J ClinEpidemiol Vol. 44, No.
0895-4356/91$3.00+ 0.00 Copyright 0 1991Pergamon Press plc
3, pp. 329-334, 1991 Printed in Great Britain. All rights reserved
INCREASED INCIDENCE OF DIABETES MELLITUS IN RELATION TO ABDOMINAL ADIPOSITY IN OLDER WOMEN SUSAN
A. KAYE,’AARONR. FOLSOM,‘**J. MICHAELSPRAFKA,’ RONALDJ. PRINEAS’ and ROBERT B. WALLACE~
‘Division of Epidemiology, School of Public Health, University of Minnesota, Stadium Gate 27, 611 Beacon St S.E., Minneapolis, MN 55455, 2Department of Epidemiology and Public Health, School of Medicine, University of Miami, P.O. Box 016069, 1550 N.W. 10th Avenue, Miami, FL 33136 and 3Department of Preventive Medicine and Environmental Health, University of Iowa, 2800 Steindler Building, Iowa City, IA 52242, U.S.A. (Received in revised form 22 March 1990; received for publication 2 October 1990)
Abstract-The relationship between body fat distribution, measured by the ratio of waist-to-hip circumferences (WHR), and the 2 year incidence of diabetes mellitus was examined in a cohort of 41,837 women aged 55-69 years. The 399 women who reported the new onset of diabetes had a significantly greater mean body mass index (kg/m2) and WHR than non-cases. After adjustment for body mass index (BMI), age and education level using multivariate logistic regression, WHR was a significant independent predictor of diabetes in a dose-response fashion. Cases were 4.6 times (95% CI = 3.8, 5.6) more likely than non-cases to be in the upper tertile of WHR and 2.2 times (95% CI = 1.8, 2.7) more likely to be in the middle tertile. Women in the highest tertiles of both WHR and BMI had a 14.4-fold (95% CI = 9.5, 21.9) higher risk of diabetes than women in the lowest tertiles. These results demonstrate that increased abdominal adiposity is a significant independent risk factor for the development of diabetes mellitus in older women. Body weight
Diabetes mellitus
Fat distribution
INTRODUCTION It is well established that increased adiposity elevates the risk of non-insulin dependent diabetes mellitus [l-3]. Recent epidemiologic and clinical studies have indicated that a preponderance of adipose tissue in the abdomen compared to the hips further elevates risks of diabetes mellitus and abnormal glucose metabolism [4-91. However, with the exception of the prospective studies conducted in Gothenburg, Sweden [lo, 111, most evidence of an association of abdominal
adiposity
*Author for correspondence.
with diabetes
mellitus
Obesity
comes from cross-sectional investigations. We report here the prospective association of body mass and abdominal adiposity to the incidence of diabetes in a sample of free-living older women. We hypothesized that abdominal adiposity, measured by the waist-to-hip circumference ratio (WHR), would be associated with increased diabetes incidence after accounting for overall body mass. DATA COLLECTIONMETHODS Participants
Participants were a random sample of all women aged 55-69 years listed on the 1985 Iowa 329
330
SUSAN A. byE
Department of Transportation driver’s license list. A total of 41,837 women responded to a January 1986 mail survey, representing a response rate of 42.7%. Information contained on the driver’s license indicated that responders were about 3 months older, 0.4 kg/m2 lighter, and slightly more likely to live in rural, less affluent counties than non-responders [12]. A second mail survey of the cohort was conducted in November 1987 to obtain, among other things, the 2-year incidence of diabetes mellitus. Of the original 41,837 women, 37,579 returned the second questionnaire, and 3 18 were ascertained to be deceased. The response rate to the follow-up mailing was 91%. Baseline questionnaire
The baseline mail questionnaire inquired about several personal characteristics, including education level, cigarette smoking, leisuretime physical activity, and prior medical history, including use of insulin and oral hypoglycemic agents. Questions on the frequency of moderate and heavy intensity leisure-time physical activities were combined to create a 3-level physical activity score (low, medium, and high). To ascertain diabetes, women were asked “Have you ever been told by a physician that you had diabetes mellitus (sugar diabetes)?“. The accuracy of self-report of diabetes was assessed by requesting confirmation from the three most recent private physicians of 44 randomly selected women who reported diabetes mellitus and 41 who reported being disease-free. Participants were asked to report current height (without shoes and to the nearest inch) and weight (without clothes and to the nearest pound). Quetelet body mass index (BMI) was calculated as weight (kg)/height (m)‘. To assess adipose tissue distribution, we enclosed a paper tape measure and written instructions with illustrations for having a friend measure circumferences, in duplicate, at the waist (one inch above the umbilicus) and hips (maximal protrusion). Waist-to-hip circumference ratio (WHR) was calculated using the average of the duplicate measurements, dividing waist circumference by hip circumference. The tape measures were sampled and were found to be accurate within 0.4% against a standard meter rule. The accuracy and reliability of measurements of circumferences, height, and weight by this protocol were good [ 131. The intraclass correlation coefficients comparing self-measurements with technician measurements were: weight, r = 0.98;
et
at.
