interassay variability and lack of standardization. There f o re, we stro n g l y advise against the use of glycosylated hemoglobin as a screening test for the early diagnosis of diabetes in patients with cystic fib ro s i s . REINHARD W. HOLL, MD CHRISTIAN BUCK, MD CHRISTINE BABKA, MD ANNA WOLF, MD ANGELIKA THON, MD F rom the Department of Pediatrics (R.W.H.), Uni- versity of Giessen, Giessen; the Department of Pedi- atrics (C.Bu., C.Ba., A.W.), the University of Ulm, Ulm; and the Medical School Hannover (A.T. ) , H a n n o v e r, Germ a n y. A d d ress correspondence to PD Dr. Reinhard W. Holl, Universitätskinderklinik Giessen, Feulgenstr. 12, D-35385 Giessen, Germ a n y. R e f e re n c e s 1 . H u n k e rt F, Lietz T, Stach B, Kiess W: Potential impact of HbA1 c d e t e rm i n a t i o n on clinical decision making in patients with cystic fib ro s i s – related diabetes (Let- ter). Diabetes Care 22:1008–1009, 1999 2 . Holl RW, Buck C, Cario H, Wolf A, Thon A, Heinze E, Kohne E, Debatin K-M: Diag- nosis of diabetes in cystic fib rosis and tha- lassemia major. Diabetes Care 2 1 : 6 7 1 – 6 7 2 , 1 9 9 8 3 . The Expert Committee on the Diagnosis and Classification of Diabetes Mellitus: R e p o rt of the Expert Committee on the Diagnosis and Classification of Diabetes Mellitus. Diabetes Care 2 0 : 1 1 8 3 – 1 1 9 7 , 1 9 9 7 4 . Lanng S, Hansen A, Thorsteinsson B, N e rup J, Koch C: Glucose tolerance in patients with cystic fib rosis: five year p rospective study. B M J 311:655–659, 1995 5 . deLuca F, Arrigo T, Nibali SC, Sferlazzas C, Gigante A, DiCesare E, Cucinotta D: Insulin secretion, glycosylated haemoglo- bin and islet cell antibodies in cystic fib ro- sis children and adolescents with diff e re n t d e g rees of glucose tolerance. H o rm Metab R e s 23:495–498, 1991 6 . Moran A, Doherty L, Wang X, Thomas W: A b n o rmal glucose metabolism in cystic fib rosis. J Pediatr 133:10–17, 1998 7 . Moran A: Highlights of the Febru a ry 1998 Consensus Conference on CFRD. Bonn, G e rm a n y, Cystic Fibrosis Foundation, 1 9 9 8 P ro115Gln P e roxisome P ro l i f e r a t o r- A c t i v a t e d R e c e p t o r- and O b e s i t y R istow et al. (1) re p o rted an activating mutation in the peroxisome pro l i f- e r a t o r-activated re c e p t o r- g e n e ( P ro115Gln PPA R - ), which was pre s e n t in 3% (4 of 121) of obese and 0% (0 of 238) of nonobese German Caucasians. These findings may have profound implica- tions, particularly if the presence of this variant and its association with obesity are c o n firmed in other populations. We perf o rmed polymerase chain re a c- t i o n – restriction fragment length polymor- phism analysis for the Pro115Gln PPA R - variant as described (1) on DNA samples f rom several independent populations, including lean and obese Caucasians fro m the Baltimore, Maryland, region; African- DIABETES CARE, VOLUME 23, NUMBER 1, JANUARY 2000 127 Letters Americans from Jackson, Mississippi, and Forsyth County, North Carolina; Pima Indians from Arizona; and Old Ord e r Amish from Lancaster County, Pennsylva- nia (Table 1). A PCR fragment corre s p o n d- ing to gastric insulinotropic peptide, which has two known restriction sites for H i n dII, was mixed with each sample as a positive control. Among a total of 983 sub- jects (1,966 alleles), the Pro115Gln variant was not detected in a single subject. These findings were unexpected because there is substantial overlap of gene pools of Caucasians from Central E u rope and the Baltimore and Amish Caucasians studied (2). The Germ a n Caucasians studied by Ristow et al. were said to be unrelated and re c ruited fro m the Nord rh e i n - Westfalen region, but they may be a genetic isolate whose gene pool does not re flect that of other Caucasian populations. Altern a t i v e l y, if the individ- uals carrying the mutation were re l a t e d , the true frequency of the Pro1 1 5 G l n P PA R - variant may have been overe s t i- mated. The absence or very low fre- quency of this variant has also been doc- umented in Danish (3) and German (4) populations. This study is the first, to our knowledge, to examine American popu- lations for this variant. In summary, the study by Ristow et al. demonstrating that the Pro1 1 5 G l n P PA R - variant is activating