Genetic Epidemiology 7:105-119 (1990)
Evolution of the Pittsburgh Studies of the Epidemiology of Insulin-Dependent Diabetes Mellitus Pittsburgh Diabetes Epidemiology and Etiology Research Group: Lewis H. Kuller, Dorothy J. Becker, Karen J. Cruickshanks, Janice S. Dorman, Mark S. Eberhardt, Allan L. Drash, Ronald E. LaPorte, Rebecca Lipton, Claudia Moy, Leslie A. O’Leary, Trevor J. Orchard, Marian Rewers, Thomas Songer, Naoko Tajima, Massimo Trucco, and Diane Wagener Department of Epidemiology, Graduate School of Public Health, University of Pittsburgh (L.H.K., J.S.D., R.E.L., R.L., C.M., L.A.O., T.J.O., T.S., M.T.), and Division of Pediatric Endocrinology, Children’s Hospital of Pittsburgh (D.J. B., A.L.D.), Pittsburgh; Department of Preventive Medicine, University of Colorado Health Science Center, Denver (K.J.C.); U.S. Centers for Disease Control, National Center for Health Statistics (M.S.E., D. W.), Hyattsville, Maryland; Department of Pediatrics, Medical Academy Poznan, Poland (M.R.); Department of fnternal Medicine, JIKEf University of School of Medicine, Tokyo, Japan (N.T.). The Pittsburgh project evaluating the epidemiology and etiology of insulin-dependent diabetes mellitus (IDDM) is currently one of the large ongoing studies of childhood diabetes. This paper traces the evolution of the project, from the initial basic epidemiologic approach in the early 1980s, to the current thrust where complex molecular genetic approaches are being incorporated into population-based research. The epidemiology models employed in the Pittsburgh project are similar to those that could be used in many areas of chronic disease research. The integration of immunogenetics into epidemiology produces a powerful approach for understanding the complex interaction of host susceptibility and environmental agents that contribute to the development of IDDM. Key words: diabetes, HLA, immunogenetics, secular trends
Received for publication March 23, 1989; revision accepted November 21, 1989. Address reprint requests to Dr. Lewis H. Kuller, Department of Epidemiology, Graduate School of Public Health, A527 Crabtree Hall, University of Pittsburgh, PA 15261.
0 1990 Wiley-Liss, Inc.
106
Kuller et al.
INTRODUCTION
The study of the epidemiology and etiology of insulin-dependent diabetes mellitus (IDDM) at the University of Pittsburgh is an example of the interaction of genetics and epidemiology to study the etiology of a disease. The studies are population-based and utilize the community of Allegheny County, Pennsylvania. Our research depends on important collaborations between epidemiologists, clinicians, and basic scientists both within Pittsburgh and worldwide. They require the application of sophisticated laboratory methodology to community epidemiological studies and have the important potential of identifying methods to prevent IDDM and its severe complications. Finally, the research program has been a major training resource in epidemiology, pediatrics, population genetics, and most recently, molecular biology. Studies of IDDM substantially changed their focus in the early and middle 1970s when a clear separation of insulin- and non-insulin-dependent diabetes (NIDDM) was firmly established. The separation was in part the result of the vastly different epidemiology of the two conditions. IDDM was separated from NIDDM by both clinical and pathologic criteria [World Health Organization, 19801. Insulin-dependent diabetes mellitus is characterized by the almost complete destruction of the insulinproducing beta cells in the pancreas leading to a marked decrease in insulin production. In contrast, with NIDDM, typically there is little beta cell destruction. Moreover, individuals with NIDDM often have normal or even “high” levels of insulin and are insulin resistant. Some individuals with NIDDM may require insulin to normalize or maintain their blood sugar level, but would not “die” immediately if insulin were taken away. Patients with IDDM, however, would rapidly expire shold they be deprived of exogenous insulin. There are marked differences in age at onset, where IDDM tends to develop in the first two decades of life, while NIDDM is generally a disease of older persons. Our initial interest in the study of IDDM was based on several observations: 1) the “incidence” of the disease appeared to be reduced in the warmer summer months [Gamble, 19801; 2) the peak age at diagnosis was during adolescence among individuals diagnosed before age 20 years [Drash et al., 19811; 3) it was less frequent in blacks than whites [LaPorte et al., 19861; and 4)there was substantial familial aggregation, but less than half of monozygotic twins were concordant for IDDM [Pyke and Nelson, 19761. We hypothesized that the etiology of insulin-dependent diabetes was due to a virus, and therefore, the research began in search of evidence of a viral etiology. We began evaluating the pattern of IDDM in the population to determine the degree to which it resembled an “infectious” disease. Our first studies were primarily descriptive epidemiology, measuring the incidence of the disease in the population and its relationship to seasonality. There was little information collected in the United States about these factors [LaPorte et al., 1981a,b; Fishbein et al., 19821. In Allegheny County we have been fortunate in that a local foundation has supported a diabetes clinic at the Children’s Hospital in Pittsburgh for many years. Sixty to 70% of all children less than 17 years of age with IDDM in the region are seen at Children’s Hospital. Because of the excellent relationships between the Children’s Hospital and the pediatricians in the community and the pediatric departments at other hospitals, it was possible in the late 1970s to identify all of the hospitalized cases of IDDM among children in Allegheny County [LaPorte et al., 1981al.
