Does the Association of Risk Factors and Atherosclerosis Change With Age?

Whisnant and coworkers have recently noted a declining impact of hypertension, smoking, and history of transient ischemic attacks as risk factors for cerebral infarction.¹ Others investigators have suggested that the relative risk of cardiovascular diseases associated with hypertension² or dyslipidemia³ may be small or negligible at older ages. However, other reports note a continued elevated relative risk of cardiovascular disease at older ages associated with cardiovascular disease risk factors.⁴⁵ The public health implications of a decline in the risk of CCVD associated with the presence of CCVD risk factors at older ages (if it truly exists) are substantial. For example, Langer and coworkers suggest that the decreased impact of hypertension in the elderly “suggests a need to reevaluate strategies for managing high blood pressure in the elderly.”² If risk factors have a declining impact with increasing age, then approaches used to reduce CCVD risk in elderly populations may need to be modified.

The baseline data from the ARIC study and CHS offer a unique opportunity to assess the potential for a differential impact of risk factors on atherosclerosis with increasing age. This opportunity is presented because (1) the age range spanned by ARIC (45 to 64 years) is contiguous with that of CHS (65 years and older), and together these studies span the age spectrum in which clinical events associated with the development of atherosclerosis become pronounced; (2) both of these large epidemiological studies assessed the thickness of the intima-media layer of the common carotid artery, which is a frequently used index of atherosclerosis in both epidemiological studies⁶⁷⁸⁹¹⁰ and clinical trials¹¹¹²¹³ ; and (3) both studies used similar methods to assess major cardiovascular risk factors. The goal of this report is to examine potential age-related differences in the magnitude of the association of major CCVD risk factors with atherosclerosis in the free-living population over the age of 45 years.

Methods

The ARIC study is a prospective study investigating the etiology and natural history of atherosclerosis and its clinical sequelae in four United States communities.¹⁴ At the baseline examination conducted from 1987 through 1989 on which the present analyses are based, cardiovascular disease risk factors were assessed in a population-based sample of approximately 15 800 adults aged 45 to 64 years from four areas in the United States. Approximately 4000 adults were enrolled in the study from each of the following areas: the northwest suburbs of Minneapolis, Minnesota; Washington County, Maryland; Forsyth County, North Carolina; and Jackson, Mississippi. The Jackson Center enrolled only black participants to study cardiovascular disease in a large black cohort. The number of participants in each race and gender group ranged from 1630 black men to 6049 white women. This report used data from the baseline assessment in ARIC.

The CHS cohort consisted of 5843 men and women, aged 65 years and older, and was drawn from four US communities: Forsyth County, North Carolina; Sacramento County, California; Washington County, Maryland; and Pittsburgh (Allegheny County), Pennsylvania.¹⁵ The cohort was identified using Medicare eligibility lists of the Health Care Finance Administration from these four communities. Eligible participants were 65 years and older at the time of examination, community dwelling, and did not require a proxy respondent at baseline. Individuals who were wheelchair-bound in the home or receiving hospice treatment, radiation therapy, or chemotherapy for cancer were excluded from the study. This report used data from the baseline assessment in CHS.

Risk factors were considered for this analysis if (1) they were generally accepted “major” risk factors for the development of atherosclerosis and (2) they were measured in a similar manner in both ARIC and CHS.¹⁴¹⁵ Hypertension was defined in both studies as a systolic blood pressure greater than 160 mm Hg, or a diastolic blood pressure greater than 95 mm Hg, or use of antihypertensive medications. Both studies defined smoking history by self-report: in ARIC a current smoker was defined by a positive response to the question, “Do you now smoke cigarettes?” whereas CHS used the question, “Have you smoked cigarettes during the last 30 days?” Past smoking was defined as noncurrent smokers who had smoked more than 400 cigarettes in their lifetime for ARIC study and more than 100 in their lifetime for the CHS. In ARIC, a participant was considered to be diabetic if the fasting (8 hours or more) blood glucose was ≥140 mg/dL (or if a fasting sample was not available, if the nonfasting blood glucose was ≥200 mg/dL), there was a self-report of a physician diagnosis of diabetes, or if the participant was currently on medication. CHS, but not ARIC, performed an oral glucose tolerance test, but these data were not used to ensure comparability between the studies. Hence, in CHS a diagnosis of diabetes was a fasting glucose ≥140 mg/dL, a self-report of a physician diagnosis of diabetes, or if the participant was currently on medication. BMI was measured as weight (in kilograms) divided by height (in meters) squared. HDL-C, LDL-c, and fibrinogen were measured in a standard and similar manner in the two studies.

