Lloyd-Jones DM, Nam B, D'Agostino, Sr RB, Levy D, Murabito JM, Wang TJ, Wilson PWF, O'Donnell CJ. Parental Cardiovascular Disease as a Risk Factor for Cardiovascular Disease in Middle-aged AdultsA Prospective Study of Parents and Offspring. JAMA. 2004;291(18):2204-2211. doi:10.1001/jama.291.18.2204
Author Affiliations: Department of Preventive Medicine and Division of Cardiology, Feinberg School of Medicine, Northwestern University, Chicago, Ill (Dr Lloyd-Jones); National Heart, Lung, and Blood Institute's Framingham Heart Study, Framingham, Mass (Drs Lloyd-Jones, Nam, D'Agostino, Levy, Murabito, Wang, Wilson, and O'Donnell); Statistics and Consulting Unit (Drs Nam and D'Agostino), Department of Epidemiology and Preventive Medicine, Boston University (Drs D'Agostino, Levy, Murabito, Wilson, and O'Donnell), and Section of General Internal Medicine, Boston University School of Medicine (Dr Murabito); Cardiology Division (Drs Wang and O'Donnell), Massachusetts General Hospital, Harvard Medical School, Boston; and National Heart, Lung, and Blood Institute, Bethesda, Md (Drs Levy and O'Donnell).
Context Whether parental cardiovascular disease confers increased risk independent
of other risk factors remains controversial. Prior studies relied on offspring
report, without complete validation of parental events.
Objective To determine whether parental cardiovascular disease predicts offspring
events independent of traditional risk factors, using a prospective design
for both parents and offspring, and uniform criteria to validate events.
Design Inception cohort study.
Setting Framingham Heart Study, a US population-based epidemiologic cohort begun
in 1948 with the offspring cohort established in 1971.
Participants All Framingham Offspring Study participants (aged ≥30 years) who
were free of cardiovascular disease and both parents in the original Framingham
Main Outcome Measures We examined the association of parental cardiovascular disease with
8-year risk of offspring cardiovascular disease, using pooled logistic regression.
Results Among 2302 men and women (mean age, 44 years), 164 men and 79 women
had cardiovascular events during follow-up. Compared with participants with
no parental cardiovascular disease, those with at least 1 parent with premature
cardiovascular disease (onset age <55 years in father, <65 years in
mother) had greater risk for events, with age-adjusted odds ratios of 2.6
(95% confidence interval [CI], 1.7-4.1) for men and 2.3 (95% CI, 1.3-4.3)
for women. Multivariable adjustment resulted in odds ratios of 2.0 (95% CI,
1.2-3.1) for men and 1.7 (95% CI, 0.9-3.1) for women. Nonpremature parental
cardiovascular disease and parental coronary disease were weaker predictors.
Addition of parental information aided in discriminating event rates, notably
among offspring with intermediate levels of cholesterol and blood pressure,
as well as intermediate predicted multivariable risk.
Conclusions Using validated events, we found that parental cardiovascular disease
independently predicted future offspring events in middle-aged adults. Addition
of parental information may help clinicians and patients with primary prevention
of cardiovascular disease, when treatment decisions may be difficult in patients
at intermediate risk based on levels of single or multiple risk factors. These
data also support further research into genetic determinants of cardiovascular
An offspring report of a positive parental history of cardiovascular
disease (CVD), particularly if premature at onset, is a widely accepted risk
factor for offspring cardiovascular events. Current guidelines1,2 recommend
consideration of a positive parental history of premature coronary heart disease
when deciding whether to initiate antihyperlipidemic or antihypertensive therapy
for primary prevention. However, the true magnitude of independent risk that
is conferred by the occurrence of parental CVD remains controversial. This
uncertainty exists in large part because available data examining the association
between parental and offspring CVD are derived from retrospective case-control
studies3- 7 or
longitudinal studies8- 20 relying
on offspring self-report, with limited or absent validation of parental events.
