Abstract
Background
No studies have compared first CVD events and non-CVD death between races in a competing risks framework, which examines risks for numerous events simultaneously.
Methods and Results
We used competing Cox models to estimate hazards for first CVD events and non-CVD death within and between races in three multi-center, NHLBI-sponsored cohorts. Of 14569 ARIC study participants aged 45–64y with mean follow up of 10.5y, 11.6% had CVD and 5.0% had non-CVD death as first events; among 4237 CHS study participants aged 65–84y and followed for 8.5y, these figures were 43.2% and 15.7%, respectively. Middle-aged blacks were significantly more likely than whites to experience any CVD as a first event; this disparity disappeared by older adulthood and after adjustment for CVD risk factors. The pattern of results was similar for MESA participants. Traditional Cox and competing risks models yielded different results for CHD risk. Black men appeared somewhat more likely than white men to experience CHD using a standard Cox model (HR 1.06; 95% CI 0.90, 1.26) whereas they appeared less likely than white men to have a first CHD event using a competing risks model (HR 0.77; 95% CI 0.60, 1.00).
Conclusions
CVD affects blacks at an earlier age than whites; this may be partially attributable to elevated CVD risk factor levels among blacks. Racial disparities in first CVD incidence disappear by older adulthood. Competing risks analyses may yield somewhat different results than traditional Cox models and provide a complementary approach to examining risks for first CVD events.
Keywords: cardiovascular diseases, epidemiology, prevention, risk factors, survival
INTRODUCTION
There are considerable racial and ethnic disparities among adults in the United States in the incidence of cardiovascular disease (CVD), which includes stroke, heart failure (HF), and coronary heart disease (CHD). Prevalence of most CVD risk factors, short-term rates of CVD events, and CVD-related mortality appear to be significantly greater among blacks than whites.1–8 Although mortality due to CVD and CHD in the United States has declined in recent decades, CHD mortality has declined more slowly among blacks than among whites, potentially resulting in growing disparities.9
Comparisons of CVD incidence between racial groups may be limited, however, because they often do not account for racial disparities in death from other causes (non-CVD death). Non-CVD death competes as an outcome against CVD and may therefore affect CVD event rates differently for groups with disparate rates of non-CVD death. In other words, individuals may not experience a CVD event for which they are destined if a non-CVD death occurs first. No studies to date have compared first CVD events and non-CVD death between races in a competing risks framework, which is a type of analysis that allows for multiple possible outcomes (such as different CVD events and non-CVD death) simultaneously and determines which of these outcomes occurs first. The competing risks framework is distinct from approaches used in typical survival analyses, which consider only a single event at a time (e.g. stroke alone or CHD alone) and assume that each event is independent of the others when, in reality, these events are not independent of one another because of common underlying risk factors. In addition, the occurrence of one CVD event, such as myocardial infarction (MI), may greatly increase the risk for subsequent CVD events, such as HF or stroke. Understanding the temporality of when CVD events occur first for different races may also have implications for prevention strategies and prognosis, since strategies for stroke and heart failure prevention may differ somewhat from strategies for MI prevention.
Using data from the Atherosclerosis Risk In Communities (ARIC) study, the Cardiovascular Health Study (CHS), and the Multi-Ethnic Study of Atherosclerosis (MESA), we estimated the competing risks for different first CVD events (fatal or non-fatal) relative to each other and to non-CVD death within and between different racial groups. Middle-aged adults were analyzed separately from older adults. In light of the greater incidence of CVD as well as greater all-cause mortality rates among blacks than among whites 1,10–13, we hypothesized that blacks would have a greater incidence of non-CVD death and CVD events as first events than whites.
METHODS
Study Participants
Participants in the present study were sampled from three NHLBI-funded multi-center longitudinal cohort studies based in the United States: 1) Atherosclerosis Risk in Communities (ARIC) cohort; 2) Cardiovascular Health Study (CHS) of older adults; and 3) Multi-Ethnic Study of Atherosclerosis (MESA). The ARIC and CHS cohorts, which were comprised of participants aged 45 to 64 and 65 to 84 at baseline, respectively, were the primary cohorts used for the present analyses. The MESA cohort, comprised of participants 45 to 84 at baseline, was divided into two age strata (45 to 64 and 65 to 84) for the present study in order to validate our analyses of ARIC and CHS participants in a contemporary cohort. Details of each cohort are published and summarized briefly below. Participants with CVD at baseline were excluded from all analyses, as were participants with incomplete baseline data for important covariates. Participants were followed longitudinally for incident CVD events and cause-specific mortality.
ARIC examined 15792 men and women (55% female, 73% white, and 27% non-white – primarily African-American) aged 45 to 64 at baseline in 1987–89.14 The present study includes 14569 participants and uses follow up data through 1999 (mean follow-up 10.5 years) from the limited access dataset.
CHS is a sample of 5888 Medicare eligible men and women aged 65 to 100 years at baseline (1989 and 1992; a supplemental cohort of black adults was added in 1992).15,16 The cohort was approximately 58% female, 84% white, and 16% African American at enrollment. The present study includes 4237 participants who were followed through June 2006 (mean follow-up 8.5 years); mean follow-up for CHS was shorter than ARIC because events were more common in the older CHS cohort, and follow-up for each participant in these analyses ended when the participant experienced a first event.
MESA is a population-based sample of 6814 men and women aged 45 to 84 years who were free of clinical CVD at baseline (2000).17 The cohort was approximately 53% female, 39% white, 27% African-American, 22% Hispanic, and 12% Asian (predominantly Chinese) at enrollment. The present study uses follow up data through August 2008. Because follow-up in MESA was shorter (5.4 years) than for the other cohorts, there were fewer events and less power to detect statistical significance.
Exposure Variables and Covariates
Race/ethnicity, sex, age, education, and income were determined based on participant responses to questionnaires. Each field center for ARIC, CHS, and MESA measured anthropometric characteristics and CVD risk factors using standardized protocols.18–20 Hypertension was defined as untreated diastolic blood pressure ≥90 mmHg, systolic blood pressure ≥140 mmHg, or use of antihypertensive medication. History of smoking, diabetes, and medication use were self-reported. Participants were classified as current, former, or never smokers based on personal interviews for ARIC and responses to questionnaires for CHS and MESA. Following a minimum of 8 hours of fasting, blood was drawn from participants and total and HDL cholesterol levels, as well as plasma glucose in blood samples were assayed at central laboratories.18–20 Diabetes was defined as a fasting plasma glucose level ≥126 mg/dL or a history of medical treatment for diabetes.
