ORIGINAL CONTRIBUTIONS. Antibodies to Glutamic Acid Decarboxylase and Diabetes Mellitus in the Multiple Risk Factor Intervention Trial

Transcription

1 American Journal of Epidemiology Vol. 140, No. 8 Copyright 1994 by The Johns Hopkins University School of Hygiene and Public Health Printed in U.S.A. All rights reserved ORIGINAL CONTRIBUTIONS Antibodies to Glutamic Acid Decarboxylase and Diabetes Mellitus in the Multiple Risk Factor Intervention Trial Paul Z. Zimmet, 1 ' 2 B. Jessica Shaten, 3 Lewis H. Kuller, 4 Merrill J. Rowley, 2 William J. Knowles, 5 and Ian R. Mackay 2 for the MRFIT Research Group Diabetes mellitus is a heterogeneous disease. The better classification of types of diabetes mellitus among adults will improve epidemiologic studies of determinants of risk factors and genetic host susceptibility. Recently, an antibody to a specific enzyme, glutamic acid decarboxylase, has been closely linked to insulin-dependent diabetes mellitus. Sera were collected at baseline between 1972 and 1974 from initially nondiabetic participants in the Multiple Risk Factor Intervention Trial. After approximately 18 years of frozen storage, the serum samples were tested for antibodies to glutamic acid decarboxylase (anti-gad) in 175 men who developed diabetes and 352 matched controls who did not develop diabetes during the 6-year follow-up. Nine of the 527 samples tested had elevated (19 or more units) titers of anti-gad. Six of the nine men with elevated anti-gad subsequently developed diabetes, and three of these six were ultimately placed on insulin therapy. These data suggest that elevated levels of anti-gad may be a prospective marker for the subsequent development of insulin-dependent diabetes mellitus. The measurement of anti-gad is relatively easy, can be performed in stored serum specimens, and may be used in epidemiologic studies to enhance the understanding of the determinants of diabetes mellitus. Am J Epidemiol 1994;140: diabetes mellitus; glutamic acid decarboxylase; insulin The specific aim of this study was to determine the prevalence of antibodies to glutamic acid decarboxylase (anti-gad) in participants in the Multiple Risk Factor Intervention Trial who subsequently devel- Received for publication November 19, 1993, and in final form July 12, Abbreviation: anti-gad, antibodies to glutamic acid decarboxylase. 1 International Diabetes Institute, Melbourne, Australia. 2 Centre for Molecular Biology and Medicine, Monash University, Melbourne, Australia. 3 Division of Biostatistics, School of Public Health, University of Minnesota, Minneapolis, MN. 4 Graduate School of Public Health, University of Pittsburgh, Pittsburgh, PA. 5 Miles Research Laboratories, West Haven, CT. Reprint requests to Dr. Paul Z. Zimmet, International Diabetes Institute, 260 Kooyong Road, Caulfield South, Victoria 3162, Australia. oped diabetes mellitus and in matched controls. The description of the trial and outcomes have been published previously (1). The participants in the Multiple Risk Factor Intervention Trial were men aged years at the time of initial screening between 1972 and They attended yearly examinations for approximately 6-7 years, and the development of diabetes was ascertained during that time. The cumulative incidence of diabetes over 5 years was 4.1 percent, with 486 incident cases observed in 11,905 men at risk at baseline (2). At present, to our knowledge, there are no prospective studies that have examined the relation between anti-gad and the risk of diabetes mellitus among adults. The stored serum samples from the Multiple 683

2 684 Zimmet et al. Risk Factor Intervention Trial provided an opportunity to do a nested case-control seroepidemiologic study of the relation between anti-gad and risk of diabetes mellitus. Diabetes mellitus is a common disorder, and when it presents in this age group, it is usually classified and treated as non-insulin-dependent diabetes mellitus (3). However, non-insulin-dependent diabetes mellitus is heterogenous (3), and while a substantial proportion of individuals are obese and hyperinsulinemic (3, 4), nonobese patients with non-insulin-dependent diabetes mellitus are often hypoinsulinemic (4-6). Many subjects in this latter category fail to maintain adequate glycemic control on conventional therapy such as diet and oral hypoglycemic agents and finally progress to insulin dependency within months or even over several years (3, 7). There is increasing evidence that many of these persons have insulin-dependent diabetes mellitus