and can influence body weight in people is important. However, this variant appears to be absent or very r a re in the American populations studied. Additional studies are re q u i red in other regions of Europe and the U.S. to furt h e r d e fine the relevance of this intere s t i n g genetic variant to susceptibility to obesity. ALAN R. SHULDINER, MD WILLIAM NGUYEN, BS W.H. LINDA KAO, PHD BROCK A. BEAMER, MD ROSS E. ANDERSEN, PHD RICHARD PRATLEY, MD FREDERICK L. BRANCATI, MD, PHD F rom the Department of Medicine (A.R.S., W. N . ) , University of Maryland School of Medicine; the D e p a rtments of Medicine (B.A.B., F.L.B.) and Epi- demiology (W.H.L.K.), Johns Hopkins University School of Medicine, Baltimore, Maryland; and the National Institute of Diabetes and Digestive and Kidney Diseases (R.P.), National Institutes of Health, Phoenix, Arizona. A d d ress correspondence to Alan R. Shuldiner, MD, Professor and Head, Division of Endocrinol- o g y, Diabetes, and Nutrition, Department of Medi- cine, University of Maryland School of Medicine, 725 W. Lombard St., Room S-422, Baltimore, MD 21201. E-mail:ashuldin@medicine.umary l a n d . e d u . A c k n o w l e d g m e n t s — This study was sup- p o rted by NIH R01 DK-49692, the American Diabetes Association, GlaxoWellcome, the B a l t i m o re Geriatrics Research and Education Clinical Center of the Baltimore Ve t e r a n s Administration Medical Center. R e f e re n c e s 1 . Ristow M, Muller- Wieland D, Pfeiffer A, K rone W, Kahn CR: Obesity associated with a mutation in a genetic regulator of adipocyte diff e rentiation. N Engl J Med 3 3 9 : 953–959, 1998 2 . C a v a l l i - S f o rza LL, Menozzi P, Piazza A: T h e H i s t o ry of Human Genes. Princeton Univer- sity Press, Princeton, NJ, 1994 3 . Elk J, Urhammer SA, Sorensen TI, Ander- sen T, Auwerx J, Pedersen O: Homozygosity of the Pro12Ala variant of the pero x i s o m e p roliferation-activated re c e p t o r- g a m m a 2 ( P PAR-gamma2): divergent modulating e ffects on body mass index in obese and lean Caucasian men. Diabetologia 4 2 : 8 9 2 – 895, 1999 4 . Mamann A, Munzberg H, Buttron P, Busing B, Hinney A, Mayer H, Siegfried W, Hebe- brand J, Greten H: Missense variants in the human peroxisome pro l i f e r a t o r- a c t i v a t e d re c e p t o r-gamma 2 gene in lean and obese subjects. Eur J Endocrinol 141:90–92, 1999 Insulin Secre t i o n , Insulin Sensitivity, and Glucose E ffectiveness in Nonobese Individuals With Va ry i n g D e g rees of Glucose To l e r a n c e A lthough it is well known that insulin s e c retion, insulin sensitivity, and glu- cose effectiveness are impaired in type 2 diabetic patients (1–5), little is known about the role of each of these fac- tors individually on the evolution of type 2 diabetes. In this context, a major issue is that hyperglycemia per se impairs insulin s e c retion and insulin sensitivity and that obesity observed in type 2 diabetic patients per se causes insulin resistance (6,7). To o v e rcome this problem, we studied 37 u n t reated nonobese subjects classified as having normal glucose tolerance (NGT) (n = 14; BMI 21.0 ± 0.5 kg/m2 [ r a n g e 18.1–23.6], mean ± SEM), impaired glu- cose tolerance (IGT) (n =12; BMI 21.3 ± 0.8 kg/m2 [17.0–26.9]), and type 2 dia- betes (n =11; fetal bovine serum 7.0 ± 0.5 mmol/l [range 5.2–9.9], BMI 20.0 ± 0.9 k g / m2 [13.6–24.5]), based on the criteria of the World Health Organization (8). They were normotensive and had norm a l renal, hepatic, and thyroid function. Insulin sensitivity and glucose eff e c t i v e n e s s w e re estimated by the minimal model a p p roach (1–4). Insulin secretion was e x p ressed as the area under the insulin c u rve between 0 and 10 min after an intra- venous glucose injection (2). After the data w e re analyzed by one-way analysis of vari- ance, Bonferroni correction was used to evaluate the diff e rences between any two of the three groups we studied (9). No signifi- cant difference was observed in BMI among the three groups. Compared with the subjects with NGT, the subjects with Table 1—Characteristics of subjects screened for Pro115Gln PPA R - n ( a l l e l e s Age ± SD F e m a l e BMI ± SD P o p u l a t i o n t y p e d ) ( y e a r s ) ( % ) ( k g / m2) C a u c a s i a n s B a l t i m o re Longitudinal 105 (210) 58.7 ± 15.3 5 3 . 