Outcomes: CVD Events and Non-Cardiovascular Deaths
Protocols and criteria for the ascertainment and diagnosis of CVD events as well as the ascertainment of deaths were largely similar in ARIC, CHS, and MESA and have been reported previously.18,21–25
Statistical Analysis
We examined competing risks for different first CVD events relative to each other and to non-CVD death for black and white men and women separately in each study, then examined racial disparities in risks for specific events within age-sex groups. The purpose of initially comparing CVD events and non-CVD death for each age-race-sex group separately was to clarify the extent to which the risk for non-CVD death “competes” against the risk for CVD within each group. Baseline characteristics were compared across race-sex groups using general linear models for continuous variables and chi-square tests for categorical variables.
In competing risk analyses, the occurrence of one type of event precludes consideration of any subsequent event. Thus, at any point in time, the overall competing risk cumulative incidence for all events is equal to the sum of the cumulative incidences for each type of event since only the first event is counted.26
For our analysis, we first estimated the Kaplan-Meier cumulative incidence for occurrence of any CVD (fatal or non-fatal) or non-CVD death at any time during the time from the baseline exam through the most recent follow-up. We then determined the first event (any CVD event vs. non-CVD death); if a CVD event occurred on the same day as death, then the CVD event was coded as occurring first. When multiple CVD events were diagnosed on the same day, for example MI and HF, we arbitrarily assigned the MI as occurring first and HF as occurring second; events rarely occurred simultaneously in the cohorts analyzed in this study and the results did not change when we altered this statistical assumption.
A competing Cox regression model was used to analyze the competing risks separately for black and white men and women. We used the data augmentation method as described by Lunn and McNeil to fit Cox proportional hazards models. The Lunn and McNeil method enables direct comparisons between failure types and allows inferences on the instantaneous rate of a particular type of failure among those still at risk in the presence of the competing risks. The estimates created by these models provide hazards ratios and competing cumulative incidences for all CVD events combined compared with non-CVD death.27 These two failure types were evaluated and no evidence of non-proportionality of hazard ratios was found.
We also used the competing risks model described by Fine and Gray, which uses competing Cox models for the subdistribution hazard to assess the association of race with the cumulative incidence of five competing outcomes: CHD (defined as non-fatal MI or CHD death), fatal/non-fatal stroke, HF, other CVD death, and other non-CVD causes of death.28 The subdistribution analysis allows direct inferences regarding the proportion experiencing a particular event over time in the presence of competing risks. A property of this approach is that the risk of a competing event (e.g. non-CVD death) can lead to a difference in the subdistribution of the event of inference (e.g. CVD events) even when the cause-specific hazards for the event of interest are the same. These models provide the hazards ratio for blacks compared with whites for a given outcome in the context of risks for all competing outcomes. The analyses were performed using SAS statistical software, R version 2.10.1 and its competing risk library cmprsk. For each cohort, models were performed unadjusted (Model 1), adjusted for age, education, and income (Model 2), and additionally adjusted for CVD risk factors including systolic blood pressure, total cholesterol, diabetes, and smoking (Model 3). Baseline age, systolic blood pressure, and total cholesterol were modeled as continuous linear covariates whereas education (less than high school, high school, college or bachelor, graduate or professional), income (<$35000, $35000 or greater), diabetes (yes, no), and smoking (never, current, former) were modeled categorically. In the adjusted regression models, the proportional hazards model was used to assess the effects of covariates on the subdistribution of CVD events in a competing risks setting.
RESULTS
Baseline Characteristics
Baseline characteristics of ARIC and CHS participants included in our analyses are shown stratified by sex and race in Tables 1A and 1B, respectively. Blacks had greater risk factor burden than whites in both age groups. Baseline characteristics among MESA participants were similar to those among ARIC and CHS participants (Supplementary Tables 1A and 1B).
Table 1A.
Baseline Characteristics of ARIC Participants by Race and Sex (N=14954)
| White Men (N=4941) | Black Men (N=1531) | P Value | White Women (N=5934) | Black Women (N=2548) | P Value | |
|---|---|---|---|---|---|---|
| Age (yrs) | 54.6 | 53.7 | <.01 | 53.9 | 53.3 | <.01 |
| Education, % | <.01 | <.01 | ||||
| High School or Less | 46.4 | 63.2 | 58.8 | 63.1 | ||
| Greater than High School | 53.6 | 36.8 | 41.2 | 36.9 | ||
| Income, % | <.01 | <.01 | ||||
| <$35,000/yr | 39.5 | 76.3 | 51.2 | 87.3 | ||
| ≥$35,000/yr | 60.5 | 23.7 | 48.8 | 12.7 | ||
| Systolic Blood Pressure, mmHg | 120.1 | 130.4 | <.01 | 116.9 | 128.0 | <.01 |
| Diastolic Blood Pressure, mmHg | 73.4 | 82.6 | <.01 | 69.6 | 78.1 | <.01 |
| Total Cholesterol, mg/dl | 210.3 | 210.8 | 0.70 | 218.1 | 217.2 | 0.36 |
| HDL Cholesterol, mg/dl | 43.1 | 50.8 | <.01 | 57.5 | 58.0 | 0.29 |
| Triglycerides, mg/dl | 146.1 | 119.1 | <.01 | 128.0 | 109.3 | <.01 |
| BMI, kg/m2 | 27.4 | 27.6 | 0.13 | 26.6 | 30.8 | <.01 |
| Smoking Status, % | <.01 | <.01 | ||||
| Never | 29.2 | 28.9 | 50.9 | 58.1 | ||
| Former | 46.3 | 33.0 | 24.2 | 17.4 | ||
| Current | 24.6 | 38.1 | 25.0 | 24.5 | ||
| Diabetes, % | 6.9 | 14.9 | <.01 | 6.4 | 18.0 | <.01 |
| Treated for hypertension, % | 18.2 | 32.9 | <.01 | 19.0 | 43.9 | <.01 |
Table 1B.