Pittsburgh IDDM Epidemiology Studies
107
Our first studies also included the development of two interlocking registries [Drash et al., 19811. One contains all children with diabetes with onset from 1965 until the present, and the second registry includes all children with IDDM admitted to Children’s Hospital in Pittsburgh. The registry from Children’s Hospital originally went back to 1965. Later, children diagnosed during 1950-1964 were added. The Children’s Hospital registry is not population-based since it includes children from outside the county. However, over half of the 1,528 children (865, or 57%) less than 20 years of age at onset in the Allegheny County registry are also in the Children’s Hospital registry. Results from the Children’s registry are likely to be generalizable to the Allegheny County population, since the characteristics of the cases seen at Children’s Hospital were quite similar to those in the county registry. Careful validation studies indicated that over ninety-five percent of all children under the age of 20 years with IDDM diagnosed in the county are included in our registries [LaPorte eta]. , 1981al. The incidence of IDDM was higher in the winter than in the spring or summer, which is consistent with previous research. Variation in time of onset was substantially greater in boys than in girls and in the younger, 5-9 year-olds, as compared to the 10-14 year age group [Fishbein et al., 19821. The incidence of IDDM was substantially higher in whites than in blacks [LaPorte et al., 1981a; Table I]. There was, however, relatively little difference in the incidence between boys and girls. The age at onset of boys as compared to girls appeared to differ. In girls the peak age at onset was 9-12 years of age, while in boys, it was 12-16 years of age (Fig. 1). This strongly suggested a relationship between sexual maturation or growth and the risk of IDDM [LaPorte et al., 1981al. We could find no evidence that there was a socioeconomic gradient in the incidence of IDDM [LaPorte et al., 19861. Thus, the lower incidence in blacks probably was not the result of lower socioeconomic status. We could not demonstrate any significant time trends in the incidence of diabetes, at least between 1965 and 1985, for either blacks or whites, boys or girls [Rewers et al., 19891. Measurement of the secondary attack rate of IDDM within families and possible relationships with genetic and environmental factors was the next step in our studies [Wagener et al., 1982a,b]. Family history data were initially collected on 1,203 consecutive admissions of IDDM cases to the Children’s Hospital of Pittsburgh between 1965 and the end of 1980. Complete data were available on a total of 1,143 families, and these were included in the analysis. The cumulative risk to the siblings of an insulin-
TABLE I. Agekacehex-Specific IDDM Annual Incidence Rate per 100,000 Persons and 95% Confidence Interval, Allegheny County, Pennsylvania, 1965 to 1980* White males rate (CI)
White females rate (CI)
Non-white males rate (CI)
Non-white females rate (CI)
5-9 10-14 15-19
8.2 (1.8) 17.8 (2.8) 21.2 (2.9) 16.7 (2.6)
Age-adjusted rate
18.1 (1.3)
6.8(1.9) 16.6 (2.8) 23.7 (3.1) 9.8 (2.0) 14.8 (1.3)
4.4 (3.8) 6.4 (4.5) 11.7 (6.0) 11.9 (6.3) 8.9 (2.7)
4.4 (3.9) 5.8 (4.3) 18.2 (7.4) 10.6 (5.7) 10.3 (2.9)
Age, years
Evolution of the Pittsburgh Studies of the Epidemiology of Insulin-Dependent Diabetes Mellitus Pittsburgh Diabetes Epidemiology and Etiology Research Group: Lewis H. Kuller, Dorothy J. Becker, Karen J. Cruickshanks, Janice S. Dorman, Mark S. Eberhardt, Allan L. Drash, Ronald E. LaPorte, Rebecca Lipton, Claudia Moy, Leslie A. O’Leary, Trevor J. Orchard, Marian Rewers, Thomas Songer, Naoko Tajima, Massimo Trucco, and Diane Wagener Department of Epidemiology, Graduate School of Public Health, University of Pittsburgh (L.H.K., J.S.D., R.E.L., R.L., C.M., L.A.O., T.J.O., T.S., M.T.), and Division of Pediatric Endocrinology, Children’s Hospital of Pittsburgh (D.J. B., A.L.D.), Pittsburgh; Department of Preventive Medicine, University of Colorado Health Science Center, Denver (K.J.C.); U.S. Centers for Disease Control, National Center for Health Statistics (M.S.E., D. W.), Hyattsville, Maryland; Department of Pediatrics, Medical Academy Poznan, Poland (M.R.); Department of fnternal Medicine, JIKEf University of School of Medicine, Tokyo, Japan (N.T.). The Pittsburgh project evaluating the epidemiology and etiology of insulin-dependent diabetes mellitus (IDDM) is currently one of the large ongoing studies of childhood diabetes. This paper traces the evolution of the project, from the initial basic epidemiologic approach in the early 1980s, to the current thrust where complex molecular genetic approaches are being incorporated into population-based research. The epidemiology models employed in the Pittsburgh project are similar to those that could be used in many areas of chronic disease research. The integration of immunogenetics into epidemiology produces a powerful approach for understanding the complex interaction of host susceptibility and environmental agents that contribute to the development of IDDM. Key words: diabetes, HLA, immunogenetics, secular trends
Received for publication March 23, 1989; revision accepted November 21, 1989. Address reprint requests to Dr. Lewis H. Kuller, Department of Epidemiology, Graduate School of Public Health, A527 Crabtree Hall, University of Pittsburgh, PA 15261.
0 1990 Wiley-Liss, Inc.
106
Kuller et al.
INTRODUCTION
The study of the epidemiology and etiology of insulin-dependent diabetes mellitus (IDDM) at the University of Pittsburgh is an example of the interaction of genetics and epidemiology to study the etiology of a disease. The studies are population-based and utilize the community of Allegheny County, Pennsylvania. Our research depends on important collaborations between epidemiologists, clinicians, and basic scientists both within Pittsburgh and worldwide. They require the application of sophisticated laboratory methodology to community epidemiological studies and have the important potential of identifying methods to prevent IDDM and its severe complications. Finally, the research program has been a major training resource in epidemiology, pediatrics, population genetics, and most recently, molecular biology. Studies of IDDM substantially changed their focus in the early and middle 1970s when a clear separation of insulin- and non-insulin-dependent diabetes (NIDDM) was firmly established. The separation was in part the result of the vastly different epidemiology of the two conditions. IDDM was separated from NIDDM by both clinical and pathologic criteria [World Health Organization, 19801. Insulin-dependent diabetes mellitus is characterized by the almost complete destruction of the insulinproducing beta cells in the pancreas leading to a marked decrease in insulin production. In contrast, with NIDDM, typically there is little beta cell destruction. Moreover, individuals with NIDDM often have normal or even “high” levels of insulin and are insulin resistant. Some individuals with NIDDM may require insulin to normalize or maintain their blood sugar level, but would not “die” immediately if insulin were taken away. Patients with IDDM, however, would rapidly expire shold they be deprived of exogenous insulin. There are marked differences in age at onset, where IDDM tends to develop in the first two decades of life, while NIDDM is generally a disease of older persons. Our initial interest in the study of IDDM was based on several observations: 1) the “incidence” of the disease appeared to be reduced in the warmer summer months [Gamble, 19801; 2) the peak age at diagnosis was during adolescence among individuals diagnosed before age 20 years [Drash et al., 19811; 3) it was less frequent in blacks than whites [LaPorte et al., 19861; and 4)there was substantial familial aggregation, but less than half of monozygotic twins were concordant for IDDM [Pyke and Nelson, 19761. We hypothesized that the etiology of insulin-dependent diabetes was due to a virus, and therefore, the research began in search of evidence of a viral etiology. We began evaluating the pattern of IDDM in the population to determine the degree to which it resembled an “infectious” disease. Our first studies were primarily descriptive epidemiology, measuring the incidence of the disease in the population and its relationship to seasonality. There was little information collected in the United States about these factors [LaPorte et al., 1981a,b; Fishbein et al., 19821. In Allegheny County we have been fortunate in that a local foundation has supported a diabetes clinic at the Children’s Hospital in Pittsburgh for many years. Sixty to 70% of all children less than 17 years of age with IDDM in the region are seen at Children’s Hospital. Because of the excellent relationships between the Children’s Hospital and the pediatricians in the community and the pediatric departments at other hospitals, it was possible in the late 1970s to identify all of the hospitalized cases of IDDM among children in Allegheny County [LaPorte et al., 1981al.
Pittsburgh IDDM Epidemiology Studies
107
Our first studies also included the development of two interlocking registries [Drash et al., 19811. One contains all children with diabetes with onset from 1965 until the present, and the second registry includes all children with IDDM admitted to Children’s Hospital in Pittsburgh. The registry from Children’s Hospital originally went back to 1965. Later, children diagnosed during 1950-1964 were added. The Children’s Hospital registry is not population-based since it includes children from outside the county. However, over half of the 1,528 children (865, or 57%) less than 20 years of age at onset in the Allegheny County registry are also in the Children’s Hospital registry. Results from the Children’s registry are likely to be generalizable to the Allegheny County population, since the characteristics of the cases seen at Children’s Hospital were quite similar to those in the county registry. Careful validation studies indicated that over ninety-five percent of all children under the age of 20 years with IDDM diagnosed in the county are included in our registries [LaPorte eta]. , 1981al. The incidence of IDDM was higher in the winter than in the spring or summer, which is consistent with previous research. Variation in time of onset was substantially greater in boys than in girls and in the younger, 5-9 year-olds, as compared to the 10-14 year age group [Fishbein et al., 19821. The incidence of IDDM was substantially higher in whites than in blacks [LaPorte et al., 1981a; Table I]. There was, however, relatively little difference in the incidence between boys and girls. The age at onset of boys as compared to girls appeared to differ. In girls the peak age at onset was 9-12 years of age, while in boys, it was 12-16 years of age (Fig. 1). This strongly suggested a relationship between sexual maturation or growth and the risk of IDDM [LaPorte et al., 1981al. We could find no evidence that there was a socioeconomic gradient in the incidence of IDDM [LaPorte et al., 19861. Thus, the lower incidence in blacks probably was not the result of lower socioeconomic status. We could not demonstrate any significant time trends in the incidence of diabetes, at least between 1965 and 1985, for either blacks or whites, boys or girls [Rewers et al., 19891. Measurement of the secondary attack rate of IDDM within families and possible relationships with genetic and environmental factors was the next step in our studies [Wagener et al., 1982a,b]. Family history data were initially collected on 1,203 consecutive admissions of IDDM cases to the Children’s Hospital of Pittsburgh between 1965 and the end of 1980. Complete data were available on a total of 1,143 families, and these were included in the analysis. The cumulative risk to the siblings of an insulin-
TABLE I. Agekacehex-Specific IDDM Annual Incidence Rate per 100,000 Persons and 95% Confidence Interval, Allegheny County, Pennsylvania, 1965 to 1980* White males rate (CI)
White females rate (CI)
Non-white males rate (CI)
Non-white females rate (CI)
5-9 10-14 15-19
8.2 (1.8) 17.8 (2.8) 21.2 (2.9) 16.7 (2.6)
Age-adjusted rate
18.1 (1.3)
6.8(1.9) 16.6 (2.8) 23.7 (3.1) 9.8 (2.0) 14.8 (1.3)
4.4 (3.8) 6.4 (4.5) 11.7 (6.0) 11.9 (6.3) 8.9 (2.7)
4.4 (3.9) 5.8 (4.3) 18.2 (7.4) 10.6 (5.7) 10.3 (2.9)
Age, years
