Fox CS, Coady S, Sorlie PD, Levy D, Meigs JB, D’Agostino RB, Wilson PWF, Savage PJ. Trends in Cardiovascular Complications of Diabetes. JAMA. 2004;292(20):2495-2499. doi:10.1001/jama.292.20.2495
Author Affiliations: National Heart, Lung,
and Blood Institute’s Framingham Heart Study (Drs Fox and Levy), Framingham,
Mass; National Heart, Lung, and Blood Institute, National Institutes of Health
(Drs Fox, Sorlie, Levy, and Savage and Mr Coady), Bethesda, Md; Brigham and
Women’s Hospital Department of Endocrinology, Diabetes, and Hypertension,
Harvard Medical School (Dr Fox), General Medicine Division, Department of
Medicine, Massachussetts General Hospital and Harvard Medical School (Dr Meigs),
and Boston University Department of Mathematics (Dr D’Agostino), Boston,
Mass; and Department of Endocrinology, Diabetes, and Medical Genetics, Medical
University of South Carolina, Charleston (Dr Wilson).
Context Despite reductions in cardiovascular disease (CVD) mortality over the
past few decades, it is unclear whether adults with and without diabetes have
experienced similar declines in CVD risk.
Objective To determine whether adults with and without diabetes experienced similar
declines in incident CVD in 1950-1995.
Design, Setting, and Participants Participants aged 45-64 years from the Framingham Heart Study original
and offspring cohorts who attended examinations in 1950-1966 (“earlier”
time period; 4118 participants, 113 with diabetes) and 1977-1995 (“later”
time period; 4063 participants, 317 with diabetes). Incidence rates of CVD
among those with and without diabetes were compared between the earlier and
Main Outcome Measures Myocardial infarction, coronary heart disease death, and stroke.
Results Among participants with diabetes, the age- and sex-adjusted CVD incidence
rate was 286.4 per 10 000 person-years in the earlier period and 146.9
per 10 000 in the later period, a 49.3% (95% confidence interval [CI],
16.7%-69.4%) decline. Among participants without diabetes, the age- and sex-adjusted
incidence rate was 84.6 per 10 000 person-years in the earlier period
and 54.3 per 10 000 person-years in the later period, a 35.4% (95% CI,
25.3%-45.4%) decline. Hazard ratios for diabetes as a predictor of incident
CVD were not different in the earlier vs later periods.
Conclusions We report a 50% reduction in the rate of incident CVD events among adults
with diabetes, although the absolute risk of CVD is 2-fold greater than among
persons without diabetes. Adults with and without diabetes have benefited
similarly during the decline in CVD rates over the last several decades. More
aggressive treatment of CVD risk factors and further research on diabetes-specific
factors contributing to CVD risk are needed to further reduce the high absolute
risk of CVD still experienced by persons with diabetes.
Marked reductions in cardiovascular disease (CVD) mortality have occurred
over the last 50 years.1- 11 It
has been reported that adults with diabetes have experienced less decline
in CVD mortality than those without diabetes.12- 15 Adults
with diabetes are at a 2- to 4-fold increased risk of CVD events relative
to those without diabetes16- 20 and
are at about a 60% increased risk of early mortality.21 However,
it is uncertain whether the lack of decline in risk among diabetes patients
actually exists, as conclusions drawn from current data are methodologically
limited due to use of self-reported diabetes status,12,15 limited
time span of data collection,12 and assessment
of cause of death via death certificate.12,13
Given the importance of understanding whether CVD risk reduction has
differentially affected adults with and without diabetes, we sought to test
whether differences in CVD events have developed over the past several decades
among those with and without diabetes in the Framingham Heart Study original
and offspring cohorts. The Framingham Heart Study provides a unique setting
in which this question can be answered because of long-term follow-up, standardized
CVD event ascertainment, and careful documentation of concomitant risk factors.
Participants for this study were drawn from the Framingham Heart Study.