height, r = 0.86; waist girth, r = 0.93; and hip girth, r = 0.96. Statistical methods
Data were analyzed as a nested case-control study [14]. Incident cases of diabetes mellitus (n = 399) were defined as women free of diabetes mellitus at baseline who reported the new onset of diabetes on the follow-up questionnaire. Controls (n = 1716) were randomly selected from the women who reported not having diabetes on either the baseline or followup questionnaire. Women with cancer at baseline were excluded from both groups because of its potential to alter adiposity. The main hypothesis tested was that abdominal adiposity was significantly associated with diabetes mellitus incidence after accounting for the known association with BMI. WHR was used as the measure of abdominal adiposity. Data were initially examined by comparing ageadjusted means of anthropometric variables using analysis of covariance. Odds ratios were calculated from stratified analyses in which anthropometric variables were trichotomized to tertile cutpoints of the entire distribution. Odds ratios were age-adjusted by the MantelHaenszel method [15] using 5 year groupings. Ninety-five per cent confidence intervals for the summary odds ratios were test based [16]. Unconditional multiple logistic regression was employed for multivariate modelling. A number of multivariate models were examined including age, BMI, WHR, physical activity (1 = high physical activity, 0 = low or medium physical activity), cigarette smoking (1 = current smoker, 0 = never or former smoker), and educational status (1 = greater than high school graduate, 0 = less than or equal to high school graduate) as predictor variables. Indicator variables were used to designate tertiles of anthropometric variables. Age, BMI and WHR were forced to enter each regression model. Covariates were entered into the regression model using backward stepwise selection and the criterion for covariate entry into the regression model wasp < 0.05. Multivariate-adjusted odds ratios and their 95% confidence intervals (CI) were computed from the logistic regression model. SAS was used for all analyses [17]. RESULTS
Study participants were aged 55-69 years at baseline. Ninety-nine per cent were white. The
Body Fat Distribution and the Incidence of Diabetes Mellitus
331
Table 1. Age-adjusted mean anthropometric characteristics of incident cases of diabetes mellitus vs controls, Iowa women, aged 55-69 years Age-adjusted mean
Variable BMI (kg/m*) Weight (kg) Waist circumference (in.) Hip circumference (in.) WHRt
Controls Cases (n = 399) (n = 1716) 31.3 80.1 39.7 44.0 0.901
26.9 69.1 34.4 41.2 0.833
Casexontrol difference* 4.4 11.0 5.3 2.8 0.068
(95% CI) (3.8, 5.0) (9.5, 12.5) (4.7, 5.9) (2.4, 2.9) (0.059, 0.077)
*All differences are p < 0.001. iWaist-to-hip circumference ratio.
baseline prevalence of a self-reported physician diagnosis of diabetes mellitus was 6.4%. The approximate 2-year incidence of diabetes mellitus was 1.0%. The validation study, comparing self-reported diabetes mellitus with physician reports, indicated probable overreporting of diabetes. Of the 44 women who reported diabetes mellitus, there was physician confirmation in 28 (64%). Of the 41 women who reported being disease-free, 100% of the self-reports were verified as accurate. At baseline, the mean age-adjusted BMI was 30.3 kg/m2 in women with prevalent diabetes vs 26.8 kg/m2 in those without diabetes (p < 0.0001). The mean WHR was 0.896 in prevalent diabetics compared to 0.834 in non-diabetics (p < O.OOOl),indicating greater abdominal adiposity in diabetics. In analyses restricted to diabetics treated with either insulin or oral hypoglycemic agents, treated diabetics had significantly (p < 0.001) higher mean age-adjusted BMI (3 1.1 kg/m*) and WHR (0.905) than nondiabetics (26.9 kg/m2 and 0.836, respectively). The prevalent cases were excluded from the follow-up. As shown in Table 1, compared to controls, incident cases of diabetes mellitus had significantly greater age-adjusted mean levels of BMI (31.3 kg/m2 vs 26.9 kg/m2), as well as greater weight, and waist and hip circumferences. The mean WHR was 7.5% higher (p < 0.001) in cases (0.901) than controls (0.833). The product-moment correlation between BMI and WHR in the controls and incident cases was Y = 0.44, between BMI and waist circumference was r = 0.83, and between waist circumference and WHR was r = 0.75. Adjusted for age, the incident diabetes mellitus cases were much more likely than controls to be in the upper tertiles of BMI, waist circumference, and WHR (Table 2). The risk of diabetes was elevated over 6-fold in the upper compared to the lower tertile of each anthropometric
variable. Women with the highest levels of education and physical activity had about a 50% reduction in risk of diabetes. Cigarette smoking was not associated with diabetes incidence. Figure 1 shows the age-adjusted odds of incident diabetes mellitus jointly by tertiles of BMI and WHR. Risk increased strikingly across each tertile of WHR, stratified by BMI tertiles, and vice versa. The odds ratios of diabetes increased 3- to 7-fold from lowest to highest WHR tertiles (within BMI tertiles) and 2- to 5-fold from lowest to highest BMI tertiles (within WHR tertiles). Women who were jointly in the highest WHR and BMI tertiles had a 15.8-fold (95% CI = 10.2,24.5) increased risk of diabetes mellitus compared to women in the lowest tertiles. A series of multivariate logistic regression models were constructed using waist circumference, BMI, and WHR, both singly and in combination, to examine which set of anthropometric variables had the greatest predictive power. Waist circumference was the best single anthropometric predictor of diabetes mellitus. However, the regression model containing both BMI and WHR was a better predictor of diabetes than any of the other models. Table 3 shows the results of the multivariate logistic regression analysis, showing the independent odds of diabetes mellitus with relation to WHR, BMI, age, and education. No interaction was found between WHR and BMI, and neither physical activity nor cigarette smoking contributed independently to the prediction of diabetes mellitus. Diabetes risk was increased 4.6-fold for women in the highest tertile of WHR and 3.1-fold for those in the highest tertile of BMI. Dose-response relationships were observed. Moreover, women in the highest tertiles of both WHR and BMI had a 14.4-fold (95% CI = 9.5, 21.9) higher risk of diabetes than women in the lowest WHR and BMI group.
