ARTICLES

Acarbose for prevention of type 2 diabetes mellitus: the STOPNIDDM randomised trial
Jean-Louis Chiasson, Robert G Josse, Ramon Gomis, Markolf Hanefeld, Avraham Karasik, Markku Laakso, for The STOP-NIDDM Trial Research Group*

Summary
Background The worldwide increase in type 2 diabetes mellitus is becoming a major health concern. We aimed to assess the effect of acarbose in preventing or delaying conversion of impaired glucose tolerance to type 2 diabetes. Methods In a multicentre, placebo-controlled randomised trial, we randomly allocated patients with impaired glucose tolerance to 100 mg acarbose or placebo three times daily. The primary endpoint was development of diabetes on the basis of a yearly oral glucose tolerance test (OGTT). Analyses were by intention to treat. Findings We randomly allocated 714 patients with impaired glucose tolerance to acarbose and 715 to placebo. We excluded 61 (4%) patients because they did not have impaired glucose tolerance or had no postrandomisation data. 211 (31%) of 682 patients in the acarbose group and 130 (19%) of 686 on placebo discontinued treatment early. 221 (32%) patients randomised to acarbose and 285 (42%) randomised to placebo developed diabetes (relative hazard 075 [95% CI 063090]; p=00015). Furthermore, acarbose significantly increased reversion of impaired glucose tolerance to normal glucose tolerance (p<00001). At the end of the study, treatment with placebo for 3 months was associated with an increase in conversion of impaired glucose tolerance to diabetes. The most frequent sideeffects to acarbose treatment were flatulence and diarrhoea. Interpretation Acarbose could be used, either as an alternative or in addition to changes in lifestyle, to delay development of type 2 diabetes in patients with impaired glucose tolerance. Lancet 2002; 359: 207277

Introduction
Type 2 diabetes mellitus is a major health problem associated with excess morbidity and mortality and results in substantial health-care costs.1 Prevalence of this disorder worldwide will double over the next 25 years.2 People who develop type 2 diabetes pass through a phase of impaired glucose tolerance. Defects in the action or secretion of insulin are the two major abnormalities leading to development of glucose intolerance.3 Resistance to insulin progressively increases when passing from normal glucose tolerance through impaired glucose tolerance to diabetes, whereas secretion of insulin gradually decreases.4 Glucose tolerance is assumed to remain normal as long as the cells can compensate for insulin resistance.3 Impaired glucose tolerance will develop only when insulin secretion fails to compensate fully for such resistance,5 resulting in postprandial hyperglycaemia. Such a mechanism could be sufficient to induce toxic effects of glucose, which further inhibit secretion and action of insulin6 and contribute to progression of impaired glucose tolerance to diabetes. Thus, any intervention in the impaired glucose tolerance phase that reduces resistance to insulin, or protects the cells, or both, should prevent or delay progression to diabetes. The -glucosidase inhibitor acarbose improves sensitivity to insulin and decreases postprandial hyperglycaemia, thereby releasing the stress on the cells.7 On the basis of these observations, we aimed to assess the effect of acarbose on conversion of impaired glucose tolerance to type 2 diabetes in the Study TO Prevent Non-insulin-dependent diabetes mellitus (STOPNIDDM) trial.8

Methods
Study design Details of the study design of the STOP-NIDDM trial have been described.8 The study was a double-blind, placebo-controlled randomised trial done in Canada, Germany, Austria, Norway, Denmark, Sweden, Finland, Israel, and Spain. Patients were recruited mainly through screening of high-risk populations, and in particular from first-degree relatives of patients with type 2 diabetes. We screened men and women aged between 40 and 70 years who had a body-mass index of between 25 and 40 kg/m2. Patients were eligible for the study if they had impaired glucose tolerancedefined as 2-h plasma glucose concentration of 78 mmol/L or greater, and less than 111 mmol/L after a 75 g glucose load.9 Patients also had to have a fasting plasma glucose concentration of 5677 mmol/L, a value that is associated with a 34-fold increase in risk of progression to diabetes.10 Eligible patients were randomly allocated to placebo or 100 mg acarbose three times daily, taken with the first bite of a meal. We used a computer program to generate the random allocation sequence, which was stratified by centre. Randomisation was done in blocks of four and six. Numbered drug containers were used to implement the random allocation process. Patients were randomised sequentially at each centre since the random code was

