Study Sample

We evaluated participants who attended the 6th quadrennial Framingham Offspring Study examination (1995 to 1998).⁷ At that examination, attendees underwent routine medical history and physical examination, anthropometry, assessment of risk factors, and testing for subclinical disease (see Subclinical Disease Measures and Subclinical Disease Score below). Height, weight, and waist circumference (WC) were routinely measured with standardized protocols. The study was approved by the Institutional Review Board at Boston Medical Center, and all participants gave written informed consent.

Of 3532 attendees, we excluded 1594 participants for the following reasons: prevalent CVD (n=415); unavailable data for: ECGs (n=6), measurement of urinary albumin excretion (n=460), ankle-brachial blood pressure (n=49), carotid ultrasound data (n=70), echocardiographic data (n=587); or a BMI <18.5 kg/m² (n=7). After exclusions, 1938 individuals (mean age, 57.4 years; 59% women) remained eligible. The clinical features and incidence of overt CVD were compared in participants excluded because of unavailable data on subclinical disease measures and those included in our sample (see Data Supplement Table I and Results).

Subclinical Disease Measures and Subclinical Disease Score

The measures and definitions of subclinical disease were selected on the basis of published literature and their availability at the 6th examination (Table 1). Sex-specific Cornell voltage criteria were applied to computerized standard 12-lead resting ECGs to evaluate left ventricular hypertrophy (LVH).⁸ Routine transthoracic echocardiography was performed, and M-mode measurements of the left ventricle were obtained with a leading-edge–to–leading-edge technique.⁹ Left ventricular ejection fraction was estimated on the basis of visual assessment of contractile performance in multiple views.⁹ Inter- and intrareader correlations of echocardiographic measurements were excellent.¹⁰ Carotid ultrasound was performed according to a standardized protocol¹¹ with a high-resolution 7.5-MHz transducer for the common carotid artery (CCA) and a 5.0-MHz transducer for the carotid bulb and internal carotid artery.¹² Intima-media thickness (IMT) was measured with gated diastolic images of the left and right carotid arteries at the level of the distal common carotid artery, the carotid artery bulb, and the proximal 2 cm of the internal carotid artery. The maximal IMT at each site was defined as the mean of the maximal IMT measured at the near and far walls of the vessel. The internal carotid artery IMT was defined as the mean of the maximal IMT measurements for the carotid artery bulb and the internal carotid artery on the right and left sides. On the basis of 25 replicate readings by 2 separate readers, intraclass correlation coefficients for the maximum internal carotid artery and common carotid artery IMT were 0.86 and 0.90, respectively.¹²

Ankle-brachial blood pressure measurements were obtained in standardized fashion with an 8-MHz Doppler pen probe and an ultrasonic Doppler flow detector (Parks Medical Electronics Inc., Aloha, Ore).¹³ Urinary albumin excretion was assessed by estimation of the albumin-to-creatinine ratio (UACR) on a spot urine sample. The urinary albumin concentration was measured with an immunoturbidometric assay (Tina-Quant Albumin Assay, Roche Diagnostics), and the urinary creatinine concentration with a modified Jaffe method. Intraassay coefficients of variation were 7.2% and 2.3%, respectively, for the 2 assays.¹⁴ The UACR measured in a spot urine sample is highly correlated with 24-hour urine albumin excretion.¹⁵

We chose to dichotomize subclinical disease measures with cut points described in the literature, even though these measures have a continuous distribution. Such an approach facilitates the assessment of prevalence of subclinical disease, and simplifies the evaluation of its prognostic importance. An abnormal test on any of the 5 available measures (Table 1) was regarded as indicative of subclinical disease. In the case of echocardiography and carotid sonography, presence of any abnormality (of the 3 potential abnormalities) was considered to reflect subclinical disease. Apart from the assessment of subclinical disease as a binary condition, we also constructed a subclinical disease score, which ranged from 0 (no subclinical disease on any test) to 5 (evidence of subclinical disease on all tests), based on the number of abnormal tests. To simplify interpretation, each of the tests was weighted equally, similar to the approach used previously by Kuller et al.²¹

Follow-Up and Outcome Events

The follow-up period was from the baseline examination (1995 to 1998) to December 31, 2005. All participants were under longitudinal surveillance for the occurrence of CVD events. An end point adjudication committee, which comprised experienced investigators, reviewed hospitalization and physician office visit records for all suspected CVD events.