3 24.8 ± 5.3 Study on Aging Johns Hopkins We i g h t 285 (570) 42.8 ± 12.0 6 5 . 3 38.8 ± 10.0 Management Center Amish, Lancaster, PA 173 (346) 45.7 ± 14.2 5 1 . 4 27.4 ± 8.9 A f r i c a n - A m e r i c a n s A t h e ro s c l e rosis Risk in 228 (456) 54.8 ± 5.9 6 0 . 0 31.4 ± 5.5 Communities Study Pima Indians A r i z o n a 192 (384) 47.5 ± 14.8 — 32.6 ± 6.5 All non-Amish subjects were unrelated; Amish subjects were not fir s t - d e g ree relatives of each other. 128 DIABETES CARE, VOLUME 23, NUMBER 1, JANUARY 2000 Letters IGT had significantly lower insulin secre- tion (3,207 ± 557 vs. 1,895 ± 319 pmol l 1 m i n 1, P = 0.027) and glucose eff e c- tiveness (0.023 ± 0.002 vs. 0.014 ± 0.002 m i n 1, P 0.001). Insulin sensitivity index was lower in subjects with IGT (0.74 ± 0.13 m i n 1 pmol l) than in those with NGT (1.05 ± 0.14 min 1 pmol l ) , but was not statistically significant (P = 0.171). In con- trast, disposition index calculated by the p roduct of insulin secretion and insulin sensitivity was significantly lower in sub- jects with IGT (1,197 ± 221) than in those with NGT (2,710 ± 371, P = 0.004). On the other hand, patients with type 2 dia- betes had significantly lower insulin secre- tion (212 ± 85 pmol l 1 m i n 1, P = 0.008) compared with subjects with IGT. Although no significant diff e rence was observed in insulin sensitivity index between subjects with type 2 diabetes and IGT (0.74 ± 0.13 vs. 1.09 ± 0.21 min 1 pmol l, P = 0.141), disposition index was s i g n i ficantly diminished in type 2 diabetic patients as compared with subjects with IGT (224 ± 89 vs. 1,197 ± 221, P = 0.022). Glucose effectiveness in type 2 diabetic patients (0.011 ± 0.001 min 1) was similar to that in subjects with IGT (0.014 ± 0.002 m i n 1, P = 0.307) but was signific a n t l y lower than that in the subjects with NGT (P 0.001). From these results, the fol- lowing may be hypothesized: 1) Impair- ments in insulin secretion and disposition index and decreased glucose eff e c t i v e n e s s , but not insulin resistance, seem to consti- tute the basic characteristics of patients with IGT or type 2 diabetes in nonobese Japanese populations. 2) Risk factors wors- ening to type 2 diabetes in subjects with IGT are associated with further impair- ments in insulin secretion and disposition index, but not associated with furt h e r derangement in glucose effectiveness in Japanese populations. ATARU TANIGUCHI, MD MITSUO FUKUSHIMA, MD MASAHIKO SAKAI, MD ITARU NAGATA, MD KENTARO DOI, MD SHOICHIRO NAGASAKA, MD KUMPEI TOKUYAMA, MD YOSHIKATSU NAKAI, MD F rom the First Department of Internal Medicine ( A . T., M.S., I.N.), Kansai-Denryoku Hospital, and the Department of Internal Medicine, Hoshida- Minami Hospital (M.F.), Osaka; the Second Depart- ment of Internal Medicine (K.D.) and the College of Medical Technology (Y.N.), Kyoto University, Kyoto; the Department of Internal Medicine (S.N.), Jichi Medical College, Tochigi; and the Laboratory of Biochemistry of Exercise and Nutrition (K.T. ) , Tsukuba University, Tsukuba, Japan. A d d ress correspondence to Ataru Ta n i g u c h i , MD, First Department of Internal Medicine, Kansai- D e n ryoku Hospital, 2-1-7, Fukushima-ku, Fuku- shima, Osaka City, Osaka 553-0003, Japan. E-mail: k 5 8 4 0 3 @ k e p c o . c o . j p . R e f e re n c e s 1 . B e rgman RN: To w a rd physiological under- standing of glucose tolerance: minimal- model approach. D i a b e t e s 3 8 : 1 5 1 2 – 1 5 2 7 , 1 9 8 9 2 . Welch S, Gebhart SSP, Bergman RN, Phillips LS: Minimal model analysis of intravenous glucose tolerance test-derived insulin sensitivity in diabetic subjects. J C l i n Endocrinol Metab 71:1508–1518, 1990 3 . Taniguchi A, Nakai Y, Fukushima M, Kawamura H, Imura H, Nagata I, Tokuyama K: Pathogenic factors re s p o n s i- ble for glucose tolerance in patients with NIDDM. D i a b e t e s 41:1540–1546, 1992 4 . Nagasaka S, Tokuyama K, Kusaka I, Hayashi H, Rokkaku K, Nakamura T, Kawakami A, Higashiyama M, Ishikawa S, Saito T: Endogenous glucose pro d u c t i o n and glucose effectiveness in type 2 diabetic subjects derived from stable-labeled mini- mal model approach. D i a b e t e s 4 8 : 1 0 5 4 – 1060, 1999 5 . Best JD, Kahn SE, Ader M, Watanabe RM, Ni T-C, Bergman