Both ARIC and CHS evaluated the IMT as an index of systemic atherosclerosis. In this role, IMT has proved to be closely related to incident coronary ischemic events,¹⁶¹⁷ and has been accepted as the primary end point in major clinical trials.¹¹¹²¹³ ARIC and CHS used similar ultrasound protocols, where the IMT of the far and near walls of the common carotid artery over a 1-cm site directly distal to the dilation of the bifurcation. ARIC evaluated the thickness in up to 11 segments 1 mm apart, allowing estimation of mean and maximum wall thicknesses. CHS drew a continuous line along visualized interfaces over this same 1-cm length of vessel, and a computer program then evaluated the distances between adjacent lines (thereby eliminating the need for operator intervention in estimating point-by-point pairs). In previous reports, CHS generally used maximum wall thickness to describe the IMT of a wall at a site, whereas ARIC generally used mean wall thickness. The choice of mean versus maximum wall thickness does not significantly affect results because the correlation between mean and maximum wall thicknesses is far above 0.9. For this report the maximum wall thickness was used to describe IMT.The mean value of the four maximum IMTs (far and near walls of the left and right common carotid arteries) was used as the outcome variable in all analyses.

The impact of select risk factors was estimated within 5-year age strata over the combined age range of the ARIC and CHS studies. In order to prevent confounding of this study from the estimation of effect within any of these strata, 103 ARIC participants aged 65 years and older at baseline were deleted from analyses (allowing the 65- to 69-year age strata to contain CHS participants only). In addition, the 35 ARIC participants aged 44 years at baseline were also deleted. To maintain a sample size sufficient to produce reliable estimates, CHS participants aged 75 years and older were considered as a single stratum. This resulted in seven contiguous age strata, four from the ARIC study (45 to 49, 50 to 54, 55 to 59, and 60 to 64) and three from CHS (65 to 69, 70 to 74, and 75 and older). The sample size in each of these strata is provided in Table 1.

In this “strata” approach, a general linear model was fit separately to the data from blacks and whites. For continuous risk factors, a single regression model was fit to all strata with parameters to estimate the regression relationship between the risk factor and IMT within each age-sex stratum. This model required 28 parameters, a slope and intercept for each of the 14 (7 ages×2 sexes) strata. A similar model was fit for the dichotomous variables, with parameters to estimate the mean IMT in the absence of the interaction and the difference in IMT attributed to the risk factor present for each of the strata. The slope parameter corresponds to the difference in IMT associated with a unit difference in the risk factor, and the parameter associated with the dichotomous factor corresponds to the difference in IMT between those with and those without the risk factor.

Differences in the estimated slope parameter between the age strata for the continuous risk factors, and the difference parameter for dichotomous risk factors, are the focus of this report. A number of statistical tests were applied to the estimated strata-specific risk factor associations. First, an age-by-sex interaction test was applied, addressing the question of whether the pattern in the IMT difference associated with the risk factor across age was different in men and women (a 6 degree of freedom test). In cases in which this test was clearly nonsignificant (P>.1), further tests focused on differences seen when men and women were pooled. Second, a test was applied to assess whether there were any differences between the seven age strata, in which the age strata were considered as categories (no trend information used, simply testing whether there are any differences between any pair of strata). This test considering strata in categorical (rather than ordinal) analysis was then applied separately for men and then for women. Next, the ordinal nature of the age strata was incorporated into the tests, and a trend test was applied in which a linear relationship between age strata and the magnitude of the estimated impact of the risk factor was assumed. Since this trend test assumed a linear relationship, it required estimating a single parameter (ie, 1 degree of freedom). This test was then applied within the ARIC and CHS cohorts separately (pooling men and women) and for men and women separately (pooling ARIC and CHS).

Results

A description of the risk factors for the two studies is provided in Table 2. Generally, differences between the two studies are large, but expected because of the age differences in the cohorts. The CHS cohort had a higher prevalence of hypertension and diabetes than their ARIC counterparts. In addition, they were less likely to be current smokers and more likely to be past smokers. The lipid profile of the CHS cohort was less conducive to atherosclerosis, with the higher average HDL-C and lower average LDL-c, than their ARIC counterparts. The average BMI was lower in CHS, while the fibrinogen levels were greater. Finally, the mean maximum IMT of the common carotid artery was thicker in CHS compared with ARIC.