Baseline Characteristics of CHS Participants by Race and Sex (N=4288)
| White Men (N=1471) | Black Men (N=258) | P Value | White Women (N=2149) | Black Women (N=410) | P Value | |
|---|---|---|---|---|---|---|
| Age (yrs) | 73.1 | 72.6 | .17 | 72.1 | 72.6 | 0.05 |
| Education, % | <.01 | <.01 | ||||
| High School or Less | 50.3 | 62.4 | 55.9 | 65.9 | ||
| Greater Than High School | 49.7 | 37.6 | 44.1 | 34.2 | ||
| Income, % | <.01 | <.01 | ||||
| < $35,000/yr | 68.9 | 85.3 | 76.5 | 91.7 | ||
| ≥$35,000/yr | 31.1 | 14.7 | 23.6 | 8.3 | ||
| Systolic Blood Pressure, mmHg | 136.0 | 139.0 | 0.03 | 135.2 | 143.1 | <.01 |
| Diastolic Blood Pressure, mmHg | 72.2 | 76.6 | <.01 | 69.3 | 74.9 | <.01 |
| Total Cholesterol, mg/dl | 198.4 | 198.8 | 0.86 | 221.6 | 215.7 | <.01 |
| HDL Cholesterol, mg/dl | 48.0 | 52.3 | <.01 | 59.3 | 61.4 | 0.02 |
| Triglycerides, mg/dl | 139.7 | 119.2 | 142.6 | 114.0 | <.01 | |
| BMI, kg/m2 | 26.4 | 26.7 | 0.24 | 26.3 | 29.5 | |
| Smoking Status, % | <.01 | |||||
| Never | 32.8 | 28.4 | 56.4 | 57.4 | 0.94 | |
| Former | 56.8 | 49.0 | 30.5 | 29.9 | ||
| Current | 10.4 | 22.6 | 13.1 | 12.8 | ||
| Diabetes, % | 15.2 | 23.9 | <.01 | 10.7 | 24.4 | <.01 |
| Treated for hypertension, % | 35.1 | 48.5 | <.01 | 38.0 | 62.1 | <.01 |
First Events Across Cohorts
The numbers of participants within each cohort experiencing first incident non-CVD death and selected CVD events are shown in Table 2 (these data stratified by race and sex are shown in Supplementary Tables 2 and 3). The competing cumulative incidences of non-CVD death and selected CVD events within each race-sex group in the ARIC cohort are shown in figures 1–4. Among 14569 ARIC participants, 5.0% experienced non-CVD death and 11.6% experienced CVD as a first event (mean follow-up 10.5 years). HF and CHD were the most common first CVD events. A greater proportion of CHS participants experienced events (mean follow-up 8.5 years); 15.7% experienced non-CVD death as a first event, while 43.2% experienced first incident CVD. CHD was the most common type of first CVD event among CHS participants. Findings among MESA participants of comparable ages to ARIC and CHS participants were similar although fewer events occurred in MESA due in part to shorter follow-up times (Supplementary Table 4).
Table 2.
First Incident Events for ARIC, CHS, YOUNGER MESA, and OLDER MESA Participants (N=23522)
| Percent of Participants within Cohort with Selected Event As First Incident Event | ||||
|---|---|---|---|---|
| ARIC (Ages 45–64; N=14954) | YOUNGER MESA (Ages 45–64; N=2418) | CHS (Age ≥ 65; N=4288) | OLDER MESA (Age ≥ 65; N=1862) | |
| Duration of Follow-up, Years | 10.5 | 5.6 | 8.5 | 5.2 |
| First Incident Event | ||||
| Non-CVD Death | 5.0 % (749) | 1.3 % (32) | 15.7 % (673) | 4.8 % (90) |
| CVD Event | 11.6 % (1731) | 4.7 % (114) | 43.2 % (1855) | 12.1 % (225) |
| CHD Death or Nonfatal MI | 3.8 % (572) | 2.8 % (68) | 17.9 % (767) | 6.0 % (112) |
| Stroke | 2.4 % (358) | 1.0 % (24) | 9.7 % (414) | 2.8 % (52) |
| Heart Failure | 3.9 % (588) | 0.7 % (18) | 9.6 % (413) | 2.2 % (41) |
| Other CVD death | 1.4 % (213) | 0.2 % (4) | 5.0 % (216) | 1.1 % (20) |
| No Events | 83.4 % (12474) | 94.2 % (2278) | 41.0 % (1760) | 83.1 % (1547) |
Figure 1.
Competing Cumulative Incidences of CVD Events and Non-CVD Death Among Black Male ARIC Participants
Figure 4.
Competing Cumulative Incidences of CVD Events and Non-CVD Death Among White Female ARIC Participants
Competing Risks for First CVD Events versus Non-CVD Death Within Groups
When comparing competing risks of first CVD events versus non-CVD death within race-sex groups of the ARIC cohort (ages 45–64 at baseline), CVD occurred three times more frequently than non-CVD death as a first event for white men, while black men were two times more likely to have their first event be CVD rather than non-CVD death (Table 3). First incident CVD was more than twice as common as non-CVD death as a first event in both black and white women in ARIC. These patterns persisted following adjustment for income and education. After further adjustment for CVD risk factors, first incident CVD was no longer significantly more likely than non-CVD death as a first event in black men and women.
Table 3.