332
SUSANA. KAm et al. Table 2. Distribution of incident diabetes mellitus cases and controls anthropometric variables, Iowa women, aged 5569 years Cases n
Variable
Controls n
Age-adjusted* odds ratio
according to
(95% CI)
Bh4I Lblm2 It 29.2
98 235
628 595 451
1.0 1.9 6.0
(1.3, 2.6) (4.4, 8.0)
Waist circumference (in.)? 37.0
42 108 242
662 587 443
1.0 2.8 8.5
(1.9, 4.0) (6.3, 11.6)
WHRt < 0.802 0.802Al.878 >0.878
45 112 233
653 580 458
2.8 7.4
(1.9, 3.9) (5.4, 10.0)
Education level High school graduate
104 174 117
309 682 703
1.0 0.8 0.5
(0.6, 1.0) (0.4, 0.7)
Physical activity Low Medium High
224 96 68
757 457 440
1.0 0.7 0.5
(0.5,0.9) (0.4,0.7)
Cigarette smoking Never Former Current
257 82 49
1113 321 230
1.0 1.1 0.9
(0.8, 1.5) (0.5, 1.8)
\“,
,,
5.5
1.0
*Age-adjusted by the Mantel-Haenszel method using 5-year groupings. The lowest category is the reference group. tGroups are tertiles. WHR = waist-to-hip circumference ratio. DISCUSSION
In this prospective study of older women, abdominal adiposity, as measured by WHR, was positively and significantly associated with self-reported incidence of diabetes mellitus in a dose-response fashion. The age-adjusted odds of diabetes mellitus was 7.4-fold higher in the highest vs the lowest tertile of WHR. After adjusting for BMI, age and education in multivariate analyses, WHR remained a significant independent predictor of diabetes mellitus.
These findings are consistent with previous investigations in both men and women examining the association of fat distribution with the prevalence of diabetes mellitus [6,8,9]. Hartz et al. [8] reported that the prevalence of diabetes in a sample of weight-conscious women was lo-fold higher in obese women with abdominal fat preponderance than in lean women with gluteal fat preponderance. Gillum [6] found that the unadjusted prevalence of diabetes for men and women in the highest quartile of a “fat distribution index” was approximately 6-fold greater than for those in the lowest
cc:::> 15.8
Table 3. Results of multivariate logistic regression analysis of risk factors for diabetes mellitus, Iowa women, aged 55-69 years
Variable
Level
Age WHR’
5 year increases Tertile 1 Tertile 2 Tertile 3 Tertile 1 Tertile 2 Tertile 3 High School
BMI* Tertiles of WHR Tertiles of BMI Fig. 1. Age-adjusted odds of self-reported incident diabetes mellitus by tertiles of BMI and WHR in Iowa women, aged 55-69 years.
Education
Odds ratio 1.1 1.0 2.2 4.6 1.0 1.2 3.1 1.0 0.7
(95% CI) (0.9, 1.3) (1.8,2.7) (3.8, 5.6) (1.0, 1.5) (2.6,3.7) (0.6,0.8)
*See Table 2 for tertile cutpoints. Lowest category is the reference group.
Body Fat Distribution and the Incidence of Diabetes Mellitus
quartile. Intra-abdominal visceral fat assessed by computed tomography has also been found to be positively associated with diabetes prevalence [9]. To date, there has only been one other incidence study conducted in women of the relation of fat distribution with diabetes. During 12 years of follow-up of a population sample of 1414 women, there was a 13.5fold increased risk of diabetes for women who were jointly in the highest quintiles of both WHR and BMI compared to women in the lowest quintiles [lo]. These results, however, were based on only 43 incident cases of diabetes mellitus. In men from the same population, the risk of diabetes during 13 years of follow-up was 16.6 times greater for those in the upper 5% of WHR distribution than for those in the lowest quartile of WHR [ll]. Surprisingly few prospective studies have reported a relation between physical inactivity and diabetes incidence [3]. In this study, the age-adjusted odds of diabetes for medium and high physical activity were 0.7 and 0.5, respectively (p < 0.05). This association was, to a large degree, explained by an inverse association between physical activity and body size. Nevertheless, because of the role of physical activity in caloric balance, the age-adjusted association of physical activity with diabetes should not be discounted. The methodologic limitations of this investigation warrant discussion. Although the recruitment survey had only a 42% response rate, there was little difference in age and BMI between respondents and non-respondents. In general, self-report of diabetes is accurate [18]. In contrast, our validation of self-report of diabetes suggested over-reporting of diabetes. However, because the validation study requested that the participant report only their three most recent physicians, the apparent over-reporting could reflect some under-ascertainment of a confirming diagnosis. The 6.4% self-reported prevalence and the 1.0% self-reported incidence of diabetes mellitus are consistent with other reports in this age-sex group [19,20]. Some women with undetected diabetes would have been missed, but the observed associations probably remain valid. Women with abdominal fat preponderance seem unlikely to have any greater propensity to report diabetes than women with gluteal fat preponderance. However, if physicians happen to look more intensively for diabetes mellitus in persons with
333
obesity, our results based on self-reports of physican diagnosis may suffer from a diagnostic detection bias. Our results also depended on self-reported weight, height, and self-measured girths. Although some bias in self-reported weight and height has been found by others [21], the accuracy of the anthropometrics in this population was good [13]. It is therefore likely that self-measurement does not invalidate the results. The association between abdominal adiposity and diabetes mellitus has biological plausibility. WHR is positively associated with both serum insulin and glucose levels [4,22]. Abdominal adipocytes have higher lipolytic activity than femoral adipocytes [7,23]. An increased flux of free fatty acids emptying directly into the portal vein affects both insulin sensitivity and glucose metabolism [22]. Abdominal adiposity has also been associated with a decreased metabolic clearance rate of insulin [24]. Women with abdominal adiposity also have higher plasma concentrations of free testosterone and lower concentrations of sex-hormone binding globulin [25,26]. Associations of body mass and fat distribution with sex hormone concentrations in postmenopausal women, submitted, 19891, both of which have been theorized to inhibit hepatic extraction of insulin [24,27]. In conclusion, abdominal adiposity was independently associated with an increased risk of diabetes in older women. Measurement of WHR could be used clinically as an easy, noninvasive method of assessing diabetes risk. Obese individuals, especially those with abdominal adiposity, should be encouraged to lose weight to reduce the risk of developing diabetes. Acknowledgements-The authors would like to thank Linh Pham, Joonho Kim, Susan Gapstur, Karen Johnson and Joyce Wenz for their contributions and Laura Kemmis for manuscript preparation. This work is supported by a grant from the National Cancer Institute (ROI CA39742).