*Members listed at end of report Research Centre, Centre Hospitalier de lUniversit de Montral, Htel-Dieu, Department of Medicine, University of Montreal, Quebec, Canada (Prof J-L Chiasson MD); St Michaels Hospital, Department of Medicine, University of Toronto, Toronto, Ontario, Canada (Prof R G Josse MBBS); Diabetes Unit, Hospital Clinic, Universitari de Barcelona, Barcelona, Spain (Prof Ramon Gomis MD); Centre for Clinical Studies GWT, Dresden Technical University, Germany (Prof M Hanefeld DSc); Institute of Endocrinology, Chaim Sheba Medical Centre, Tel Hashomer, Israel (Prof A Karasik MD); University of Kuopio, Kuopio, Finland (Prof M Laakso MD) Correspondence to: Prof Jean-Louis Chiasson, Research Centre, Centre Hospitalier de lUniversit de Montral, Htel-Dieu, 3850 St Urbain Street, Room 8-202, Montreal, Quebec H2W 1T7, Canada (e-mail: jean.louis.chiasson@umontreal.ca)

2072

THE LANCET Vol 359 June 15, 2002 www.thelancet.com

For personal use. Only reproduce with permission from The Lancet Publishing Group.

ARTICLES

stratified by centre. Random codes were concealed in a three-part container label that was separated from the box and stored in the event that investigators needed to know the randomisation status of the patient. An independent statistician, who was a member of the data safety and quality review committee, generated the allocation sequence; enrolment and randomisation was handled at the sites. To keep the known gastrointestinal side-effects of acarbose (flatulence, diarrhoea, or abdominal cramps) to a minumum, the drug was started at 50 mg per day, and increased gradually to a maximum of 100 mg three times daily or to the maximum tolerated dose. We instructed patients on a weight-reduction or weight-maintenance diet, and encouraged them to exercise regularly. All participants met with a dietician before randomisation and yearly thereafter. Patients also completed a 3-day nutritional diary at time of eating and recorded their physical activities during the 3 days (2 weekdays, 1 weekend day) in the last month before each yearly visit. All patients were seen every 3 months by the co-ordinating nurse for pill count and distribution, documentation of adverse events, and measurement of fasting plasma glucose, and were examined by the investigator every 6 months. All patients remained in the study until the last randomised patient had been treated for 3 years; mean follow-up time was 33 years (SD 115). At the end of treatment, all patients who had not developed diabetes were followed up on placebo in a single-blind fashion for 3 months, after which outcome measures were repeated. The protocol was approved by appropriate Institutional Review Boards and every patient signed an informed consent form. An independent committee (the data safety and quality review committee), who were aware of the randomisation status, oversaw the safety of the study and the quality of data, and did the interim analysis.8 The primary endpoint was development of diabetes, defined as a plasma glucose concentration of 111 mmol/L or greater 2 h after taking 75 g glucose based on one oral glucose tolerance test (OGTT). All patients had their tolerance to 75 g oral glucose tested yearly, but, if the plasma glucose concentration was 70 mmol/L or greater during fasting at any 3-month visit, the patient was automatically scheduled for an OGTT. Patients who reached the primary endpoint remained in the study, continued on the study drug in a masked fashion, and had a yearly OGTT, unless they had to be put on additional antidiabetic drugs. Laboratory analyses The 2-h plasma glucose concentration was measured in local laboratories with the glucose oxidase or hexokinase method. The glycosylated haemoglobin (HbA1c), lipid profile, and plasma insulin were measured in two central laboratories, one in Toronto, Canada, and one in Dresden, Germany. HbA1c was measured by highperformance liquid chromatography,11 and serum insulin by a highly specific immunoradiometric assay with a twosite monoclonal antibody.12 We measured total concentrations of cholesterol, triglycerides, and HDL-cholesterol in serum enzymatically.13 If the concentration of triglicerides was less than 451 mmol/L, we calculated LDL-cholesterol mathematically.14 We cross-checked measurements of glucose, HbA1c, insulin, and lipids every 4 months for all participating laboratories.8 Statistical analysis Sample calculation has been described previously.8 On the assumption of a conversion rate of 7% per year, a 36%