Offspring report of parental history may be highly unreliable,21- 25 in
part due to recall bias, and may lead to inflated estimates of risk associated
with parental CVD. The effect of parental CVD on offspring risk across strata
of individual risk factors is also not well understood.
We sought to determine whether validated parental occurrence of CVD
is an independent, prospective predictor of offspring cardiovascular events.
Accurate estimates of the relationship between parental and offspring cardiovascular
events could provide stronger evidence for the clinical emphasis on prevention
in patients with a positive parental history. The Framingham Heart Study is
uniquely suited to perform such an analysis given its longitudinal follow-up
and exhaustive documentation of events in both parents and offspring.
The Framingham Heart Study was established in 1948, when 5209 residents
of Framingham, Mass, aged 28 to 62 years, were enrolled in a prospective epidemiologic
cohort study. Members of this original cohort have undergone follow-up evaluations
every 2 years. In 1971, an additional 5124 participants (offspring of original
cohort subjects and their spouses) were enrolled in the Framingham Offspring
Study (offspring cohort). These participants have undergone follow-up evaluations
every 4 years. Study design and entry criteria for both cohorts have been
detailed elsewhere.26,27All participants
have provided written informed consent at each examination, and all study
protocols have been approved by the Institutional Review Board of Boston University
School of Medicine.
We included all Framingham Offspring Study participants who were aged
30 years or older and free of prevalent CVD and for whom both parents were
followed up in the original cohort. Parental subjects were followed up from
1948 to 2001, and offspring participants were followed up from 1971 to 2001.
Given the structure of follow-up examinations, we elected to study the 8-year
incidence of cardiovascular events in the offspring cohort. We focused on
atherosclerotic cardiovascular events since atherosclerosis is believed to
be the mechanism underlying familial aggregation. Using previously published
Framingham Heart Study criteria28 to validate
parental and offspring events, we defined a cardiovascular event as the occurrence
of coronary death, myocardial infarction, coronary insufficiency, angina pectoris,
atherothrombotic stroke, intermittent claudication, or cardiovascular death.
Hard coronary heart disease events were defined as coronary death, myocardial
infarction, or hospitalized coronary insufficiency only. Premature parental
CVD was defined as the occurrence of a validated parental event prior to an
offspring baseline examination and before age 55 years in a father or age
65 years in a mother. These age cut points were drawn from the recommendations
of the National Cholesterol Education Program Third Adult Treatment Panel
(ATP-III)1 and Seventh Joint National Committee
on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure
We used pooled logistic regression analyses to examine the risks of
incident offspring cardiovascular events associated with positive parental
CVD over 8 years after an examination. This person-time method of pooling
person-examinations accounts for time-dependent covariance of risk factors
and parental cardiovascular events and has been shown to provide valid estimates
of effect similar to using time-dependent Cox analyses. This is a robust assumption,
as demonstrated by D'Agostino et al.29 Participants
contributed a mean of 2.2 person-examination cycles. For all logistic regression
analyses, the reference group consisted of offspring participants with no
parental CVD prior to the time of the offspring examination. Odds ratios (ORs)
and 95% confidence intervals (CIs) were calculated in unadjusted analyses,
as well as after adjustment for offspring age, age with other individual categories
of cardiovascular risk factors, and all other risk factors in the model. These
analyses were repeated examining parental nonpremature CVD and parental hard
coronary heart disease as risk factors for offspring CVD. Risk factor covariates,
chosen a priori for inclusion in the models and updated using pooled logistic
analyses, were: offspring age, systolic blood pressure, total cholesterol/high-density
lipoprotein cholesterol (HDL-C) ratio, body mass index, presence of diabetes,
current smoking, and use of antihypertensive drug therapy. In the multivariable
models, there was only 1 significant first-order interaction (between maternal
CVD and antihypertensive therapy in men). Inclusion of the interaction term
did not alter the substance of our findings, so it was excluded from the models.
