Healthy Lifestyle Change and Subclinical Atherosclerosis in Young Adults: Coronary Artery Risk Development in Young Adults (CARDIA) Study

Abstract

Background

The benefits of healthy habits are well-established, but it is unclear whether making health behavior changes as an adult can still alter coronary artery disease risk.

Methods and Results

The Coronary Artery Risk Development in Young Adults (CARDIA) prospective cohort study (n = 3538) assessed 5 healthy lifestyle factors (HLFs) among young adults between ages 18–30 (Year 0 baseline) and 20 years later (Year 20): not overweight/obese, low alcohol intake, healthy diet, physically active, nonsmoker. We tested whether change from Year 0 to 20 in a continuous composite HLF score (HLF change; range: −5 to +5) is associated with subclinical atherosclerosis [coronary artery calcification (CAC) and carotid intima-media thickness (IMT)] at Year 20, after adjustment for demographics, medications, and baseline HLFs. By Year 20, 25·3% of the sample improved (HLF change ≥ +1); 40·4% deteriorated (had fewer HLFs); 34·4% stayed the same; 19·2% had CAC (>0). Each increase in HLFs was associated with reduced odds of detectable CAC (OR = .85, 95% CI: .74 – .98) and lower IMT (carotid bulb β = −.024, p = 0.001), and each decrease in HLFs was predictive to a similar degree of greater odds of CAC (OR = 1.17, 95% CI: 1.02 – 1.33) and greater IMT (β = +.020, p < 0.01).

Conclusions

Healthy lifestyle changes during young adulthood are associated with decreased, and unhealthy lifestyle changes with increased risk for subclinical atherosclerosis in middle age.

Introduction

Coronary heart disease (CHD) is the leading cause of mortality^1–3,4 and behavioral risk factors heighten the odds of developing CHD.^5–8 Cigarette smoking, poor quality diet, physical inactivity, excessive alcohol consumption, and obesity all are major preventable causes of CHD and premature mortality.^5–8

Substantial epidemiological evidence associates the sustained presence of healthy lifestyle factors with reduced risk of myocardial infarction (MI) and CHD mortality.^9–12 Consequently, many advocate a public health policy of primordial prevention that preserves low risk from youth onward by preventing the development of risk factors.^13–15 However, the population prevalence of a behavioral low risk profile is extremely low.^{11, 16} Behavioral Risk Factor Surveillance data show that only 5% of adults meet heart-healthy standards for all of three healthy behaviors: physical activity, fruit and vegetable consumption, and nonsmoking.¹⁶ The fact that most people reach young adulthood with at least one unhealthy behavior makes it essential to learn whether lifestyle changes in adulthood may still affect cardiovascular health.

We used data from the Coronary Artery Risk Development in Young Adults (CARDIA) study to determine whether health behavior changes made earlier in adulthood are associated with the burden and extent of subclinical atherosclerosis in middle age. We predicted that changes in healthy lifestyle behaviors would be related to the odds of coronary artery calcium (CAC) and carotid artery intima-media thickness (IMT), both markers of subclinical cardiovascular disease that predict future cardiovascular events.^{17, 18}

Methods

Cohort Description and Selection

CARDIA, a longitudinal investigation of cardiovascular disease risk factors in a young adult population, has been described in detail elsewhere.¹⁹ The cohort of 5,115 adults (51% of eligible persons contacted) was recruited from four metropolitan areas in the U.S. (Birmingham, Alabama; Chicago, Illinois; Minneapolis, Minnesota; and Oakland, California). Participants, between ages 18 to 30 at Year 0 (1985–1986), were reexamined at Years 2 (1987–88), 5 (1990–91), 7 (1992–93), 10 (1995–96), 15 (2000–01), and 20 (2005–06) with follow-up rates examined of 91%, 86%, 81%, 79%, 74%, and 72% respectively. Follow-up rates for vital status are >90% in each year. Institutional review boards at each site approved the protocol and all participants provided informed consent.

Study Sample

Those who were not examined at Year 20 were excluded. Of the 3549 participants examined at Years 0 and 20, we excluded: an individual who underwent gender reassignment (n=1) and women pregnant at Year 0 (n=4) or Year 20 (n=6) leaving 3538 participants for the primary analysis. The sample included 646 black men, 1000 black women, 889 white men, and 1003 white women.