risk reduction, and a drop-out rate of 10%, we estimated that 600 patients would be needed in each of the two treatment groups for a two-tailed of 005 and a 1 of 90%. Analysis was by intention to treat. We assessed the difference between the two groups with Students t test for independent samples. The primary variable was time to development of diabetes, for which we used survival analysis to compare the two treatment groups. Formal analysis was done with Coxs proportional-hazards model, which was stratified by geographic region with treatment group as a covariate. We also used the Kaplan-Meier method to calculate the probability of survival outcome. We assessed change in bodyweight from baseline to endpoint with ANCOVA. A person (WT) from the data safety and quality review committee did the interim analysis.8 The first analysis was done after all participants had been followed up for 1 year, and then every 6 months for a maximum of five analyses. Differences were judged significant if p was 000001, 0001, 0008, 0023, and 0043 for the first to fifth analyses, respectively. Role of the funding source The study sponsor (Bayer AG) provided the infrastructure and staff for data monitoring and collection, but had no role in study design, data analysis, data interpretation, or in the writing of the report.

Results
The first patient with impaired glucose tolerance was enrolled in December, 1995, and the last in July, 1998. The study was completed in August, 2001 after a mean follow-up of 33 years (SD 115). Figure 1 shows the trial profile. Of the 1429 patients who were randomised (714 to acarbose and 715 to placebo), 567 (40%) were from Canada, 382 (27%) from Germany and Austria, 334 (24%) from the Nordic countries, and 73 (5%) each from Spain and Israel. We excluded 17 patients (eight on
14 742 assessed for eligibility 13 313 excluded (did not meet eligibility criteria) 1429 randomised

714 allocated to acarbose 32 excluded 8 did not meet criteria for IGT 24 had no postrandomisation data 211 discontinued early

715 allocated to placebo 29 excluded 9 did not meet criteria for IGT 20 had no postrandomisation data 130 discontinued early

682 analysed Figure 1: Trial profile

IGT=impaired glucose tolerance

686 analysed

THE LANCET Vol 359 June 15, 2002 www.thelancet.com

2073

For personal use. Only reproduce with permission from The Lancet Publishing Group.

ARTICLES

Acarbose (n=682) Placebo (n=686) Sex Men Women Age (years) White Weight (kg) Body-mass index (kg/m2) Waist circumference (cm) Plasma glucose (mmol/L) Fasting 2-h Plasma insulin (pmol/L) Fasting 2-h Serum lipids (mmol/L) Total cholesterol HDL-cholesterol LDL-cholesterol Triglycerides Blood pressure (mm Hg) Systolic Diastolic Smoking
Data are mean (SD) or number (%).

Acarbose

Placebo

Hazard ratio (95% CI)

Favours Favours acarbose placebo

Overall 221/682 (32%) 285/686 (42%) 075 (063090) 00015

329 (48%) 353 (52%) 543 (79) 664 (97%) 876 (153) 310 (43) 1021 (117) 623 (050) 926 (106) 9934 (5764) 60637 (43746) 576 (104) 119 (032) 366 (091) 207 (110) 1314 (163) 828 (94) 79 (12%)

344 (50%) 342 (50%) 546 (79) 670 (98%) 871 (141) 309 (42) 1022 (112) 624 (053) 925 (101) 9813 (5678) 59799 (41438) 561 (099) 117 (033) 354 (090) 207 (117) 1309 (162) 820 (93) 99 (14%)

Age (years) 55 128/370 (35%) 147/354 (42%) 079 (06210) 00559 >55 92/311 (30%) 137/329 (42%) 070 (053091) 00084 Sex Men 111/329 (34%) 144/344 (42%) 077 (060099) 00382 Women 110/353 (31%) 141/342 (41%) 071 (056092) 00089 Body-mass index (kg/m2) 30 132/361 (37%) 163/368 (44%) 077 (061097) 00269 <30 89/321 (28%) 121/318 (38%) 070 (053092) 00115 00 05 10 15 20

Figure 2: Effect of acarbose on development of diabetes

Data were calculated with Coxs proportional-hazard model adjusted for age, sex, and body-mass index.