from the beginning (5-8), albeit in a slowly evolving form; this contrasts with the acute and dramatic clinical presentation of insulin-dependent diabetes mellitus usually seen in childhood and adolescence (9). Insulin-dependent diabetes mellitus, in most cases, is the result of autoimmune destruction of the pancreatic islet beta cells (10). The specific antigens that are responsible are still unknown, but the major serologic markers of insulin-dependent diabetes mellitus include islet cell antibodies (11), insulin antibodies (12), and anti-gad (13-15). We have recently described a readily applicable radioimmunoprecipitation assay for anti-gad (15) that may permit early diagnosis and better discrimination of insulin-dependent diabetes mellitus from questionable cases of non-insulin-dependent diabetes mellitus (5, 6, 16). Hagopian et al. (8) have reported similar observations. MATERIALS AND METHODS The Multiple Risk Factor Intervention Trial was a clinical trial for the primary prevention of coronary heart disease. The men were considered to be at risk for the development of coronary heart disease because of their history of cigarette smoking, serum cholesterol, and high diastolic blood pressure (1). Approximately 360,000 men were screened for entry into the trial in 22 clinical centers, and 12,866 men aged years were included. The men were randomized into two groups: Special Intervention (n = 6,428) and Usual Care (n = 6,438). A detailed description of the trial and outcomes has been published (1). The screening for eligibility for the Multiple Risk Factor Intervention Trial occurred at three visits (1). At the second screening visit, a blood sample was taken after a requested overnight fast. The blood was shipped to a central laboratory at the Institute of Medical Sciences, San Francisco, California, for biochemical analyses including levels of serum glucose and lipoproteins. An aliquot of the fasting blood serum from the second screening examination was stored at -50 to -70 C. The potential participants were given a 75-g oral glucose load at the second screening, and the serum glucose was measured 1 hour later. Men who had a clinical history of diabetes mellitus were excluded from the trial. A few participants with a fasting serum glucose of more than 140 mg/dl (7.8 mmol/liter) or a 1-hour serum glucose of more than 300 mg/dl (16.7 mmol/liter) were included in the trial, but are not included in this analysis. Despite these exclusions, there were probably a few men in the study who would have been classified as having diabetes mellitus if a 2-hour postload serum glucose rather than a 1-hour measurement had been done according to the World Health Organization (17) and National Diabetes Data Group (18) criteria. At the second screening visit, the height and weight were measured. Men with a body weight greater than 150 percent of desirable weight were excluded from the trial for which desirable weight was defined as 0.9 of the average weight for men of the same height in the National Health Examination Survey (19). Body

3 Glutamic Acid Decarboxyiase and Diabetes Mellitus 685 mass index (kg/m 2 ) was also used as a measure of obesity. A participant was considered to have a parental history of diabetes if at the second screening he indicated, in a self-administered questionnaire, that either his mother or father or both had diabetes. Autoantibodies to anti-gad were measured by a radioimmunoprecipitation assay using radiolabeled porcine brain GAD that was purified by affinity chromatography. The assay has been reported in detail elsewhere (15). The results were expressed as units derived from percentage of the radioactivity precipitated by the test sera and that precipitated by a reference serum designated to contain 100 units of activity of anti-gad. As described elsewhere (20), the coefficients of variation for interassay and intraassay replicates for a strongly positive serum were 9 and 13 percent, respectively, and the interassay coefficient of variation for a weakly positive serum was 14 percent. From results for normal sera (mean, 8.2 units; standard deviation, 2.7 units), the upper normal limit of anti-gad activity was set at 18 units (mean + 3 standard deviation) (15, 20). The serum insulin measurements were performed in the laboratory of Dr. Dorothy Becker, Children's Hospital of Pittsburgh, Pittsburgh, Pennsylvania (21). Serum insulin was measured using a standard double antibody charcoal assay. Previous studies (21) have demonstrated the stability