The estimated overall effect of the risk factors is shown by race in Table 3. The magnitude of the estimated effects for all risk factors was many times the estimated standard errors, suggesting that each of these is (on average) an important risk factor for atherosclerosis as indexed by IMT (for example, hypertensive participants had walls 41.5±7.1 μm thicker than normotensive participants, a difference nearly six times its standard error). Among white participants there was some evidence (P<.05) that the magnitude of the impact of “ever smoking,” HDL-C, fibrinogen, and BMI differed between men and women, in each case with larger effects in men than women. However, the goal of this article is to assess whether these overall effects differ across the spectrum of the age of the participants.

Figs 1 and 2 show the estimated difference in IMT thickness associated with the presence (Fig 1) or 2-SD change (Fig 2) for eight risk factors within age-sex strata for whites and blacks. Table 4 provides probability values for differences between the strata across age that are discussed below. The following discussion is ordered by those factors where there was a trend for an increasing impact of the risk factor with age (smoking), those where the response was mixed (HDL-C), those where the impact decreased (diabetes), and, finally, those with little impact (hypertension, LDL-c, fibrinogen, and BMI).

For both current and ever smoking, the difference in IMT was greater among older white participants compared with their younger white counterparts (P=.010 for current smoking and P=.024 for past smoking). As seen in Fig 1, current or ever smoking exposure was associated with a small and nonsignificant difference in IMT among participants aged 45 to 49 years (≈16 μm), but increased to a more than 30-μm difference by age 60 to 64 years. Above that age, there was little change in the magnitude of the association. This increase across the younger ages, but plateauing for older ages, is shown by a highly significant trend in the ARIC cohort (P=.001) but no trend in the CHS cohort (P=.20 for current smoking and P=.27 for ever smoking). There was little evidence of a differential effect by gender (P=.44 and P=.087) and little evidence of a change across ages in the impact of smoking exposure on IMT among blacks (P>.05).

For whites (see Fig 2), IMT difference associated with HDL-C increased among older participants (P=.015). Among white participants aged 45 to 49 years, an approximate 2-SD increase in HDL-C (33.8 mg/dL) was associated with IMT differences of −43 μm in men and −30 μm in women; while for those older than 75 years the same difference in HDL-C was associated with IMT differences of −62 μm and −69 μm, respectively. For blacks (Fig 2), there was evidence of a different pattern across ages for men and women (P_{age-by-sex interaction}=.022). Among black women, the magnitude of the HDL-C–IMT association declined among the older age strata; a 2-SD difference in HDL-C was associated with a difference of approximately −40 μm for ages 45 to 49 years and 50 to 54 years. For ages 55 to 59 years, the same difference in HDL-C was associated with a difference in IMT of −97 μm; however, for age strata above 60 years there was a reasonably consistent decrease in the magnitude of the association with HDL-C, and in fact it became positive in the oldest age strata (≈36 μm per 2-SD change). There was not a significant trend in either direction for black men (P=.17).

For whites, the estimated trend for the difference in IMT associated with diabetes decreased among the older age strata (P=.003). For white men, diabetics aged 45 to 49 years had average IMTs 123 μm greater than nondiabetics, and the difference for white women aged 45 to 49 years was 106 μm. The difference in IMT between diabetics and nondiabetics tended to decrease at older ages, and the differences for those older than age 75 years were 9 μm for men and 71 μm for women. This larger decrease in the magnitude of the association of diabetes and HDL-C for men than for women resulted in an interaction that suggested a differential effect by gender (P_{age-by-sex interaction}=.15). When sex-specific trends between diabetes and the magnitude of the IMT difference were tested, the decreased impact of diabetes was clear in men (P=.003) and not present in women (P=.21). There was no evidence of a change in the magnitude of the association with diabetes across the age strata for blacks.

For hypertension (Fig 1), LDL-c (Fig 1), fibrinogen (Fig 2), and BMI (Fig 2), there was little evidence that the estimated IMT difference changed across the age strata for either whites or blacks (P>.05).