Table 1: Baseline characteristics of intention-to-treat population

acarbose and nine on placebo) because they did not have impaired glucose tolerance (2-h plasma glucose at screening was 111 mmol/L or greater, or less than 78 mmol/L). We excluded another 44 patients (24 acarbose and 20 placebo) because they had no valid postrandomisation data. Table 1 shows the baseline characteristics of the analysed patients. Mean fasting plasma glucose concentration at baseline was 62 mmol/L (SD 05), with a mean 2-h post-75 g glucose of 93 mmol/L (10). 78 (11%) of 686 patients on placebo and 57 (8%) of 682 on acarbose had a plasma glucose concentration of 70 mmol/L or greater during fasting and a concentration of less than 111 mmol/L 2 h after consumption of 75 g glucose orally, and would have been classified as having diabetes in accordance with the new diagnostic criteria if diagnosis had been confirmed by a second abnormal fasting plasma glucose on a separate day.15 629 (46%) patients had hypertension and 789 (58%) had dyslipidaemia. These patients presented with all the characteristics of the metabolic syndrome and were thus at very high risk of developing type 2 diabetes and macrovascular disease.16 We measured the concentration of autoantibodies to glutamic acid decarboxylase and
Acarbose (n=714) Placebo (n=715) All adverse events Gastrointestinal adverse events Flatulence Diarrhoea Abdominal pain Other Withdrawn consent Loss to follow-up Non-compliance Death Moved Protocol violation Other Total
Values are number (%).

tyrosine phosphatase at baseline in 728 (53%) patients using a new combined radiobinding assay;17 only 16 (2%) of the 728 were positive, suggesting that inclusion of type1 diabetes that is slow to develop was not an issue in our trial.18 Almost a quarter of patients discontinued early (figure 1), of whom almost a half (48%) discontinued during the first year. The commonest single cause of early discontinuation was gastrointestinal side-effects (table 2). All these patients were assessed at 3 years to measure endpoint variables, including an OGTT. 43 (3%) of the 1429 patients did not have any follow-up measurement of endpoints, but their inclusion did not affect the results (p=00258). Patients assigned to acarbose received a mean daily dose of 194 mg (SD 87). These patients were 25% less likely to develop diabetes than those on placebo (figures 2 and 3). Inclusion of the 17 patients who did not have impaired glucose tolerance did not affect the efficacy of acarbose (p=00016). This effect was noted at 1 year and persisted throughout the study. Mean bodyweight decreased from 876 kg (SD 152) to 871 kg (153) during the study in patients given acarbose and increased from 870 kg (141) to 873 kg (152) in those on placebo (difference 077 kg [95% CI 001154], p=00184). Analysis with Coxs proportional-hazards model, with weight change as a covariate, showed that weight loss contributed to the decreased risk of diabetes (p<00001), but that treatment with acarbose reduced risk of diabetes even after adjustment for change in weight (p=00063). The beneficial effect of acarbose was consistent irrespective of age, sex, and body-mass index (figure 2).
100 095 090 085 080 075 070 065 060 055 050 045 040
Acarbose Placebo p=00022

136 (19%)* 93 (13%) 67 (9%) 39 (5%) 23 (3%) 9 (1%) 29 (4%) 18 (3%) 10 (1%) 6 (1%) 3 (<1%) 3 (<1%) 6 (1%) 211 (30%)

37 (5%) 18 (3%) 5 (1%) 6 (1%) 4 (1%) 7 (1%) 55 (8%) 17 (2%) 4 (1%) 3 (<1%) 3 (<1%) 2 (<1%) 8 (1%) 130 (18%)

Cumulative probability

Patients at risk Acarbose 682 655 628 612 531 523 515 497 463 447 432 349 268 212 Placebo 686 671 655 640 512 505 497 470 434 427 414 331 255 208

Table 2: Reasons for premature discontinuation

Figure 3: Effect of acarbose and placebo on cumulative probability of remaining free of diabetes over time

2074

THE LANCET Vol 359 June 15, 2002 www.thelancet.com

For personal use. Only reproduce with permission from The Lancet Publishing Group.

0 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 0 90 10 0 00 11 00 12 00 13 00

Days after randomisation

ARTICLES

Glucose tolerance No data Study drug Acarbose (n=682) Placebo (n=686) 26 (4%) 18 (3%) NGT 241 (35%) 212 (31%) IGT DM

194 (28%) 221 (32%) 171 (25%) 285 (42%)

Values are number of patients (%). NGT=normal glucose tolerance. IGT=impaired glucose tolerance. DM=diabetes mellitus. *In patients who had not converted to diabetes.