of insulin in frozen samples over long periods. The mean serum insulin at baseline for 175 diabetics in the study was 24.8 /xu/liter with a standard deviation of 12.5 (table 1). The participants in the trial attended annual clinic visits the morning after an overnight fast. A blood sample was taken at each annual visit, and the serum glucose concentration was measured. An annual medical history was also completed. Participants were questioned about whether a physician had told them that they had diabetes mellitus and whether they were using insulin or oral hypoglycemic agents either concurrently or within the previous 12 months. At the sixth annual examination, a repeat 75-g oral glucose challenge and a 1-hour serum glucose measurement were included, as well as the fasting glucose level. No stored blood samples were available except from the second screen baseline examination (2). In this report, the onset of diabetes was defined to have occurred in 1) the first of 2 consecutive years in which fasting serum glucose values exceeded 140 mg/dl (7.8 mmol/liter); 2) the year in which the fasting serum glucose exceeded 140 mg/dl if, in the next year, the participant reported using insulin or oral hypoglycemic agents; or 3) the year in which the participant first re- TABLE 1. Characteristics of diabetes and controls in the Multiple Risk Factor Intervention Trial Diabetes Study, Diabetics First control* Second control Mean SDt Mean SD Mean SD No4 Age (years) % black BMIt (kg/m 2 ) Baseline serum glucose fasting (mg/dl) Glucose 1 hour after 75-g load Fasting insulin baseline (nu/ml) * Matched for baseline blood glucose with diabetic and secondary on body mass index, t SD, standard deviation; BMI, body mass index. $ Includes only those with sera available for analysis.

4 686 Zimmet et al. ported using insulin or oral hypoglycemic agents in the preceding 12 months. Because of the requirements for two serum glucose levels of 140 mg/dl or more, the incidence of diabetes could only be determined for 5 years of follow-up, i.e., an elevated serum glucose in both the fifth and sixth years of follow-up was required for incident diabetes in year 5 (2). The epidemiology of diabetes mellitus within the Multiple Risk Factor Intervention Trial cohort has recently been published (2). For the current nested case-control study, we identified 179 incident diabetics among both the Special Intervention and Usual Care groups as our defined cases and matched them with two controls. (Selection was determined by ability to match and by presence of stored sera.) The first control was matched by randomization date (within 30 days), study group (Usual Care or Special Intervention), clinical center (one of 22), and fasting glucose (within 5 mg/dl (0.3 mmol/liter)). If more than one control was available, the participant with a body mass index closest to that of the case was selected. The second control was matched only by randomization date, study group, and clinic. The sera from 175 of the diabetics and 177 of the controls in one group and 175 of the controls in the other group were available for analysis of anti-gad and serum insulin levels (table 1). Associations between anti-gad and the development of diabetes were assessed using Cox proportional hazards regression models, stratified by case-control triad (this method is equivalent to fitting logistic regression models for the presence of diabetes). The relative risk for developing diabetes among those with anti-gad of 19 units or more versus those with anti-gad of 18 units or less was estimated by the odds ratio after adjustment for age; age and other baseline coordinates; and age, other baseline covariates, and insulin. We do not know whether the diabetics using insulin were insulin dependent or whether some of the diabetics not using insulin at the end of the follow-up subsequently became insulin dependent. RESULTS The baseline characteristics of the 175 diabetics and of controls are shown in table 1. Their mean age was 46 years at baseline, 89 percent were white, 8 percent were black, and 3 percent were from other ethnic groups. The average baseline fasting insulin level of the 175 men who developed diabetes was 24.8 /luaiter, the level for the controls matched for blood glucose was 21.7, and that for the second controls was 18.6 (table 1). The overall prevalence of a high level (19 or more units) of anti-gad in the tested stored sera was nine of 527 subjects, or 1.7 percent. Six of the nine (66.7 percent) with a high level of anti-gad