Author + information
- Received February 18, 2004
- Revision received May 3, 2004
- Accepted May 18, 2004
- Published online September 1, 2004.
- Irene Meissner, MD⁎,⁎ (, )
- Bijoy K. Khandheria, MD†,
- Sheldon G. Sheps, MD‡,
- Gary L. Schwartz, MD‡,
- David O. Wiebers, MD⁎,
- Jack P. Whisnant, MD§,
- Jody L. Covalt, RN∥,
- Tanya M. Petterson, MS¶,
- Teresa J.H. Christianson, BS¶ and
- Yoram Agmon, MD†
- ↵⁎Reprint requests and correspondence:
Dr. Irene Meissner, Department of Neurology, Mayo Clinic, 200 First Street SW, Rochester, Minnesota 55905
Objectives The goal of this study was to investigate whether complex aortic atherosclerosis is associated with increased risk of vascular events in a non-selected population.
Background In selected high-risk patients, aortic atherosclerosis is associated with increased risk of vascular events.
Methods We describe the relationship between simple versus complex (>4-mm thick or mobile debris) aortic atherosclerotic plaques and vascular events during follow-up in a random sample of 585 persons (age ≥45 years) using 1993 to 2000 data from the Stroke Prevention: Assessment of Risk in a Community (SPARC), a prospective population-based longitudinal study.
Results At five-year median follow-up (range, 0.5 to 6.5 years), cardiac events (death, non-fatal myocardial infarction, coronary revascularization, heart failure associated with coronary artery disease) and cerebrovascular events (ischemic fatal and non-fatal strokes, transient ischemic attacks) had occurred in 95 subjects and 41 subjects, respectively. Age, male gender, prior coronary artery disease, higher pulse pressure, and diabetes were significant cardiovascular predictors. Age, prior myocardial infarction, and a history of atrial fibrillation were significant cerebrovascular predictors. Simple aortic plaques (253 persons) were not independently associated with either cardiac or cerebrovascular events. Complex plaques (44 persons) were marginally associated with cardiac events, adjusting for age and gender (hazard ratio [HR], 2.28; 95% confidence interval [CI], 1.11 to 4.68; p = 0.053 for two degrees of freedom [complex and simple plaques vs. no plaques]) but not after adjusting for additional clinical risk factors (HR, 1.22; 95% CI, 0.57 to 2.62; p = 0.64). Complex plaques were associated with cerebrovascular events only univariately.
Conclusions Aortic atherosclerotic plaques are not associated with future cardiac or cerebrovascular events. Aortic atherosclerosis may not be an independent risk factor for vascular events in the general population.
High-resolution imaging afforded by transesophageal echocardiography (TEE) has facilitated the diagnosis and severity grading of aortic atherosclerosis. With the paradigm shift from secondary to primary prevention of stroke, this tool is increasingly applied in the evaluation of risk of stroke and cardiovascular disease. Aortic atherosclerosis is well known to increase with advancing age. Despite the many studies reporting an association between aortic atherosclerosis and increased cardiovascular and stroke morbidity and mortality (1–3), this relationship remains undefined because of the highly selected nature of TEE patients, most with prior strokes, and the variable adjustment for the effects of age and other comorbid conditions.
The Stroke Prevention: Assessment of Risk in a Community (SPARC) cohort was established in 1993 to identify risk factors for cerebral ischemia and cardiovascular disease in the population. The cohort is unique in that it provides comprehensive TEE data on the prevalence of aortic atherosclerosis in a cohesive random sample of the population with comprehensive follow-up (4,5). This report assesses the impact of aortic atherosclerosis as an independent risk factor for subsequent cardiovascular and cerebrovascular events in a population-based sample.
The resources of the Rochester Epidemiology Project in Rochester, Minnesota, were used to enumerate the Olmsted County population 45 years old or older. The study design and initial results of the first phase of SPARC have been previously published (4,5). Of the 1,475 residents initially sampled, 230 were ineligible by predetermined exclusion criteria (terminal illness, dementia precluding informed consent, severe disability, or esophageal disease precluding TEE), and 657 declined to participate. The final SPARC sample consisted of 588 randomly sampled subjects (about one-half of those eligible), who consented to multimodality testing, including medical record review, home medical interview, TEE, carotid ultrasonography, and repeated blood pressure measurements. End point definitions were prespecified before initiation of the study, and updated information was obtained from the medical records of these subjects five years after the original SPARC assessment. Three of the 588 patients declined authorization for medical research under Minnesota statute 144.335, leaving 585 patients available for follow-up; TEE was successful in 579 of these patients.
A comparison of comorbid conditions, age, and gender among participants and a random sample of eligible non-participants (decliners) demonstrated no significant differences, thus confirming that participants were a representative sample of the population (5).