Table 3: Effect of acarbose and placebo on deterioration or improvement* of glucose tolerance in patients with IGT

Body-mass index (as did weight change) significantly affected development of diabetes (p=00066), whereas age and sex did not (p=08634 and p=04385, respectively). None of the three covariates, however, modified the positive effect of acarbose treatment (p=00012, p=00015, and p=00016, respectively). 2 years after the study started, the diagnostic criteria were modified in two ways: fasting plasma glucose was decreased to 70 mmol/L, and whichever variable was used, it had to be confirmed on a separate day.15 Thus, if we use fasting plasma glucose of 70 mmol/L or greater on two consecutive visits as the criterion, 117 (17%) patients developed diabetes in the acarbose group compared with 178 (26%) in the placebo group (hazard ratio 068 [95% CI 054085], p=00010), resulting in an absolute reduction of 87% and a relative reduction of 324%. Furthermore, if we use any two positive OGTTs with a 2-h plasma glucose of 111 mmol/L or greater, 105 (15%) patients converted to diabetes in the acarbose group compared with 165 (24%) in the placebo group (hazard ratio 064 [0498108129], p=00003) for an absolute reduction of 87% and a relative reduction of 364%. Based on one abnormal plasma glucose concentration 2 h after 75 g glucose load, cumulative incidence of diabetes was 221 (324%) in the acarbose-treated group versus 285 (415%) in the placebo group. Incidence of the disorder was thus 101 cases per 1000 person-years in the acarbose group and 121 cases per 1000 person-years in the placebo group, with a risk difference of 91% over 33 years. Treatment with acarbose not only decreased progression to diabetes by 25%, but also improved glucose tolerance in patients who reverted to normal glucose tolerance (table 3). The probability of reverting to normal glucose tolerance over time was significantly
Acarbose (n=714) Adverse events Gastrointestinal Flatulence Diarrhoea Abdominal pain Dyspepsia Nausea Constipation Gastroenteritis Gastritis Anorexia General symptoms Cardiovascular Respiratory Musculoskeletal Metabolic and nutritional Nervous Urogenital Skin Haematological and lymphatic Endocrine Patients with adverse events
Values are number (%).

higher in patients on acarbose than in those on placebo (p<00001). At the end of treatment, all patients were given placebo in a single-blind fashion for 3 months, after which all outcome measures were repeated, including an OGTT. During this placebo-treatment period, incidence of diabetes in patients who had not converted (also excluding those who discontinued prematurely) was higher in the group originally assigned to acarbose (47 of 306), than in the group first randomised to placebo (21 of 199). The most common side-effects were gastrointestinal symptoms, which were more frequent in those given acarbose than in those given placebo (p<00001; table 4), but were judged to be mild to moderate in severity. The data safety and quality review committee deemed that no serious adverse events were related to the study drug.

Discussion
Our results have shown that pharmacological intervention with acarbose in patients with impaired glucose tolerance can delay progression to type 2 diabetes. The risk of progression to diabetes over 33 years was reduced by 25%. Furthermore, acarbose increases the probability that impaired glucose tolerance will revert to normal glucose tolerance over time. Although changes in lifestyle are effective in prevention of type 2 diabetes,1921 our findings and those of the US Diabetes Prevention Program (DPP)21 provide evidence that pharmacological intervention can also delay progression of impaired glucose tolerance to diabetes.22 The magnitude of the benefit of changes in lifestyle in the Finnish Diabetes Prevention Study19 and in the DPP21 was higher than the benefit we recorded with acarbose (58% vs 25%). However, we are probably underestimating the efficacy of acarbose since our results are diluted by the 211 (30%) patients randomised to acarbose who discontinued prematurely and therefore did not take the study drug throughout most of the study. Furthermore, if we analyse the results of only patients whose diabetes was confirmed by a second OGTT or by two measurements of fasting plasma glucose, the number of events decreases, and, although the absolute risk reduction (87%) does not change, the relative risk reduction is increased to 361% and 324%, respectively. Such a risk reduction is very similar to the magnitude observed with metformin in the DPP (31%) and the lifestyle change in the Da Qing study (36%).20 In the Finnish, DPP, and Chinese studies, interventions consisted of individualised intensified programmes, and as such, these trials could not be double-blind. Such bias could have affected the outcome of the trial since either diet or exercise reduced progression of impaired glucose tolerance to diabetes by 36%, whereas diet plus exercise had no additional effect. Yearly incidence of diabetes in our impaired glucose tolerance population on placebo was 124%a much higher incidence than that suggested by epidemiological data (1572%)23 and the Finnish Study (6%),19 but lower than the Da Qing study (157%).20 The higher incidence of diabetes in our impaired glucose tolerance population than in those of epidemiological data and the Finnish study was probably due, at least in part, to the requirement of a fasting plasma glucose of 56 mmol/L or greater, since this requirement has been shown to increase conversion to diabetes by a factor of 34.10 It was probably also partly due to use of only one OGTT for diagnosis of diabetes. In fact, incidence of diabetes