Selection criteria and study design have been described previously.22,23 The standard clinic examination included
an interview, physical examination, and laboratory tests. Cardiovascular events
were documented throughout follow-up by daily hospital and death surveillance.
We selected participants aged 45 to 64 years from 4 original cohort
examinations, approximately 12 years apart (1950-1955, 1962-1966, 1977-1979,
and 1986-1990), and 2 offspring examinations, 12 years apart (1979-1983 and
1991-1995). Participants could contribute information at more than 1 examination
provided they reached the next examination free of a CVD event. For example,
a 50-year-old participant with diabetes attending an examination in 1950 could
contribute follow-up information for the next 12 years. If this participant
was free of CVD in 1962, he/she could provide additional follow-up information
for the assessment of CVD events.
Study participants were classified as belonging to 2 groups: an earlier
period (examinations attended in the 1950s and 1960s) and a later period (examinations
attended in the 1970s, 1980s, and 1990s). These 2 periods formed the basis
for comparison of CVD incidence rates among participants with and without
diabetes. Participants were followed up for CVD events for up to 12 years.
The early period contributed 4118 participants (55 385 person-years of
follow-up) and the later period contributed 4063 participants (44 073
person-years of follow-up). A total of 779 original cohort participants were
part of the later period. Cardiovascular disease events were accrued until
December 31, 2000.
The Boston Medical Center Institutional Review Board approved the study,
and all participants gave written informed consent.
Cardiovascular disease events were defined as recognized myocardial
infarction, coronary heart disease death, and stroke. A panel of 3 physicians
reviewed each CVD event according to preestablished criteria.24
Diabetes was diagnosed as either fasting plasma glucose level of at
least 126 mg/dL (7.0 mmol/L) (offspring examinations), nonfasting plasma glucose
level of at least 200 mg/dL (11.1 mmol/L) (cohort examinations), or treatment
with insulin or an oral hypoglycemic agent. Participants with a history of
ketoacidosis or age at onset of younger than 30 years were excluded (n = 16).
Age- and sex-adjusted incidence rates for CVD events (per 10 000
person-years) were calculated for each period; standard errors and 95% confidence
intervals (CIs) were computed using the bootstrap bias-corrected and accelerated
method,25 using SAS software, version 8.2.26 Period-specific incidence rates were compared by
calculating the percentage decline among participants with and without diabetes
between the earlier and later periods. Proportional hazards models were used
to examine whether the hazard ratio for diabetes as a CVD risk factor changed
by period and included adjustment for age, sex, systolic blood pressure, hypertension
treatment, current smoking, total cholesterol level, and body mass index.
Time periods were pooled in the proportional hazards models and the interaction
of time period and diabetes was used to test the null hypothesis of a constant
hazard ratio for diabetes among time periods (ie, no interaction between the
time period and diabetes). Secondary analyses were performed by sex and age.
A 2-tailed P<.05 was considered statistically
significant for all analyses.
Since diabetes was defined differently in the offspring and original
Framingham cohorts, a sensitivity analysis was performed to examine the effect
of changing the offspring cohort’s diabetes definition (fasting glucose
level ≥126 mg/dL) to at least 140 mg/dL and at least 160 mg/dL.
To reduce the influence of aging in our closed cohort and to ensure
that the age distributions across all decades overlapped, we restricted our
analysis to participants between ages 45 and 64 years at index examinations.
There were 4005 nondiabetic participants in the earlier time period
and 113 participants who had diabetes, compared with 3746 and 317 participants
without and with diabetes in the later time period. Participants with diabetes
tended to be older, have higher blood pressure, and were more likely to be
obese. In the early compared with the later period, participants with and
without diabetes both experienced significant declines in systolic blood pressure
and total cholesterol level (Table 1).
