Lupus (1999) 8, 351±355

ß 1999 Stockton Press. All rights reserved 0961-2033/99 $12.00

http://www.stockton-press.co.uk/lup

PAPER

Loss of the female advantage with cardiovascular disease for

women with diabetes
C Leibson1*
1
Department of Health Sciences Research, Mayo Clinic Foundation, Rochester, MN, USA

Background: The relative risk of cardiovascular disease (CVD) associated with diabetes is greater
for women than men, and diabetic women did not experience temporal declines in CVD mortality
observed for the general population and diabetic men.
Objective: To examine sex differences in CVD risk factors for persons with diabetes over time.
Design: Population-based historical cohort study.
Methods: The provider-linked medical records of all Rochester, MN, residents assigned a clinical
diagnosis of diabetes were reviewed to con®rm case status and assign diagnosis date. Data on
fasting glucose, obesity, persistent proteinuria, smoking, hypertension, and dyslipidemia were
obtained at diagnosis for con®rmed incidence cases.
Results: There were 1330 diabetes cases 1970 ± 1989. Compared to men, women at diagnosis were
older and more likely hypertensive, had similar levels of fasting glucose and persistent proteinuria,
and less likely to smoke. Among persons diagnosed at younger ages, women were more likely than
men to be obese. Comparison with published data for the Rochester population revealed the excess
obesity and hypertension associated with diabetes were highest for women < age 55 y. Temporal
trends in CVD risk factors did not differ between diabetic men and women.

Keywords: diabetes; obesity; epidemiology; hypertension; cardiovascular disease

Introduction why the excess risk of CVD mortality associated with

diabetes is greater for women than men, and why the
excess risk increased for women.
Persons with diabetes are at increased risk of
This study uses the previously established 1970 ±
cardiovascular disease (CVD) morbidity and mortal-
1989 Rochester diabetes incidence cohort to compare
ity; the excess risks are greater for women than
men and women at diagnosis for the prevalence of
men.1 ± 3 The disparity is especially high for younger
CVD risk factors and to examine temporal trends in
women. The risk of premature ischemic heart disease
these risk factors. Some of the data are preliminary.
death associated with diabetes is nine-fold for women
and only four-fold for men;3 of all women who die
from heart disease before age 55, approximately one-
third have diabetes.4 Methods
Temporal trends reveal that the prevalence of
diabetes has increased for both women and men; the Rochester, MN, is relatively isolated from other urban
increase is due to rising incidence5,6 rather than centers and is home to the Mayo Clinic, a large
improved survival; and declines in CVD mortality tertiary care medical center. Therefore, local residents
observed for the general population over the past three receive essentially all medical care from very few
decades were shared more by diabetic men than providers. Also, information from every contact for
diabetic women.5,7 Thus, it is important to investigate Mayo patients, including hospital, of®ce, or ER visits;
correspondence; death certi®cate; and autopsy is
contained within a unit medical record. An indexing
system that codes and enters diagnoses assigned at
each encounter into continuously updated computer
*Correspondence: C Leibson, Department of Health Sciences ®les has been expanded to include non-Mayo
Research, Mayo Clinic Foundation, Rochester, MN55905, USA. providers of care to local residents.8