Placebo (n=715) 426 (60%) 196 (27%) 123 (17%) 89 (12%) 62 (9%) 39 (5%) 35 (5%) 36 (5%) 19 (3%) 3 (<1%) 444 (62%) 287 (40%) 282 (39%) 277 (39%) 231 (32%) 221 (31%) 198 (28%) 168 (24%) 42 (6%) 31 (4%) 675 (95%)

597 (83%) 486 (68%) 229 (32%) 125 (17%) 53 (7%) 36 (5%) 27 (4%) 31 (4%) 18 (3%) 1 (<1%) 415 (58%) 220 (31%) 229 (32%) 243 (34%) 222 (31%) 195 (27%) 178 (25%) 148 (21%) 32 (4%) 28 (4%) 698 (98%)

Table 4: Frequency of adverse events

THE LANCET Vol 359 June 15, 2002 www.thelancet.com

2075

For personal use. Only reproduce with permission from The Lancet Publishing Group.

ARTICLES

was reduced to 79% if a second OGTT was used to confirm the diagnosis. The higher incidence of diabetes in the Da Qing study than in other studies could be due to ethnic differences.24 Our results also suggest that acarbose effectively reduced risk of developing diabetes irrespective of age, sex, and body-mass index. Thus, all patients with impaired glucose tolerance could benefit from acarbose treatment, especially since the drug has no toxic effects.25 When we discontinued acarbose at the end of the study, incidence of diabetes increased in the group initially assigned to acarbose. Caution is needed in interpretation of this observation, since the analysis was done on a subgroup of the intent-to-treat population, which excluded those who had already converted to diabetes and those who had discontinued prematurely. Therefore, the most that can be said is that acarbose treatment should probably be continued to maintain the preventive effect of the study drug on development of diabetes. What are the possible mechanisms by which acarbose prevents diabetes? Results of many studies10,26 have shown that the higher the 2-h plasma glucose concentration after a 75 g glucose load in patients with impaired glucose tolerance, the higher the conversion rate to diabetes. Acarbose significantly decreases the postprandial rise in plasma glucose.7 This effect of acarbose could decrease toxic effects of glucose and thus delay conversion of impaired glucose tolerance to diabetes. It could also possibly explain, at least in part, why acarbose treatment was associated with an increased reversion to normal glucose tolerance. We have previously shown that acarbose could decrease resistance to insulin in patients with impaired glucose tolerance.7 Since such resistance is an important factor in development of type 2 diabetes, such an effect could have a role in delaying progression of impaired glucose tolerance to diabetes in our study. The small weight loss in the acarbose group probably contributed to delaying onset of diabetes.1921 However, acarbose was still effective after adjustment for weight loss (p=00063). The molecular mechanisms still need investigation. Our results suggest that 11 patients with impaired glucose tolerance would need to be treated for 33 years to prevent one event of development of diabetes. Lifestyle modification has already been shown to prevent type 2 diabetes. Our results show that intervention with acarbose is also effective. Whether these two treatment options can be used together remains to be determined. Nevertheless, recommendations for screening and treatment of impaired glucose tolerance should now be reassessed.
Contributors
All authors particpated in development of study design, implementation of the study, and in interpretation of the results.