In the earlier period, the age- and sex-adjusted incidence rate for
CVD among participants with diabetes was 286.4 per 10 000 person-years
compared with 146.9 in the later period, a 49.3% decline (Table 2). Among participants without diabetes, the age- and sex-adjusted
incidence rate for CVD in the earlier period was 84.6 per 10 000 person-years
compared with 54.3 in the later period, a 35.4% decline. Participants with
diabetes experienced a greater absolute decline than those without diabetes
(13.9%), but the 95% CI (−21.6% to 37.1%) suggests that the decline
was not significantly different among those with and without diabetes. Women
with diabetes experienced a 52.9% decline in CVD incidence rates (95% CI,
17.2%-88.6%) compared with a 48.4% decline (95% CI, 33.4%-63.3%) among women
without diabetes. Similar trends were observed among men: those with diabetes
experienced a 45.8% decline in CVD incidence rate (95% CI, 8.2%-83.5%) and
those without diabetes experienced a 29.6% decline (95% CI, 17.5%-41.7%).
Period-specific hazard ratios for diabetes as a CVD risk factor were
computed (Table 3). The multivariable-adjusted
hazard ratio for diabetes as a CVD risk factor decreased slightly from the
earlier to the later period (2.68 to 1.96). However, the interaction terms
between time period and diabetes were not significant (P = .15). Results were similar when sex-specific analyses
were performed (data not shown).
Sensitivity analyses were conducted in which the definition of diabetes
was adjusted to a fasting glucose level of at least 140 mg/dL or at least
160 mg/dL in the offspring sample. The time period × diabetes
interaction term remained statistically nonsignificant, with P values for all models greater than.65 (data not shown).
Adults with diabetes have experienced a 50% reduction in the rate of
incident CVD, although persons with diabetes have remained at a consistent,
approximate 2-fold excess for CVD events compared with those without diabetes.
Adults without diabetes have had a smaller but statistically similar 35% reduction
in CVD event rates. Patients with diabetes have benefited in a similar manner
to those without diabetes during the decline in CVD rates in the US population
over the last several decades. Although gains have been made, substantial
opportunity remains for additional progress to reduce the high absolute risk
of CVD events in persons with diabetes.
The results of our study differ from those previously published, which
have suggested that adults with diabetes have experienced less declines in
CVD risk than those without diabetes.12- 15 Differences
in our findings may be attributed to the longer duration of follow-up, and
comparison groups composed of older as well as more contemporary data and
a different outcome measure. For instance, Gu et al12 compared
CVD mortality rates between 1971-1975 and 1982-1984. Our earlier time period
contains data from the 1950s, and our later time period uses data collected
as recently as 2000, allowing a much longer period over which to detect declines
in CVD event rates. In addition, other studies have relied on self-reported
physician diagnosis of diabetes12,15 or
medical record review for mention of diabetes,14 whereas
our diabetes diagnosis is derived from routine screening by use of glucose
measures and direct questioning of participants. In addition, other studies
that have reported less declines in CVD risk for those with diabetes have
relied on death certificates to obtain cause of death.12- 15 Death
certificates have been shown to overestimate deaths attributed to coronary
heart disease by more than 24%.27 Furthermore,
reporting of diabetes on death certificates is not random.28,29 Thus,
if a death certificate of a descendent with diabetes is more likely to record
a CVD death, the risk of CVD death among persons with diabetes may be inflated.
Since coding practices of death certificates change over time, heightened
awareness of diabetes as a CVD risk factor may have led to increases in the
attribution of CVD deaths to diabetes in more recent periods.