Baseline transthoracic echocardiography and multiplane TEE (6) were conducted according to standard practice guidelines using commercially available ultrasonographic instruments. The ascending aorta, aortic arch, and descending aorta were imaged in short- and long-axis views. Atherosclerosis was defined as irregular intimal thickening with increased echogenicity. Atherosclerotic plaques were defined as complex in the presence of protruding atheroma more than 4-mm thick (4,7–9), mobile debris, or plaque ulceration, and as simple for plaques lacking these morphologic features in any aortic segment.
Cardiac events included death, non-fatal myocardial infarction, heart failure associated with coronary artery disease (CAD), and coronary revascularization (bypass surgery, angioplasty). Cerebrovascular events referred to deaths related to cerebrovascular disease (CVD), ischemic strokes, and transient ischemic attacks (TIA).
Survival free of cardiac events or cerebrovascular events was estimated by the Kaplan-Meier product-limit method, starting at the initial SPARC visit. Patients were censored at the last medical visit if they were lost to follow-up (two patients), at the date of the medical record abstraction, or at the date of death if death was not due to CAD or CVD. Cumulative incidence (10) was calculated to account for the competing risks of cardiac event, cerebrovascular event, or death. Expected survival was calculated using the cohort method and the Minnesota State Life Tables (11).
The Cox proportional hazards regression was used to model the role of clinical risk factors for cardiovascular and cerebrovascular events. Univariate models, age- and gender-adjusted models, and, finally, stepwise and backward multivariate models, adjusting for age and male gender, were used to determine the final subset of risk factors for each end point. A p value of 0.05 was used to determine if a variable would enter or leave the model. All the variables presented in Table 1,except the aortic plaque variables (assessed at a later stage of the analysis) and total cholesterol (data available in a subset of subjects), were considered for both the univariate and stepwise modeling. Pulse pressure was highly correlated with systolic and diastolic blood pressure, resulting in problems of multicolinearity. Thus, two versions of the stepwise process were run, one version with pulse pressure as the blood pressure variable and one version with the systolic and diastolic blood pressure variables. Only the pulse pressure results are reported (the same variables were in the models using systolic and diastolic blood pressure). The functional forms of all the continuous variables were examined, and no transformations were deemed necessary. Two-way interactions in the final stepwise models were explored.
The relationships among aortic atherosclerosis and the CAD and CVD end points were examined univariately (unadjusted), adjusting for age and gender, and adjusting for all variables in the respective risk factor models. Atherosclerosis was defined using all three segments (ascending, arch, or descending) and again separately using only two segments (ascending or arch). All interactions among the atherosclerosis and risk factor variables were assessed to determine if they jointly entered the model. The proportional hazards assumption was examined for the atherosclerosis variable and for the final multivariate models for each end point.
Table 1summarizes the atherosclerosis risk factors, comorbid conditions, and atherosclerotic plaques present at baseline and used in the stepwise age- and gender-adjusted multivariate analysis. Baseline was defined as conditions that were present at or before the first SPARC visit. Mean age was 66.9 ± 13.3 years. Simple and complex aortic plaques were detected in at least one of the three aortic segments in 253 (43.7%) and 44 (7.6%) persons, respectively. Cardiac events occurred in 95 persons (31 myocardial infarctions, 40 cardiac failures, 16 coronary revascularizations, 8 cardiac deaths), and cerebrovascular events occurred in 41 persons (20 strokes, 21 TIA). Eighteen persons had both a cardiac event and a cerebrovascular event. Nineteen persons died before having either; their data were censored. Of 585 persons, 448 had no qualifying event and were alive at the date of censor. Two cases were lost to follow-up and were censored. Mean follow-up was five years (0.5 to 6.5 years) with a total follow-up of 2,902 person-years.
Cardiac end point
The cumulative incidence of cardiac end points was 3.2% (one year), 6.8% (two years), 10.1% (three years), 12.8% (four years), and 15.6% (five years). Results of the univariate analyses are presented in Table 2.Each 10-year increase in age increased the hazard two-fold. The Kaplan-Meier product-limit estimates by age quartiles (Fig. 1)illustrate this univariate association. Table 3presents the same variables adjusted for age and gender. As reported in Table 4,increasing age, male gender, prior CAD, and higher pulse pressure were jointly associated with cardiac events. A mild departure from the proportional hazards assumption (p = 0.03) indicated that the hazard of CAD with increased pulse pressure may decrease over time compared with the baseline hazard.
