Diabetes Incidence in Rochester, Minnesota Burke et al. Impact of Case Ascertainment on Recent Trends in Diabetes Incidence in Rochester, Minnesota

Transcription

1 American Journal of Epidemiology Copyright 2002 by the Johns Hopkins Bloomberg School of Public Health All rights reserved Vol. 155, No. 9 Printed in U.S.A. Diabetes Incidence in Rochester, Minnesota Burke et al. Impact of Case Ascertainment on Recent Trends in Diabetes Incidence in Rochester, Minnesota James P. Burke, 1 Peter O Brien, 1 Jeanine Ransom, 1 Pasquale J. Palumbo, 2 Eva Lydick, 3 Barbara P. Yawn, 4 L. Joseph Melton III, 1 and Cynthia L. Leibson 1 The authors previously reported an increase in the incidence of diabetes mellitus among Rochester, Minnesota, residents during the period This study provides updated rates from data collected through Trends in diabetes surveillance, i.e., the proportion of residents who had a blood glucose measurement in each year between 1987 and 1994, are also provided. The authors reviewed medical records to identify residents aged 30 years or more who first met National Diabetes Data Group criteria for diabetes between January 1, 1970, and December 31, Age- and sex-adjusted incidence rates were calculated for successive quinquennia (5-year periods), and Poisson regression was used to test for an effect of calendar year; calendar period ( vs ) was added to assess whether the association with calendar year varied in the most recent quinquennium. Altogether, 1,992 Rochester residents first met National Diabetes Data Group criteria for diabetes between 1970 and The age-adjusted incidence per 100,000 person-years increased 67% for males (7.0 vs ) and 42% for females (225.4 vs ) between and Calendar year (p < 0.001) and calendar period (p = 0.0) were significant, suggesting that rates accelerated during The proportion of residents with at least one blood glucose measurement per year was unchanged (p = 0.181) from 1987 to 1994, while the incidence of diabetes increased (p = 0.033). Thus, the authors conclude that the increase in diabetes incidence accelerated over the last quinquennium and was not due to increased surveillance. Am J Epidemiol 2002;155: diabetes mellitus, non-insulin-dependent; incidence Received for publication April 18, 2001, and accepted for publication November 27, Abbreviations: NDDG, National Diabetes Data Group; NHIS, National Health Interview Survey. 1 Department of Health Sciences Research, Mayo Clinic, Rochester, MN. 2 Department of Medicine, Mayo Clinic, Scottsdale, AZ. 3 SmithKline Beecham Pharmaceuticals, Philadelphia, PA. 4 Olmsted Medical Center, Rochester, MN. Reprint requests to Dr. James P. Burke, Department of Health Sciences Research, Harwick 6, Mayo Clinic, 200 First Street SW, Rochester, MN ( jburke@mayo.edu). A number of studies have shown that the prevalence of diabetes is increasing in the United States (1, 2), but trends in prevalence fail to distinguish between increases in relative survival and increases in the rate at which new cases of disease develop. For purposes of devising and implementing appropriate intervention strategies, it is necessary to make such a distinction. Unfortunately, there are few opportunities to evaluate trends in diabetes incidence. Data from the National Health Interview Survey (NHIS) indicated no change in the incidence of diabetes between 1968 and 1992 (2), but a subsequent report found a 48 percent increase between 1980 and 1994 (3). The NHIS identifies incident cases annually by asking respondents whether they were first assigned a clinical diagnosis of diabetes in the previous year. This approach is problematic because of limitations of self-reporting and changes in clinical criteria over time. To address these limitations, researchers in the San Antonio Heart Study obtained data on glucose levels at baseline and follow-up for Mexican Americans and non- Hispanic Whites residing in San Antonio, Texas. They found a threefold increase in the incidence of type 2 diabetes over the subsequent 7 8 years between 1979 and 1988, although the trend reached statistical significance only among Mexican Americans (4). Several limitations of the NHIS data were also addressed with retrospective review of laboratory glucose values in the predominately White population of Rochester, Minnesota. The Rochester study applied standardized glycemic criteria for diabetes and found that the age- and sex-adjusted incidence of diabetes increased significantly between and (5). Although a comparison of patient characteristics as of the incidence date suggested that the rise in diabetes incidence was not an artifact of earlier detection of milder cases (5, 6), the question of whether observed trends in incidence were accompanied by increases in diabetes ascertainment and/or detection remains unanswered. In the present study, we updated the Rochester incidence rates from through the year 1994 to test whether the previously observed increase in diabetes incidence has continued at the same rate or has accelerated, leveled off, or declined. In addition, we examined trends in passive sur- 859