subsequently developed diabetes and, of these six, three reported the use of insulin in years 4, 5, or 6. These six comprise 3.4 percent of the 175 diabetics, while the frequency in nondiabetics was three in 352 (0.9 percent). Age-adjusted relative risk for developing diabetes was 4.38 (95% confidence interval ) for those with anti-gad of 19 or more versus those with anti-gad levels of less than 19 or missing. After further adjustment by body mass index, fasting glucose, glucose 1 hour after a 75-g load, race, and insulin, the relative risk was 3.56 (95% confidence interval ) (table 2). The characteristics of the nine men who had a raised level of anti-gad (19 units or more) is shown in table 3. Two of three men who did not develop diabetes reported a family history of diabetes. For the six men who developed diabetes, the diagnosis was made in year 1 (one case), year 2 (two cases), year 4 (two cases), and year 5 (one case). The three who developed diabetes and used insulin had the highest levels of anti- GAD 104, 90, and 89 units. The four with the highest levels of anti-gad (104, 90, 89, and 55 units) reported no parental history of diabetes, whereas three of the four with the

5 Glutamic Acid Decarboxylase and Diabetes Mellitus 687 TABLE 2. Percentage of Multiple Risk Factor Intervention Trial participants with elevated anti-gad* by diabetic status, Anti-GAD <1Oor missingt No. No. % No. % No. % Diabetes Controls * CO CM Total 527 Anti-GAD, antibodies to glutamic acid decarboxylase. t Twelve samples were missing. $ Two participants negative for diabetes had a positive family history lowest positive levels of anti-gad (19, 21, 21, and 31 units) all reported a parental history of diabetes. The three anti-gadpositive men who did not develop diabetes had the lowest levels of anti-gad (19, 21, and 21 units), and two of these reported a parental history of diabetes. Three of the six men (50 percent) with elevated anti-gad who developed diabetes were using insulin at some time between years 4 and 6. Among all men in this study who developed diabetes, 21 (12 percent) were using insulin at some time in years 4-6. Thus, among diabetics, the likelihood of using insulin was 8.4 times higher among those with elevated levels of anti- GAD compared with those who did not have elevated levels. DISCUSSION The results of this nested case-control study of the Multiple Risk Factor Intervention Trial participants has demonstrated that an elevated serum level of anti-gad precedes the clinical recognition of diabetes mellitus. The serum samples used for the analysis of anti-gad had been stored at -50 to -70 C for about 18 years. (Because cases and controls were matched by randomization date, length of storage was identical for the two groups.) Thereafter, the sera were thawed prior to shipment to Australia for the analysis for anti-gad. Thus, anti-gad are very stable in stored serum, and nested case-control studies using stored serum specimens should provide the opportunity to evaluate further the proportion of subjects with anti-gad who later develop diabetes mellitus, especially insulin-dependent diabetes. A much higher proportion of the diabetics with elevated anti-gad used insulin compared with diabetics with borderline or low levels of anti-gad, supporting the hypothesis that elevated anti-gad may be a prospective marker for insulin-dependent diabetes mellitus (13-15). A limitation of this study was the inability to determine whether incident diabetic subjects were truly insulin dependent as opposed to insulin treated, since the final classification of the exact type of diabetes is not available. The fact that three men were on insulin therapy favors a diagnosis of insulin-dependent diabetes mellitus in these three, and the other diabetics with elevated anti-gad may also have come to require insulin. The follow-up period in the Multiple Risk Factor Intervention Trial may not have been sufficient to determine insulin dependency. Evidence from other studies raises this possibility (5-8, 16). The prevalence of insulin-dependent diabetes mellitus as well as anti-gad positivity would be lower in the Multiple Risk Factor Intervention Trial than that found in other populations of persons with diabetes because the selection criteria for the Multiple Risk Factor Intervention Trial included hypertension and elevated cholesterol levels, resulting in a high proportion of obese or overweight men and favoring the selection of potential cases of non-insulindependent diabetes mellitus (2).