Univariately, atherosclerosis grade (using all three segments or two of the three segments) was associated with cardiac events. The hazard ratios (HR) (using all three segments) were 3.41 (95% confidence interval [CI], 2.04 to 5.69) for simple atherosclerosis versus no atherosclerosis and 9.53 (95% CI, 5.10 to 17.8) for complex atherosclerosis versus no atherosclerosis (Table 5).The Kaplan-Meier product-limit estimates of atherosclerosis grade (Fig. 2)demonstrate this univariate association with cardiac events. After adjustment for age and gender, the HR for simple atherosclerosis versus no atherosclerosis did not differ significantly from 1.0 (both 95% CIs contain 1.0) (Table 5), although complex aortic atherosclerotic plaques had a marginally greater hazard of cardiac events compared with the absence of plaques (Table 5). After adjustment for additional clinical risk factors in the multivariate analysis, complex atherosclerosis using either definition was not associated significantly with an increased risk of cardiac events (Table 5).
Cerebrovascular end point
The cumulative incidence of cerebrovascular end points was 1.7% (one year), 3.1% (two years), 4.4% (three years), 5.3% (four years), and 6.9% (five years). Results of the univariate analyses are presented in Table 2. Each 10-year increase in age increased the hazard of a cerebrovascular event by 95%. Kaplan-Meier product-limit curves illustrate this univariate association of age with cerebrovascular events (Fig. 1). Table 3presents the same variables adjusted for age and gender. Older age, prior myocardial infarction, and a history of atrial fibrillation were predictors of cerebrovascular events in a multivariate model (Table 4).
Univariately, atherosclerosis grade (using either definition) was associated with cerebrovascular events as illustrated by the Kaplan-Meier product-limit curves (Fig. 2) (HRs reported in Table 2). After adjustment for age and gender, neither simple nor complex atherosclerosis remained significantly associated with outcome (Table 5). After additional adjustment for other clinical risk factors, atherosclerosis was not significantly associated with outcome (Table 5).
The proportional hazards assumption was met in the univariate models of simple and complex atherosclerosis versus no atherosclerosis for each end point, which indicates that subjects who had atherosclerosis were not more likely to experience earlier events than were subjects without atherosclerosis (data not shown).
In this population-based study, a detailed analysis of multiple variables led to the following conclusions:
1. Although highly significant in the univariate analysis, aortic atherosclerosis of any severity, after adjustment for age and other risk factors, was not an independent predictor of cardiac events or stroke.
2. Several atherosclerotic risk factors (age, male gender, higher pulse pressure, and diabetes) are significant independent predictors of cardiac events.
3. Age, prior myocardial infarction, and atrial fibrillation are predictive of cerebrovascular events.
In that atherosclerosis is a diffuse process, it is not surprising that multiple studies have found an association between aortic atherosclerosis and CAD. Extrapolation of these studies to the general population is misleading. Study limitations include the highly selected nature of patient groups having TEE and coronary angiography for specific clinical indications (12–17), limiting inclusion to patients with atrial fibrillation and other cardioembolic risks (7,8,18), case-control studies with nondescript or selected “control” populations, and inconsistent or absent adjustment for risk factors (1,2,18–20). Most studies described patients with prior embolic events, predominantly strokes. In contrast, the current study enabled us to assess aortic anatomy and multiple risk factors in a large non-referred population representative of the general population in a well-defined geographic area.
Several authors have shown a strong association between aortic atherosclerosis and other risk factors for stroke. Age, hypertension, atrial fibrillation, carotid artery disease, and diabetes mellitus are well-known risk factors for cardiac events and stroke (21–23). A recent study from our group confirmed the strong association of aortic atherosclerosis with hypertension and CAD in the general population (9,24). These findings challenge previously drawn conclusions about the cause-and-effect relationship of aortic atherosclerosis and clinical outcome. In the Stroke Prevention in Atrial Fibrillation (SPAF) III trial, complex aortic atherosclerosis emerged as a significant independent predictor for stroke without adjusting for age. Age has been shown to be a strong independent predictor of aortic atherosclerosis (8), which suggests that atherosclerosis may be a marker of aging rather than a true risk factor for stroke. One autopsy study concluded that the presence of moderate or severe aortic atherosclerosis did not predict ischemic stroke subtype (25). Complex atherosclerosis is a high-risk marker in high-risk patients, but it is also a marker of many other risk factors for stroke, so a cause-and-effect relationship cannot be established.
Our previous retrospective analysis of the relationship between prior CAD and TEE-assessed atherosclerosis in the SPARC population showed a strong relationship even after adjusting for age (24). The current prospective analysis had the advantage that the CAD event was known to have occurred after the TEE assessment. The prospective analysis examined atherosclerosis in all patients who had CAD, including those who died because of it (a small percentage of total CAD events). Of necessity, the retrospective analysis examined atherosclerosis in patients with CAD before assessment who survived long enough to enter our study. Death rates in the overall SPARC I cohort, including those with or without atherosclerosis, were about one-half of what was expected for the Minnesota white population. A comparison of respondents and non-respondents in an Olmsted County study of osteoporosis found similar death rates (26). Because expected rates are based on the entire population, including those too ill to be eligible, this finding is not surprising (their survival is poor) and is a concern common observational studies and clinical trials. It is unclear what impact using a “healthier” population had on the study of the relationship between atherosclerosis and subsequent CAD or CVD. Considered jointly, the retrospective and prospective studies suggest that although atherosclerosis increases with age (as does the risk of CAD), it does not in itself result in an increased risk for CAD.