Acknowledgments
The STOP-NIDDM trial was funded by an unrestricted research grant from Bayer AG. We thank the coordinating nurses and dieticians in all centres, Susanne Bordeleau-Chnier for preparing the report and illustrations, and Michael Hummel (Schwabing Hospital, Munich, Germany) for measurement of autoantibodies; Bayer AG for provision of staff and infrastructure for site monitoring and data collection, especially M Seger, S Demas, M Bungert, D Petzinna, C Pinol, G Morera, K Mattila, P Valtonen, S Tal; and Glen Saunders, who did all the statistical analysis.

References
1 2 Harris MI. Diabetes in America: Epidemiology and scope of the problem. Diabetes Care 1998; 21 (suppl): C11C14. King H, Aubert RE, Herman WH. Global burden of diabetes, 19952025: prevalence, numerical estimates, and projections. Diabetes Care 1998; 21: 141431. DeFronzo RA, Ferrannini E. Insulin resistance: a multifaceted syndrome responsible for NIDDM, obesity, hypertension, dyslipidemia, and atherosclerotic cardiovascular disease. Diabetes Care 1991; 14: 17394. Tripathy D, Carlsson M, Almgren P, et al. Insulin secretion and insulin sensitivity in relation to glucose tolerance: lessons from the Botnia study. Diabetes 2000; 49: 97580. Beck-Nielsen H, Groop LC. Metabolic and genetic characterization of prediabetic states: sequence of events leading to non-insulin-dependent diabetes mellitus. J Clin Invest 1994; 94: 171421. Rossetti L, Giaccari A, DeFronzo RA. Glucose toxicity. Diabetes Care 1990; 13: 61030. Chiasson J-L, Josse RG, Leiter LA, et al. The effect of acarbose on insulin sensitivity in subjects with impaired glucose tolerance. Diabetes Care 1996; 19: 119194. Chiasson J-L, Gomis R, Hanefeld M, Josse RG, Karasik A, Laakso M. for the STOP-NIDDM trial. An international study on the efficacy of an -glucosidase inhibitor to prevent type 2 diabetes in a population with impaired glucose tolerance: rationale, design, and preliminary screening data. Diabetes Care 1998; 21: 172025. WHO Study Group. Diabetes mellitus. WHO Tech Rep Ser 1985; 727: 7113. Saad MF, Knowler WC, Pettitt DJ, Nelson RG, Mott DM, Bennett PH. The natural history of impaired glucose tolerance in the Pima Indians. N Engl J Med 1988; 319: 150006. Cole RA, Soeldner JS, Dunn PJ, Bunn HF. A rapid method for the determination of glycosylated hemoglobin using high pressure liquid chromatography. Metabolism 1978; 27: 289301. Sobey WJ, Beer SF, Carrington CA, et al. Sensitive and specific two-site immunoradiometric assays for human insulin, proinsulin, 6566 split and 32-33 split proinsulins. Biochem J 1989; 260: 53541. Warmick GR, Bendersen J, Albers JJ. Dextran sulfate-Mg2+ precipitation procedure for quantitation of high-density lipoprotein cholesterol. Clin Chem 1982; 28: 137988. Friedwald WT, Levy RJ, Frederickson DS. Estimation of concentration of low-density lipoprotein cholesterol in plasma without the use of the preparative ultracentrifuge. Clin Chem 1972; 18: 499502. The Expert Committee on the Diagnosis and Classification of Diabetes Mellitus. Report of the expert committee on the diagnosis and classification of diabetes mellitus. Diabetes Care 1997; 20: 118397. Mykknen L, Kuusisto J, Pyrl K, Laakso M. Cardiovascular disease risk factors as predictors of type 2 (non-insulin-dependent) diabetes mellitus in elderly subjects. Diabetologia 1993; 36: 55359. Dittler J, Seidel D, Schenker M, Ziegler AG. GADIA2-combi determination as first-line screening for improved prediction of type 1 diabetes in relatives. Diabetes 1998; 47: 59297. Zimmet P, Turner R, McCarty D, Rowley M, Mackay I. Crucial points at diagnosis: type 2 diabetes or slow type 1 diabetes. Diabetes Care 1999; 22 (suppl): B59B63. Tuomilehto J, Lindstrom J, Eriksson JG, et al. Prevention of type 2 diabetes mellitus by changes in lifestyle among subjects with impaired glucose tolerance. N Engl J Med 2001; 344: 134350. Pan XR, Li G-W, Hu Y-H, et al. The Da Qing IGT and Diabetes Study: effects of diet and exercise in preventing NIDDM in people with impaired glucose tolerance. Diabetes Care 1997; 20: 53744. Knowler WC, Barrett-Connor E, Fowler SE, et al, for The Diabetes