Assessment of Demographic and Healthy Lifestyle Factors

Participants fasted for 12 hours and avoided smoking and heavy physical activity for 2 hours before each clinical exam. Three seated blood pressure (BP) measurements were obtained with random-zero sphygmomanometer; the average of the 2^nd and 3^rd readings was used. Serum total cholesterol was determined by enzymatic procedures using the ABA Biochromatic instrument.²⁰ Body weight was measured to the nearest 0.2 lb, with participants wearing light clothing and no shoes. Height was measured without shoes to the nearest 0.5 cm. Body mass index (BMI) was computed as weight in kilograms divided by height in meters squared (kg/m²). Age, sex, race, education, and medical information (current treatment and history) were ascertained through questionnaires. At Y0, participants brought in a list of all medications used; at Y20 they physically brought in their medications. Moderate and vigorous physical activity over the previous 12 months²¹ and cigarette smoking were assessed by validated interviewer-administered questionnaires. “Current smoking” was defined as smoking at least 5 cigarettes per week almost every week. Weekly alcohol (beer, wine, and liquor) intake, assessed by questionnaire, was converted to milliliters of alcohol per day.²²

Diet history was obtained at Years 0 and 20 using the validated²³ CARDIA Diet History, an interviewer-administered quantitative food frequency questionnaire.^24, Approximately 700 food items were used to develop a dietary assessment tool suitable for various populations and ethnic groups.^{23, 24} Participants were asked to recall their usual dietary intake for the previous 28 days and report the frequency, amount, and method of food preparation.

Definition of Healthy Lifestyle Factor Scores

The 5 HLFs measured at Years 0 and 20 were: 1) not currently smoking cigarettes; 2) average physical activity ≥60^th percentile at Year 0 by race and sex);¹¹ 3) BMI <25.0 kg/m²; 4) alcohol intake ≤15 g/day (women) or ≤30 g/day (men); and 5) healthy diet reflecting low daily intake of saturated fatty acids and high intake of potassium, calcium, and fiber. For the dietary assessment, participants were assigned a score of 1 to 5 according to a race-sex-specific quintile of each of the four nutrients, with 5 representing the more favorable quintile. The four scores were summed, and a score ≥ 60^th percentile at Year 0 by race and sex was considered to be a healthy diet. A Healthy Lifestyle Factor (HLF) score (range: 0 to 5), summing the number of HLFs present, was computed for each participant at Years 0 and 20, following the method used by Liu et al.²⁵ We also calculated a score representing change in HLF from Year 0 to Year 20 (HLF change; range: −5 to +5).

Measurement of Coronary Artery Calcium (CAC) and Carotid Intima–Media Thickness (IMT)

CAC was measured at Year 20 by computed tomography (CT) of the chest.^{11, 26} Electron beam CT and multidetector CT scanners were used to obtain 40 contiguous 2.5–3-mm-thick transverse images from the root of the aorta to the apex of the heart in 2 sequential scans. Participants were scanned over a hydroxylapatite phantom to allow monitoring of image brightness and noise and adjust for scanner differences. Data from both scans were transmitted electronically to the CARDIA CT Reading Center. A calcium score in Agatston units was calculated for each calcified lesion and the scores were summed across all lesions within a given artery and across all arteries (left anterior descending, left main, circumflex, and right coronary) to obtain the total calcium score for the participant. Because CAC is uncommon before midlife,^{27, 28} detectable CAC (>0) was used as an outcome.

High-resolution B-mode ultrasound was used to acquire a longitudinal image of the common carotid, two of the carotid artery bulb, and two of the internal carotid artery above the bulb on the right and left sides.²⁹ Measurements of the maximal carotid IMT were made at a central reading center by readers blinded to all clinical information. The maximum IMT of the common carotid was defined as the mean of the carotid IMT of the near and far wall on both the left and right sides, with 1 to 4 measurements available for the common carotid and 1 to 8 for the carotid artery bulb and internal carotid artery. Carotid IMT was analyzed as a continuous measure.