Discussion

Not surprisingly, there was a clear relationship between the presence or level of risk factors and increased IMT, where adverse levels of risk factors were associated with increased IMT (see Table 3).¹⁸¹⁹²⁰ However, the focus of this report is on the consistency of the risk factor/atherosclerosis relationship across age strata. On the basis of the decreasing relationship between clinical events and risk factors with increasing age,¹²³ we had hypothesized a decreasing relationship between risk factors and IMT at older ages. In neither whites nor blacks, however, was there a consistent pattern of either increasing or decreasing association with risk factors across the age spectrum. Rather, most risk factors appeared to have a relatively consistent magnitude of association across age strata. The few risk factors that did appear to have a differential magnitude of association by age did not differ in a consistent pattern, with greater IMT differences at older ages for current smoking and ever smoking among whites, greater magnitude of association of HDL-C at older ages for whites, smaller magnitude of association of HDL-C at older ages for black women, and smaller magnitude of association of diabetes at older ages for whites (particularly men). These results are somewhat counterintuitive findings and warrant reconciliation; that is, Why do some risk factors increase in their impact with age, others are stable, and still others decrease?

One important aspect not addressed in these analyses is the confounding of age with the duration of exposure to the risk factors. Current cigarette smoking is the risk factor found most clearly to increase in magnitude of association with increasing age, but age and duration of exposure are highly confounded. Since nearly all smokers begin smoking below the age of 20 years,²¹ exposure to smoking increases proportionally with age in current smokers. That past smoking shows a pattern of increasing and then plateauing impact with increasing ages is also consistent with expected exposure patterns. The median age of stopping smoking is 45 years.²² At younger ages, past smokers are more likely to have recently quit, and the risk of being a past smoker will resemble that seen for current smokers. However, at older ages the time of quitting is more likely to be remote, and once a smoker quits additional exposure is not acquired. Alternatively, it is possible that the change in the magnitude of the association could be associated to the minor difference in the definition of “past” smoking between the two studies, where ARIC required 400 cigarettes, and CHS 100 cigarettes, during the participant’s lifetime. However, we feel this difference in definitions is relatively minor and unlikely to explain the plateauing. However, because of the former reasons, it is thus reasonable to expect the association with past smoking to increase at young ages and plateau at older ages. Exposure for participants with low HDL-C or elevated LDL-c is also potentially confounded with age (dyslipidemia acquired at a “young” age and endured to the time of evaluation), and as such age is also a surrogate for exposure time. Hence, age is likely a surrogate for exposure for the lipid and smoking measurements. This is in contrast to the relationship between age and either hypertension or diabetes, risk factors that tend to be acquired at much older ages. For example, it is possible that a 70-year-old diabetic participant has not been exposed to diabetes longer than a 50-year-old diabetic patient. In addition, survival bias may remove participants from this cross-sectional analysis who have had these risk factors for a more extended period. Although the data on time of acquisition of elevated fibrinogen levels and increased BMI are less clear, a similar process may be occurring for these factors. It thus appears that the association of risk factors that tend to be acquired at a young age (implying that age is a surrogate for exposure) increase at older ages, but the associations with risk factors acquired at older ages to be stable or decrease with advancing age, perhaps because age is not necessarily a surrogate for exposure.

We can only offer speculation as to why we observe a stable or increasing association between risk factors with atherosclerosis with increasing age, while others have observed a decreasing association between risk factors and clinical events at older ages.¹²³ While it is difficult to reconcile these discordant observations, several explanations may be offered. First, the decline in risk factor impact on clinical events is traditionally described by ratio measures (odds ratios, hazard ratios, etc). If the impact of a risk factor is reflected by a constant difference in the distribution (such as for hypertension), but the prevalence of the risk factor increases with age (such as for hypertension), the ratio of those with the disease to those without disease will decrease.²³ For example, if odds of events in hypertensive participants at age 45 to 49 years is 0.2 and for normotensive participants is 0.1, then the resulting odds ratio is 2.0. But if these odds increase in a parallel manner to 0.4 and 0.3, respectively, for ages older than 75 years, the odds ratio will decrease to 1.25. As such, the equal impact of a risk factor on the underlying atherosclerosis, but the increase in the prevalence of the risk factor and clinical events, could lead to a decrease in the odds ratio. Even if the relative risk decreased with age, efficacious treatments may reduce the number of events among the elderly because of the overall increase in rates. The finding that most risk factors have a relatively constant association with atherosclerosis across the age spectrum may thus be consistent with decreasing odds ratios associated with the risk factor at older ages. It is important to note that the observed decrease in the relative importance of risk factors at older ages among the elderly is consistent with a constant absolute difference associated with the risk factor. Second, in ARIC and CHS the IMT measurements may reflect subclinical atherosclerosis estimated with a reduced survival bias; with only a small proportion of the population eventually advancing to clinical events and subsequently removed from the evaluation. As such, survival bias may have a smaller impact (few people die with “advanced subclinical” disease) in the assessment of changes in IMT than for clinical disease. Because this bias is playing a smaller role, we may be observing a more representative view of the changes in risk factor association over a broad age range. Finally, it is possible that survival selection in the ARIC and CHS cohorts may result in a cohort able to develop and maintain atherosclerosis but to not suffer clinical events.