A key question raised by prior studies is whether the presence of diabetes
reduces the benefit of advances in CVD prevention and treatment. Our data
do not support this claim. When comparing CVD risk factors from the earlier
vs the later period, we demonstrate significant declines for important CVD
risk factors, including systolic blood pressure and total cholesterol. These
findings occur in the setting of data from clinical trials demonstrating significant
benefits of CVD risk factor reduction among diabetics. The UK Prospective
Diabetes Study showed that blood pressure control reduced the risk of death
from diabetes.30 Results from the Heart Outcomes
Prevention Evaluation (HOPE) and MICRO-HOPE studies showed that ramipril reduced
CVD events by 25% among diabetics.31 A multifaceted
intervention regarding CVD risk factor reduction among patients with type
2 diabetes reduced CVD events by 50%.32 Among
large clinical trials, subgroup analyses of patients with diabetes have demonstrated
higher absolute reductions in CVD outcomes compared with those without diabetes,
including blood pressure33 and lipid control.34,35 Thus, data from clinical trials provide
the mechanism by which adults with and without diabetes can experience reductions
in CVD incidence.
Despite significant declines in CVD risk associated with diabetes, adults
with diabetes are still at an approximate 2-fold risk of CVD events compared
with those without diabetes. Ongoing efforts remain necessary to promote aggressive
CVD risk reduction among adults with diabetes. Data from the Third National
Health and Nutrition Examination Survey show that adults with diabetes are
not treated optimally.36 Eighteen percent of
participants had poor glycemic control (hemoglobin A1c >9.5%),
34.3% had blood pressure greater than 140/90 mm Hg, and 58% had low-density
lipoprotein cholesterol levels greater than 130 mg/dL. Given that the prevalence
of diabetes is increasing,37- 42 it
is critical that efforts be made to implement findings from clinical trials
to promote CVD risk factor reduction.
Some limitations in our data exist. Our study sample is not nationally
representative, nor is it ethnically diverse. However, the relations of risk
factors to cardiac outcomes observed in Framingham have been validated in
several ethnically and geographically diverse cohorts and were found to be
applicable.43,44 We were unable
to rely on a standard definition of diabetes in both of our study periods.
The earlier period is composed of participants with diabetes diagnosed predominantly
by nonfasting glucose samples, whereas the later period consists of diabetes
diagnoses made by both nonfasting and fasting glucose samples. We have tried
to circumvent this issue by conducting a sensitivity analysis. Given that
our data are not substantially different, we do not believe that this difference
in diabetes diagnosis can fully account for our findings.
We report a 50% reduction in the rate of incident CVD events among adults
with diabetes. Adults with and without diabetes have benefited similarly during
the decline in CVD rates in the US population over the last several decades.
However, the absolute risk of CVD among those with diabetes remains 2-fold
greater compared with persons without diabetes. Both aggressive treatment
of conventional CVD risk factors and further research on diabetes-specific
factors contributing to CVD risk are needed to further reduce the high absolute
risk of CVD still experienced by persons with diabetes.
Corresponding Author: Caroline S. Fox, MD,
MPH, Framingham Heart Study, 73 Mt Wayte Ave, Suite 2, Framingham, MA 01702-5827
Author Contributions: Dr Fox had full access
to all of the data in the study and takes responsibility for the integrity
of the data and the accuracy of the data analysis.
Study concept and design: Fox, Sorlie, Savage.
Acquisition of data: Levy, D’Agostino,
Analysis and interpretation of data: Fox, Coady,
Sorlie, Meigs, D’Agostino, Wilson, Savage.
Drafting of the manuscript: Fox.
Critical revision of the manuscript for important
intellectual content: Coady, Sorlie, Levy, Meigs, D’Agostino,
Statistical analysis: Fox, Coady, Sorlie, D’Agostino,
Obtained funding: Levy.
Administrative, technical, or material support:
Meigs, D’Agostino, Savage.
Study supervision: Levy, Meigs, Wilson.
Funding/Support: This work was supported by
the National Heart, Lung, and Blood Institute’s Framingham Heart Study
(N01-HC-25195). Dr Meigs is supported by an American Diabetes Association
Career Development Award.
Role of the Sponsor: Drs Fox, Sorlie, and Savage
and Mr Coady, as employees of the National Heart, Lung, and Blood Institute,
contributed to the design and conduct of the study, the collection, analysis,
and interpretation of the data, and the preparation, review, and approval
of the manuscript.