2 860 Burke et al. veillance, i.e., the proportion of all local residents aged 30 years or more who had blood glucose values measured over this time period, which might explain temporal changes in the observed incidence of diabetes. MATERIALS AND METHODS Epidemiologic studies of diabetes are possible in Rochester (which had a population of 85,806 in the 2000 US Census) because the community is relatively isolated from other metropolitan areas and is home to few medical care providers, primarily only the Olmsted Medical Center and the Mayo Clinic, one of the world s largest private medical centers. Therefore, nearly all of the medical care received by local residents is provided by either the Mayo Clinic, together with its two affiliated hospitals, or the Olmsted Medical Center, with its one affiliated hospital. In addition, every Mayo patient since 1907 has been assigned a unique identifier, and all information obtained from every Mayo contact is contained within a single dossier for each patient (7). This includes hospital inpatient and outpatient care, emergency room visits, office and nursing home visits, and all laboratory results, as well as death certificate and autopsy information. The diagnoses assigned at each visit are coded and entered into continuously updated computer files. Under the auspices of the Rochester Epidemiology Project, the diagnostic index and medical-record linkage were expanded to include the small number of other providers of medical care to local residents, primarily the Olmsted Medical Center. The result was linkage of all inpatient and outpatient medical records from all sources of medical care available to and utilized by members of this geographically defined population (8). The resources of the Rochester Epidemiology Project were used to construct the Rochester incident diabetes cohort. As described elsewhere (5, 9), case assignment was determined through retrospective review of each patient s complete (hospital and ambulatory) medical records by trained nurse abstractors, under the direction of an endocrinologist (P. J. P.). The records were reviewed for all laboratory glucose values (available from 1930 to the present) and for evidence of any antidiabetes medication taken over the entire duration of residence in the community. Methods of ascertainment were consistent throughout the study period. Standardized case criteria were applied using National Diabetes Data Group (NDDG) recommendations i.e., two consecutive fasting plasma glucose measurements of 140 mg/dl (7.8 mmol/liter) or both 1- and 2-hour glucose measurements of 200 mg/dl (11.1 mmol/liter) obtained during a standard oral glucose tolerance test (10). Using the method of West (11), adjustments were made for temporal changes in laboratory methods. Persons who failed to meet the above glycemic criteria but used oral antidiabetes agents or insulin for at least 2 weeks or until death also qualified as cases. The abstractor determined the date on which each individual first met the criteria and verified Rochester residency on that date. Case subjects were required to have been a Rochester resident 1 year prior to diagnosis. The study was approved by the institutional review boards of the Mayo Clinic and the Olmsted Medical Center, and in accordance with a Minnesota state statute (8), individuals who refused authorization for use of their medical records in research (n 72) were excluded. Although laboratory glucose values have been available in the medical records since 1930, they were not available electronically prior to Because it was not feasible to manually review the hard-copy records of all Rochester residents, the review for all years was limited to potential cases. Potential cases were identified by generating a list from the Rochester Epidemiology Project diagnostic index of all Rochester residents ever assigned any diabetes-related diagnosis (e.g., elevated glucose level, hyperglycemia, impaired glucose tolerance, rule out diabetes, diabetic retinopathy, etc.) since 1945 (9). This preliminary step recognized that diabetes is a chronic progressive condition and was based on the assumption that individuals whose laboratory glucose values met NDDG criteria would, at some point in their medical history, receive a diagnosis of diabetes or a diabetes-like condition, e.g., elevated glucose. This assumption was examined in a recent study that identified all 7,192 Rochester residents who died on or after age 30 years during the time period (12); the median duration of medical history available for review was 43 years (interquartile range, years), and 25 percent (n 1,788) had been assigned at least one of the diabetes diagnoses under consideration. The likelihood of such a diagnosis did not change as a function of calendar year of death, providing confidence that the proportion of residents considered as potential cases in constructing the Rochester incident diabetes cohort was high and stable over time. Analysis of temporal trends in the present study was limited to adult members of the incident cohort i.e., all identified individuals who first met NDDG criteria between January 1, 1970, and December 31, 1994, and who were aged 30 years or more and residing in Rochester at the time they first met the criteria. Characteristics of the incident cases were summarized for successive quinquennia ( , , , , and ) using descriptive statistics. Incidence rates were calculated for each sex and age group (30 44, 45 54, 55 64, 65 74, 75 84, and 85 years) for successive quinquennia, with newly diagnosed cases used as the numerator and Rochester person-years at risk as the denominator. The denominator was based on decennial census data with linear interpolation for intercensus years (13). Incidence rates were age- and sex-adjusted by the direct method to the White population of the United States as of Poisson regression was used to evaluate associations between incidence and age, sex, and calendar year of incidence. The significance of two-way interaction terms and higher-order polynomials was examined. To determine whether incidence of diabetes in the latest quinquennium increased at a faster rate, continued to increase at the same rate, leveled off, or declined, we added calendar period ( vs ) to the final model as an independent variable. If the period variable contributed significantly to the model, this would indicate that there was a change in the incidence trend in the last quinquennium compared with the earlier time period.