6 7

9 10

11

12

13

14

15

16

The STOP-NIDDM Trial Research Group:

CanadaS Arndt, H Conter, G A Costain, D Finegood, H C Gerstein, J A Hunt, C J Joyce, D Lau, L A Leiter, H Lochnan, S M Ludwig, P Maheux, N W Rodger, S A Ross, E A Ryan, T M S Wolever, J-F Yale. GermanyP Algenstaedt, UR Flsch, T Linn, H-J Lddeke, S Matthaei, C May, R Paschke, K Rett, W Scherbaum, G Schernthaner. AustriaH Schmechel, E Standl, T Temelkova-Kurktschiev, H Weiss, J Ziegelasch. IsraelH Bar On, O Cohen, M Kidron, M Maislos, I Raz, P Segal, T Shoster. Nordic countriesD Andersson, M Carstensen, S Efendic, J Eriksson, B Forsn, T Forsen, K Furuseth, K Hermansen, A B Hertzenberg Faehn, A Latva-Nevala, A Pulkkinen, S Vaaler, M Vanhala, P Vanhala, M Viitaniemi, A Wajngot. Spain: I Conget, A Costa.

17

18

19

20

Conflict of interest statement

None declared

21

2076

THE LANCET Vol 359 June 15, 2002 www.thelancet.com

For personal use. Only reproduce with permission from The Lancet Publishing Group.

ARTICLES

Prevention Program Research Group. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med 2002; 346: 393403. 22 Bailey CJ. The diabetes prevention program: headline results. Br J Diabetes Vasc Dis 2001; 1: 6264. 23 Harris MI. Impaired glucose toleranceprevalence and conversion to NIDDM. Diabetic Med 1996; 13 (suppl): S9S11. 24 Chen H-S, Shaw C-K, Tseng W-P, Chen H-I, Lee M-L. Prevalence of diabetes mellitus and impaired glucose tolerance in Aborigines

and Chinese in eastern Taiwan. Diabetes Res Clin Practice 1997; 38: 199205. 25 Chiasson J-L, Josse RG, Hunt JA, et al. The efficacy of acarbose in the treatment of patients with non-insulin-dependent diabetes mellitus: a multicenter controlled clinical trial. Ann Intern Med 1994; 121: 92835. 26 Heine RJ, Nijpels G, Mooy JM. New data on the rate of progression of impaired glucose tolerance to NIDDM and predicting factors. Diabetic Med 1996; 13 (suppl): S12S14.

Clinical picture
Rare site for an intramuscular lipoma
K Shiraki, M Kamo, T Sai, R Kamo

We examined an 8-year-old girl who had a right medial canthus mass present for 3 years. Computed tomography showed the mass to be enveloped by an enlarged ring-shaped medial rectus muscle, and a biopsy specimen showed mature adipose tissue. 6 years later, because of the increased, elevated appearance of the medial canthus, surgical removal was attempted. Preoperative magnetic resonance imaging showed the mass to be enveloped by medial rectus muscle (figure, A, B), with part of the muscle entrapped inside the mass. The mass was of the same signal intensity as that of the orbital adipose tissue on T1-weighted and T2-weighted images; signal-intensity was decreased on a fat-suppressed T1weighted image (figure, C), and gadolinium enhancement was lacking. Histopathologic examination again showed mature adipose tissue infiltrating striated muscle fibres (figure, D). We know of no other report of intramuscular lipoma in a muscle as small as the external ocular muscle.
Department of Ophthalmology, Osaka City Kita Municipal Hospital (M Kamo MD, T Sai MD); Department of Ophthalmology (K Shiraki MD), Department of Dermatology (R Kamo MD), Osaka City University Graduate School of Medicine, Abeno-ku, Osaka 545-8585 Japan

THE LANCET Vol 359 June 15, 2002 www.thelancet.com

2077

For personal use. Only reproduce with permission from The Lancet Publishing Group.