A multivariable risk score was calculated using weighted coefficients
for each covariate, with the exception of parental CVD information. Offspring
participants were stratified into quintiles of predicted multivariable risk
for CVD. Eight-year event rates were then compared for participants in each
quintile, as well as across clinical strata of individual risk factors, according
to presence or absence of parental CVD, using χ2 tests. C statistics were calculated for the model before and after
inclusion of parental CVD information. All statistical analyses were performed
using SAS statistical software version 8 (SAS Institute, Cary, NC). A 2-tailed P value less than .05 was defined as statistically significant.
The study sample included 1128 men and 1174 women offspring participants
who were free of CVD at a mean age of 44 years. Characteristics of these individuals
are shown in Table 1. During follow-up,
164 men and 79 women had incident cardiovascular events, of which 14 (5.7%)
were coronary deaths (11 [4.5%], sudden; 3 [1.2%], nonsudden), 5 (2.1%) were
other cardiovascular deaths, 76 (31.3%) were nonfatal myocardial infarction
or coronary insufficiency, 71 (29.2%) were angina pectoris, 39 (16.0%) were
stroke, and 38 (15.6%) were intermittent claudication. There was a similar
distribution of event types among parents.
The ORs for offspring cardiovascular events associated with premature
parental CVD are shown in Table 2.
After adjustment for offspring age, the ORs associated with parental CVD were
2.6 (95% CI, 1.7-4.1) for men and 2.3 (95% CI, 1.3-4.3) for women. Additional
adjustment for individual risk factors did not substantially attenuate the
risk associated with parental CVD further (Table 2). After multivariable adjustment for offspring age and all
other risk factors, parental occurrence of CVD remained a significant predictor
of offspring events in men. In women, the multivariable association was of
borderline statistical significance. In age-adjusted and multivariable-adjusted
analyses, the ORs for offspring CVD were higher for premature (Table 2) than for nonpremature parental CVD (Table 3).
The occurrence in 1 or more parent of premature hard coronary disease
only was less strongly associated with risk for CVD, with multivariable ORs
of 1.7 (95% CI, 1.0-2.8) for men and 1.2 (95% CI, 0.5-2.5) for women offspring.
Because 1737 of the 2302 offspring were part of a sibship, a given parent's
event may have been counted more than once in these analyses. When we restricted
the analysis to 1 sibling per family (the oldest), the results were not changed
When we stratified participants by levels of individual risk factors,
parental information added substantially to discrimination of observed 8-year
event rates. Table 4 shows the
8-year cardiovascular event rates for offspring men and women, separately
and combined, without and with premature parental CVD. The overall event rate
was 44/1000 (4.4%) over 8 years. Although absolute event rates were lower
for women than they were for men, the overall pattern of effect was similar.
When we stratified by offspring age, parental CVD was associated with significantly
higher offspring event rates with a 3-fold difference for offspring aged 30
to 59 years, and a 2-fold difference for older offspring.
Parental CVD increased offspring risk across all strata of total cholesterol
levels and blood pressure measurements. The greatest relative differences
in event rates were observed among participants with borderline levels of
cholesterol and high-normal blood pressure. Premature parental CVD increased
risk in these intermediate subsets to the point that the 10-year risks of
CVD would exceed 10%, a threshold for treatment used in some clinical guidelines.
Premature parental CVD was also associated with increased risk among both
smokers and nonsmokers, as well as among those who did not have diabetes.
However, parental CVD did not significantly increase risk among diabetic participants,
who already had substantially higher event rates.
Eight-year cardiovascular event rates are displayed in Figure 1 for offspring according to quintile of predicted multivariable
risk. Observed 8-year event rates increased markedly and in a stepwise fashion
from lowest to highest quintile. At very low and very high predicted risk,
the increase in event rates associated with the presence of premature parental
CVD was modest: offspring with favorable risk factor profiles were not at
substantially increased risk despite parental CVD, and offspring with very
unfavorable risk factor profiles remained at high risk even in the absence
of parental CVD. In the intermediate quintiles, premature parental CVD was
associated with significantly higher cardiovascular event rates (Figure 1).