3 Diabetes Incidence in Rochester, Minnesota 861 To address whether temporal trends in diabetes incidence were accompanied by changes in surveillance, we used Poisson regression to test for a significant effect of calendar year on the percentage of local residents who had a blood glucose measurement in each year during the period As we noted above, laboratory glucose values were not available electronically for the entire population prior to For comparison purposes, we determined whether there was an increase in diabetes incidence over this same time period by limiting the Poisson regression analysis of incidence described above to the calendar years and testing for a significant effect of calendar year. RESULTS A total of 1,992 Rochester residents aged 30 years or more were identified as first meeting NDDG criteria for diabetes between January 1, 1970, and December 31, Characteristics of the cases at the time of identification are provided for successive quinquennia in table 1. In particular, glucose values at the time of diabetes identification did not change significantly over time. In addition, the proportion of cases who were male did not change significantly over time, but there was a significant decline in age at identification among both males ( p 0.028) and females ( p < 0.001) that was most apparent in the latest quinquennium. The incidence of diabetes in Rochester during the period is presented by sex and age group in table 2. There was an increase in diabetes incidence over time that was most apparent at younger ages. In males, the ageadjusted incidence rate increased 67 percent, from 7.0 per 100,000 person-years in to per 100,000 person-years in In females, the age-adjusted incidence rate increased 42 percent, from per 100,00 person-years to per 100,000 person-years, over the same time period. The overall age- and sex-adjusted incidence of diabetes increased 54 percent, from per 100,000 person-years in to per 100,000 person-years in The results of Poisson regression analysis of incidence rates for the period revealed that diabetes incidence rates differed between males and females ( p 0.001) and increased with increasing age ( p < 0.001). However, the effect of age was nonlinear ( p < for age squared), and it differed between the sexes ( p < for the interaction between sex and age and p for the interaction between sex and age squared). There was also a significant increase in diabetes incidence with increasing calendar year ( p < 0.001). The absence of any significant interaction between calendar year and sex suggests that temporal trends were similar for males and females. However, consistent with the findings from table 2, the temporal trends were not constant across all age groups ( p < for the interaction between age and calendar year). To facilitate interpretation of this complex model, we show in figure 1 the model-predicted incidence of diabetes for males and females at three ages (50, 60, and 70 years) and for the calendar years 1970 and Within each calendar year, incidence rates were higher for males than for females, and the between-sex difference was greater at older ages. Incidence rates increased with each decade of age, but the difference in incidence was greater between ages 50 and 60 years than between ages 60 and 70 years. At each age and for both sexes, the incidence rates were higher in 1990 than in The increase in incidence between 1990 and 1970 was greatest at younger ages. To determine whether the incidence of diabetes in the last quinquennium had increased at a faster rate, increased at the same rate, leveled off, or decreased, we included calendar period ( vs ) as an independent factor in the previous Poisson regression model. The effect of calendar period was positive and significant ( p 0.0), indicating that the increase in incidence occurred at a faster rate during the latest quinquennium. TABLE 1. Characteristics of diabetic residents of Rochester, Minnesota, aged 30 years at the time they met National Diabetes Data Group criteria for diabetes mellitus, by 5-year calendar period, Quinquennium (n = 273) (n = 290) (n = 336) (n = 460) (n = 633) 50.6 Male sex (%) age (years) Males* Females* glucose level (mg/dl) Males Females * p < 0.05 for trend over the five quinquennia., standard deviation.