Statistical Analyses

Multiple Imputation of Missing Data

Most variables used in our analyses had less than 2% missingness, and the percent missing of our primary outcomes ranged from 8%–14%. Even though the fraction of missing values in our data set was relatively small, a complete-case analysis resulted in discarding 1270 observations. Compared with participants who had no missing values at year 20, those with missing values were somewhat younger and less educated (both by 10 months) and were more likely to be Black, male, and smokers. To prevent reducing generalizability by discarding those with incomplete data, we imputed missing values in the primary cohort of 3538 using the sequential regression imputation approach³⁰ implemented in IVEware (software from the Survey Research Center, Institute for Social research, University of Michigan). Five datasets were generated using data from all 7 exams, resulting in complete Year 0 and Year 20 data for the sample of N = 3538. Each completed data set was analyzed separately and results were combined using Rubin’s rules.³¹

Sensitivity Analyses

A complete case analysis excluded participants with missing values on any variable in our model, resulting in an analysis of 2268 cases. Additional sensitivity analyses included baseline education, a potential confounder, in both the models that used imputed data and those based on the sample of 2268 with complete data.

Primary Analyses

Baseline (Year 0) characteristics were calculated for the six HLF groups (HLF 0 through 5), and trends across the groups were tested using linear or logistic regression. We used logistic regression to estimate the odds of CAC (>0) and linear regression to estimate carotid IMT expressed as a continuous variable at Year 20. All models included age, sex, and race, Year 0 HLF score and HLF change score (Y20 HLFs – Y0 HLFs). All analyses were conducted with SAS statistical software version 9.2.

Results

Sample Demographics

TABLE 1 shows baseline (Year 0) characteristics as a function of number of healthy lifestyle factors. Supplementary Table 1 compares the baseline characteristics of those included in the analysis to those who were not included. HLFs were inversely related to total cholesterol, blood pressure, BMI, triglycerides, lipids, and fasting glucose. The sample (N = 3538) included approximately equal proportions of men and women, and black and white participants. Mean age at Year 0 was 25.1 years (SD = 3.6) and mean education 14.0 years (SD=2.2). The modal number of HLFs was three (36% of the sample); only 8% had all five HLFs at Year 0.

Table 1.

Baseline Characteristics of CARDIA Participants by Number of Healthy Lifestyle Factors at Year 0

# HLFs at Y01	5	4	3	2	1	0	p-value
# of people	283	826	1278	882	238	31
%	8	23	36	25	7	1
Women (%)	59	58	56	56	55	45	.18a
Black (%)	41	43	45	51	53	48	<.0001a
Total Chol.	171	175	176	181	184	179	<.0001b
DBP (mmHg)	67	69	68	69	70	71	<0.01b
SBP (mmHg)	108	109	110	111	113	116	<.0001b
BMI	22	23	24	27	28	27	<.0001b
Triglycerides (mg/DL)	59	66	69	80	94	123	<.0001b
HDL (mg/DL)	56	55	54	51	50	52	<.0001b
LDL (mg/DL)	103	107	109	114	116	102	<.0001b
Glucose (mg/DL)	81	82	82	83	84	85	.0001b
Diabetes (%)	0.4%	0.2%	0.3%	0.7%	2.0%	0%	0.01a

¹HLF: healthy lifestyle factor; Y0: Year 0

^aLogistic regression;

^bLinear regression.

HLF Change and Coronary Artery Calcium (CAC)

At year 20, 19.2% of the sample had CAC>0 (13.3% CAC: 1–100; 3.4% CAC: 101–400; 2.5% CAC>400). TABLE 2 reports results of the logistic regression model estimating the odds of having CAC at Year 20. In addition to Y0 HLF score, the following variables were included as covariates: age, sex, race, education, and use of hypertension, cholesterol lowering, or diabetes medication. Results indicate that being older, male, white, less educated, using hypertension or cholesterol lowering medication, or having fewer HLFs at baseline were all associated with greater odds of having CAC at Year 20. Importantly, after adjustment for these factors, a change in HLF was associated with significantly altered odds of having CAC at Year 20, OR = .85 (95% CI: .77 – .94). A 1-HLF increase was associated with a 15% reduction in odds of detectable CAC. The relation between HLF change and odds of detectable CAC was graded: those with a greater increase in HLFs had a proportionally lower prevalence of CAC; those with a greater decrease in HLFs had proportionally higher prevalence of CAC and greater odds of CAC (see FIGURE 1). The pattern of results was consistent when models were run using non-imputed data (Supplementary Table 2)

Table 2.