In a separate analysis, we stratified offspring according to their predicted
absolute 10-year risk for hard coronary events, as estimated by the ATP-III
risk equations.1 Premature parental CVD presence
vs absence was associated with increased events among offspring predicted
to be at low risk by ATP-III (n = 3093; 54/1000 vs 18/1000 events over 8 years,
respectively; P<.001). There was a nonsignificant
difference in the ATP-III intermediate-risk group (n = 153; 200/1000 vs 138/1000,
respectively; P = .33), but there were few participants,
particularly women, in the ATP-III intermediate-risk (n = 9 women) and high-risk
(n = 29 women) groups due to the young age of the offspring sample.
Inclusion of premature parental CVD as a covariate altered the c statistic (area under the receiver operating characteristic
curve) for our multivariable model predicting offspring CVD from 0.80 to 0.81
for men and from 0.81 to 0.82 for women. When men and women were combined,
inclusion of premature parental CVD in the multivariable model altered the c statistic from 0.82 to 0.83. The attributable risk percentages
for premature parental CVD were 29.0% in men and 20.6% in women offspring;
for nonpremature parental CVD, they were 20.5% and 5.8%, respectively.
Using a prospective design to ascertain both parental and offspring
events, we found that the occurrence of parental CVD is an independent predictor
of offspring cardiovascular events in middle-aged men and women. After adjustment
for other risk factors, premature CVD in at least 1 parent was associated
with a significant doubling in cardiovascular risk for men and a 70% increase
(nonsignificant) in risk for women over 8 years. Premature parental CVD was
found to discriminate risk best among offspring with intermediate levels of
cardiovascular risk as predicted by individual traditional risk factors or
multivariable risk equations.
We focused on a cohort of middle-aged offspring because parental CVD
is likely to be a greater factor in determining relative CVD risk in younger
than older individuals, as our findings confirm (Table 4). Adjustment for offspring age markedly attenuated the association
of parental and offspring CVD. This may reflect the fact that older subjects
also have older parents, who have had a longer lifespan in which to experience
a cardiovascular event unrelated to familial aggregation. Multivariable adjustment
for age and other traditional risk factors further attenuated the risk associated
with parental CVD, and these covariates accounted for the majority of the
crude risk associated with parental CVD. A significant, but modest, residual
risk remained associated with parental occurrence of CVD, suggesting that
parental CVD is a clinically important aggregate marker of both heritable
risk factors and as yet unmeasured genetic risk factors.
These results should help to inform clinicians and patients about use
of parental history in risk stratification and treatment decisions. It should
be noted that the validation of parental events available for this study is
not available in most clinical settings. Nevertheless, our results shed important
light on the true magnitude of the association between offspring and parental
CVD. It is likely that clinicians place greater importance on positive parental
CVD in younger compared with older patients, which appears appropriate in
terms of relative risk. However, at present it is difficult for clinicians
and patients to know how to assess and incorporate into clinical practice
the risk associated with parental CVD, independent of shared risk factors.
As expected, we observed a substantial, stepwise increase in the incidence
of CVD across quintiles of multivariable risk as estimated by multivariable
equations using a combination of traditional, modifiable risk factors (Figure 1). Recent studies confirm that these
traditional risk factors are present in almost all patients who develop CVD,30,31 and that they account for the majority
of risk.32,33 Thus, clinicians
and patients must continue to focus on proven lifestyle and drug therapies
to modify traditional risk factors and reduce risk.
The addition of parental data may aid in discriminating risk most among
men and women at intermediate levels of predicted risk. Knowledge of parental
CVD status may not change the magnitude of risk substantially for those at
very high or very low predicted risk. For patients with intermediate predicted
risk, however, the additional information provided by positive parental CVD
may change the posttest probability enough to consider altering the treatment
of the patient. Treatment decisions in current guidelines are based on absolute
risk levels. We demonstrate significantly higher 8-year absolute rates of
CVD among participants with borderline cholesterol or blood pressure levels
and premature compared with no parental CVD (Table 4). It is precisely these patients in whom decisions about
lifestyle modification or drug treatment may be the most difficult. Thus,
our data support the emphasis placed by the ATP-III1 and
JNC 7 guidelines on ascertainment of parental history2 to
help guide treatment decisions for primary prevention. Furthermore, our findings
support consideration of clinical trials to assess the benefits of aggressive
risk-factor modification in intermediate risk patients with parental CVD.