Downloaded from lup.sagepub.com at UNIV OF MICHIGAN on April 19, 2015

 Cardiovascular disease for women with diabetes
C Leibson
352
Identi®cation of diabetes cases over time was addressed by testing for interactions
between sex and year of diagnosis.
Univariate statistics were used to compare esti-
The complete community-based medical records,
mates of the prevalence of hypertension and body
including glucose values, of all Rochester residents
mass index at diabetes diagnosis with published
assigned a clinical diagnosis of diabetes or diabetes-
estimates for a random sample of the Rochester
like condition 1945 ± 1990 were reviewed to con®rm
population prospectively surveyed in 1986
that individuals met National Diabetes Data Group
(n ˆ 2122).13
criteria9 and determine residency. Adjustments were
made for temporal changes in laboratory glucose
methods.10 Persons treated with insulin or oral agents
also quali®ed as cases, irrespective of glucose values.
A detailed description of case ascertainment is Results
provided elsewhere.5
The 1970 ± 1989 Rochester diabetes incidence cohort
consisted of 1330 individuals age  30 y at diagnosis
Data on CVD risk factors (50% male). The proportion of individuals for whom
data at diagnosis were available was very high for
The present study includes 1970 ± 1989 diabetes each risk factor under consideration (Table 1), with
incidence cases with age at diagnosis  30 y. The the exception that 33% of individuals were missing
records of all cases were reviewed for fasting blood triglyceride and cholesterol values at diagnosis  two
glucose; obesity, that is body mass index (weight in years; these latter factors were excluded from
kilograms=height in meters2)  27.3 for women and analyses.
 27.8 for men;11 and prevalence of persistent Mean age at diagnosis was greater for women than
proteinuria at diagnosis. Persistent proteinuria was men (63.8  14.1 vs 60.4  12.7, P < 0.01). Mean
de®ned as  2 consecutive urinalyses with grade 1 or fasting glucose values did not differ between women
greater protein and no subsequent negative values; and men (220 91 mg=dl vs 222 88 mg=dl,
exceptions were a single negative value in a string of P ˆ 0.78). As shown in Table 1, diabetic women
positives and values following dialysis or transplant.12 were less likely than men to smoke. There were no
As part of a pilot study, a random sample of 400 of differences between women and men in the proportion
the 1970 ± 1989 incidence cases were identi®ed to with prevalent persistent proteinuria or in the propor-
assess the feasibility of obtaining additional CVD risk tion that were obese. The likelihood of having two
factors. Records were reviewed at diagnosis two blood pressures that quali®ed as hypertensive was
years for evidence of hypertension, that is two signi®cantly higher for women than men. When
ambulatory blood pressure readings  140 mm Hg persons on anti-hypertensives were included, 85% of
systolic and=or 90 mm Hg diastolic or evidence of women and 75% of men (P ˆ 0.02) were hyperten-
antihypertensive medication, as determined by a sive.
neurologist; smoking status; and all cholesterol and Tests for interactions in multivariable regression
triglyceride values. analyses revealed differences in smoking between
women and men were less at younger ages; and at
younger ages, women were more often obese than
men. The prevalence of hypertension was greater in
Statistical analysis
obese than non-obese persons (OR ˆ 3.74, 95%
CI ˆ 2.22 ± 6.32, P < 0.01). After adjusting for age
Comparisons between sexes and, within each sex, and obesity, the effect of female gender on hyperten-
between calendar periods (1970 ± 79 vs 1980 ± 89) are sion no longer reached signi®cance (OR ˆ 1.56, 95%
presented as means  standard deviation (s.d.) for CI ˆ 0.92 ± 2.64, P ˆ 0.09).
continuous variables or as odds ratios (OR) and 95% Between 1970 ± 79 and 1980 ± 89, there were
con®dence intervals (CI) for dichotomous variables. signi®cant declines in the prevalence of persistent
Tests for signi®cant differences used Student's t (or proteinuria (53% for women, 50% for men) and
rank sum) and chi-square. Multivariable regression signi®cant increases in the prevalence of obesity (16%
was used to estimate the independent effects of sex for both women and men) (Table 2). The prevalence
and year of diagnosis, after adjusting for age and other of hypertension declined 2% for women and increased
CVD risk factors as appropriate. The question of 6% for men; the prevalence of smoking declined 10%
whether temporal trends differed between the sexes for women and 19% for men; none of the differences

Downloaded from lup.sagepub.com at UNIV OF MICHIGAN on April 19, 2015

 Cardiovascular disease for women with diabetes
C Leibson
353
Table 1 The prevalence of cardiovascular disease risk factors at diagnosis of diabetes by sex among members of the 1970 ± 1989
Rochester, MN, diabetes incidence cohort*
No. of records % of reviewed % of data available
CVD risk factor reviewed records with data available that are positive for risk factor OR (95% CI) P-value

Qualifying blood pressures{

(Dx 2 y)
Women 186 97% (181=186) 83% (150=181) 1.93 (1.18 ± 3.16) < 0.01
Men 205 98% (200=205) 72% (143=200)
Obesity
(Dx 6 months)
Women 660 99% (652=660) 59% (386=652) 1.08 (0.87 ± 1.35) 0.47
Men 670 99% (664=670) 57% (380=664)
Persistent proteinuria{
(prior to Dx)
Women 624 99% (620=624) 6.0% (37=620) 0.90 (0.57 ± 1.42) 0.64
Men 627 99% (620=627) 6.6% (41=620)
Smoking status{
(Dx 2 y)
Women 186 91% (169=186) 19% (32=169) 0.41 (0.26=0.67) < 0.01
Men 205 95% (194=205) 36% (70=194)

*Age at diagnosis  30 y, n ˆ 1330 y.

{400 individuals were randomly selected from the 1330 for a pilot study. Nine refused consent for medical records research (53% male, mean age ˆ 64 13
for women, 61 13 for men, P ˆ 0.02).
{Data were collected for a separate study that was limited to 1970 ± 1989 incidence cases with age ˆ 40 y at diagnosis (n ˆ 1251).