REFERENCES Lew EA, Garfinkel L. Variations in mortality by weight among 750,000 men and women. J Chron Dis 1979; 32: 563-576.
Seidell JC, Bakx KC, Deurenberg P, van den Hoogen HJM, Hautvast JGAJ, Stijnen T. Overweight and chronic illness: A retrospective cohort study with a follow-up of 617 years in men and women of initially 20-50 years of age. J Chron Dis 1986; 39: 585-593. West KM. The Epidemiology of Diabetes and Its Vascular Lesions. New York: Elsevier North-Holland Inc.; 1978: 231-248. Kalkhoff RK, Hartz AJ, Rupley D, Kissebah AH, Kelber S. Relationship of body fat distribution to blood pressure, carbohydrate tolerance, and plasma
334
5.
6.
7.
8.
9.
10.
11.
12.
13.
14. 15.
16.
SUSANA. KAYEet lipids in healthy obese women. J Lab CBn Med 1983; 102: 621-627. Vague J. The degree of masculine differentiation of obesities: A factor determining predisposition to diabetes, atherosclerosis, gout, and uric calculous disease. Am J CBn Nutr 1956; 4: 20-34. Gillum RF. The association of body fat distribution with hypertension, hypertensive heart disease, coronary heart disease, diabetes and cardiovascular risk factors in men and women ages 18-79 years. J Cluon Dis 1987; 40: 421428. Kissebah AH, Vydelingum N, Murray R et al. Relation of body fat distribution to metabolic complications of obesity. J Clln Endocrlaol Metab 1982; 54: 254-260. Hartz AJ, Rupley DC, Kalkhoff RD, Rimm AA. Relationship of obesity to diabetes: Influence of obesity level and body fat distribution. Prev Med 1983; 12: 351-357. Sparrow D, Borkon GA, Gerzof SG, Wisniewski C, Silbert CK. Relationship of fat distribution to glucose tolerance. Results of computed tomography in male participants of the Normative Aging Study. Diabetes 1986; 35: 411-415. Lundgren H, Bengtsson C, Blohme G, Lapidus L, Sjostrom L. Adiposity and adipose tissue distribution in relation to incidence of diabetes in women: results from a prospective population study in Gothenburg, Sweden. Int J Obesity 1989; 13: 413-423. Ohlson LO, Larsson B, Svardsudd K et al. The influence of body fat distribution on the incidence of diabetes mellitus. 13.5 years of follow-up of the participants in the study of men born in 1913. Diabetes 1985; 34: 105551058. Folsom AR, Prineas RJ, Kaye SA, Soler JT. Body fat distribution and self-reported prevalence of hypertension, heart attack, and other- heart disease in older women. Int J Eoidemiol 1989: 18: 361-367. Kushi LH, Kay> SA, Folsom AR, Soler JT, Prineas RJ. Accuracy and reliability of self-measurement of body girths. Am J Epidemiol 1988; 128: 740-748. Mantel N. Synthetic retrospective studies and related topics. Biometrics 1973; 29: 479486. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Nat1 Cancer Inst 1959; 22: 719-748. Miettinen OS. Estimability and estimation in casereferent studies. Ant J Epidemiol 1976; 103: 251-253.
al.
17. SAS User’s Guide: Statistica. Cary, NC.: SAS Institute, Inc.; 1985. 18. Harlow SD, Linet MS. Agreement between questionnaire data and medical records: the evidence for accuracy of recall. Am J Epidemlol 1989; 129: 233-248. 19. Prevalence of Diagnosed Diabetes, Undiagnosed Mab&es, and Impaired Glucose Tolerance in Adults 20-74 Years of Age. DHEW publication No. (PHS) 87-1687 (Vital and health statistics; Series 11; No. 237). Washington, D.C.: U.S. Government Printing Office: National Center for Health Statistics; 1987; 28. 20. Diabetes in America. Diabetes Data Compiled 19&1, NIH Publication No. 85-1468. Washington, DC.: U.S. Department of Health and Human Services. 1985: IV-1 to IV-II. 21. Stewart AW, Jackson RT, Ford MA, Beaglehole R. Underestimation of relative weight by use of selfreported height and weight. Am J Epidemioll987; 125: 122-126. 22. Kissebah AH, Evans DJ, Peiris A, Wilson CR. Endocrine characteristics in regional obesities: Role of sex steroids. In: Vague J, Ed. Metabolic Complications of Human Obesities. New York: Elsevier Science; 1985; 115-130. 23. Smith U, Hammersten J, Bjomtorp P, Kral JG. Regional differences and effect of weight reduction on human fat cell metabolism. Ear J Clin Invest 1979; 9: 327-332. 24. Peiris AN, Struve MF, Kissebah AH. Relationship of body fat distribution to the metabolic clearance of insulin in premenopausal women. Int J Obesity 1987; 11: 581-589. 25. Evans DJ, Hoffman RG, Kalkhoff RK, Kissebah AH. Relationship of androgenic activity to body fat topography, fat cell morphology, and metabolic aberrations in premenopausal women. J Clin Endocriool Metab 1983; 57: 304-309. 26. Kaye SA, Folsom AR, Soler JT, Prineas RJ, Potter JD. Associations of body mass and fat distribution with sex hormone concentrations in postmenopausal women. Int J Epidemiol 1991; in press. 27. Peiris AN, Mueller RA, Struve MF, Smith GA, Kissebah AH. Relationship of androgenic activity to splanchnic insulin metabolism and peripheral glucose utilization in premenopausal women. J Clin Endocrinol Metab 1987; 64: 162-169.