Healthy Lifestyle Factor (HLF) Change Score Predicting Detectable CAC at Y20^a

Variable	OR	Lower CI	Upper CI	p-value
Age (years)	1.11	1.08	1.15	<.0001
Male	3.25	2.66	3.96	<.0001
Black	.73	.59	.90	.004
Education	.94	.91	.98	.004
Hypertension medication	1.78	1.41	2.27	<.0001
Cholesterol lowering medication	1.93	1.43	2.61	<.0001
Diabetes medication	1.31	.88	1.96	.19
Y0 HLF score	.74	.66	.84	<.0001
HLF change (Y20-Y0)	.85	.77	.94	.00

^aCovariates included in the model were Age, Sex, Race, maximum attained education, HLF score at Year 0, and hypertension medication, cholesterol lowering medication, and diabetes medication at Year 20.

^bResults are based on analyses combined across 5 multiply imputed data sets.

Figure 1.

Graded relationship observed between HLF change (from Year 0 to Year 20) and incidence of CAC at Year 20. Green shows those who increased HLFs. Gray shows those whose HLFs stayed the same. Red shows those who decreased HLFs. Checked section signifies CAC between 1–100; Dotted section signifies CAC >100. Adjusted for age, race, sex, and baseline number of HLFs; based on imputed data.

Assessment of HLF change in these models assumes a uniform linear effect for both healthy (positive) and unhealthy (negative) changes in lifestyle. To test this assumption, we also ran the above model with number of healthy and unhealthy HLF changes entered as independent variables. The results confirmed a uniform linear effect: each additional healthy HLF change reduced the odds of detectable CAC (OR = .85, 95% CI: .74 – .98) and each additional unhealthy HLF change increased the odds of detectable CAC (OR = 1.17, 95% CI: 1.02 – 1.33) Note that the effect associated with healthy change was nearly identical in magnitude (OR = 1/.85 = 1.18) to that associated with unhealthy change (OR = 1.17).

HLF Change and Carotid Intima-media Thickening (IMT)

At year 20, 6.8% of the sample had common carotid IMT>1; 45.5% had carotid bulb IMT >1; 12.3% had internal carotid >1%. Linear regression adjusting for age, sex, race, education, use of hypertension, cholesterol lowering, or diabetes medication, and Year 0 HLF score, showed that an increase in HLFs was associated with lower carotid IMT at Year 20: common carotid IMT, β = −.008, p = .001; carotid bulb IMT, β = −.022, p < 0.0001; internal carotid IMT, β = −.014, p = 0.002. (See TABLE 3). The pattern of results was consistent when models were run using non-imputed data. (Supplementary Table 3)

Table 3.

Healthy Lifestyle Factor (HLF) Change Score Predicting Carotid IMT at Y20^a

	Common IMT			Bulb IMT			Internal IMT
Variable	Regression Coefficient	SE	p-value	Regression Coefficient	SE	p-value	Regression Coefficient	SE	p-value
Age (years)	0.008	0.001	<0.0001	0.011	0.001	<0.0001	0.007	0.001	<0.0001
Male	0.048	0.004	<0.0001	0.094	0.009	<0.0001	0.051	0.008	<0.0001
Black	0.062	0.005	<0.001	0.000	0.010	0.97	0.011	0.008	<0.17
Education (years)	-0.002	0.001	0.07	−0.009	0.002	<0.0001	−0.004	0.002	0.01
Hypertension medication*	0.030	0.006	<0.0001	0.058	0.015	0.0001	0.039	0.010	0.0002
Cholesterol medication*	0.015	0.009	0.09	−0.002	0.018	0.91	0.022	0.013	0.09
Diabetes medication*	0.037	0.011	0.001	0.071	0.025	0.005	0.015	0.020	0.45
Y0 HLF score	−0.016	0.002	<0.0001	−0.025	0.006	<0.0001	−0.022	0.004	<0.0001
HLF change (Y20-Y0)	−0.008	0.002	0.0009	−0.022	0.005	<0.0001	−0.014	0.004	0.002

^aCovariates included in the model were Age, Sex, Race, maximum attained education, HLF score at Year 0, and hypertension medication, cholesterol lowering medication, and diabetes medication at Year 20. Results are based on analyses combined across 5 multiply imputed data sets.