RN: Role of glucose eff e c- tiveness in the determination of glucose tol- erance. Diabetes Care 19:1018–1030, 1996 6 . Rossetti L, Giaccari A, DeFronzo RA: Glu- cose toxicity. Diabetes Care 1 3 : 6 1 0 – 6 3 0 , 1 9 9 0 7 . Taniguchi A, Nakai Y, Doi K, Fukuzawa H, Fukushima M, Kawamura H, To k u y a m a K, Suzuki M, Fujitani J, Tanaka H, Nagata I: Insulin sensitivity, insulin secretion, and glucose effectiveness in obese subjects: a minimal model analysis. M e t a b o l i s m 4 4 : 1397–1400, 1995 8 . World Health Organization: Diabetes Melli - tus: Report of a WHO Study Gro u p . G e n e v a , World Health Org., 1985 (Tech. Rep. Ser. , no. 727) 9 . Winer BJ: Statistical Principles in Experimen - tal Design. 2nd ed. New York, McGraw-Hill, 1971, p. 514–603 Late-Onset Tro g l i t a z o n e - I n d u c e d Hepatic Dysfunction R e c e n t l y, Iwase et al. (1) re p o rted a case of liver dysfunction occurring after 19 months of troglitazone therapy. Because it was thought before this re p o rt that the risk of liver dysfunction with tro g l i- tazone after 12 months was negligible, we wish to re p o rt another patient who took t roglitazone intermittently and developed hepatic dysfunction after 18 months. A 76-year-old white man with type 2 diabetes, ischemic heart disease (post angioplasty and stent placement), hyper- tension, degenerative joint disease, benign p rostatic hypert ro p h y, dyslipidemia, and g a s t roesophageal re flux disease had tro g l i- tazone 400 mg daily added to his re g i m e n of glimeperide 4 mg daily and metform i n 500 mg b.i.d. because of poor glycemic c o n t rol (HbA1 c 8.9% [normal 4–6%]). The other medicines he used were aspirin and pravastatin. After 3 months of triple oral therapy, his HbA1 c level dropped to 6.3%, and, after 6 months, to 6.0%. After 6 months, the patient’s HbA1 c began to rise: 7.1% at 9 months, 7.9% at 1 year, and 9.6% at 18 months. Liver function tests were normal until 18 months, when his aspartate amino- transferase (AST) was found to be 64 (nor- mal 0–37 U/l) and alanine aminotrans- ferase (ALT) 68 (normal 7–56 U/l). The t roglitazone regimen was discontinued, and testing for hepatitis B and C, h e m a c h romatosis, autoimmune liver dis- ease, and gallbladder disease were nega- tive. Two months after discontinuing t roglitazone, the patient’s AST and ALT had decreased to 50 and 64 U/l, and, after 3 months, had re t u rned to normal at 33 and 36 U/l, re s p e c t i v e l y. His AST and ALT have remained normal since then and he has continued to take pravastatin, aspirin, m e t f o rmin, and glimeperide. When the patient was told to discon- tinue troglitazone, he admitted that he had been taking it only interm i t t e n t l y. He esti- mated that he took the drug regularly at first, but after the first 6 months, he took the drug only once or twice weekly on aver- age. He gave the following three reasons for his lack of compliance: a lack of funds, a fear of liver disease, and a tendency to avoid taking drugs whenever possible. This case, like the case described by Iwase et al., illustrates that the hepatic dysfunction caused by troglitazone can occur after 12 months and further sup- p o rts the U.S. Food and Drug Administra- t i o n ’s current recommendation that quar- terly liver function tests should be obtained when troglitazone utilization extends beyond 1 year (2). In this case, could the onset of hepatic dysfunction have been delayed because the DIABETES CARE, VOLUME 23, NUMBER 1, JANUARY 2000 129 Letters d rug was being taken only interm i t t e n t l y after the first 6 months, and the estimated total load presented to the liver would be equivalent to the exposure at 8 months in a compliant patient? We doubt this, since t ro g l i t a z o n e ’s hepatic effects are thought to be idiosyncratic and there f o re the total e x p o s u re should be irre l e v a n t . DAVID S.H. BELL, MB FERNANDO OVALLE, MD F rom the Division of Endocrinology and Metabo- lism, Department of Medicine, School of Medicine, University of Alabama, Birmingham, Alabama. A d d ress correspondence to David S.H. Bell, MB, 1808 7th Ave. S., Birmingham, AL 35294. D.S.H.B. and F.O. have served on an advisory panel for Sankyo Parke-Davis and have received con- sulting fees, re s e