The outcome for the present investigation was the occurrence of a first overt CVD event on follow-up, defined as a composite of coronary heart disease (recognized or unrecognized myocardial infarction, angina pectoris, coronary insufficiency, or coronary heart disease death), cerebrovascular disease (stroke or transient ischemic attack), congestive heart failure, and intermittent claudication. The diagnostic criteria for CVD events have been detailed elsewhere.²²

Statistical Analyses

Participants were categorized into 3 BMI groups (normal weight: BMI, 18.5 to <25.0 kg/m²; overweight: BMI, 25.0 to <30 kg/m²; obese: BMI, ≥30.0 kg/m²)²³; and into 2 WC groups (normal WC, <88 cm [women] or <102 cm [men]; increased WC, ≥88 cm (women) or ≥102 cm [men]).²⁴ First, we determined the prevalence of different measures of subclinical disease in each BMI/WC category. Age- and sex-adjusted logistic regression models were constructed to assess the cross-sectional associations of overweight, obesity, and increased WC with the prevalence of subclinical disease (a score of 1 or more), with participants who had normal BMI or normal WC, respectively, as referent.

Next, we calculated age- and sex-adjusted incidence rates of CVD for the BMI and WC categories overall, and stratifying by presence of subclinical disease. We used age- and sex-adjusted Cox regression to assess the risk of a first CVD event in participants with overweight or obesity (participants with normal BMI as referent), and in individuals with increased WC (participants with normal WC as referent) after confirmation that the assumption of proportionality of hazards was met. We did not evaluate multivariable models that incorporated known risk factors (such as blood pressure, diabetes mellitus, and serum cholesterol) because these factors are strongly associated with overweight and obesity, and therefore are on the causal path from excess adiposity to subclinical disease to overt CVD; our objective was not to assess if subclinical disease predicted CVD above and beyond traditional risk factors. We performed the following analyses of the relations of BMI or WC categories, subclinical disease, and CVD incidence hierarchically: A, with no adjustment for subclinical vascular disease (models included only age, sex, and BMI or WC category); B, with adjustment for subclinical disease modeled as a dichotomous variable (models included presence [score≥1] versus absence of subclinical disease [score=0]) in addition to age, sex, and BMI or WC category); C, with adjustment for subclinical disease score as an ordinal variable (models included subclinical disease score, age, sex, and BMI or WC category); and D, with stratification of the BMI or WC category by presence versus absence of subclinical vascular disease, adjusting for age and sex.

To assess the incremental utility of subclinical disease for prediction of CVD risk, we calculated the C-statistic for models A through C after incorporation of BMI as a continuous variable. The C-statistic from a Cox model is conceptually analogous to the area under a receiver-operating characteristic curve estimated for logistic models. This value estimates the probability that the evaluated model correctly identifies the individual with disease from 2 randomly paired individuals (one with and one without disease).

We examined effect modification by tests of the significance of the following 2-way interaction terms that incorporated subclinical disease and age, sex, and BMI/WC category. In secondary analyses, we repeated our analyses after adjustment for smoking in addition to age and sex. We also evaluated multivariable models that related individual measures of subclinical vascular disease to the risk of CVD. Only measures with a continuous distribution were included in these analyses. A 2-sided probability value <0.05 was considered statistically significant. All analyses were performed with SAS 9.1 (SAS Institute, Cary, NC).

The authors had full access to and take full responsibility for the integrity of the data. All authors have read and agree to the manuscript as written.