Competing Cumulative Incidences and Hazard Ratios for First CVD Events Compared with Non-CVD Death Within Race-Sex Groups (Lunn and McNeil Method)
| ARIC Participants (N=14569; 10.5 Years Mean Follow-Up)
|
CHS Participants (N=4237; 8.5 Years Mean Follow-Up)
|
|||||||
|---|---|---|---|---|---|---|---|---|
| Men | Women | Men | Women | |||||
|
| ||||||||
| Black (n= 1394) | White (n= 5211) | Black (n= 2228) | White (n= 5736) | Black (n= 249) | White (n= 1461) | Black (n= 391) | White (n= 2136) | |
|
|
|
|||||||
| Non-CVD Death | 11.8% | 7.5% | 5.1% | 3.7% | 30.7% | 18.1% | 9.9% | 13.5% |
|
|
|
|||||||
| Any CVD Event | 23.9 | 18.9 | 16.5 | 8.8 | 42.0 | 48.7 | 38.1 | 39.0 |
|
|
|
|||||||
| CHD Death or Nonfatal MI | 6.2 | 7.7 | 3.3 | 2.7 | 17.3 | 23.1 | 13.4 | 14.6 |
|
|
|
|||||||
| Stroke | 5.6 | 2.7 | 4.3 | 1.5 | 6.2 | 8.7 | 9.3 | 10.2 |
|
|
|
|||||||
| HF | 7.3 | 6.1 | 7.2 | 3.7 | 10.0 | 9.9 | 10.8 | 8.9 |
|
|
|
|||||||
| Other CVD Death | 4.8 | 2.5 | 1.7 | 0.9 | 8.4 | 7.0 | 4.5 | 5.3 |
|
Model 1: Unadjusted
| ||||||||
| Hazards Ratio for CVD vs Non-CVD Death Within Group | 2.02 (1.65, 2.48) | 2.97 (2.60, 3.39) | 3.18 (2.55, 3.97) | 2.29 (1.94, 2.70) | 1.94 (1.39, 2.70) | 2.70 (2.35, 3.11) | 3.93 (2.71, 5.71) | 2.86 (2.52, 3.26) |
|
| ||||||||
|
Model 2: Adjusted for Age, Education, and Income
| ||||||||
| Hazards Ratio for CVD vs Non-CVD Death Within Group | 2.15 (1.49, 3.10) | 3.93 (2.94, 5.24) | 3.57 (2.45, 5.20) | 2.94 (2.10, 4.12) | 2.95 (1.66, 5.21) | 3.10 (2.44, 3.95) | 8.83* (3.72, 21.0) | 3.23 (2.68, 4.13) |
|
| ||||||||
|
Model 3: Adjusted for Age, Education, Income, SBP, TC, Diabetes and Smoking
| ||||||||
| Hazards Ratio for CVD vs Non-CVD Death Within Group | 1.53 (0.90, 2.60) | 3.08 (2.19, 4.33) | 1.35 (0.80, 2.29) | 2.04 (1.36, 3.07) | 2.05 (0.57, 7.38) | 3.10 (1.86, 5.19) | 14.3* (3.54, 58.1) | 4.54 (2.98, 6.90) |
The marked increase in hazards ratios for CVD events vs. non-CVD death with adjustment within older black women in CHS is of uncertain significance and should be interpreted cautiously given the wide confidence intervals.
In the CHS cohort (ages 65 and older at baseline), black and white men and women were significantly more likely to have their first event be CVD rather than non-CVD death in unadjusted and demographic-adjusted analyses (Table 3). After further adjustment for CVD risk factors, CVD was no longer significantly more likely than non-CVD death as a first event in black men. The trends observed within race-sex groups in ARIC and CHS were similar in comparably aged MESA participants (Supplementary Table 4).
Comparison of competing risks of first CVD events and non-CVD death between blacks and whites
When directly comparing blacks and whites in the ARIC cohort, black men were nearly twice as likely as white men to experience non-CVD death as a first event and 1.3 times as likely as white men to experience first incident CVD (Table 4). Black men in ARIC remained more likely than whites to experience CVD as a first event after adjustment for demographics, but this association was attenuated to statistical insignificance. However, after additional adjustment for CVD risk factors, black men actually became significantly less likely than white men to experience CVD as a first event. CHD was the most common first CVD event among white men, whereas HF was the most common among black men. Black men were significantly more likely than white men to experience stroke, HF, or other CVD death as a first event. White men were more likely than black men to have CHD as a first event, but this distinction was of borderline statistical significance. When a sensitivity analysis of the ARIC cohort was performed using traditional Cox modeling to examine first CVD events, the results differed from the competing risks model (Table 5), especially with regard to risks for CHD. The hazards ratios for black men compared with white men for specific CVD events were somewhat greater in the traditional Cox model than in the competing risks model, particularly with regard to CHD; black men were estimated to be more likely than white men to experience CHD using a standard Cox model (HR 1.06; 95% CI 0.90, 1.26) whereas they appeared less likely than white men to have a first CHD event using a competing risks model (HR 0.77; 95% CI 0.60, 1.00). In a supplementary analysis in which each covariate was sequentially added for adjustment to the Fine and Gray competing risks model (Supplementary Table 5), the hazards ratio for any CVD event for black men compared with white men was 1.15 prior to adjustment for baseline risk factors but subsequently decreased to 0.99, 0.99, 0.92, and 0.85 after systolic blood pressure, total cholesterol, diabetes, and smoking were sequentially added as adjustment covariates.
Table 4.
Hazard Ratios for CVD Events and Non-CVD Death for Blacks Compared with Whites (Fine and Gray Method)
| ARIC Participants (N=14569; 10.5 Years Mean Follow-Up) | CHS Participants (N=4237; 8.5 Years Mean Follow-Up) | |||
|---|---|---|---|---|
| Men | Women | Men | Women | |
| Hazards Ratio for Selected Event (Black vs. White; White as Referent) | Hazards Ratio for Selected Event (Black vs. White; White as Referent) | Hazards Ratio for Selected Event (Black vs. White; White as Referent) | Hazards Ratio for Selected Event (Black vs. White; White as Referent) | |
| Model 1: Unadjusted | ||||
| Non-CVD Death | 1.87 (1.53, 2.29) | 1.39 (1.10, 1.77) | 1.32 (0.99, 1.76) | 0.68 (0.47, 0.97) |
| CHD Death or Nonfatal MI | 0.77 (0.60, 1.00) | 1.24 (0.94, 1.65) | 0.75 (0.55, 1.02) | 0.95 (0.70, 1.28) |
| Stroke | 2.18 (1.63, 2.90) | 2.91 (2.15, 3.92) | 0.61 (0.34, 1.07) | 0.97 (0.68, 1.38) |
| HF | 1.17 (0.91, 1.49) | 2.07 (1.65, 2.59) | 1.07 (0.69, 1.64) | 1.15 (0.81, 1.64) |
| Other CVD Death | 1.85 (1.34, 2.56) | 2.20 (1.42, 3.42) | 1.29 (0.81, 2.06) | 0.82 (0.48, 1.40) |
| Any CVD Event (Overall) | 1.27 (1.11, 1.45) | 2.03 (1.76, 2.34) | 0.88 (0.71, 1.08) | 1.07 (0.89, 1.30) |
| Model 2: Adjusted for Age, Education, and Income | ||||
| Any CVD Event | 1.15(1.00, 1.32) | 1.82 (1.56, 2.14) | 0.87 (0.70, 1.08) | 0.99 (0.82, 1.20) |
| Model 3: Adjusted for Age, Education, Income, SBP, TC, Diabetes and Smoking | ||||
| Any CVD Event | 0.85 (0.73, 0.98) | 1.28 (1.08, 1.51) | 0.80 (0.64, 0.99) | 0.87 (0.72, 1.05) |
Table 5.