Table 2 The prevalence of cardioavscular disease risk factors at diagnosis of diabetes by sex and decade of diagnosis among members
of the 1970 ± 1989 Rochester, MN, diabetes incidence cohort*
Women Men

CVD risk factor 1970 ± 1979 1980 ± 1989 OR (95% CI) P-value 1970 ± 1979 1980 ± 1989 OR (95% CI) P-value

% with qualifying blood 84% (61=72) 82% (89=109) 0.80 (0.36 ± 1.80) 0.59 69% (63=91) 73% (80=109) 1.23 (0.66 ± 2.50) 0.52
pressures values (Dx 2 y)
% obese (Dx 2 y) 54% (152=280) 63% (234=372) 1.43 (1.04 ± 1.96) 0.03 52% (144=275) 61% (236=389) 1.40 (1.03 ± 1.92) 0.02
% with persistent proteinuria{ 8.5% (23=272) 4.0% (14=348) 0.45 (0.23 ± 0.90) 0.02 9.3% (24=257) 4.7 (17=363) 0.48 (0.25 ± 0.91) 0.02
(prior to Dx)
% currently smoking (Dx 2 y) 20% (12=59) 18% (20=110) 0.87 (0.39 ± 1.93) 0.73 40% (34=84) 33% (36=110) 0.72 (0.40 ± 1.29) 0.26
Mean age, y (s.d.) (at Dx) 64.9 (13.2) 62.9 (14.7) ± 0.06 60.1 (12.7) 60.6 (12.7) ± 0.65
Mean fbg, mg=dl (s.d.) (at Dx) 222 (97) 219 (88) ± 0.66 226 (93) 219 (84) ± 0.36

*Age at diagnosis  30 y, n ˆ 1330.

{400 individuals were randomly selected from the 1,330 for a pilot study. Nine refused consent for medical records research (53% male, mean
age ˆ 64 13 for women, 61 13 for men, P ˆ 0.02).
{Data were collected for a separate study that was limited to 1970 ± 1989 incidence cases with age ˆ 40 y at diagnosis (n ˆ 1251).

reached statistical signi®cance. There were no sig- Age- and sex-speci®c estimates of mean body mass
ni®cant differences between time periods in age or index (BMI) for all members of the 1970 ± 1989
fasting glucose for either women or men. diabetes incidence cohort reveal the difference
Multivariable regression analyses were performed between diabetic women and men was greatest for
for women and men separately to estimate the effect persons diagnosed before age 45 y (mean
of year of diagnosis on the prevalence of CVD risk BMI ˆ 35.2  10.1 vs 30.2  7.2, difference ˆ 4.5,
factors, adjusted for age. The results were unchanged P < 0.01) (Figure 1). Figure 1 also provides published
from the ®ndings in Table 2. When women and men estimates for the 1986 Rochester population, age
were included together in the models, tests for  45.13 The greatest difference in BMI between
interactions revealed no difference between the sexes diabetic persons and the general population occurred
with respect to temporal trends in the prevalence of for women age < 55 y (mean BMI ˆ 32.2 vs 24.8,
CVD risk factors (data available on request). difference ˆ 7.4). A decline in BMI with increasing

Downloaded from lup.sagepub.com at UNIV OF MICHIGAN on April 19, 2015

 Cardiovascular disease for women with diabetes
C Leibson
354
remained high in every age group (OR for age
> 74 y ˆ 13.6, 95%CI ˆ 4.3 ± 42.9) compared to the
odds associated with diabetes for men (OR for
age < 55 ˆ 7.4, 95%CI ˆ 4.5 ± 12.3; OR for age > 74
ˆ 3.8, 95%CI ˆ 1.7 ± 8.7).