0895-4356/91$3.00+ 0.00 Copyright 0 1991Pergamon Press plc
3, pp. 329-334, 1991 Printed in Great Britain. All rights reserved
INCREASED INCIDENCE OF DIABETES MELLITUS IN RELATION TO ABDOMINAL ADIPOSITY IN OLDER WOMEN SUSAN
A. KAYE,’AARONR. FOLSOM,‘**J. MICHAELSPRAFKA,’ RONALDJ. PRINEAS’ and ROBERT B. WALLACE~
‘Division of Epidemiology, School of Public Health, University of Minnesota, Stadium Gate 27, 611 Beacon St S.E., Minneapolis, MN 55455, 2Department of Epidemiology and Public Health, School of Medicine, University of Miami, P.O. Box 016069, 1550 N.W. 10th Avenue, Miami, FL 33136 and 3Department of Preventive Medicine and Environmental Health, University of Iowa, 2800 Steindler Building, Iowa City, IA 52242, U.S.A. (Received in revised form 22 March 1990; received for publication 2 October 1990)
Abstract-The relationship between body fat distribution, measured by the ratio of waist-to-hip circumferences (WHR), and the 2 year incidence of diabetes mellitus was examined in a cohort of 41,837 women aged 55-69 years. The 399 women who reported the new onset of diabetes had a significantly greater mean body mass index (kg/m2) and WHR than non-cases. After adjustment for body mass index (BMI), age and education level using multivariate logistic regression, WHR was a significant independent predictor of diabetes in a dose-response fashion. Cases were 4.6 times (95% CI = 3.8, 5.6) more likely than non-cases to be in the upper tertile of WHR and 2.2 times (95% CI = 1.8, 2.7) more likely to be in the middle tertile. Women in the highest tertiles of both WHR and BMI had a 14.4-fold (95% CI = 9.5, 21.9) higher risk of diabetes than women in the lowest tertiles. These results demonstrate that increased abdominal adiposity is a significant independent risk factor for the development of diabetes mellitus in older women. Body weight
Diabetes mellitus
Fat distribution
INTRODUCTION It is well established that increased adiposity elevates the risk of non-insulin dependent diabetes mellitus [l-3]. Recent epidemiologic and clinical studies have indicated that a preponderance of adipose tissue in the abdomen compared to the hips further elevates risks of diabetes mellitus and abnormal glucose metabolism [4-91. However, with the exception of the prospective studies conducted in Gothenburg, Sweden [lo, 111, most evidence of an association of abdominal
adiposity
*Author for correspondence.
with diabetes
mellitus
Obesity
comes from cross-sectional investigations. We report here the prospective association of body mass and abdominal adiposity to the incidence of diabetes in a sample of free-living older women. We hypothesized that abdominal adiposity, measured by the waist-to-hip circumference ratio (WHR), would be associated with increased diabetes incidence after accounting for overall body mass. DATA COLLECTIONMETHODS Participants
Participants were a random sample of all women aged 55-69 years listed on the 1985 Iowa 329
330
SUSAN A. byE
Department of Transportation driver’s license list. A total of 41,837 women responded to a January 1986 mail survey, representing a response rate of 42.7%. Information contained on the driver’s license indicated that responders were about 3 months older, 0.4 kg/m2 lighter, and slightly more likely to live in rural, less affluent counties than non-responders [12]. A second mail survey of the cohort was conducted in November 1987 to obtain, among other things, the 2-year incidence of diabetes mellitus. Of the original 41,837 women, 37,579 returned the second questionnaire, and 3 18 were ascertained to be deceased. The response rate to the follow-up mailing was 91%. Baseline questionnaire
The baseline mail questionnaire inquired about several personal characteristics, including education level, cigarette smoking, leisuretime physical activity, and prior medical history, including use of insulin and oral hypoglycemic agents. Questions on the frequency of moderate and heavy intensity leisure-time physical activities were combined to create a 3-level physical activity score (low, medium, and high). To ascertain diabetes, women were asked “Have you ever been told by a physician that you had diabetes mellitus (sugar diabetes)?“. The accuracy of self-report of diabetes was assessed by requesting confirmation from the three most recent private physicians of 44 randomly selected women who reported diabetes mellitus and 41 who reported being disease-free. Participants were asked to report current height (without shoes and to the nearest inch) and weight (without clothes and to the nearest pound). Quetelet body mass index (BMI) was calculated as weight (kg)/height (m)‘. To assess adipose tissue distribution, we enclosed a paper tape measure and written instructions with illustrations for having a friend measure circumferences, in duplicate, at the waist (one inch above the umbilicus) and hips (maximal protrusion). Waist-to-hip circumference ratio (WHR) was calculated using the average of the duplicate measurements, dividing waist circumference by hip circumference. The tape measures were sampled and were found to be accurate within 0.4% against a standard meter rule. The accuracy and reliability of measurements of circumferences, height, and weight by this protocol were good [ 131. The intraclass correlation coefficients comparing self-measurements with technician measurements were: weight, r = 0.98;
et
at.