This SPARC study is the largest prospective population-based TEE study published to date. However, size limitations imposed by the use of an invasive study in a relatively healthy population sample may have limited the ability of the SPARC study to detect a statistically significant hazard for the less prevalent risk factors. With a larger sample size, the wide CIs around the adjusted HR for simple and complex atherosclerosis (three segments) would narrow, but would still likely be centered around 1.0. Nevertheless, we cannot exclude the possibility that the hazard due to complex atherosclerosis compared with no atherosclerosis may be as high as 2.6 and 3.9 for CAD and CVD, respectively. The unselected population base and comprehensive aortic anatomical data are useful in determining important trends in outcome. Furthermore, our unadjusted estimates of the hazard of CAD or CVD reflect those of other studies that failed to adjust for age and common comorbid conditions, giving pause to the common adage that atherosclerosis is strongly related to these events. Because of the anticipated small number of strokes in this cohort, a combined end point for TIAs and strokes was used; however, because TIA is a well-recognized significant risk factor for stroke, these data provide valuable information about the predictive impact of aortic atherosclerosis in this group.
Plaques in any location of the aorta were included in the analyses. Their inclusion is a potential limitation for the cerebrovascular outcome analysis, because plaques were most common in the descending aorta, which is not an expected pathophysiologic substrate for stroke. The location of aortic atherosclerosis has not been shown to correlate clearly with the site of an embolism. In SPAF III, all embolic events were cerebral, yet aortic atherosclerotic plaques were most common in the descending aorta (8,27), suggesting that plaques in the descending aorta are high-risk markersfor stroke and, possibly, direct sources of (retrograde) embolism (28). Limiting our analysis to atherosclerosis found in the ascending aorta and aortic arch alone, we found similar results, although adjusted HRs for CAD and CVD events were higher.
Serum lipid levels, available in only a subgroup of subjects (Table 1), did not affect the relationship between atherosclerosis and subsequent stroke or cardiac events. These findings are consistent with data from a large echocardiographic study published in 1999 that found no association between hyperlipidemia and aortic atherosclerosis (8).
Aortic atherosclerosis is increasingly detected with more frequent use of TEE. It does not appear to be an independent risk factor or a risk marker for vascular events in the population. This study has important ramifications for the management of subjects who have aortic atherosclerosis, and it underscores the importance of aggressive treatment of high blood pressure, diabetes, and known cardiovascular risk factors.
Further follow-up of this cohort may identify other mechanisms or risk factors (e.g., dyslipidemia, atypical atherosclerotic risk factors, genetic factors) that play a role in the development of cardiovascular or cerebrovascular events.
Supported, in part, by research grant NS-06663 from the National Institute of Neurological Disorders and Stroke.
- Abbreviations and acronyms
- coronary artery disease
- confidence interval
- cerebrovascular disease
- hazard ratio
- Stroke Prevention in Atrial Fibrillation
- Stroke Prevention: Assessment of Risk in a Community
- transesophageal echocardiography
- transient ischemic attack
- Received February 18, 2004.
- Revision received May 3, 2004.
- Accepted May 18, 2004.
- American College of Cardiology Foundation
- Montgomery D.H.,
- Ververis J.J.,
- McGorisk G.,
- Frohwein S.,
- Martin R.P.,
- Taylor W.R.
- Blackshear J.L.,
- Pearce L.A.,
- Hart R.G., et al
- Agmon Y.,
- Khandheria B.K.,
- Meissner I,
- et al.
- Technical Report Series No. 63: Expected survival based on hazard rates (update). Rochester, MN: Department of Health Science Research, Mayo Clinic, 1999,
- Therneau T.,
- Offord J.
- Fazio G.P.,
- Redberg R.F.,
- Winslow T.,
- Schiller N.B.
- Tribouilloy C.,
- Peltier M.,
- Colas L.,
- Rida Z.,
- Rey J.L.,
- Lesbre J.P.
- Di Tullio M.R.,
- Sacco R.L.,
- Savoia M.T.,
- Sciacca R.R.,
- Homma S.
- Wolf P.A.,
- D'Agostino R.B.,
- Belanger A.J.,
- Kannel W.B.
- Pearson T.A.,
- Blair S.N.,
- Daniels S.R,
- et al.,
- for the Population Science Committee of the American Heart Association