Discussion

Figure 1 A comparison of mean body mass index (BMI) for members of This study compares the prevalence of CVD risk
the 1970 ± 1989 Rochester, MN, diabetes incidence cohort (n ˆ 1316) with
mean BMI for a random sample of the 1986 Rochester, MN, population, factors at diagnosis of diabetes between women and
by sex and age group. men, over time, and with published estimates for
selected risk factors in the 1986 Rochester population.
With the exception of smoking, CVD risk factor levels
for women were either similar to or greater than levels
for men. The high levels of CVD risk compared to the
Rochester population were especially high for young
diabetic women.
The CVD risk factor estimates for diabetic women
and men reported here agree with those from other
studies.4,14 ± 16 The unique contribution of this study is
the comparison of CVD risk factors between diabetic
women and men over time. Data on temporal trends
Figure 2 A comparison of the prevalence of hypertension in a random are needed to inform investigations of why diabetic
sample of individuals (n ˆ 391) identi®ed from the 1970 ± 1989 Rochester, women did not exhibit declines in CVD mortality that
MN, diabetes incidence cohort with the prevalence of hypertension in a
random sample of the 1986 Rochester, MN, population (n ˆ 2122), by sex were experienced by the general population and
and age group. Data for the Rochester population were obtained from diabetic men in recent decades.5,7 The absence of
Phillips SJ. Mayo Clin Proc 1988; 63: 691. Hypertension in the diabetes any signi®cant sex by calendar year interactions in the
study was de®ned as the presence of any two qualifying outpatient blood
pressure values recorded in the medical record at the time of present study argues against between-sex differences
diagnosis two years. Hypertension in the Rochester study was de®ned in temporal trends in CVD risk factors as an
as two qualifying in-home blood pressure values obtained prospectively in explanation.
a standardized manner. Qualifying blood pressures were de®ned as
> ˆ 140 mm Hg systolic and=or > ˆ 90 mm Hg diastolic in both studies. The present study is limited by the fact that data
were collected retrospectively. Criteria for determin-
ing which and how frequently individuals were
age was most apparent for diabetic women. For measured were not standardized. Data on several
persons diagnosed after age 74 y, the mean BMI was CVD risk factors, for example exercise, insulin levels,
similar for women (26.2 5.2) and men (26.7  3.7, waist-to-hip ratios, were unavailable. This limitation
P ˆ 0.34); and the difference in mean BMI between is especially problematic for triglyceride and choles-
diabetic and Rochester women aged > 74 y terol valuesÐthe association between diabetes and
(mean ˆ 23.3, difference ˆ 2.9) was less than half dyslipidemia is well recognized; elevated triglycerides
that between younger and older diabetic women and LDL to HDL cholesterol ratios are associated
(difference ˆ 6.0). with increased risk of CVD among persons with
The 1986 Rochester survey also collected data on diabetes;17,18 and a few studies have reported
the prevalence of hypertension.13 As shown in Figure between-sex differences in these factors and their
2, levels of hypertension in the Rochester population associated risks.15,19,20 The extent to which compar-
were much lower than levels for diabetic individuals, isons between the sexes and over time in the present
for both sexes and at every age. Among Rochester study were confounded by differences in unmeasured
residents, the prevalence of hypertension was less for risk factors is unclear. It is also not known whether
women than for men at younger ages and greater for diabetic women and men differed with respect to
women than for men at older ages; this pattern was not temporal trends in these unmeasured factors.
observed for persons with diabetes. The odds The possibility that the loss of the female
associated with diabetes were highest for women advantage in diabetes is related to differences between
age < 55 y (OR ˆ 19.7, 95%CI ˆ 9.7 ± 40.2) and women and men in the relative odds of obesity and