Hazard Ratios among ARIC Participants (N=14569; 10.5 Years Mean Follow-Up) for CVD Events and Non-CVD Death for Blacks Compared with Whites: Competing Risks Method and Standard Cox Method
| Men | Women | |||
|---|---|---|---|---|
| Competing Risks Method (Fine and Gray Method) | Standard Cox Model Method | Competing Risks Method (Fine and Gray Method) | Standard Cox Model Method | |
| Hazards Ratio for Selected Event (Black vs. White; White as Referent) | Hazards Ratio for Selected Event (Black vs. White; White as Referent) | Hazards Ratio for Selected Event (Black vs. White; White as Referent) | Hazards Ratio for Selected Event (Black vs. White; White as Referent) | |
| Model 1: Unadjusted | ||||
| Non-CVD Death | 1.87 (1.53, 2.29) | 1.76 (1.49, 2.07) | 1.39 (1.10, 1.77) | 1.59 (1.33, 1.90) |
| CHD Death/Nonfatal MI | 0.77 (0.60, 1.00) | 1.06 (0.90, 1.26) | 1.24 (0.94, 1.65) | 1.72 (1.40, 2.10) |
| Stroke | 2.18 (1.63, 2.90) | 2.37 (1.84, 3.04) | 2.91 (2.15, 3.92) | 2.92 (2.23, 3.82) |
| HF | 1.17 (0.91, 1.49) | 1.24 (1.02, 1.50) | 2.07 (1.65, 2.59) | 2.14 (1.78, 2.57) |
| Other CVD Death | 1.85 (1.34, 2.56) | N/A | 2.20 (1.42, 3.42) | N/A |
| Any CVD Event (overall) | 1.27 (1.11, 1.45) | 1.29 (1.14, 1.46) | 2.03 (1.76, 2.34) | 2.07 (1.81, 2.37) |
| Model 2: Adjusted for Age, Education, and Income | ||||
| Any CVD Event | 1.15(1.00, 1.32) | 1.19 (1.04, 1.38) | 1.82 (1.56, 2.14) | 1.70 (1.46, 1.98) |
Among female ARIC participants, blacks were more than twice as likely as whites to experience first incident CVD and were significantly more likely to experience each type of CVD event. Black women were also 1.4 times more likely than white women to experience non-CVD death as a first event. These patterns were largely maintained after adjustment for demographic factors and after additional adjustment for CVD risk factors. Black women in ARIC appeared to be at substantially higher risk for CHD than white women in a sensitivity analysis using traditional Cox modeling (HR 1.72; 95% CI 1.40, 2.10); however, in competing Cox analyses this risk was substantially lower and non-significant (HR 1.24; 95% CI 0.94, 1.65). In a supplementary analysis in which each covariate was sequentially added to the adjusted model (Supplementary Table 5), black women were significantly more likely than their white counterparts to have a first CVD event in the unadjusted model and each subsequent model in which adjustment covariates were sequentially added.
In the older CHS cohort, when comparing black and white participants, black women were significantly less likely than white women to experience non-CVD death (Table 4). Black men and women had consistently lower cumulative incidences of each type of CVD event than their white counterparts, although none of these differences were statistically significant in unadjusted models or after adjustment for demographics. After additional adjustment for CVD risk factors, however, black men in CHS were significantly less likely than white men to experience a first CVD event. CHD was the most common type of first incident CVD event for each race-sex group. Findings by age, race, and sex were similar in MESA participants (Supplementary Table 6). A sensitivity analysis using traditional Cox modeling to examine first CVD events yielded results that differed somewhat from the competing risks model (Supplementary Table 7).
DISCUSSION
Principal Findings
In this competing risks analysis that analyzes multiple outcomes simultaneously and assesses whether participants first experience a CVD event (fatal or non-fatal) or non-CVD death, CVD was the most likely first event rather than non-CVD death in each age, sex, and race group studied. The burden of CVD events and non-CVD death appears to affect blacks earlier in life than whites; black men and women aged 45 to 64 at baseline (ARIC) were significantly more likely than their white counterparts to experience CVD or non-CVD death as a first event, but there was no consistent or significant racial disparity in CVD events among participants aged 65 and over at baseline (CHS). Among participants aged 45 to 64 at baseline, blacks were more likely than whites to experience a stroke but less likely to experience CHD, and black men were more likely than their white counterparts to have first incident HF and other CVD not due to CHD.
Implications
The competing risks framework enables assessment of the order and burden of CVD events and non-CVD deaths and is a potentially valuable method for examining racial disparities in incident CVD, especially given the substantially greater rates of non-CVD death among blacks compared with whites.29–31 This greater risk of non-CVD death among blacks indicates a greater burden of competing risk among blacks than among whites – in other words, a greater proportion of blacks at risk for CVD events may not actually experience a CVD event because they die of a non-CVD cause first. While study participants who die of a non-CVD cause without having experienced a prior CVD event would be censored in traditional analyses, leading to potential over-estimation of risk for CVD, these individuals are still incorporated into competing risks analyses.
This study demonstrates that a competing risks analysis may yield somewhat different results than a traditional Cox analysis of identical data. Whereas middle-aged black men in ARIC were somewhat more likely than their white counterparts to experience CHD as a first event in a traditional Cox analysis, they actually were less likely to experience CHD as a first event in a competing risks analysis (Table 5). This ability to understand risks for multiple different outcomes at a given point in time provided by a competing risks analysis may be clinically useful because it affords patients and clinicians a sense of real-life risks for first events. In an era in which health care information is increasingly accessible, a competing risks framework can therefore provide meaningful information to both patients and clinicians regarding comparative risks for different events at various stages of life.