height, r = 0.86; waist girth, r = 0.93; and hip girth, r = 0.96. Statistical methods
Data were analyzed as a nested case-control study [14]. Incident cases of diabetes mellitus (n = 399) were defined as women free of diabetes mellitus at baseline who reported the new onset of diabetes on the follow-up questionnaire. Controls (n = 1716) were randomly selected from the women who reported not having diabetes on either the baseline or followup questionnaire. Women with cancer at baseline were excluded from both groups because of its potential to alter adiposity. The main hypothesis tested was that abdominal adiposity was significantly associated with diabetes mellitus incidence after accounting for the known association with BMI. WHR was used as the measure of abdominal adiposity. Data were initially examined by comparing ageadjusted means of anthropometric variables using analysis of covariance. Odds ratios were calculated from stratified analyses in which anthropometric variables were trichotomized to tertile cutpoints of the entire distribution. Odds ratios were age-adjusted by the MantelHaenszel method [15] using 5 year groupings. Ninety-five per cent confidence intervals for the summary odds ratios were test based [16]. Unconditional multiple logistic regression was employed for multivariate modelling. A number of multivariate models were examined including age, BMI, WHR, physical activity (1 = high physical activity, 0 = low or medium physical activity), cigarette smoking (1 = current smoker, 0 = never or former smoker), and educational status (1 = greater than high school graduate, 0 = less than or equal to high school graduate) as predictor variables. Indicator variables were used to designate tertiles of anthropometric variables. Age, BMI and WHR were forced to enter each regression model. Covariates were entered into the regression model using backward stepwise selection and the criterion for covariate entry into the regression model wasp < 0.05. Multivariate-adjusted odds ratios and their 95% confidence intervals (CI) were computed from the logistic regression model. SAS was used for all analyses [17]. RESULTS
Study participants were aged 55-69 years at baseline. Ninety-nine per cent were white. The
Body Fat Distribution and the Incidence of Diabetes Mellitus
331
Table 1. Age-adjusted mean anthropometric characteristics of incident cases of diabetes mellitus vs controls, Iowa women, aged 55-69 years Age-adjusted mean
Variable BMI (kg/m*) Weight (kg) Waist circumference (in.) Hip circumference (in.) WHRt
Controls Cases (n = 399) (n = 1716) 31.3 80.1 39.7 44.0 0.901
26.9 69.1 34.4 41.2 0.833
Casexontrol difference* 4.4 11.0 5.3 2.8 0.068
(95% CI) (3.8, 5.0) (9.5, 12.5) (4.7, 5.9) (2.4, 2.9) (0.059, 0.077)
*All differences are p < 0.001. iWaist-to-hip circumference ratio.
baseline prevalence of a self-reported physician diagnosis of diabetes mellitus was 6.4%. The approximate 2-year incidence of diabetes mellitus was 1.0%. The validation study, comparing self-reported diabetes mellitus with physician reports, indicated probable overreporting of diabetes. Of the 44 women who reported diabetes mellitus, there was physician confirmation in 28 (64%). Of the 41 women who reported being disease-free, 100% of the self-reports were verified as accurate. At baseline, the mean age-adjusted BMI was 30.3 kg/m2 in women with prevalent diabetes vs 26.8 kg/m2 in those without diabetes (p < 0.0001). The mean WHR was 0.896 in prevalent diabetics compared to 0.834 in non-diabetics (p < O.OOOl),indicating greater abdominal adiposity in diabetics. In analyses restricted to diabetics treated with either insulin or oral hypoglycemic agents, treated diabetics had significantly (p < 0.001) higher mean age-adjusted BMI (3 1.1 kg/m*) and WHR (0.905) than nondiabetics (26.9 kg/m2 and 0.836, respectively). The prevalent cases were excluded from the follow-up. As shown in Table 1, compared to controls, incident cases of diabetes mellitus had significantly greater age-adjusted mean levels of BMI (31.3 kg/m2 vs 26.9 kg/m2), as well as greater weight, and waist and hip circumferences. The mean WHR was 7.5% higher (p < 0.001) in cases (0.901) than controls (0.833). The product-moment correlation between BMI and WHR in the controls and incident cases was Y = 0.44, between BMI and waist circumference was r = 0.83, and between waist circumference and WHR was r = 0.75. Adjusted for age, the incident diabetes mellitus cases were much more likely than controls to be in the upper tertiles of BMI, waist circumference, and WHR (Table 2). The risk of diabetes was elevated over 6-fold in the upper compared to the lower tertile of each anthropometric
variable. Women with the highest levels of education and physical activity had about a 50% reduction in risk of diabetes. Cigarette smoking was not associated with diabetes incidence. Figure 1 shows the age-adjusted odds of incident diabetes mellitus jointly by tertiles of BMI and WHR. Risk increased strikingly across each tertile of WHR, stratified by BMI tertiles, and vice versa. The odds ratios of diabetes increased 3- to 7-fold from lowest to highest WHR tertiles (within BMI tertiles) and 2- to 5-fold from lowest to highest BMI tertiles (within WHR tertiles). Women who were jointly in the highest WHR and BMI tertiles had a 15.8-fold (95% CI = 10.2,24.5) increased risk of diabetes mellitus compared to women in the lowest tertiles. A series of multivariate logistic regression models were constructed using waist circumference, BMI, and WHR, both singly and in combination, to examine which set of anthropometric variables had the greatest predictive power. Waist circumference was the best single anthropometric predictor of diabetes mellitus. However, the regression model containing both BMI and WHR was a better predictor of diabetes than any of the other models. Table 3 shows the results of the multivariate logistic regression analysis, showing the independent odds of diabetes mellitus with relation to WHR, BMI, age, and education. No interaction was found between WHR and BMI, and neither physical activity nor cigarette smoking contributed independently to the prediction of diabetes mellitus. Diabetes risk was increased 4.6-fold for women in the highest tertile of WHR and 3.1-fold for those in the highest tertile of BMI. Dose-response relationships were observed. Moreover, women in the highest tertiles of both WHR and BMI had a 14.4-fold (95% CI = 9.5, 21.9) higher risk of diabetes than women in the lowest WHR and BMI group.