a rch grant support, and honoraria for speaking engagements from Sankyo Parke-Davis. R e f e re n c e s 1 . Iwase M, Yamaguchi M, Yoshinari M, Oka- mura C, Hirahashi T, Tsuji H, Fujishima M: A Japanese case of liver dysfunction after 19 months of troglitazone tre a t m e n t . Diabetes Care 22:1382–1384, 1999 2 . Parke-Davis, Division of Wa rn e r- L a m b e rt : Rezulin Package Insert. Morris Plains, NJ, 1 9 9 9 G l y b u r i d e - I n d u c e d Hemolysis in M y e l o d y s p l a s t i c S y n d ro m e G lyburide, also known as gliben- clamide, is a widely used sulfonylure a to treat patients with type 2 diabetes. Hemolytic anemia is an extremely rare side e ffect of which there have been only a few re p o rts (1–3). We describe a patient with myelodysplastic syndrome who pre s e n t e d with glyburide-induced hemolysis. A 68-year-old man with a long history of type 2 diabetes presented with left foot cellulitis of 1 week’s duration. This patient was known to have slowly pro g re s s i v e pancytopenia for 2 years, for which no work-ups had been perf o rmed. His med- ications included the following: glyburide, 5 mg per day, which he had taken for m o re than 1 year; buformine, 150 mg per day; and boglibose, 0.6 mg per day. He was afebrile, and the physical examination was normal, except for localized cellulitis on his left foot, for which he was start e d on intravenous antibiotics. The laboratory studies revealed a white blood cell count of 3.3 1 09/ l , hemoglobin 8.4 g/dl, platelet count 138 1 09/l, reticulocyte count 3.7%, mean cor- puscle volume 96 fl, moderate anisocyto- sis, fasting plasma glucose 89 mg/dl, H b A1 c 5.2%, lactate dehydrogenase 192 IU/l, total bilirubin 0.9 mg/dl, and hapto- globin 1 1 mg/dl. Red cell glucose-6- phosphate dehydrogenase level was ade- quate. Cold agglutinin test, Ham’s test, and sugar water test were normal. Both dire c t and indirect Coombs’ tests were negative. Red cell resistance to osmolarity was mildly low (Parpart ’s method). Urinalysis demonstrated no urobilinogen. Endo- scopic studies did not reveal gastro i n t e s t i- nal bleeding. Ultrasonography of the abdomen showed no splenomegaly. The result of bone marrow aspiration was equivocal. On the basis of pre s u m p t i v e glyburide-induced hemolysis, glyburide was discontinued and the patient was switched to subcutaneous insulin on the seventh day. There a f t e r, his hemoglobin level increased to 11.1 g/dl, re t i c u l o c y t e count decreased to 1.3%, and anisocytosis d i s a p p e a red pro m p t l y. He was discharg e d with insulin therapy after 1 month in the hospital, which is when the cellulitis re s o l v e d . T h ree months later, his hemoglobin level was 10.0 g/dl and his haptoglobin remained low. Repeated bone marro w aspiration confirmed the diagnosis of myelodysplastic syndro m e . We conclude that this patient devel- oped glyburide-induced hemolysis super- imposed on red cell fragility secondary to an underlying bone marrow disord e r. T h e re have been several re p o rts of hemoly- sis caused by sulfonylureas, most of which have been considered immune-mediated (1,2,4). Our case points to the possibility that glyburide could cause hemolysis by a non–immune-medicated mechanism. It is i m p o rtant to be aware of this potential side e ffect of glyburide in light of this medica- t i o n ’s widespread prescription, even though such a side effect is rare . HIROSHI NOTO, MD KAZUHISA TSUKAMOTO, MD SATOSHI KIMURA, MD F rom the Department of Diabetes and Metabolism, Tokyo University Hospital, Tokyo, Japan. A d d ress correspondence to Hiroshi Noto, MD, D e p a rtment of Diabetes and Metabolism, Tokyo Uni- versity Hospital, 7-3-1 Hongo, Bunkyo-ku, To k y o 113-8655, Japan. E-mail: noto-tky@umin.ac.jp. R e f e re n c e s 1 . Nataas OB, Nesthus I: Immune haemolytic anaemia induced by glibenclamide in selective IgA defic i e n c y. B M J 2 9 5 : 3 6 6 – 367, 1987 2 . Abbate SL, Hoogwerf BJ: Hemolytic ane- mia associated with sulfonylurea use. D i a - betes Care 13:904–905, 1990 3 . Meloni G, Meloni T: Glyburide-induced acute haemolysis in a G6PD-defic i e n t patient with NIDDM. Br J Haematol 9 2 : 159–160, 1996 4 . Kopicky JA, Packman CH: The mechanism of sulfonylurea-induced immune hemoly- sis: case re