Cross-Sectional Prevalence of Subclinical Vascular Disease by Degree of Generalized and Abdominal Obesity

Forty percent of overweight and nearly 50% of obese participants had evidence of subclinical disease in at least 1 test (as compared with 29% of individuals with normal BMI; Table 2). On a parallel note, 45% of participants with increased WC had evidence of subclinical disease (as compared with 32% of participants with normal WC). The prevalence of subclinical disease increased with BMI and WC category (Figure).

• Download figure
• Open in new tab
• Download powerpoint

Sex-specific prevalences of subclinical disease in individuals by BMI (normal, 18.5 to <25.0 kg/m²; overweight, 25.0 to <30.0 kg/m²; obese, ≥30.0 kg/m²), and WC (normal: <88 cm for women, <102 cm for men; increased: ≥88 cm for women, ≥102 cm for men).

In age- and sex-adjusted logistic regression models, overweight and obesity were significantly associated with greater prevalences of increased LV mass, increased carotid artery IMT, and carotid artery stenosis (Table 3). Obesity, but not overweight, was significantly associated with electrocardiographic LVH. Overall, overweight was associated with a 1.7-fold odds of having at least 1 subclinical disease abnormality compared with the normal BMI group, and obesity was associated with a 2.8-fold odds.

Similarly, increased WC was significantly associated with electrocardiographic LVH, increased LV mass, and increased carotid artery IMT and carotid artery stenosis (Table 3). Overall, participants with increased WC had a 1.7-fold odds of presence of at least 1 subclinical disease abnormality compared with individuals with normal WC. The results shown in Table 3 remained robust on additional adjustment for smoking (data not shown).

Prognostic Significance of Subclinical Disease by Degree of Generalized and Abdominal Obesity

Upon follow-up (mean 7.2 years), 139 individuals (7.1%; 46% of events in women) developed a first overt CVD event (recognized myocardial infarction, n=27; unrecognized myocardial infarction, n=2; angina pectoris, n=48; coronary insufficiency, n=4; coronary heart disease death, n=3; stroke, n=14; transient ischemic attack, n=13; congestive heart failure, n=19; intermittent claudication, n=9). The age- and sex-adjusted incidence of overt CVD rose across BMI categories and with increased WC (Table 4). The presence of subclinical disease was associated with an increased CVD event rate in all groups. Interestingly, overweight participants with subclinical disease experienced CVD incidence rates comparable to obese participants with subclinical disease. Also, the incidence rates of overt CVD in participants with obesity but with no subclinical disease were similar to rates in participants with normal BMI but with subclinical disease (P=0.79). A trend existed toward lower rates of overt CVD in individuals with increased WC who had no subclinical disease compared participants with normal WC but with subclinical disease (P=0.06).

In age- and sex-adjusted models, obesity was borderline significantly associated with an increased CVD risk in models that did not adjust for presence of subclinical disease (Table 5, BMI Model A). Overweight was not significantly associated with CVD incidence. On adjustment for the presence of subclinical disease, the association of obesity with CVD risk was attenuated and became statistically nonsignificant (Table 5, BMI Model B). In these models, the presence of subclinical disease was associated with a >2-fold risk of overt CVD (Table 5, BMI Model B). In models adjusted for the subclinical disease score (from 0 to 5), each point increase in the score was associated with a 70% increased risk of CVD (Table 5, BMI Model C). In these models, the associations of overweight and obesity with overt CVD remained statistically nonsignificant.

In models that stratified BMI groups by presence versus absence of subclinical vascular disease (Table 5, BMI Model D), overweight or obese individuals with subclinical disease experienced a 2.6-fold risk of overt CVD, compared with participants with a normal BMI without subclinical disease (referent). Furthermore, presence of subclinical disease was a significant predictor of overt CVD also in the individuals with a normal BMI. Conversely, overweight or obese people without any subclinical disease did not experience a statistically significant increased risk of overt CVD risk compared with the referent group.