Understanding the temporality of CVD events – particularly first-occurring events – is desirable in light of the high level of morbidity and functional decline that follow incident CVD. Survivors of MI, for instance, have between a 1.5 and 15 times greater likelihood of illness and death than the general population.32 Similarly, after a first stroke, over 20% of U.S. men and women die within one year and approximately 50% die within five years; strokes also impose significant disability on patients and resulting burden on patient caregivers.1,33 The burden of mortality after HF diagnosis is also quite significant; approximately 20% of U.S. adults die within one year of new incident HF, and approximately three-fourths of men and women under 65 years old diagnosed with HF die within eight years.1 In light of this considerable morbidity and mortality, it is critical to understand which events are most likely to occur first for population sub-groups and to prioritize the prevention of common first events in these groups.
The differences we observed in common first CVD events by race, sex, and age have potential implications for CVD prevention and intervention. For instance, our finding that middle-aged blacks were more likely than whites to have strokes as first events is particularly worrisome given their high prevalence of hypertension and comparatively high mortality following strokes; 42% of black women and 34% of black men die within five years of their first stroke, compared to 32% of white women and 32% of white men.1 The greater likelihood of strokes as first events among middle-aged black men compared with whites may reflect disparities in access to health care, such as poorer control of hypertension among black men, which have been observed in large-scale national databases.1,34
Given the discrepancies we observed in specific types of first incident CVD across race-sex-age groups, we hypothesize that it may be beneficial to prioritize disease-specific preventive therapies differently depending on which CVD events are most likely to occur first for particular race-sex-age groups. For instance, for subgroups especially at risk for stroke and HF as first events, efforts focused on blood pressure control through lifestyle and medications may be warranted. Likewise, for groups at risk for any type of first CVD events – whether CHD, HF, or stroke – emphasis on lifestyle modification is certainly warranted. Future studies examining the impact, in terms of events prevented and years of life gained, of targeted strategies aimed at preventing specific first CVD events are warranted.
Our finding that middle-aged black men were more likely than their white counterparts to have CVD as a first event in unadjusted models but actually became less likely to have a first CVD event after adjustment for demographics and risk factors suggests that the earlier burden of CVD among middle-aged black men is largely attributable to their greater CVD risk factor burden. A supplementary analysis in which each covariate was sequentially added to the adjusted model confirmed this. These findings affirm the importance of CVD prevention and risk factor control, especially among groups that the burden of CVD affects earliest, such as middle-aged black men and women.
The significant differences we found in first CVD events by race occurred in middle-aged study participants only. This cannot be attributed only to the larger sample size of our middle-aged study population by comparison to our older study population, as the trends in our older study population suggest that older blacks are equally – if not less – likely than older whites to experience CVD as a first event. Rather, it is possible that disparities in first incident CVD between blacks and whites disappear between middle age and older adulthood because relatively more blacks at high risk of CVD die prior to older adulthood, as suggested by our competing risk methodologies. It is also possible that because the CHS cohort was older at inception than the contemporaneous ARIC cohort, CVD risk factor management may have been applied differently across races for CHS participants in the years preceding their enrollment in CHS (i.e., during their middle-aged years) than it was for the middle-aged ARIC participants following their enrollment in the cohort. Finally, different levels of disparities in baseline CVD risk factor burden between blacks and whites in ARIC compared with CHS may also be responsible for our finding that significant racial disparities in first CVD events existed only for ARIC participants. However, racial disparities in baseline CVD risk factor levels were largely similar in ARIC and CHS.
Limitations
Some limitations of our study should be acknowledged. First, the mean periods of follow-up ranged between 5 and 10 years across studies. Although longer follow-up certainly would have resulted in more incident events, the number of events in ARIC, CHS, and (for some non-stratified analyses) MESA provided sufficient power to detect statistically significant differences in CVD events and non-CVD death within and between groups. Each study had its own protocols for adjudicating outcomes and measuring covariates; however, protocols were similar across cohorts, in part because many of the same investigators were involved in each study and the MESA protocol was developed using earlier protocols as a guide. Events, including deaths, were adjudicated in each cohort, thereby minimizing reporting bias that may occur in studies that rely solely on death certificates for fatal events data. Furthermore, ARIC and CHS cohorts (baseline 1987–1989) were essentially contemporaneous and provided a substantial majority of our data (in the form of events). The MESA participants, who were initially examined in 2000, may have been subjected to more intensive preventive therapy as a result of trends in treatment prevalence as well as intensive surveillance for subclinical CVD. Although we used the MESA cohort to validate results from ARIC and CHS participants in a more contemporary cohort and the trends by race and sex were largely similar, it is certainly possible that different trends in risk factor development and differential application of preventive therapies, particularly statins, across races in recent years would yield results different than those observed in the present study. Another potential limitation of this study is that the black participants of the MESA and ARIC cohorts may be insufficiently representative of blacks in the US population, particularly urban-dwellers living in poverty. Despite these potential limitations, our findings provide important insights into differences in first incident CVD by race, sex, and age, and therefore highlight which specific types of CVD might be prioritized for prevention in various race, sex, and age groups.
Supplementary Material
Figure 2.
Competing Cumulative Incidences of CVD Events and Non-CVD Death Among White Male ARIC Participants
Figure 3.
Competing Cumulative Incidences of CVD Events and Non-CVD Death Among Black Female ARIC Participants
Clinical Summary.
In this manuscript, we used three multi-center, ethnically diverse, NHLBI-sponsored cohort studies to compare differences between black and white study participants in competing risks for specific first cardiovascular disease (CVD) events relative to each other and to non-CVD death. The competing risks framework is distinct from approaches used in typical survival analyses, which consider only a single event at a time (e.g. stroke alone or CHD alone) and assume that each event is independent of the others when, in reality, these events are not independent of one another because of common underlying risk factors. We found that CVD was the most likely first event rather than non-CVD death in black and white adults and that middle-aged blacks were significantly more likely than whites to experience any CVD as a first event, though this disparity disappeared by older adulthood and after adjustment for CVD risk factors. A sensitivity analysis of the identical data using traditional Cox modeling rather than the competing risks methodology yielded different results for CHD risk. Competing risks models estimate which specific events are more likely to occur first for various populations; this ability to understand risks for multiple different outcomes at a given point in time may be clinically useful because it affords patients and clinicians a more accurate sense of real-life risks for first events.