4 862 Burke et al. TABLE 2. Incidence of diabetes mellitus among Rochester, Minnesota, residents aged 30 years, by age group, sex, and time period, Quinquennium Sex and age group (years) Rate 95% CI No. Rate 95% CI No. Rate 95% CI No. Rate 95% CI No. Rate 95% CI No. 86.7, , , , , , , , , , , , , , 1, , , , , , , , , , , , Men 30 44* 45 54* * , , , , , Subtotal 99.2, , , , , , , , , , , , , , , , , , , , , , , , , Women 30 44* 45 54* 55 64* , , , , , , , , , , Subtotal Total * p < 0.05 for temporal trend. Per 100,000 person-years. CI, confidence interval. Directly age-adjusted to the 1980 US White population. Directly age- and sex-adjusted to the 1980 US White population. Figure 2 shows the age-adjusted incidence of diabetes in both males and females for the time period , along with the age-adjusted proportions of local residents who had at least one blood glucose measurement in each year during the same time period. Over this time period, there was no significant effect of calendar year ( p 0.181), adjusted for age and sex, on the percentage of local residents with at least one laboratory glucose measurement taken at the Mayo Clinic or the Olmsted Medical Center. The percentage of local residents with at least one blood glucose measurement in each year averaged approximately 37 percent annually for males and 44 percent annually for females. In each year, 90 percent of individuals with at least one blood glucose determination had three or fewer measurements per year; the proportion with only one value was unchanged over time (i.e., 71 percent, 69 percent, 66 percent, 69 percent, 70 percent, 69 percent, 65 percent, and 70 percent, successively, in the years ). However, during the same time period, there was a significant effect of calendar year ( p 0.033), adjusted for age and sex, on diabetes incidence. This indicates that while surveillance did not change over this time period, the incidence of diabetes increased significantly. DISCUSSION A continually rising trend in the incidence of diabetes was observed among individuals aged 30 years or more in the Rochester, Minnesota, population from 1970 through This trend was significant in both sexes and was greatest in younger age groups. Further analyses indicated that the trend in Rochester accelerated in the last quinquennium studied, , as compared with the previous time period, The recent rise in diabetes incidence in Rochester between 1970 and 1994 may have two possible explanations. One is an increase in surveillance over this time period. A second explanation is that there was an actual increase in the rate of diabetes, probably due to changes in environmental factors (e.g., obesity). The first possibility is improbable, since the proportion of individuals receiving at least one blood glucose measurement each year did not increase. Thus, increased surveillance is an unlikely explanation for the observed increase in diabetes incidence. Instead, a possible explanation for the rise in diabetes incidence may be the marked increase in obesity in the general population, which has been observed in a number of studies (4, 14, 15). In a previous article (16), we provided data on body mass index (weight (kg)/height (m) 2 ) from a prospective survey of Rochester residents aged 45 years in A subsequent survey of Rochester residents was performed in In both surveys, weight and height were measured in a sample of the population identified by random digit dialing. There was a significant increase in mean body mass index between 1986 and 1996 for both male and female residents. The difference was most apparent in the younger age groups (Dr. Irene Meissner, Mayo Clinic, personal communication, 2001). These observations regarding temporal increases in body mass index within the Rochester popula-

5 Diabetes Incidence in Rochester, Minnesota 863 FIGURE 1. Predicted incidence of adult-onset diabetes mellitus among Rochester, Minnesota, males and females aged 50, 60, and 70 years for 1970 and tion generally are consistent with our recent report of temporal increases in body mass index among members of the Rochester diabetes cohort, as well as with the age-specific temporal increases in the rate of new cases of diabetes observed in this population (see figure 1). Other than the studies in Rochester, there have been few opportunities, using standardized glycemic criteria, to examine trends in the incidence of diabetes in this country. One such study in San Antonio, Texas, found an increase in the incidence of type 2 diabetes from 1979 through 1988 in Mexican Americans and non-hispanic Whites who participated in the San Antonio Heart Study (4). Participants in the study were aged years and were