The results of analyses to assess the risk of overt CVD in participants with increased WC paralleled those for the BMI categories (Table 5). When stratified by presence of subclinical disease (Table 5, WC Model D), individuals with increased WC who had subclinical disease experienced a 2.8-fold risk of overt CVD, compared with participants with normal WC and no subclinical disease. In contrast, individuals with increased WC but without subclinical disease had risk of overt CVD comparable to the referent group. The results shown in Table 5 remained essentially unchanged on additional adjustment for smoking (data not shown).

Additional Analyses

The C-statistic for a model that incorporated age, sex and BMI (modeled as a continuous variable) was 0.70. When subclinical disease was added as a dichotomous variable, the C-statistic increased to 0.72 (P=0.02 for comparison with the model without subclinical disease). Similarly, when subclinical disease score was incorporated as an ordinal variable, the C-statistic was 0.73 (P=0.005 for comparison with the model without subclinical disease). With regard to the prognostic impact of subclinical disease, we did not observe effect modification by age, sex, or BMI/WC category (P>0.40 for all interaction terms).

In secondary analyses to relate individual subclinical disease measures to CVD risk, LV mass indexed by height, carotid artery IMT, ankle-brachial index, and UACR were positively associated with overt CVD, whereas fractional shortening was inversely related. The ECG Cornell voltage was not associated with overt CVD (Table II in the online Data Supplement).

In analyses to assess potential selection bias that resulted from exclusion of participants with unavailable subclinical disease data, we observed that excluded participants were 3 years older, more likely to be men, and had a greater mean BMI (Table I in the online Data Supplement). However, the age- and sex-adjusted CVD incidence rate in excluded participants (8.0; 95% confidence interval [CI], 6.4 to 9.6) did not differ significantly (P=0.31) from the incidence rate in our study sample (7.0; 95% CI, 5.8 to 8.2).

Principal Findings

In the present investigation, we comprehensively characterized the extent and prognostic importance of subclinical disease by the degree of generalized or abdominal obesity. Several important observations in our study should be noted. First, two fifths of overweight people and nearly half of participants with generalized or abdominal obesity had evidence of subclinical disease in at least 1 of the 5 tests evaluated. Second, individuals with overweight, obesity, or increased WC and evidence of subclinical disease experienced a 2.6- to 2.8-fold risk of developing overt CVD compared with people with normal BMI or WC and no subclinical disease. Third, nearly one third of individuals with a normal BMI or normal WC in our middle-aged sample had evidence of subclinical disease. Notably, the presence of subclinical disease was associated with more than twice the risk of development of overt CVD even in individuals with a normal BMI or WC. Fourth, upon adjustment for presence of subclinical disease, the associations of obesity or increased WC with CVD risk were attenuated, which suggests that subclinical disease may be a key contributor to the cardiovascular risk associated with obesity. In the present study, subclinical disease accounted for 35% and 40% of the increased risk associated with overweight and obesity, respectively, on the basis of the relative decrease of the β coefficients (for overweight and obesity) in models with and without subclinical disease. Fifth, the incorporation of subclinical disease measures in models that contained data on age, sex, and BMI resulted in a significant but modest increase in the C-statistic. This emphasizes that our findings are more important from a pathophysiological point of view than from the perspective of risk prediction. Finally, no evidence was present of effect modification by BMI category of the relation of subclinical disease to overt CVD (similar effect sizes for the association of subclinical disease to overt CVD regardless of BMI), which indicates that subclinical disease portends an increased risk of overt CVD across the BMI strata.