p o rt and review of the literature . Am J Hematol 23:283–288, 1986 E ffects of Exposure at an Altitude of 3,000 m on P e rf o rmance of Glucose Meters S elf-monitoring of blood glucose is m a n d a t o ry for type 1 diabetic p a t i e n t s who participate in sports to adjust insulin dose and carbohydrate ingestion (1). Sports also include activities p e rf o rm e d at moderately high altitudes, such as hiking or skiing. Capillary blood glucose monitors (BGMs) have been shown to underestimate blood glucose values at an altitude of 2,244 m (2) and at a simulated altitude of 2,000 m with t e m p e r a t u re and humidity kept constant (3). The aim of the present study was to assess the accuracy of two BGMs at a moderately high altitude in which changes in temperature, humidity, and p O2 can result in errors in blood glucose d e t e rmination (2). Two BGMs, the LifeScan One Touch II (OT) (Ortho Diagnostics, Milpitas, CA) and the Glucometer Elite II (GE) (Bayer Diagnostics, Brussels, Belgium), were tested during a study on the effects of acute exposure at an altitude of 3,000 m and exercise on blood pre s s u re and albu- min excretion rate in six type 1 diabetic p a t i e n t s . All subjects (four men and two women) were free of disease-related com- plications and in good and stable glycemic control (GHb 6.8 ± 1.1%). All subjects gave their informed and written consent to participate in the study pro t o- col. All subjects were investigated both at 130 DIABETES CARE, VOLUME 23, NUMBER 1, JANUARY 2000 Letters sea level and after ascent by car and cable car to the Angelo Mosso Institute at Col d’Olen (2,950 m altitude), Gressoney La Trinité, Italy. At sea level and at a moderately high altitude, BGM reliability at diff e rent blood glucose levels was tested, and blood glu- cose was assessed in fasting and re s t i n g conditions at 7:00 A.M.; at 10:00 A.M. b e f o re an in-field exercise test; and imme- d i a t e l y, 5 min, and 15 min after the exer- cise stopped. Capillary glucose was simul- taneously assessed with the OT and the GE. Both of these BGMs measure capillary blood glucose through the glucose oxi- d a s e - p e roxidase reaction. BGMs were cali- brated at the beginning of each test ses- sion. A venous blood sample was simulta- neously drawn from the contralateral antecubital vein in a sodium fluoride tube, centrifuged, and stored at 20°C. Plasma glucose was assayed with the glucose oxi- dase method (GO) within 3 days. This last assessment was taken as a re f e re n c e method. Statistical analysis compare d BGM capillary glucose values and GO plasma glucose values for each blood col- lection time. Measurement linearity was tested with Pearson’s correlation coeff i- cient. The mean of the diff e rences between the BGM and GO results re p resents the mean bias between the methods with accuracy expressed as percent error (PE): PE (%) = BMG – GO 1 0 0 % G O The level of statistical significance was c o n s i d e red to be P 0 . 0 5 . The GE and OT measurements had a good correlation with plasma glucose both at moderately high altitude and at sea level. P e a r s o n ’s correlation coefficients were 0.960 and 0.981 for the GE and 0.946 and 0.985 for the OT at sea level and at moderately high altitude, re s p e c t i v e l y. Biases between plasma glucose and BGM measure m e n t s w e re as follows: for the GE, 5.9 ± 27.3 at sea level and 4.9 ± 21.6 at moderately high altitude; for the OT, 4.5 ± 32.3 at sea level and 13.3 ± 22.7 at moderately high alti- tude. At sea level, both the GE and the OT tended to underestimate glucose values (NS); at moderately high altitude, the GE tended to overestimate and the OT tended to underestimate glucose values (NS). Mean PEs between plasma glucose and BGM m e a s u rements were 0.5 (OT) and 4.1 (GE) at sea level and 6.4 (OT) and 8.1 (GE) at moderately high altitude. PE tended to be higher for both BGMs at moderately high altitude (NS). Figures 1 and 2 show the bias between the single measurements with both devices at sea level and at moderately high altitude, re s p e c t i v e l y. At moderately high altitude (Fig. 2), the tendency of the GE to o v e restimate was more evident for low ( 100 mg/dl) and intermediate (100–200 mg/dl) blood glucose values, whereas the OT