Acknowledgments
ACKNOWLEDGEMENTS AND FUNDING SOURCES
The Atheroscleroris Risk in Communities Study (ARIC) is conducted and supported by the NHLBI in collaboration with the ARIC Study Investigators. This Manuscript was prepared using a limited access dataset obtained from the NHLBI and does not necessarily reflect the opinions or views of the ARIC or the NHLBI.
This research was supported by contracts N01-HC-95159 through N01-HC-95169 from the National Heart, Lung, and Blood Institute. The authors thank the other investigators, the staff, and the participants of the MESA study for their valuable contributions. A full list of participating MESA investigators and institutions can be found at http://www.mesa-nhlbi.org.
Work on this manuscript was supported by contracts: University of Alabama at Birmingham, Coordinating Center, N01-HC-95095; University of Alabama at Birmingham, Field Center, N01-HC-48047; University of Minnesota, Field Center, N01-HC-48048; Northwestern University, Field Center, N01-HC-48049; Kaiser Foundation Research Institute, N01-HC-48050; Wake Forest University, N01-HC-45205; New England Medical Center, N01-HC-45204 from the National Heart, Lung and Blood Institute.
Dr. Lloyd-Jones and this work were supported in part by R21 HL085375. Dr. Lloyd-Jones had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Footnotes
DISCLOSURES: None.
References
- 1.Lloyd-Jones D, Adams R, Brown T, Carnethon M, Dai S, De Simona G, Ferguson TB, Ford E, Furie K, Gillespie C, Go A, Greenlund K, Haase N, Hailpern S, Ho PM, Howard V, Kissela B, Kittner S, Lackland D, Lisabeth L, Marelli A, McDermott MM, Meigs J, Mozaffarian D, Mussolino M, Nichol G, Roger VL, Rosamond W, Sacco R, Sorlie P, Stafford R, TOM T, Wasserthiel-Smoller S, Wong ND, Wylie-Rosett J. Heart Disease and Stroke Statistics – 2010 Update. A Report from the American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Circulation. 2010;121:e46–e215. doi: 10.1161/CIRCULATIONAHA.109.192667. [DOI] [PubMed] [Google Scholar]
- 2.Hozawa A, Folsom AR, Sharrett AR, Chambless LE. Absolute and attributable risks of cardiovascular disease incidence in relation to optimal and borderline risk factors: comparison of African American with white subjects: Atherosclerosis Risk in Communities Study. Arch Intern Med. 2007;167:573–579. doi: 10.1001/archinte.167.6.573. [DOI] [PubMed] [Google Scholar]
- 3.Watkins LO. Epidemiology and burden of cardiovascular disease. Clin Cardiol. 2004;27:III2–III6. doi: 10.1002/clc.4960271503. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Mensah GA, Mokdad AH, Ford ES, Greenlund KJ, Croft JB. State of disparities in cardiovascular health in the United States. Circulation. 2005;111:1233–1241. doi: 10.1161/01.CIR.0000158136.76824.04. [DOI] [PubMed] [Google Scholar]
- 5.Centers for Disease Control and Prevention (CDC) Prevalence of stroke: United States, 2005. MMWR Morb Mortal Wkly Rep. 2007;56:469–474. [PubMed] [Google Scholar]
- 6.Pleis JR, Lucas JW. Summary health statistics for U.S. adults: National Health Interview Survey, 2007. Vital Health Stat 10. 2009;240:1–159. [PubMed] [Google Scholar]
- 7.Hertz RP, Unger AN, Cornell JA, Saunders E. Racial disparities in hypertension prevalence, awareness and management. Arch Intern Med. 2005;165:2098–2104. doi: 10.1001/archinte.165.18.2098. [DOI] [PubMed] [Google Scholar]
- 8.National Institute of Diabetes and Digestive and Kidney Diseases. National Diabetes Statistics and Fact Sheet: General Information and National Estimates on Diabetes in the United States, 2005. Bethesda, Md: US Department of Health and Human Services, National Institutes of Health; 2005. [Google Scholar]
- 9.Cooper R, Cutler J, Desvigne-Nickens P, Fortmann S, Friedman L, Havlik R, Hogelin G, Marler J, McGovern P, Morosco G, Mosca L, Pearson T, Stamler J, Stryer D, Thom T. Trends and disparities in coronary heart disease, stroke, and other cardiovascular diseases in the United States: Findings of the national conference on cardiovascular disease prevention. Circulation. 2000;102:3137–3147. doi: 10.1161/01.cir.102.25.3137. [DOI] [PubMed] [Google Scholar]
- 10.Manton KG, Patrick CH, Johnson KW. Health differentials between blacks and whites: recent trends in mortality and morbidity. The Milbank Quarterly. 1987;65:129–199. [PubMed] [Google Scholar]
- 11.Subramanian SV, Chen JT, Rehkopf DH, Waterman PD, Krieger N. Racial disparities in context: a multilevel analysis of neighborhood variations in poverty and excess mortality among black populations in Massachusetts. Am J Pub Health. 2005;l95:375. doi: 10.2105/AJPH.2003.034132. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Alexander M, Grumbach K, Selby J, Brown AF, Washington E. Hospitalization for congestive heart failure. Explaining racial differences. JAMA. 1995;274:1037–42. [PubMed] [Google Scholar]
- 13.Bahrami H, Kronmal R, Bluemke DA, Olson J, Shea S, Liu K, Burke GL, Lima JA. Differences in the incidence of congestive heart failure by ethnicity: the multi-ethnic study of atherosclerosis. Arch Intern Med. 2008;168:2138–2145. doi: 10.1001/archinte.168.19.2138. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.The Atherosclerosis Risk in Communities (ARIC) Study: design and objectives. The ARIC investigators. Am J Epidemiol. 1989;129:687–702. [PubMed] [Google Scholar]