There are two statistical issues that warrant comment. In general, more differences in the magnitude of the association between risk factors were found for the white, as compared with black, participants. While it may be true that more differential effects are present in whites (relative to blacks), we feel that a more likely explanation for this observation is the increased power to detect differences in whites that was provided by a larger sample size. In addition, there could be concerns that the large number of statistical tests presented in Table 4 could give rise to concerns of spurious findings through multiple testing. The broad scope of these probability values was provided for completeness and for further examination by the readers. However, we do not feel that multiple testing is a large problem for two reasons. First, although there are many probability values in Table 4, few are actually interpreted in the text or in our decisions of how to interpret the data. Specifically, we first examined the probability value for interaction with sex to assess whether it is reasonable to pool the genders. If there was no evidence, then the overall tests were the next probability values assessed. These are the primary tests of the hypotheses of the article, and the majority of the interpretation is made on these relatively few statistical tests (hence, reducing concerns from multiple testing). The other probability values are offered to describe the specific effects or to describe the consistency of the effect in subgroups. Second, multiple testing concerns are usually an issue when there are many “significant” findings (of which some subgroup happened by chance alone). In our case, the primary finding of the paper was a lack of significant findings, and, as such, spurious findings are perhaps of less concern. Nevertheless, we did examine eight risk factors in two ethnic groups (a total of 16 evaluations), and as such we should expect approximately one association to have occurred by chance alone (1/16≈1/20=0.05).

There are a number of shortcomings of this report. A number of assumptions were required to combine the data of the ARIC and CHS studies. Although we discuss the relationship between risk factors and IMT as an index of atherosclerosis, the relationship of IMT or atherosclerosis may be different in younger than older populations. Of particular concern is the possibility that any trend (or lack of trend) in the association across age is confounded with study. For example, a positive association would be present if there were no association with age within either study but a difference between studies where the risk factors play a larger role in CHS (perhaps because of a difference in cohorts). There is no statistical method to remove this potential bias. However, we (subjectively) feel that this is unlikely because of the consistency (lack of large discontinuous “jumps” in effect size) of the relationship between age and risk factors between the oldest ARIC participants (60 to 64 years) and the youngest CHS participants (65 to 69 years). Second, while the conduct and interpretation of ultrasound exams were similar, they were not identical. Differences in the ultrasound examination may be influencing results in complex ways. Similar concerns exist for the measurements of risk factors that were assessed using similar (but not identical) methods. It is also possible that a survival bias is acting to reduce those older participants with more advanced atherosclerosis, potentially producing biased estimates of the effect of risk factors (particularly in the elderly). Finally, as for all cross-sectional analyses, we are reporting only the changing in the magnitude of the estimated association between the risk factors and IMT, and inferences to causation are always problematic in cross-sectional analyses.

In conclusion, the association of most major risk factors with atherosclerosis was relatively constant over 5-year age strata from 45 to 49 years to older than 75 years. This consistency is in contrast to reported diminished associations with clinical events at advanced ages, and may reflect more complete and unbiased ascertainment as well as assessment of absolute rather than relative risks. This relatively consistent association of risk factors to atherosclerosis may possibly support the continued importance of investigating efforts to reduce atherosclerosis through risk factor control in the elderly population.