6 688 Zimmet et al. o CO o>1 CM N- tn jnit en Ai O> Q) lev # Q (3 i. c a c 13 art a. tlontrl c nterv o u (D 0) IE " O c 1stles of i S s o BLE 3. ing glucose il at year 6 (mg/dl) ) ast evf u_ icose seline ting glu 1 at ba: (mg/dl Fas leve CD C ISUill *o O v) 2,«Q_ ing o g ^ cn"a).<3 = in n T- ' a -Q s 1 to tn II <D d) <" GAD lev units) ± ~ -a 1 a articipar nli-gad 219 ui - n O S O i- on i- r ) Z O) (A (A (O ) O O J C O I O O O W C M T - E 7, >- s ff T3 (O II? Q -a < s An important, unanswered question is whether adult-onset diabetics who have an elevated level of anti-gad have an unusual type of diabetes or merely late-onset insulin-dependent diabetes mellitus with a slower evolution than that of the insulindependent diabetes mellitus usually diagnosed during childhood or adolescence (5 8). Patients with insulin-dependent diabetes mellitus, especially those diagnosed during childhood, usually have specific genetic polymorphisms of the class I major histocompatibility locus, especially of the DQ alpha and beta chains (22). Adult-onset diabetics with a high level of anti-gad will need to be characterized to determine whether they have the same genetic polymorphisms that specify high susceptibility to insulin-dependent diabetes mellitus. This study has shown that the measurement of anti-gad is feasible in relatively large population samples and, especially, from stored sera. Therefore, it is possible to determine the risk of development of diabetes, as well as the characteristics of adult onset diabetics with high levels of anti- GAD. Ascertainment of the characteristics of these diabetic patients with regard to insulin dependency, geographic distribution, familial aggregation or heredity, and association with specific risk factors including body mass index and fat distribution will be of considerable importance. An important limitation of this study is the relatively few men with elevated anti-gad antibody, accounting for the instability of the relative risk estimates. We used 3 standard deviations above mean to classify an individual as having high anti-gad (&: 19). At 2 standard deviations (15 or more), there would have been one more diabetic and two more controls identified. The use of anti-gad antibody measurements in epidemiologic studies of diabetes mellitus may help to define further different types of diabetes mellitus and improve our understanding of environmental and genetic determinants. The anti-gad antibody measure is not primarily a screening test for diabetes mellitus. It may improve our discrimination of the types of diabetes. The

7 Glutamic Acid Decarboxylase and Diabetes Mellitus 689 failure in the past to separate insulin-dependent from non-insulin-dependent diabetes limited earlier studies of diabetes. The number of cases in the Multiple Risk Factor Intervention Trial is small, and replication of the results in other studies is necessary. Specifically, studies of insulin-dependent adult-onset diabetics, as well as evaluations in different race, sex, and ethnic groups, will be very important. Finally, identification of persons who are likely to develop latent autoimmune diabetes as adults may be of considerable clinical significance, first, because insulin therapy is the preferred treatment and can be instituted early, and second, because there is now increasing evidence that early treatment with insulin can arrest the autoimmune process and allow preservation of beta-cell function (23). Indeed, early treatment of these persons with insulin or immunomodulatory drugs could greatly postpone the time of onset of diabetes. ACKNOWLEDGMENTS Supported by grants from the Government Employees Medical Research Fund (Dr. Zimmet), the National Institutes of Health, and National Heart, Lung, and Blood Institute contract HC The authors acknowledge the efforts of the many Multiple Risk Factor Intervention Trial investigators who collected these data. (For a complete list'of all Multiple Risk Factor