332
SUSANA. KAm et al. Table 2. Distribution of incident diabetes mellitus cases and controls anthropometric variables, Iowa women, aged 5569 years Cases n
Variable
Controls n
Age-adjusted* odds ratio
according to
(95% CI)
Bh4I Lblm2 It 29.2
98 235
628 595 451
1.0 1.9 6.0
(1.3, 2.6) (4.4, 8.0)
Waist circumference (in.)? 37.0
42 108 242
662 587 443
1.0 2.8 8.5
(1.9, 4.0) (6.3, 11.6)
WHRt < 0.802 0.802Al.878 >0.878
45 112 233
653 580 458
2.8 7.4
(1.9, 3.9) (5.4, 10.0)
Education level High school graduate
104 174 117
309 682 703
1.0 0.8 0.5
(0.6, 1.0) (0.4, 0.7)
Physical activity Low Medium High
224 96 68
757 457 440
1.0 0.7 0.5
(0.5,0.9) (0.4,0.7)
Cigarette smoking Never Former Current
257 82 49
1113 321 230
1.0 1.1 0.9
(0.8, 1.5) (0.5, 1.8)
\“,
,,
5.5
1.0
*Age-adjusted by the Mantel-Haenszel method using 5-year groupings. The lowest category is the reference group. tGroups are tertiles. WHR = waist-to-hip circumference ratio. DISCUSSION
In this prospective study of older women, abdominal adiposity, as measured by WHR, was positively and significantly associated with self-reported incidence of diabetes mellitus in a dose-response fashion. The age-adjusted odds of diabetes mellitus was 7.4-fold higher in the highest vs the lowest tertile of WHR. After adjusting for BMI, age and education in multivariate analyses, WHR remained a significant independent predictor of diabetes mellitus.
These findings are consistent with previous investigations in both men and women examining the association of fat distribution with the prevalence of diabetes mellitus [6,8,9]. Hartz et al. [8] reported that the prevalence of diabetes in a sample of weight-conscious women was lo-fold higher in obese women with abdominal fat preponderance than in lean women with gluteal fat preponderance. Gillum [6] found that the unadjusted prevalence of diabetes for men and women in the highest quartile of a “fat distribution index” was approximately 6-fold greater than for those in the lowest
cc:::> 15.8
Table 3. Results of multivariate logistic regression analysis of risk factors for diabetes mellitus, Iowa women, aged 55-69 years
Variable
Level
Age WHR’
5 year increases Tertile 1 Tertile 2 Tertile 3 Tertile 1 Tertile 2 Tertile 3 High School
BMI* Tertiles of WHR Tertiles of BMI Fig. 1. Age-adjusted odds of self-reported incident diabetes mellitus by tertiles of BMI and WHR in Iowa women, aged 55-69 years.
Education
Odds ratio 1.1 1.0 2.2 4.6 1.0 1.2 3.1 1.0 0.7
(95% CI) (0.9, 1.3) (1.8,2.7) (3.8, 5.6) (1.0, 1.5) (2.6,3.7) (0.6,0.8)
*See Table 2 for tertile cutpoints. Lowest category is the reference group.
Body Fat Distribution and the Incidence of Diabetes Mellitus
quartile. Intra-abdominal visceral fat assessed by computed tomography has also been found to be positively associated with diabetes prevalence [9]. To date, there has only been one other incidence study conducted in women of the relation of fat distribution with diabetes. During 12 years of follow-up of a population sample of 1414 women, there was a 13.5fold increased risk of diabetes for women who were jointly in the highest quintiles of both WHR and BMI compared to women in the lowest quintiles [lo]. These results, however, were based on only 43 incident cases of diabetes mellitus. In men from the same population, the risk of diabetes during 13 years of follow-up was 16.6 times greater for those in the upper 5% of WHR distribution than for those in the lowest quartile of WHR [ll]. Surprisingly few prospective studies have reported a relation between physical inactivity and diabetes incidence [3]. In this study, the age-adjusted odds of diabetes for medium and high physical activity were 0.7 and 0.5, respectively (p < 0.05). This association was, to a large degree, explained by an inverse association between physical activity and body size. Nevertheless, because of the role of physical activity in caloric balance, the age-adjusted association of physical activity with diabetes should not be discounted. The methodologic limitations of this investigation warrant discussion. Although the recruitment survey had only a 42% response rate, there was little difference in age and BMI between respondents and non-respondents. In general, self-report of diabetes is accurate [18]. In contrast, our validation of self-report of diabetes suggested over-reporting of diabetes. However, because the validation study requested that the participant report only their three most recent physicians, the apparent over-reporting could reflect some under-ascertainment of a confirming diagnosis. The 6.4% self-reported prevalence and the 1.0% self-reported incidence of diabetes mellitus are consistent with other reports in this age-sex group [19,20]. Some women with undetected diabetes would have been missed, but the observed associations probably remain valid. Women with abdominal fat preponderance seem unlikely to have any greater propensity to report diabetes than women with gluteal fat preponderance. However, if physicians happen to look more intensively for diabetes mellitus in persons with
333
obesity, our results based on self-reports of physican diagnosis may suffer from a diagnostic detection bias. Our results also depended on self-reported weight, height, and self-measured girths. Although some bias in self-reported weight and height has been found by others [21], the accuracy of the anthropometrics in this population was good [13]. It is therefore likely that self-measurement does not invalidate the results. The association between abdominal adiposity and diabetes mellitus has biological plausibility. WHR is positively associated with both serum insulin and glucose levels [4,22]. Abdominal adipocytes have higher lipolytic activity than femoral adipocytes [7,23]. An increased flux of free fatty acids emptying directly into the portal vein affects both insulin sensitivity and glucose metabolism [22]. Abdominal adiposity has also been associated with a decreased metabolic clearance rate of insulin [24]. Women with abdominal adiposity also have higher plasma concentrations of free testosterone and lower concentrations of sex-hormone binding globulin [25,26]. Associations of body mass and fat distribution with sex hormone concentrations in postmenopausal women, submitted, 19891, both of which have been theorized to inhibit hepatic extraction of insulin [24,27]. In conclusion, abdominal adiposity was independently associated with an increased risk of diabetes in older women. Measurement of WHR could be used clinically as an easy, noninvasive method of assessing diabetes risk. Obese individuals, especially those with abdominal adiposity, should be encouraged to lose weight to reduce the risk of developing diabetes. Acknowledgements-The authors would like to thank Linh Pham, Joonho Kim, Susan Gapstur, Karen Johnson and Joyce Wenz for their contributions and Laura Kemmis for manuscript preparation. This work is supported by a grant from the National Cancer Institute (ROI CA39742).