tended to underestimate mainly high blood glucose values (NS). In our study, BGM perf o rmance was similar and good at sea level. At a moder- ately high altitude, a tendency to overe s t i- mate blood glucose for the GE and to u n d e restimate for the OT was observ e d . The overestimation for the GE involved mainly low ( 100 mg/dl) and interm e d i- ate (100–200 mg/dl) blood glucose val- ues. This could present a problem in the p resence of symptoms suggesting hypo- Figure 1—Relationship between plasma and capillary glucose at sea level. Figure 2—Relationship between plasma and capillary glucose at moderately high altitude. LifeScan One Touch II Glucometer Elite II LifeScan One Touch II Glucometer Elite II DIABETES CARE, VOLUME 23, NUMBER 1, JANUARY 2000 131 Letters glycemia and normal blood glucose val- ues. The OT tended to undere s t i m a t e mainly high blood glucose values, although its perf o rmance with low to i n t e rmediate values was good. The pre s- ent study assessed the accuracy of two BGMs at a moderately high altitude in which changes in temperature, humidity, and pO2 can result in errors in blood glu- cose determination (2). Our results are consistent with previous studies (2,3). The decrease in pO2 could alter the sec- ond phase of the chromogen reaction and u n d e restimate blood glucose values (4); on the other hand, an increase in atmos- pheric pre s s u re could overestimate blood glucose values (5). In our study, minimal o v e restimation by the GE at low interm e- diate blood glucose values at moderately high altitude cannot be explained by the a l t e red pO2. An increase in hematocrit, which is known to alter blood glucose m e a s u rements with BGMs (6), may also occur after prolonged exposure to high altitude or as a consequence of dehydra- tion. Although our study did not deter- mine hematocrit, the exercise test was s h o rt, and the patients were instructed to drink according to their thirst during the 3,000-m exposure; there f o re, dehydration was not likely to have occurred. In con- clusion, BGM perf o rmance is similar and good at sea level. At a moderately high altitude similar to that experienced during winter skiing or summer hiking, a ten- dency to overestimate low to norm a l blood glucose values for the GE and to u n d e restimate high blood glucose values for the OT was observed. The bias is not clinically meaningful for either BGM, both of which can be safely used by dia- betic patients during exposure to moder- ately high altitudes. Some care in the eval- uation of low and intermediate blood glu- cose values measured with the GE is n e v e rtheless re c o m m e n d e d . ORIANA PECCHIO, MD SIMONA MAULE, MD MARCO MIGLIARDI, MD MARINA TRENTO, BSC MASSIMO VEGLIO, MD F rom the Italian Alpine Club Medical Commission ( O . P.); the Department of Internal Medicine (M.T. ) , University of Turin; the S. Giovanni Battista Hospital (S.M.); and the Department of Endocrinology (M.M., M . V.), Mauriziano Umberto I Hospital, Turin, Italy. A d d ress correspondence to Dr. M. Veglio, via Mancini 15, 10131 Torino, Italy. E-mail: veglio@ o n w. n e t . R e f e re n c e s 1 . H o rton ES: Role and management of exer- cise in diabetes mellitus. Diabetes Care 1 1 : 201–211, 1988 2 . G i o rdano BP, Trash W, Hollenbaugh L, Dube WP: Perf o rmance of seven blood glucose testing systems at high altitude. Diabetes Educ 15:444–448, 1989 3 . Gautier JF, Bigard AX, Douce P, Duvallet A, Cathelinau G: Influence of simulated alti- tude on the perf o rmance of five blood glu- cose meters. Diabetes Care 1 9 : 1 4 3 0 – 1 4 3 3 , 1 9 9 6 4 . B a rnett C, Ryan F, Ballonoff L: Effect of altitude on the self monitoring of blood glucose (SMBG) (Abstract). Diabetes 3 6 (Suppl.):117A, 1987 5 . Piepmeier EH, Hammett-Stabler C, Price ME, Kemper GB, Davis MG: Atmospheric p re s s u re effects on glucose monitoring devices (Letter). Diabetes Care 1 8 : 4 2 3 – 4 2 4 , 1 9 9 5 6 . B a rreau PB, Buttery JE: Effect of hematocrit concentration on blood glucose value d e t e rmined on Glucometer II. Diabetes Care 11:116–118, 1988 C O M M E N T S A N D R E S P O N S E S Deterioration of Glycemic Contro l After Long-Te rm Treatment Wi t h Troglitazone in Nonobese Type 2 Diabetic Patients T roglitazone is an oral antidiabetic