We next ran the above model with number of healthy and unhealthy HLF changes entered as independent covariates. Results confirmed a uniform linear effect, such that healthy HLF change predicted significantly lower carotid IMT at Year 20 on two of the three indicators: common carotid IMT, β = −.005, p = .12; carotid bulb IMT, β = −.024, p = 0.001; internal carotid IMT, β = −.015, p < 0.01. Conversely, negative HLF change predicted significantly higher carotid IMT at Year 20 on all three indicators: common carotid IMT, β = +.009, p = .001; carotid bulb IMT, β = +.020, p < 0.01; internal carotid IMT, β = +.012, p < 0.05. Note that effects associated with healthy lifestyle change were nearly identical in magnitude to those associated with unhealthy change. Figure 2 shows the graded relationship between HLF change from Year 0 to Year 20 and incidence of Bulb IMT above the 80^th percentile at Year 20.

Figure 2.

Graded relationship observed between HLF change (from Year 0 to Year 20) and incidence of Bulb IMT above the 80^th percentile at Year 20. Green shows those who increased HLFs. Gray shows those whose HLFs stayed the same. Red shows those who decreased HLFs. Checked section signifies bulb IMT between 80^th – 90^th percentile. Dotted section signifies IMT >90^th percentile. Adjusted for age, race, sex, and baseline number of HLFs; based on imputed data.

Individual HLFs

We also ran separate models predicting Year 20 CAC > 0 and IMT in the top 80% for each individual HLF where the HLF was treated either as a continuous variable (Table 4) or as a categorical variable (Supplementary Tables 4–7). Heavier smoking at Year 0 predicted increased odds of having CAC or IMT at Year 20, and achieving or maintaining nonsmoking was associated with reduced odds. Having a higher BMI at Year 0 predicted increased odds of having CAC and IMT at Year 20. Maintaining a non-overweight body weight was associated with reduced odds of CAC and IMT, and gaining more weight gain was associated with increased odds of having common and internal IMT in the top 80%. Neither baseline nor change in physical activity and alcohol intake were significantly associated with CAC or IMT, and healthy change in diet was only associated with reduced odds of having Year 20 bulb IMT in the top 80%.

Table 4.

Odds ratios^*, 95% CI, and p-values for continuous measures of Healthy Lifestyle Factor components predicting Year 20 CAC and Carotid IMT.

HLF component	Variable	CAC > 0	Common IMT, top 80%	Bulb IMT, top 80%	Internal IMT, top 80%
Smoking
	Y0 Cigarettes/d	1.04 (1.02, 1.05), <0.0001	1.02 (1.01, 1.04), 0.004	1.02 (1.01, 1.04), 0.002	1.01 (1.00, 1.03), 0.05
	Y20-Y0 cigarettes/d	1.01 (0.99, 1.03), 0.18	1.01 (1.00, 1.03), 0.14	1.02 (1.00, 1.03), 0.02	1.00 (0.98, 1.02), 0.76
High physical activity
	Y0 PA score (per 100 PA units)	1.00 (0.96, 1.04), 0.86	0.98 (0.94, 1.02), 0.33	0.99 (0.96, 1.03), 0.78	0.97 (0.93, 1.01), 0.15
	Y20-Y0 PA change (per 100 PA units)	0.99 (0.95, 1.02), 0.43	0.99 (0.96, 1.03), 0.73	0.99 (0.95, 1.02), 0.44	0.96 (0.92, 1.00), 0.04
BMI
	Y0 BMI	1.05 (1.03, 1.08), <0.0001	1.10 (1.08, 1.12), <0.0001	1.04 (1.02, 1.06), 0.0002	1.06 (1.04, 1.07), <0.0001
	Y20-Y0 BMI	1.01 (0.99, 1.03), 0.50	1.03 (1.01, 1.05), 0.003	0.99 (0.97, 1.01), 0.41	1.02 (1.01, 1.04), 0.006
Alcohol
	Y0 alcohol (per 10 g)	1.05 (1.00, 1.11), 0.06	0.97 (0.91, 1.02), 0.22	1.01 (0.95, 1.06), 0.82	0.96 (0.91, 1.02), 0.24
	Y20-Y0 alcohol (per 10 g)	1.03 (0.99, 1.07), 0.15	0.96 (0.92, 1.01), 0.15	1.01 (0.96, 1.05), 0.80	1.01 (0.96, 1.06), 0.77
Healthy diet
	Y0 diet score	1.01 (0.96, 1.06), 0.68	1.02 (0.97, 1.06), 0.50	0.96 (0.92, 1.00) 0.07	0.98 (0.94, 1.03), 0.44
	Y20-Y0 diet score	1.01 (0.96, 1.07), 0.62	1.01 (0.97, 1.05), 0.59	0.95 (0.92, 0.99), 0.02	1.02 (0.98, 1.06), 0.36

^*adjusted for age, sex, race, maximum attained education, Y20 hypertension medication, Y20 cholesterol lowering medication, and Y20 diabetes medication.

To learn whether any single HLF accounted for the associations between the combined HLF change score and Year 20 CAC and IMT, we removed each HLF one at a time and reexamined these associations (Table 5). The residual composite HLF change score remained significantly associated with Year 20 CAC and IMT.

Table 5.

Odds ratio^* (p-value) for HLF change score when 1 HL factor is omitted

	CAC > 0	Bulb IMT, top
Omitted HL factor	OR (p-value)	OR (p-value)
Not smoking	0.88 (0.02)	0.87 (0.01)
High physical activity	0.84 (0.003)	0.78 (<0.0001)
BMI < 25.0	0.82 (0.0002)	0.85 (0.002)
Low alcohol	0.82 (0.0004)	0.81 (<0.0001)
Healthy diet	0.76 (<0.0001)	0.79 (<0.0001)

^*adjusted for age, race, sex, and 4 indicator variables representing the 5 possible baseline HLF scores (0, 1, 2, 3, or 4)

Discussion

Change in Healthy Lifestyle Factors from young adulthood to middle age is significantly associated with the presence and extent of subclinical atherosclerosis measured after 20 years of follow-up. The association held after adjustment for demographics, medications, and the baseline number of HLFs in young adulthood. The effect also was evident for two indicators of subclinical disease: CAC and carotid IMT, both predictive of future coronary events.^{17, 18} Further, the relation between HLF change and subclinical atherosclerosis was graded, such that adding HLFs was associated with improved CAC and IMT outcomes, whereas reducing HLFs was associated with poorer outcomes.

Associations between health behaviors and subclinical atherosclerosis may shed light on pathways toward the development of CHD, coronary events, and mortality. HLFs, assessed at a single time point, have been predictive of cardiovascular and all-cause mortality in several studies.^{10, 32–34} Individual HLFs (BMI, smoking, alcohol consumption, poor diet and physical inactivity) are also associated with CAC and/or IMT.^35–38 Using the same combined HLF score as this study, Liu and colleagues found that having more HLFs in young adulthood was associated with having a low cardiovascular disease risk profile in middle age (i.e., untreated cholesterol <200 mg/dL, untreated blood pressure <120/<80 mm Hg, no history of diabetes mellitus or myocardial infarction).²⁵

The stark reality, however, is that few adults have multiple HLFs. Indeed, our observation that only 10% of the young adults in our sample had all 5 HLFs is consistent with earlier estimates.³⁹ In terms of individual health behaviors, fewer than 25% of U.S. adults meet dietary guidelines⁹ and 25% report no leisure time physical activity.¹⁰ Presently, 21% of adults in the U.S.⁴⁰ report cigarette smoking.⁴¹ Suboptimal diet and a sedentary behavior pattern are pervasive, and risk behaviors tend to cluster^{11–13, 16} heightening disease risk.¹⁷ The fact that most people reach young adulthood already having acquired at least one unhealthy behavior raises a pressing question for clinical practice policy: does healthy lifestyle change in adulthood actually reduce the risk of cardiovascular disease? If poor health habits acquired by early adulthood have already done their damage, subsequent intervention to promote healthy lifestyle change might be futile.

Only a few studies have prospectively examined whether change in HLFs during adulthood is related to cardiovascular morbidity and mortality or to early markers of coronary disease. The Atherosclerosis Risk in Communities Study (ARIC) study found fewer cardiovascular events and lower all-cause mortality among those who adopted a healthy lifestyle in middle age.⁴² Hu et al. found that reduced smoking and improved diet explained a significant portion of the decline in coronary disease incidence in the Nurses’ Health Study cohort, whereas increased obesity apparently slowed the decline.⁴³ Gregg et al. found decreased all-cause and cardiovascular mortality among elderly women who increased physical activity.⁴⁴ Quitting smoking during young and middle adulthood has been associated with diminished incidence of CAC,⁴⁵ and increased obesity has been linked to increased incidence of CAC⁴⁶ and IMT.⁴⁷ In the only prior study to relate naturalistic, bidirectional change in a healthy lifestyle factor to measures of atherosclerosis, Buscemi et al.⁴⁶ observed that carotid IMT decreased among those who lost weight and increased for those who gained weight over a 10 year period. In a clinical trial, Ornish and colleagues randomized adults with bioverified coronary artery disease (CAD) to either intensive lifestyle intervention (low-fat vegetarian diet, stopping smoking, stress management training, and moderate exercise), as compared to usual care control. The investigators observed decreased stenosis and fewer cardiac events over a 5 year follow up period among treated participants, versus increased stenosis and events for controls.^48–50 The present results are consistent with prior findings and suggest that, even after reaching young adulthood, making changes in lifestyle behaviors can still impact the odds of coronary atherosclerosis. The graded, linear association we observed between changes in HLFs and presence and extent of subclinical disease suggests that making healthy lifestyle changes during early or midlife might be able to reverse or change the natural progression of CAD. Making changes in smoking and body weight appears to have greatest impact, and changing other HLFs including physical activity, diet and alcohol intake combines in a graded manner to affect the odds of subclinical atherosclerosis in midlife. It is encouraging to note that the healthy behavior changes that participants made naturally and that were associated with reduced odds of CAC and IMT were much less extreme and demanding than those required by the Ornish program.

Our findings give grounds for optimism to the vast majority of adults who have already acquired behavioral risk factors for chronic disease by young adulthood. More than 25% of our sample spontaneously made healthy lifestyle changes that were associated with lowered odds of subclinical atherosclerosis. The proportion of adults that makes healthy lifestyle changes can very likely be increased by implementing health care reimbursement for behavior change interventions or other health promotion policies. At the same time, our findings also convey a cautionary note to those who reach young adulthood free of lifestyle risk factors. Those individuals were not necessarily free of cardiovascular risk 20 years later, indicating that primordial prevention is not a panacea. Rather, the tendency to adopt unhealthy lifestyles after young adulthood was substantial, greater than the tendency to acquire healthier habits, and associated to the same degree with altered odds of developing CAC and IMT. These observations suggest that health behavior changes in adulthood, whether positive or negative, continue to have important consequences for later health outcomes.

A strength of our study is the relatively young age of our large biracial cohort (N = 3538; mean age 25.1 years at Year 0), and the excellent retention through 20 year follow-up. The finding of a robust, bidirectional, graded effect of healthy lifestyle change across both markers of subclinical disease (CAC and IMT) also is a major strength. The results lend weight to the recommendation that counseling for healthy lifestyle change is clinically indicated both to promote change from unhealthy to healthy lifestyle and to prevent loss of existing healthy habits.⁴⁷

Conclusion

Independent of the healthy lifestyle profile present in young adulthood, making subsequent changes in health behaviors is linked to alterations in the burden of subclinical atherosclerosis in middle age. Adopting a healthier lifestyle during early or midlife may reduce the risk of developing CAD; conversely discontinuing healthy behaviors may increase risk. Promotion of healthy lifestyles is essential, not only primordially at early stages of life, but also among adults to control and potentially even reverse the natural progression of CAD.