Selected Abbreviations and Acronyms

Appendix A1

ARIC

University of North Carolina at Chapel Hill: Phyllis Johnson, Marilyn Knowles, Catherine Paton; University of North Carolina, Forsyth County: Dawn Scott, Nadine Shelton, Carol Smith, Pamela Williams; University of Mississippi Medical Center, Jackson: Virginia Overman, Stephanie Parker, Liza Sulivan, Cora Walls; University of Minnesota, Minneapolis: Greg Feitl, Chris Humkins, Ellie Justiniano, Laur Kemmis; Johns Hopkins University, Baltimore: Joel Hill, Kathleen Hunt, Mary Hurt, Joan Nelling; University of Texas Medical School: Valarie Stinson, Pam Pfile, Hogan Pham, Teri Trevino; The Methodist Hospital, Atheroslerosis Clinical Laboratory, Houston: Wanda R. Alexander, Doris J. Harper, Charles E. Rhodes, Selma M. Soyal; Bowman Gray School of Medicine, Ultrasound Reading Center, Winston-Salem, NC: Anne Safrit, Melanie Wilder, Linda Allred, Carolyn Bell; University of North Carolina, Chapel Hill, Coordinating Center: Ho Kim, Charmaine Marquis, Stephen Noga, Joy Rollins.

CHS

Forsyth County, NC—Bowman Gray School of Medicine of Wake Forest University: Sharon Jackson, Alan Elster, Walter H. Ettinger, Curt D. Furberg, Gerardo Heiss, Dalane Kitzman, Margie Lamb, David S. Lefkowitz, Mary F. Lyles, Cathy Nunn, Ward Riley, John Chen, Beverly Tucker; Forsyth County, NC—Bowman Gray School of Medicine-EKG Reading Center: Farida Rautaharju, Pentti Rautaharju; Sacramento County, CA—University of California, Davis: William Bommer, Charles Bernick, Andrew Duxbury, Mary Haan, Calvin Hirsch, Lawrence Laslett, Marshall Lee, John Robbins, Richard White; Washington County, Md—The Johns Hopkins University: M. Jan Busby-Whitehead, Joyce Chabot, George W. Comstock, Adrian Dobs, Linda P. Fried, Joel G. Hill, Steven J. Kittner, Shiriki Kumanyika, David Levine, Joao A. Lima, Neil R. Powe, Thomas R. Price, Jeff Williamson, Moyses Szklo, Melvyn Tockman; MRI Reading Center-Washington County, Md—The Johns Hopkins University: R. Nick Bryan, Norman Beauchamp, Carolyn C. Meltzer, Naiyer Iman, Douglas Fellows, Melanie Hawkins, Patrice Holtz, Michael Kraut, Grace Lee, Larry Schertz, Cynthia Quinn, Earl P. Steinberg, Scott Wells, Linda Wilkins, Nancy C. Yue; Allegheny County, PA—University of Pittsburgh: Diane G. Ives, Charles A. Jungreis, Laurie Knepper, Lewis H. Kuller, Elaine Meilahn, Peg Meyer, Roberta Moyer, Anne Newman, Richard Schulz, Vivienne E. Smith; Echocardiography Reading Center (Baseline)—University of California, Irvine: Hoda Anton-Culver, Julius M. Gardin, Margaret Knoll, Tom Kurosaki, Nathan Wong; Echocardiography Reading Center (Follow-up)—Georgetown Medical Center: John Gottdiener, Eva Hausner, Stephen Kraus, Judy Gay, Sue Livengood, Mary Ann Yohe, Retha Webb; Ultrasound Reading Center—Geisinger Medical Center: Daniel H. O’Leary, Joseph F. Polak, Laurie Funk; Central Blood Analysis Laboratory—University of Vermont: Edwin Bovill, Elaine Cornell, Mary Cushman, Russell P. Tracy; Respiratory Sciences—University of Arizona-Tucson: Paul Enright; Coordinating Center—University of Washington, Seattle: Alice Arnold, Annette L. Fitzpatrick, Bonnie K. Lind, Richard A. Kronmal, Bruce M. Psaty, David S. Siscovick, Lynn Shemanski, Lloyd Fisher, Will Longstreth, Patricia W. Wahl, David Yanez, Paula Diehr, Maryann McBurnie; NHLBI Project Office: Diane E. Bild, Robin Boineau, Peter J. Savage, Patricia Smith.

Atherosclerosis Risk in Communities (ARIC) and Cardiovascular Health Study (CHS) Investigators are listed in the “Appendix.”

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

Footnotes