REFERENCES Lew EA, Garfinkel L. Variations in mortality by weight among 750,000 men and women. J Chron Dis 1979; 32: 563-576.
Seidell JC, Bakx KC, Deurenberg P, van den Hoogen HJM, Hautvast JGAJ, Stijnen T. Overweight and chronic illness: A retrospective cohort study with a follow-up of 617 years in men and women of initially 20-50 years of age. J Chron Dis 1986; 39: 585-593. West KM. The Epidemiology of Diabetes and Its Vascular Lesions. New York: Elsevier North-Holland Inc.; 1978: 231-248. Kalkhoff RK, Hartz AJ, Rupley D, Kissebah AH, Kelber S. Relationship of body fat distribution to blood pressure, carbohydrate tolerance, and plasma
334
5.
6.
7.
8.
9.
10.
11.
12.
13.
14. 15.
16.
SUSANA. KAYEet lipids in healthy obese women. J Lab CBn Med 1983; 102: 621-627. Vague J. The degree of masculine differentiation of obesities: A factor determining predisposition to diabetes, atherosclerosis, gout, and uric calculous disease. Am J CBn Nutr 1956; 4: 20-34. Gillum RF. The association of body fat distribution with hypertension, hypertensive heart disease, coronary heart disease, diabetes and cardiovascular risk factors in men and women ages 18-79 years. J Cluon Dis 1987; 40: 421428. Kissebah AH, Vydelingum N, Murray R et al. Relation of body fat distribution to metabolic complications of obesity. J Clln Endocrlaol Metab 1982; 54: 254-260. Hartz AJ, Rupley DC, Kalkhoff RD, Rimm AA. Relationship of obesity to diabetes: Influence of obesity level and body fat distribution. Prev Med 1983; 12: 351-357. Sparrow D, Borkon GA, Gerzof SG, Wisniewski C, Silbert CK. Relationship of fat distribution to glucose tolerance. Results of computed tomography in male participants of the Normative Aging Study. Diabetes 1986; 35: 411-415. Lundgren H, Bengtsson C, Blohme G, Lapidus L, Sjostrom L. Adiposity and adipose tissue distribution in relation to incidence of diabetes in women: results from a prospective population study in Gothenburg, Sweden. Int J Obesity 1989; 13: 413-423. Ohlson LO, Larsson B, Svardsudd K et al. The influence of body fat distribution on the incidence of diabetes mellitus. 13.5 years of follow-up of the participants in the study of men born in 1913. Diabetes 1985; 34: 105551058. Folsom AR, Prineas RJ, Kaye SA, Soler JT. Body fat distribution and self-reported prevalence of hypertension, heart attack, and other- heart disease in older women. Int J Eoidemiol 1989: 18: 361-367. Kushi LH, Kay> SA, Folsom AR, Soler JT, Prineas RJ. Accuracy and reliability of self-measurement of body girths. Am J Epidemiol 1988; 128: 740-748. Mantel N. Synthetic retrospective studies and related topics. Biometrics 1973; 29: 479486. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Nat1 Cancer Inst 1959; 22: 719-748. Miettinen OS. Estimability and estimation in casereferent studies. Ant J Epidemiol 1976; 103: 251-253.
al.
17. SAS User’s Guide: Statistica. Cary, NC.: SAS Institute, Inc.; 1985. 18. Harlow SD, Linet MS. Agreement between questionnaire data and medical records: the evidence for accuracy of recall. Am J Epidemlol 1989; 129: 233-248. 19. Prevalence of Diagnosed Diabetes, Undiagnosed Mab&es, and Impaired Glucose Tolerance in Adults 20-74 Years of Age. DHEW publication No. (PHS) 87-1687 (Vital and health statistics; Series 11; No. 237). Washington, D.C.: U.S. Government Printing Office: National Center for Health Statistics; 1987; 28. 20. Diabetes in America. Diabetes Data Compiled 19&1, NIH Publication No. 85-1468. Washington, DC.: U.S. Department of Health and Human Services. 1985: IV-1 to IV-II. 21. Stewart AW, Jackson RT, Ford MA, Beaglehole R. Underestimation of relative weight by use of selfreported height and weight. Am J Epidemioll987; 125: 122-126. 22. Kissebah AH, Evans DJ, Peiris A, Wilson CR. Endocrine characteristics in regional obesities: Role of sex steroids. In: Vague J, Ed. Metabolic Complications of Human Obesities. New York: Elsevier Science; 1985; 115-130. 23. Smith U, Hammersten J, Bjomtorp P, Kral JG. Regional differences and effect of weight reduction on human fat cell metabolism. Ear J Clin Invest 1979; 9: 327-332. 24. Peiris AN, Struve MF, Kissebah AH. Relationship of body fat distribution to the metabolic clearance of insulin in premenopausal women. Int J Obesity 1987; 11: 581-589. 25. Evans DJ, Hoffman RG, Kalkhoff RK, Kissebah AH. Relationship of androgenic activity to body fat topography, fat cell morphology, and metabolic aberrations in premenopausal women. J Clin Endocriool Metab 1983; 57: 304-309. 26. Kaye SA, Folsom AR, Soler JT, Prineas RJ, Potter JD. Associations of body mass and fat distribution with sex hormone concentrations in postmenopausal women. Int J Epidemiol 1991; in press. 27. Peiris AN, Mueller RA, Struve MF, Smith GA, Kissebah AH. Relationship of androgenic activity to splanchnic insulin metabolism and peripheral glucose utilization in premenopausal women. J Clin Endocrinol Metab 1987; 64: 162-169.