Comparison With Previous Studies

Several previous studies have established that subclinical disease is more common in obese and overweight people compared with individuals with a normal BMI.^3–6 Also, obesity has been shown to track with progression of subclinical coronary atherosclerosis.²⁵ However, prior studies focused typically on single measures of subclinical disease, such as LVH,³ LV dysfunction,⁴ carotid atherosclerosis,⁵ or microalbuminuria.⁶ The present study used a panel of 5 tests to characterize the prevalence of subclinical disease in different vascular territories and target organs. One prior study examined the extent of subclinical disease in different vascular compartments in obese individuals with type 2 diabetes mellitus.²⁶ In that investigation, BMI was a strong determinant of vascular stiffness, but no significant associations were observed between BMI and carotid artery IMT, extent of carotid artery plaques, coronary calcium score, or aortic calcium score. However, that study was based on a small sample (n=52) of individuals with diabetes mellitus and obesity, with BMI values that ranged from 27 to 49 kg/m², which suggests that the statistical power to detect modest associations was limited and could explain the lack of associations in that study.

Subclinical disease has been established to be a powerful predictor of future overt CVD in elderly individuals^19,27 and in individuals with diabetes mellitus.²⁸ However, to our knowledge, the prognostic significance of subclinical disease in overweight and obese individuals has not been studied before. Investigators from the Cardiovascular Health Study formulated a subclinical disease index based on several tests: presence of low ankle-brachial blood pressure, carotid artery stenosis, increased carotid artery IMT, major ECG abnormalities, abnormal wall motion or ejection fraction on echocardiography, and positive responses to the Rose angina or claudication questionnaire in the absence of history of angina or claudication.²¹ In the present study, we modified this index by incorporation of echocardiographic LVH because it is more prevalent in the community compared with electrocardiographic LVH²⁹ and by addition of microalbuminuria, which is an indicator of endothelial dysfunction and/or target organ damage.^15,30

Distribution of Subclinical Disease in Obesity

In the present study, obesity was associated with a greater prevalence of electrocardiographic LVH, increased LV mass, increased carotid artery IMT, and carotid artery stenosis. These findings are consistent with prior studies that have demonstrated associations of obesity and electrocardiographic³¹ or echocardiographic³ LVH, increased carotid artery IMT,⁵ and carotid artery stenosis.⁵

Some of the subclinical disease measures in the present study were not significantly associated with obesity. However, all components of subclinical disease were selected a priori on the basis of a review of the literature. A potential reason for lack of association between low ankle-brachial blood pressure and obesity (or increased WC) may be the low prevalence of this abnormality in our middle-aged study sample, which limits our statistical power to detect modest associations. With regard to the lack of association of microalbuminuria and obesity in the present investigation, it is noteworthy that prior studies that examined this relation have yielded disparate results; some studies reported a direct association between BMI and urinary albumin,⁶ others noted an inverse association,³² and some others failed to observe any significant relation.³³ One potential explanation for the lack of association between microalbuminuria and obesity in the present study may be the indexation of urine albumin excretion by urinary creatinine concentrations for calculation of UACR. Obese individuals may have a higher fat-free mass,³⁴ which may be associated with greater urinary creatinine and thereby lower the proportion of obese individuals with microalbuminuria as assessed with the UACR. Another reason may be that we evaluated UACR on a single occasion, and urinary albumin excretion is variable within individuals. Misclassification that arose from such intraindividual variability may have reduced our ability to detect associations.

Strengths and Limitations

The strengths of the present study include the large community-based sample, the longitudinal surveillance for CVD events blinded to subclinical disease and BMI/WC status, and the comprehensive assessment of subclinical disease. Nonetheless, several limitations of the present study must be acknowledged. We did not evaluate coronary artery calcification as a marker of subclinical disease because such imaging was not performed at the baseline examination. Our definition of the subclinical disease score is arbitrary, and not all investigators may agree on which components should be included. Also, because our study sample was comprised of white and middle-aged participants, the generalizability of our findings to other ethnicities or age groups is unknown. Furthermore, to maximize statistical power and to reduce multiple statistical testing, we evaluated only a composite CVD outcome. Last, the exclusion of a large proportion of individuals because of unavailable subclinical disease measures is an unavoidable limitation of large epidemiological studies coupled with the requirement for availability of data for all 5 tests of subclinical disease measures. However, this means that our results must be interpreted with caution with regard to their generalizability.