Our data also suggest that incorporation of validated parental data
into multivariable functions predicting 10-year absolute risks for cardiovascular
or coronary disease1,34,35 may
increase the predictive accuracy of such functions,36 although
perhaps only to a small extent. The overall c statistic for our multivariable
model increased from 0.82 to 0.83 by adding parental CVD information. It has
proved difficult to make marked improvements in risk stratification over and
above incorporation of traditional risk factors, even with novel markers such
as C-reactive protein.37- 39
Ongoing work is examining the utility of including parental occurrence
of CVD or other novel risk markers in updated Framingham coronary risk functions.
The addition of parental information to traditional risk factors must certainly
be considered in the development of these risk functions. However, in the
formulation of updated risk functions, a number of novel risk markers should
also be considered. It may well be that inclusion of such markers will improve
risk prediction better than incorporation of family history.
There is ample evidence for familial aggregation of traditional risk
factors, as well as associations of parental and sibling CVD with adverse
lipids and other risk factors in offspring.40- 42 Parental
history has also been associated with novel markers of inflammation,43 lipoprotein(a) and fibrinogen,43- 45 and
measures of subclinical atherosclerosis.46- 50 Familial
aggregation of these risk factors suggests genetic influence on the causal
pathways of familial cardiovascular risk and explains the attenuation in ORs
that we observed after multivariable adjustment. The significant residual
risk associated with parental history after multivariable adjustment suggests
that additional pathways for familial clustering merit further investigation.
Numerous case-control studies3- 7 have
reported approximately 2- to 5-fold higher prevalence of a positive familial
history among subjects with manifest CVD than among control subjects. Large,
prospective cohort studies8- 20 also
have generally found a positive association between self-reported parental
or familial history and multivariable-adjusted relative risks for offspring
CVD, with estimates ranging from 0.8 to 2.2. Our results shed further light
on the true magnitude of this association.
Our study benefited from several unique methodologic strengths. First,
as this was a prospective study, parental events were ascertained and validated
independently and prior to the occurrence of offspring events, avoiding recall
bias and inaccuracy inherent in offspring self-report.21- 25 Both
parental and offspring events were validated using consistent, standardized
definitions after review of all medical records by a panel of 3 physicians.28 Many studies have relied on death certificate data
to identify or confirm parental events, which may lead to substantial overdiagnosis
of coronary disease, especially in older decedents.51
Our study also benefited from the long duration of follow-up of both
the parental and offspring family members. Shorter studies would miss parental
events that have not yet occurred. At this point, ascertainment of premature
events in parents is complete, and ascertainment of later parental events,
near-complete. Our study design allowed for updating of parental history information
and repeated measurements of offspring risk factors independent of parental
history. Given these strengths, our estimates of the magnitude of the association
between parental and offspring CVD may represent the most accurate published
The Framingham cohorts are almost exclusively white, which may limit
the generalizability of our findings to other ethnic groups. Rates of use of preventive medications were fairly low, but could
have contributed to reduction in risk associated with parental CVD if subjects
with parental CVD were more likely to receive them. If anything, this would
bias our results toward the null. We had relatively little power to find an
independent relationship in women and to comment on any potential differences
between paternal and maternal CVD as risk factors. Apparent differences in
the magnitude of effect between offspring men and women may reflect the low
number of cases and lower prevalence of risk factors in women. Finally, an
unvalidated offspring report of parental history may not have the same clinical
utility as the validated parental events used in the current study. However,
risk estimates from studies using offspring self-report fall within the bounds
of our study. With the current study design, we attempted to determine the
most accurate possible representation of the association between parental
and offspring CVD.