d rug used to treat type 2 diabetic patients with insulin re s i s t a n c e . Troglitazone improves overall insulin sen- sitivity in the liver and skeletal muscles, which are the largest consumers and metabolizers of glucose in the body (1–3). Recent re p o rts showed that troglitazone is also effective in nonobese type 2 diabetic patients whose hyperglycemia could not be controlled with sulfonylurea therapy (4,5). However, we aware that in some patients in whom adequate glycemic con- t rol is obtained during the first several months of troglitazone treatment, their glycemic control deteriorates several months later. We assume that two distinct g roups of type 2 diabetic patients exist who respond diff e rently to long-term administration of troglitazone, one gro u p that maintains a steady response and another group that has a decre a s i n g response after certain periods. In this s t u d y, we re t rospectively examined 20 patients with type 2 diabetes who were t reated with troglitazone for 12 months and whose HbA1 c levels had improved by 1% with troglitazone by month 6. In 8 of the 20 patients (40%), HbA1 c levels increased by 0.5% after 6–9 months despite continuous tro g l i t a z o n e t reatment (group P). In contrast, the rest of the patients experienced steady glycemic c o n t rol with 0.5% of HbA1 c flu c t u a t i o n ( g roup G). During the first 5 months, H b A1 c levels decreased from means ± SEM 9.1 ± 0.9 to 7.3 ± 0.8% in group P and fro m 8.5 ± 1.2 to 6.8 ± 0.8% in group G, re s p e c- t i v e l y. No significant diff e rences were evi- dent between the two groups re g a rding the d e c rease in HbA1 c during the first 5 months (Fig. 1). From month 6 onward, HbA1 c l e v- els in group P climbed gradually by 0.2% a month up to the baseline level at month 12, but HbA1 c levels were stable in group G t h roughout the treatment period. A signifi- cant diff e rence in HbA1 c levels was evident during months 6–12 (P 0 . 0 5 ) . Among clinical characteristics, gro u p P had a significantly lower BMI (22.5 ± 3.2 vs. 26.7 ± 4.1 kg/m2, P 0.05) and s i g n i ficantly lower fasting insulin levels (4.9 ± 3.6 9 vs. 9.3 ± 4.9 µU/ml, P 0.05). Of the 10 patients with a BMI of 25 kg/m2 (80%), 8 exhibited deteriora- tion of glycemic contro l . In this study, we re p o rt a group of patients who showed a renewed decline in glycemic control after long-term tre a t m e n t with troglitazone. These results seemed to suggest a secondary failure of tro g l i t a z o n e . Our study demonstrates that this drug is indeed useful for a long-standing obese i n s u l i n - resistant diabetes but not for a nonobese type 2 diabetes. YUKO MURASE, MD TAKANOBU WAKASUGI, MD KUNIMASA YAGI, MD HIROSHI MABUCHI, MD F rom the Department of Internal Medicine (Y. M . , T. W.), Fukui Perfectural Hospital; and the Second D e p a rtment of Internal Medicine (K.Y., H.M.), Kanazawa University, Ishikawa, Japan. A d d ress correspondence to Yuko Murase, MD, the Second Department of Internal Medicine, Kanazawa University, 13-1 Takara-machi, Kanazawa, Ishikawa 920-0934, Japan. E-mail: diabe@med. k a n a z a w a - u . a c . j p . 132 DIABETES CARE, VOLUME 23, NUMBER 1, JANUARY 2000 Letters R e f e re n c e s 1 . Suter SL, Nolan JJ, Wallace P, Gumbiner B, Olefsky JM: Metabolic effects of new oral hypoglycemic agent CS-045 in NIDDM subjects. Diabetes Care 15:193–203, 1992 2 . O’Rourke CM, Davis JA, Saltiel AR, Corn i- celli JA: Metabolic effects of troglitazone in the Goto-Kakizaki rat, a non-obese and n o rmolipidemic rodent model of nonin- sulin-dependent diabetes mellitus. M e t a b - o l i s m 46:192–198, 1997 3 . Troglitazone Study Group: The metabolic e ffects of troglitazone in non-insulin dependent diabetes (Abstract). D i a b e t e s 4 6 (Suppl. 1):149A, 1997 4 . Mori K: The effect of troglitazone in combi- nation with sulfonylurea in non-insulin dependent diabetes mellitus (Abstract). J Japan Diabetes Soc 41 (Suppl. 1):361, 1998 5 . H o rton ES, Venable TC, Whitehouse F, the Troglitazone Study Group, Ghazzi MN, Whitcomb RW: Troglitazone in combina- tion with sulfonylurea re s t o res glycemic c o n t rol in patients with type 2 diabetes. Diabetes Care 21:1462–1469, 1998Figure 1—Change from baseline in HbA1c. Values are means ± SEM.