Intervention Trial investigators, see JAMA 1982; 248: ). The authors thank Ray Spark for technical assistance and S. Fournel for assistance in preparation of the manuscript. REFERENCES 1. Multiple Risk Factor Intervention Trial Research Group. Multiple Risk Factor Intervention Trial: risk factor changes and mortality results. JAMA 1982;248: Shaten BJ, Kuller LH, Smith GD, et al. Risk factors for the development of type II diabetes among men enrolled in the usual care group of the Multiple Risk Factor Intervention Trial. Diabetes Care 1993;16: Harris MI, Zimmet P. Classification of diabetes mellitus and other categories of glucose intolerance. In: Keen H, DeFronzo R, Alberti KGMM, et al., eds. The international textbook of diabetes mellitus. London, England: John Wiley & Sons, 1992: Hamman RF. Genetic and environmental determinants of non-insulin-dependent diabetes mellitus (NIDDM). Diabetes Metab Rev 1992;8: Tuomi T, Groop LC, Zimmet PZ, et al. Antibodies to glutamic acid decarboxylase reveal latent autoimmune diabetes mellitus in adults with a non-insulin-dependent onset of disease. Diabetes 1993;42: Zimmet PZ, Tuomi T, Mackay IR, et al. Latent autoimmune diabetes in adults (LADA)-The role of antibodies to glutamic acid decarboxylase in diagnosis. Diabetic Med 1994;ll: Groop LC, Bottazzo GF, Doniach D. Islet cell antibodies identify latent type 1 diabetes in patients aged years at diagnosis. Diabetes 1986;35: Hagopian WA, Karlsen AE, Gottsater A, et al. Quantitative assay using recombinant human islet glutamic acid decarboxylase (GAD65) shows that 64K autoantibody positivity at onset predicts diabetes type. J Clin Invest 1993;91: Karjalainen J, Salmela P, Ilonen J, et al. A comparison of childhood and adult type I diabetes mellitus. N Engl J Med 1989;320: Hagopian W, Lernmark A. Autoimmune diabetes mellitus. In: Rose NE, Mackay IR, eds. The autoimmune diseases II. New York, NY: Academic Press, Inc., 1992: Bottazzo GF, Florin-Christensen A, Doniach D. Islet cell antibodies in diabetes mellitus with autoimmune polyendocrine deficiencies. Lancet 1974;2: Palmer JP, Asplin CM, Clemons P, et al. Insulin antibodies in insulin-dependent diabetes before insulin treatment. Science 1983;222: Baekkeskov S, Nielsen JH, Marner B, et al. Autoantibodies in newly diagnosed diabetic children immunoprecipitate human pancreatic islet cell proteins. Nature (Lond) 1982;298: Atkinson M4, MacLaren NK, Scharp DW, et al M, autoantibodies as predictors of insulindependent diabetes. Lancet 1990;335: Rowley MJ, Mackay IR, Chen Q-Y, et al. Antibodies to glutamic acid decarboxylase discriminate major types of diabetes mellitus. Diabetes 1992;41: Scott R, Willis J, Brown L, et al. Antibodies to glutamic acid decarboxylase (GAD) predict insulin deficiency in adult onset diabetes mellitus. Diabetes 1993;42(Suppl. l):220a. 17. World Health Organization. Diabetes mellitus: report of a WHO Study Group. Technical Report Series 727. Geneva, Switzerland: World Health Organization, National Diabetes Data Group. Classification and diagnosis of diabetes mellitus and other categories of glucose intolerance. Diabetes 1979; 28: Roberts J. National Center for Health Statistics. Weight by height and age of adults, US, (Vital and health statistics, Series 11, no 14). Washington, DC: US GPO, (Public Health Service publication no. 1000).

8 690 Zimmet et al. 20. Chen Q-Y, Byrne GC, Jones T, et al. Antibodies MRFIT. Ann Epidemiol 1994;4:40-5. to glutamic acid decarboxylase in Australian 22. Dorman JS, LaPorte RE, Trucco M. Genetics of children with insulin dependent diabetes mellitus diabetes. Genes and environment. (Review), and their first degree relatives. Pediatr Res 1993: Baillieres Clin Endocrinol Metab 1991;5: 34: Orchard TJ, Eichner J, Kuller LH, et al. Insulin 23. Skyler JS, Marks JB. Immune intervention in as a predictor of coronary heart disease: inter- type I diabetes mellitus. Diabetes Rev 1993;1: action with apo E phenotype. A report from