randomly selected from low-, middle-, and high-income neighborhoods in San Antonio. Participants who were nondiabetic at baseline and returned for follow-up 7 8 years later were examined for diabetes. Standardized criteria for identifying cases of diabetes were used at both examinations. The percentage of individuals who developed type 2 diabetes in the subsequent 7- to 8- year follow-up period increased significantly, from 5.7 percent for Mexican Americans first enrolled in 1979 to 15.7 percent for those first enrolled in In non-hispanic Whites, there was a borderline-significant increase from 2.6 percent to 9.4 percent over the same time period. This trend FIGURE 2. Age-adjusted incidence of adult-onset diabetes mellitus and age-adjusted proportion of the population with at least one blood glucose measurement in each year, Rochester, Minnesota,

6 864 Burke et al. remained significant for Mexican Americans and borderlinesignificant for non-hispanic Whites after adjustment for a number of baseline risk factors, including obesity. The diabetes case criteria in both the San Antonio Heart Study and the present investigation were standardized over time, although we did not prospectively examine the population as was done in the San Antonio study. These two studies differ from almost all other reports of secular trends in diabetes incidence, in which case identification is based on a self-report of having received a diagnosis of diabetes from a clinician in the past year. Data from the sequential NHIS surveys revealed little change between 1968 and 1992 in the annual rate of noninstitutionalized US adults aged 18 years reporting having received a first diagnosis of diabetes in the previous year (2). These trends differ from a subsequent NHIS report that revealed a significant increase in both crude and age-adjusted diabetes incidence rates for the time period (17). The disparate trends could reflect either a sudden increase in diabetes prevalence in or a failure to detect an increase in the earlier NHIS because of the 1979 introduction of more stringent NDDG recommendations for assigning a clinical diagnosis of diabetes (9, 19). The extent to which the present study findings are generalizable to the entire US population is limited by the fact that the Rochester population was 95 percent White during the years of the study. Our study was also limited by the fact that, to minimize misclassification and to ensure a standardized definition throughout the study period, the criteria we employed for ascertaining incident diabetes cases were intentionally strict. As a result, individuals whose glucose values never met NDDG criteria but did meet more recently introduced American Diabetes Association criteria (18) would have been excluded. Individuals who never had a diagnosis suggestive of diabetes in the Rochester Epidemiology Project diagnostic index during their entire period of residence in the community but who would have qualified if prospectively screened also were not included. Cross-sectional studies that prospectively screen communitydwelling persons suggest that a substantial proportion of such individuals qualify as diabetic on the basis of oral glucose tolerance testing or fasting glucose testing, even though they report no prior diagnosis of diabetes (1). However, because diabetes is a chronic, progressive disease, many such individuals ultimately come to clinical attention, as was demonstrated by our study of the proportion of deceased Rochester residents with any mention of diabetes or diabetes-like conditions in their medical history (12). In addition, a significant (35 45 percent) and consistent proportion of Rochester residents had at least one blood glucose determination each year. While a large proportion of Rochester residents have a blood glucose measurement every few years, the increase in diabetes incidence observed in the most recent quinquennium, , could still be an underestimate. With each update of the Rochester diabetes cohort, we discover new cases that, on the basis of review of the patient s glucose values, actually qualified as cases in earlier time periods. This discrepancy results from the fact that potential cases for each update of the Rochester diabetes cohort are identified from the list of residents with any prior clinical mention of diabetes or diabetes-like conditions through the end of the study period (see Materials and Methods). Thus, with each update, some individuals who have qualified as diabetes cases on the basis of glycemic criteria but have not yet been assigned a clinical diagnosis are missing from the list of potential cases. For example, in the most recent update prior to this one (5), we