Downloaded from lup.sagepub.com at UNIV OF MICHIGAN on April 19, 2015

 Cardiovascular disease for women with diabetes
C Leibson
355
hypertension is encouraging from a public health 3 Will JC, Casper M. The contribution of diabetes to early deaths from
ischemic heart disease: US gender and racial comparisons. Am J Public
perspectiveÐobesity and hypertension are modi®able Health 1996; 86: 576 ± 579.
risk factors. However, temporal trends in CVD risk 4 DeStefano F, Ford ES, Newman J et al. Risk factors for coronary heart
factors observed in this study were less encouraging. disease mortality among persons with diabetes. Ann Epidemiol 1993;
3: 27 ± 34.
Although there were marked declines in persistent 5 Leibson CL, O'Brien PC, Atkinson E et al. Relative contributions of
proteinuria and smaller declines in smoking, the incidence and survival to increasing prevalence of adult-onset diabetes
prevalence of obesity increased over time and the mellitus: A population-based study. Am J Epidemiol 1997; 146:
12 ± 22.
prevalence of hypertension remained unchanged. The 6 Anonymous. Trends in the prevalence and incidence of self-reported
latter ®nding contrasts with the declines in hyperten- diabetes mellitusÐUnited States, 1980 ± 1994. Mor Mortal Wkly Rep
sion for both sexes and all age groups during the 1997; 46: 1014 ± 1018.
7 Gu K, Cowie C, Harris M. Diabetic adults experienced smaller
1970s and 1980s for the US population generally21 declines in heart disease mortality than nondiabetic adults in the U.S.
and occurred despite obvious room for improvement. population, 1971 ± 93. Diabetes 1998; 47: 41A.
Over half (52% of men and 65% of women) of 8 Melton LJ. History of the Rochester Epidemiology Project. Mayo Clin
Proc 1996; 71: 266 ± 274.
diabetic individuals with no treatment for hyperten- 9 National Diabetes Data Group Investigators. Classi®cation and
sion in our study had two outpatient blood pressure diagnosis of diabetes mellitus and other categories of glucose
readings that quali®ed as hypertensive. These ®ndings intolerance. Diabetes 1979; 28: 1039 ± 1057.
10 West KM. Standardization of de®nition, classi®cation, and reporting in
have important clinical implications. High blood diabetes-related epidemiologic studies. Diabetes Care 1979; 2: 65 ± 76.
pressure increases the risk of death from coronary 11 Kuczmarski RJ, Flegal KM, Campbell SM et al. Increasing prevalence
heart disease among young diabetics more than three- of overweight among US adults. The National Health and Nutrition
Examination Surveys, 1960 to 1991. JAMA 1994; 272: 205 ± 211.
fold;4 and control of hypertension markedly reduces 12 Larson TS, Leibson CL, Santanello N et al. Trends in persistent
CVD complications and mortality among persons with proteinuria in adult onset diabetes mellitus: A population-based study.
diabetes.22 Kidney Int 1999, submitted.
13 Phillips SJ, Whisnant JP, O'Fallon WM et al. A community blood
In conclusion, the relative burden of CVD risk pressure survey: Rochester, Minnesota, 1986. Mayo Clin Proc 1988;
factors at diabetes diagnosis is greater for women than 63: 691 ± 699.
men, especially at younger ages. This ®nding is 14 Sprafka JM, Bender AP, Jagger HG. Prevalence of hypertension and
associated risk factors among diabetic individuals. The Three-City
consistent with the high excess risk of CVD death Study. Diabetes Care 1988; 11: 17 ± 22.
observed for premenopausal diabetic women.3,4 But 15 Sprafka JM, Xue S, Bushhouse SA et al. Cardiovascular disease risk
between-sex differences in temporal trends in mortal- factors and glucose tolerance. The Wadena City Health Study. Ann
Epidemiol 1992; 2: 647 ± 656.
ity reported earlier for the Rochester diabetes cohort5 16 Chaturvedi N, Fuller JH, The WHO Multinational Study Group.
were not accompanied by between-sex differences in Mortality risk by body weight and weight change in people with
temporal trends in CVD risk factors at diagnosis. NIDDM: The WHO Multinational Study of Vascular Disease in
Diabetes. Diabetes Care 1995; 18: 766 ± 774.
17 Uusitupa MI, Niskanen LK, Siitonen O et al. Five-year incidence of
atherosclerotic vascular disease in relation to general risk factors,
insulin level, and abnormalities in lipoprotein composition in non-
Acknowledgements insulin-dependent diabetic and nondiabetic subjects. Circulation 1990;
82: 27 ± 36.
18 Uusitupa MI, Niskanen LK, Siitonen O et al. Ten-year cardio-
Funding was provided by SmithKline Beecham vascular mortality in relation to risk factors and abnormalities
Pharmaceuticals and NIH grant AG08729. Data in lipoprotein composition in type 2 (non-insulin-dependent)
diabetic and non-diabetic subjects. Diabetologia 1993; 36: 1175 ±
collection and analysis were performed by Susan 1184.
Helling, Kristine Otto-Higgins, and Jeanine Ransom. 19 Haffner SM, Miettinen H, Stern MP. Relatively more atherogenic
coronary heart disease risk factors in prediabetic women than in
prediabetic men. Diabetologia 1997; 40: 711 ± 717.
20 Howard BV, Cowan LD, Go O et al. Adverse effects of diabetes on
References multiple cardiovascular disease risk factors in women: the Strong Heart
Study. Diabetes Care 1998; 21: 1258 ± 1265.
1 Folsom AR, Vitelli LL, Lewis CE et al. A prospective study of 21 Mulrow PJ. Detection and control of hypertension in the
coronary heart disease in relation to fasting insulin, glucose, and population: the United States experience. Am J Hypertens 1998; 11:
diabetes: The Atherosclerosis Risk in Communities (ARIC) Study. 744 ± 746.
Diabetes Care 1997; 20: 935 ± 942. 22 U.K. Prospective Diabetes Study Group. Tight blood pressure control
2 Wolf PA, Kannel WB, Verter J. Current status of stroke risk factors. In: and risk of macrovascular and microvascular complications in type
Barnett HJM (ed) Neurologic Clinics. Saunders: Philadelphia, 1983, pp 2 diabetes: U.K. Prospective Diabetes Study 38. Br Med J 1998; 317:
317 ± 343. 703 ± 713.

Downloaded from lup.sagepub.com at UNIV OF MICHIGAN on April 19, 2015