- 15.Fried LP, Borhani NO, Enright P, Furberg CD, Gardin JM, Kronmal RA, Kuller LH, Manolio TA, Mittelmark MB, Newman A. The Cardiovascular Health Study: design and rationale. Ann Epidemiol. 1991;1:263–76. doi: 10.1016/1047-2797(91)90005-w. [DOI] [PubMed] [Google Scholar]
- 16.Tell GS, Fried LP, Hermanson B, Manolio TA, Newman AB, Borhani NO. Recruitment of adults 65 years and older as participants in the Cardiovascular Health Study. Ann Epidemiol. 1993;3:448–50. doi: 10.1016/1047-2797(93)90062-9. [DOI] [PubMed] [Google Scholar]
- 17.Bild DE, Detrano R, Peterson D, Guerci A, Liu K, Shahar E, Ouyang P, Jackson S, Saad M. Ethnic differences in coronary calcification: the Multi-Ethnic Study of Atherosclerosis (MESA) Circulation. 2005;111:1313–20. doi: 10.1161/01.CIR.0000157730.94423.4B. [DOI] [PubMed] [Google Scholar]
- 18.Detrano R, Guerci A, Carr J, Bild D, Burke G, Folsom A, Liu K, Shea S, Szklo M, Bluemke D, O’Leary D, Tracy R, Watson K, Wong N, Kronmal R. Coronary Calcium as a Predictor of Coronary Events in Four Racial or Ethnic Groups. NEJM. 2008;358:1336–45. doi: 10.1056/NEJMoa072100. [DOI] [PubMed] [Google Scholar]
- 19.National Heart, Lung, and Blood Institute Atherosclerosis Risk in Communities (ARIC) Study. Operations Manual 7: Blood Collecting and Processing. Bethesda: National Heart, Lung, and Blood Institute; 1988. [Google Scholar]
- 20.Cushman M, Cornell ES, Howard PR, Bovill EG, Tracy RP. Laboratory methods and quality assurance in the Cardiovascular Health Study. Clin Chem. 1995;41:264–70. [PubMed] [Google Scholar]
- 21.Bahrami H, Bluemke DA, Kronmal R, Bertoni AG, Lloyd-Jones DM, Shahar E, Szklo M, Lima AC. Novel Metabolic Risk Factors for Incident Heart Failure and Their Relationship With Obesity: The MESA (Multi-Ethnic Study of Atherosclerosis) Study. JACC. 2008;51:1775–1783. doi: 10.1016/j.jacc.2007.12.048. [DOI] [PubMed] [Google Scholar]
- 22.Ives DG, Fitzpatrick AL, Bild DE, Psaty BM, Kuller LH, Crowley PM, Cruise RG, Theroux S. Surveillance and ascertainment of cardiovascular events: The Cardiovascular Health Study. Ann Epidemiol. 1995;5:278–285. doi: 10.1016/1047-2797(94)00093-9. [DOI] [PubMed] [Google Scholar]
- 23.Psaty BM, Kuller LH, Bild D, Burke GL, Kittner SJ, Mittelmark M, Price TR, Rautaharju PM, Robbins J. Methods of assessing prevalent cardiovascular disease in the Cardiovascular Health Study. Ann Epidemiol. 1995;5:270–277. doi: 10.1016/1047-2797(94)00092-8. [DOI] [PubMed] [Google Scholar]
- 24.Borrell LN, Diez Roux A, Rose K, Catellier D, Clark BL. Neighbourhood characteristics and mortality in the Atherosclerosis Risk in Communities Study. Int J Epidemiol. 2003;33:398–407. doi: 10.1093/ije/dyh063. [DOI] [PubMed] [Google Scholar]
- 25.Schulz R, Beach SR, Ives DG, Martire LM, Ariyo AA, Kop WJ. Association between depression and mortality in older adults: the Cardiovascular Health Study. Arch Intern Med. 2000;160:1761–8. doi: 10.1001/archinte.160.12.1761. [DOI] [PubMed] [Google Scholar]
- 26.Satagopan JM, Ben-Porat L, Berwick M, Robson M, Kutler D, Auerbach AD. A note on competing risks in survival data analysis. Br J Cancer. 2004;91:1229–1235. doi: 10.1038/sj.bjc.6602102. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Lunn M, McNeil D. Applying Cox regression to competing risks. Biometrics. 1995;51:524–532. [PubMed] [Google Scholar]
- 28.Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. Journal of the American Statistical Association. 1999;94:496–509. [Google Scholar]
- 29.Wong MD, Shapiro MF, Boscardin WJ, Ettner SL. Contribution of major diseases to disparities in mortality. NEJM. 2002;347:1585–1592. doi: 10.1056/NEJMsa012979. [DOI] [PubMed] [Google Scholar]
- 30.Gornick ME, Eggers PW, Reilly TW, Mentnech RM, Fitterman LK, Kucken LE, Vladeck BC. Effects of Race and Income on Mortality and Use of Services among Medicare Beneficiaries. NEJM. 1996;335:791–799. doi: 10.1056/NEJM199609123351106. [DOI] [PubMed] [Google Scholar]
- 31.Murray CJL, Kulkarni SC, Michaud C, Tomijima N, Bulzacchelli MT, Iandiorio TJ, Ezzati M. Eight Americas: Investigating Mortality Disparities across Races, Counties, and Race-Counties in the United States. PLoS Med. 2006;3:e260. doi: 10.1371/journal.pmed.0030260. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Thom TJ, Kannel WB, Silbershatz H, D’Agostino RB. Cardiovascular disease in the United States and preventive approaches. In: Fuster V, Alexander RW, O’Rourke RA, editors. Hurst’s The Heart, Arteries, and Veins. 10. New York, NY: McGraw-Hill; 2001. pp. 3–7. [Google Scholar]
- 33.Rigby H, Gubitz G, Phillips S. A Systematic Review of Caregiver Burden Following Stroke. Int J Stroke. 2009;4:285–292. doi: 10.1111/j.1747-4949.2009.00289.x. [DOI] [PubMed] [Google Scholar]
- 34.Ong KL, Cheung BM, Man YB, Lau CP, Lam KS. Prevalence, awareness, treatment, and control of hypertension among United States adults 1999–2004. Hypertension. 2007;49:69–75. doi: 10.1161/01.HYP.0000252676.46043.18. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.