reported that the numbers of cases identified for the time periods , , , and were 270, 288, 334, and 437, respectively. While the present update detected a relatively small number of previously unidentified cases (i.e., three, two, and two) for the periods , , and , respectively, 23 new cases were identified for the time period The age- and sex-adjusted annual incidence rate reported in the previous update was 303 per 100,000 population (95 percent confidence interval: 274, 332), as compared with the 316 per 100,000 (95 percent confidence interval: 286, 345) reported here. Thus, it is likely that the next update of the cohort through 1999 will identify additional cases for the time period and that the apparent downturn from 1993 to 1994 (figure 2) is partly an artifact of our method of case identification. In conclusion, we observed a significant increase in the incidence of diabetes among residents of Rochester, Minnesota, aged 30 years or more during the period This increase accelerated in the last quinquennium and was seen in both sexes, most dramatically in younger adults. An increase in diabetes surveillance is an unlikely explanation for this increased incidence. These results, combined with those of other studies (4, 5), provide further (albeit indirect) evidence that the epidemic of obesity in the United States is contributing to an epidemic of diabetes. The implications of a further increase in diabetes incidence are likely to be substantial, including increases in the incidence of cardiovascular disease and other complications related to diabetes. ACKNOWLEDGMENTS This study was supported by grants AG08729 and AR30582 from the National Institutes of Health and by SmithKline Beecham Pharmaceuticals (Philadelphia, Pennsylvania). The authors acknowledge the contributions of the nurse abstractors on this project, Kristine Otto-Higgins and Sue Helling. REFERENCES 1. Harris MI, Flegal KM, Cowie CC, et al. Prevalence of diabetes, impaired fasting glucose, and impaired glucose tolerance in U.S. adults: the Third National Health and Nutrition Examination Survey, Diabetes Care 1998;21:

7 Diabetes Incidence in Rochester, Minnesota Kenny SJ, Aubert RE, Geiss LS. Prevalence and incidence of non-insulin-dependent diabetes. In: Harris MI, Cowie CC, Reiber G, et al, eds. Diabetes in America. 2nd ed. Washington, DC: US GPO, 1995: Trends in the prevalence and incidence of self-reported diabetes mellitus United States, MMWR Morb Mortal Wkly Rep 1997;46: Burke JP, Williams K, Gaskill SP, et al. Rapid rise in the incidence of type 2 diabetes from 1987 to Arch Intern Med 1999;159: Leibson CL, O Brien PC, Atkinson E, et al. Relative contributions of incidence and survival to increasing prevalence of adult-onset diabetes mellitus: a population-based study. Am J Epidemiol 1997;146: Larson TS, Santanello N, Shahinfar S, et al. Trends in persistent proteinuria in adult-onset diabetes mellitus: a populationbased study. Diabetes Care 2000;23: Kurland LT, Molgaard CA. The patient record in epidemiology. Sci Am 1981;245: Melton LJ III. History of the Rochester Epidemiology Project. Mayo Clin Proc 1996;71: Palumbo PJ, Elveback LR, Chu CP, et al. Diabetes mellitus: incidence, prevalence, survivorship, and causes of death in Rochester, Minnesota, Diabetes 1976;25: National Diabetes Data Group. Classification and diagnosis of diabetes mellitus and other categories of glucose intolerance. Diabetes 1979;28: West KM. Standardization of definition, classification, and reporting in diabetes-related epidemiologic studies. Diabetes Care 1979;2: Goraya TY, Jacobsen SJ, Belau PG, et al. Validation of death certificate diagnosis of out-of-hospital coronary heart disease deaths in Olmsted County, Minnesota. Mayo Clinic Proc 2000;75: Bergstralh EJ, Offord KP, Chu C-P, et al. Calculating incidence, prevalence and mortality rates for Olmsted County, Minnesota: an update. (Technical report no. 49). Rochester, MN: Mayo Clinic Foundation, Kuczmarski RJ, Flegal KM, Campbell SM, et al. Increasing prevalence of overweight among U.S. adults: the National Health and Nutrition Examination Surveys, 1960 to JAMA 1994;272: Mokdad AH, Serdula MK, Dietz WH, et al. The spread of the obesity epidemic in the United States, JAMA 1999;282: Leibson CL. Loss of the female advantage with cardiovascular disease for women with diabetes. Lupus 1999;8: Update: prevalence of overweight among children, adolescents, and adults United States, MMWR Morb Mortal Wkly Rep 1997;46: Otto-Higgins KA, Leibson C, Dinneen S, et al. The effect of changing diagnostic criteria for diabetes on clinical practice. (Abstract). Diabetes 1999;48(suppl 1):A 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: