|Home | About | Journals | Submit | Contact Us | Français|
While trials of lifestyle interventions generally focus on cardiovascular disease risk factors rather than hard clinical outcomes, 10-year coronary heart disease (CHD) risk can be estimated from the Framingham risk equations. Our objectives were to study the effect of two multi-component lifestyle interventions on estimated CHD risk relative to advice alone and to evaluate if there are differences in the effects of the lifestyle interventions among subgroups defined by baseline variables.
810 healthy adults with untreated pre- or stage I hypertension were randomized to one of 3 intervention groups: an “advice only” group, an “established” group (EST) which used established lifestyle recommendations for blood pressure control (sodium reduction, weight loss, increased physical activity), or an “established plus DASH” group (EST+DASH) that combined established lifestyle recommendations with the DASH diet. The primary outcome was 10-year CHD risk estimated from follow-up data collected at 6 months. A secondary outcome was 10-year CHD risk at 18 months. Of the 810 participants, 62% were female, and 34% were black. Mean age was 50 years, mean systolic/diastolic blood pressure was 135/85 mmHg, and median baseline Framingham risk was 1.9%. The relative risk ratio comparing 6-month to baseline Framingham risk was 0.86 (95% CI: 0.81, 0.91, P <0.001) in EST and 0.88 (95% CI: 0.83, 0.94, P <0.001) in EST+DASH relative to advice alone. Results were virtually identical in sensitivity analyses, in each major subgroup, and at 18 months.
The observed reductions of 12-14% in estimated CHD risk are substantial and, if achieved, should have important public health benefits.
Adverse lifestyles have a major impact on the burden of coronary heart disease (CHD). Physical inactivity,1 smoking,2 obesity,3 and several aspects of diet are the underlying causative factors for CHD and its risk factors. Blood pressure (BP), a major risk factor for CHD,4 is strongly associated with obesity,5 physical inactivity,6 sodium intake,7 and alcohol intake.8 Obesity,9 physical inactivity,10 and smoking11 are the major modifiable risk factors for diabetes mellitus, another important cause of CHD.12 Serum cholesterol, an established risk factor for CHD,13 is also influenced by diet and physical activity.14
Lifestyle modification is a critical component of population-based strategies to prevent CHD. Because of logistic considerations, trials of lifestyle modification rarely have sufficient power to detect intervention effects on clinical outcomes such as CHD events. Instead, the outcome variables of such trials are often CHD risk factors. Trials of single15, 16 and multiple lifestyle interventions17 reveal that lifestyle change can have substantial effects on CHD risk factors such as BP.
CHD risk prediction equations, such as the Framingham Heart Study equations,14, 18 provide an opportunity to estimate the effect of lifestyle modification on CHD risk. The current Framingham risk equation estimates 10-year CHD risk from a series of non-modifiable (sex and age) and modifiable (BP, serum lipids, and smoking) risk factors.14
We conducted an analysis of data from the PREMIER trial which tested the effects of two multi-component lifestyle interventions on BP relative to a control group.17 We hypothesize that relative to the control group, both lifestyle interventions would reduce the estimated 10-year CHD risk.
PREMIER was a National Heart, Lung, and Blood Institute-sponsored, multi-center, 3 group, parallel-arm randomized trial conducted in the United States.17 The methods and main results have been published.17, 19 Participants provided written, informed consent while institutional review boards at each center reviewed and approved the protocol.
Participants were 810 adults with pre- or stage 1 hypertension, who met Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure (JNC) criteria for a 6-month trial of non-pharmacologic BP treatment.20 Other inclusion criteria were age ≥ 25 years; a body mass index of 18.5 to 45.0 kg/m2; systolic BP of 120-159 mmHg; and diastolic BP of 80-90 mmHg.17, 19 Main exclusion criteria were use of anti-hypertensive medications or weight loss drugs; diabetes (glucose ≥ 126 mg/dL); BP-related target organ damage (cardiovascular event, congestive heart failure, or angina); cancer or treatment for cancer in the last 2 years; consumption of > 21 alcoholic drinks/week; and pregnancy or lactation.17, 19
Randomization was stratified by site and hypertension status using a block size of 24 to promote balance of treatment assignment.17, 19 Intervention staff were masked to follow-up data, and staff collecting follow-up data were masked to intervention assignment.19 Participants received their BP measurement results at baseline and 6 months.19
Over 6 months, the “advice only” control group received printed educational information and brief advice on lifestyle modifications at one individual, 30-minute session at randomization.21 Participants in EST and EST+DASH received an intensive behavioral intervention and had in-person contacts with an interventionist at four individual and 14 group sessions.21 The EST component consisted of individualized advice regarding established recommendations22 including physical activity (≥ 180 minutes/week), weight loss, and caloric, alcohol, sodium (≤ 2400 mg/day), and total (≤ 30% of calories) and saturated (≤ 10% of calories) fat intake.21 In addition to the EST recommendations, the EST+DASH group also received advice regarding the DASH diet which focused on the intake of less total (≤ 25% of calories) and saturated (≤ 7% of calories) fat and the consumption of 9-12 servings/day of fruits, and vegetables and 2-3 servings/day of low-fat dairy products.16, 21
The effects of the interventions on the primary outcome variable of the trial, i.e. systolic BP change from randomization to 6 months, have been published.17 Relative to the “advice only” group, mean systolic BP was 3.7 mmHg (P<0.001) lower in the EST group and 4.3 mmHg (P<0.001) lower in the EST+DASH group.17 There was no statistically significant systolic BP difference between EST and EST+DASH.17
BP and weight were measured in a standardized fashion by trained, certified observers using a random zero sphygmomanometer.19 From fasting blood specimens, glucose, total cholesterol and high density lipoprotein (HDL) cholesterol were directly measured.19 Smoking status, medication use, and demographic variables were obtained by questionnaire.19 We defined diabetes as a fasting glucose of ≥ 140 mg/dl because the Framingham risk equations were developed using this cutpoint.18
Our primary outcome was the change in estimated 10-year CHD risk at 6 months compared to baseline using the sex-specific Framingham risk equations.14 The current equation estimates CHD risk from age, total cholesterol, HDL cholesterol, and BP (each entered as continuous variables) and from anti-hypertensive medication use and smoking status (each entered as binary variables).14
Consistent with the main results of the PREMIER trial, we chose the 6 month follow-up visit as the time of primary outcome assessment because clinical guidelines20 recommended consideration of pharmacologic therapy for those with a BP of ≥ 140/90 mmHg after 6 months of lifestyle therapy.19 Adherence to lifestyle interventions is also optimized at 6 months.19
We analyzed our data using STATA version 9.2 (STATACorp, College Station, Texas). The distributions of baseline characteristics of the PREMIER participants were examined for each intervention group. Means and standard deviations (SD) were calculated for continuous variables, and proportions were calculated for categorical variables. Median Framingham risk scores at baseline and 6 months were calculated for each intervention group. In regression analysis, the difference between the logarithm of estimated 10-year CHD risk at the 6-month visit and the logarithm of 10-year CHD risk at the baseline visit, as the response variable, was regressed on indicators of the two behavioral interventions and on indicators of the clinical sites to evaluate the effect of the intervention groups while adjusting for site. For ease of interpretation, we use the ratio of 6 month to baseline 10-year CHD risk comparing the intervention groups (EST and EST+DASH) to the reference group, “advice only,” termed the “relative risk ratio.”
We used multiple imputation by chained equations (ice) and generated 10 imputations to replace missing values for systolic and diastolic BP at 6 months, total and HDL cholesterol at baseline and 6 months, and anti-hypertensive medication use at 6 months using the “ice” command in STATA.23 First, we evaluated the assumption of missingness at random by attempting to predict missingness with known covariates24 using multiple logistic regression and calculating the area under the ROC curve. We then developed regression models to predict the variables with missing data.
In exploratory analyses, we evaluated for interactions between the interventions and subgroups defined by baseline factors (sex, baseline 10-year CHD risk (< 10% or ≥ 10%), age (<60 years or ≥ 60 years), hypertension status (hypertension or pre-hypertension), and total cholesterol (<5.2 mmol/L or ≥ 5.2 mmol/L). We allowed different intervention effects among subgroups by using models which included a main effect for the subgroup and interactions between the subgroup and the interventions with adjustment for site.
In sensitivity analyses, we calculated the relative risk ratios for those with a complete set of covariates using the current14 and older18 versions of the Framingham risk equations. Compared to the older equations,18 the updated equations include anti-hypertensive medication use, exclude diabetes, and use continuous rather than categorical measures for BP and lipids.14 We also carried out an analysis for participants with complete data at baseline and 18 months.
A P value < 0.05 was considered statistically significant. No adjustment for multiple comparisons was made.
The authors had full access to and take full responsibility for the integrity of the data. All authors have read and agree to the manuscript as written.
Participants (N=810) were randomized into three intervention groups, and 704 had a complete set of covariates at baseline and 6 months for Framingham risk score calculation (Figure 1). Three participants were missing total and HDL cholesterol at baseline, while 77 were missing these at 6 months, and 74 participants were missing BP at 6 months. Forty-eight of those participants with missing data were missing both BP and cholesterol measurements at 6 months. Thus, we imputed 10-year CHD risk for a total of 106 (13.1%) participants. Table 1 displays baseline data of the 810 participants. Median baseline 10-year CHD risk was slightly higher in EST+DASH (2.30%) than in “advice only” (1.77%) and EST (1.92%). Among the 704 participants with complete covariates, there were slightly more smokers in the EST group, and systolic BP was slightly higher in EST and EST+DASH groups compared to “advice only.”
At 6 months, median 10-year CHD risk decreased in each of the randomized groups, and the median 10-year CHD risk was higher in the EST+DASH group than in the EST and “advice only” groups (Figure 2). BP, total cholesterol, and HDL cholesterol decreased in all groups. The decrease in 10-year CHD risk over 6 months was significantly greater in EST and EST+DASH relative to “advice only,” and there was no difference in the change in 10-year CHD risk between EST and EST+DASH (Table 2). The relative risk ratios comparing EST and EST+DASH to “advice only” were 0.86 (95% confidence interval: 0.81, 0.91, P value <0.001) and 0.88 (95% confidence interval: 0.83, 0.94, P value <0.001), respectively. Risk reductions were similar in subgroups, and interaction tests were non-significant (Table 2). Results were similar with the inclusion of the 704 participants with complete data using both the current14 and older Framingham risk equations.18
At 18 months, 654 participants had complete data for calculation of the 10-year CHD risk. Among these participants, baseline systolic BP was slightly higher in EST and EST+DASH compared to the “advice only” group. Comparing 18-month to baseline 10-year CHD risk and adjusting for site, the relative risk ratios were 0.92 (95% CI: 0.86, 0.98, P <0.001) and 0.92 (95% CI: 0.87, 0.98, P <0.001) for EST and EST+DASH, respectively, relative to “advice only.”
In individuals with pre- or stage 1 hypertension, two multi-component behavioral interventions, EST+DASH and EST, significantly reduced the estimated 10-year CHD risk by 12% and 14%, respectively, compared to “advice only.” Results were similar across subgroups defined by baseline variables, and improvements in 10-year CHD risk were maintained at 18 months.
The two behavioral interventions (EST and EST+DASH) had similar effects on CHD risk. One possibility is that participants received an inadequate dose of the DASH diet, as evidenced by lack of full adherence to DASH recommendations for fruit and vegetable intake.17 An alternative explanation is subadditivity;25 specifically, the combined effects of two interventions when implemented together are less than the sum of the two when implemented separately. Subadditivity can occur when there is reduced adherence in the combined intervention or when the interventions work through similar mechanisms to achieve improvements in CHD risk factors.
Despite the intuitive appeal and public health relevance of estimated CHD risk, few studies of lifestyle interventions have used change in CHD risk as an outcome variable. In a recent randomized trial of 315 participants in Canada with 10-year CHD risk of ≥ 10%, a lifestyle intervention (health report card and telephone counseling on smoking, exercise, nutrition and stress) reduced 10-year CHD risk at 1 year of follow-up by approximately 2% compared to a usual care group.26 A smaller randomized trial (N=75) reported a similar but non-significant decrease in 10-year CHD risk at 16 weeks for a nutrition program combined with exercise relative to the nutrition program alone, but loss to follow-up was high (36%).27 On pre-post analysis, one observational study with a median follow-up time of 8 months found a non-significant increase in estimated CHD risk with dietary advice given to patients without CHD in an urban clinic in the United Kingdom.28 Another uncontrolled, longitudinal study of multiple lifestyle changes (diet, stress management, and aerobic exercise) in participants with CHD risk factors from the Windber Coronary Artery Disease Reversal (CADRe) program reported a non-significant 6.8% decrease in estimated CHD risk over 1 year for participants at-risk for cardiovascular disease.29
A major strength of our study is its large and diverse study population. While the study was not powered to examine the subgroups, results were consistent across subgroups suggesting broad applicability of trial interventions. Second, the trial has high internal validity as evidenced by high rates of data collection during follow-up. In our study, only 13% of participants had missing data at 6 months, and missing data were imputed using multiple imputation. Third, data collectors were trained, and BP and serum cholesterol were measured directly in a standardized fashion.19 Lastly, the Framingham risk functions have been validated in whites and blacks in the United States.30
Our study also has limitations. Our results may underestimate the magnitude of the effect of the EST and EST+DASH interventions on CHD risk because there was incomplete adherence17 and because the “advice only” group made lifestyle changes perhaps because of high motivation17, 31 or a Hawthorne effect related to data collection visits.32 Second, smoking status was not available at 6 months, but the use of baseline smoking status carried forward is reasonable because the interventions did not include advice on smoking cessation. Third, the current Framingham risk equations14 do not include diabetes, but we obtained similar results when using the older equations which include diabetes status.18 Finally, the Framingham risk equations may overestimate absolute risk in some populations;33 for this reason, we emphasize relative risk reductions. Of note, the baseline CHD risk was highest in the “EST+DASH” group, and the 6-month CHD risk was highest in the EST+DASH group with the “advice only” and EST groups having similar CHD risks at 6 months. It is possible that the smaller decrease in 10-year CHD risk in the “advice only” group relative to the EST and EST+DASH groups may have occurred because of a “floor” below which the risk could not decrease. However, there are populations in which actual CHD risk is extremely low.34, 35 Also, 10-year CHD risk estimated by the Framingham equations can be <1% using biologically-plausible values of the variables included in the equations.
Future research should focus on understanding the individual components of the behavioral interventions that are most effective in decreasing CHD risk. For example, an analysis of the effect of individual PREMIER lifestyle changes on BP at 6 months showed that decreased urinary sodium, improved fitness, and low total fat intake were associated with a decrease in systolic BP before controlling for weight loss; in that study, it was concluded that error in measurement of dietary and urinary sodium might account for the loss of their statistical significance after the inclusion of weight in regression models.36 Also, a comparison of CHD incidence with the change in 10-year CHD risk would be useful to validate the change in 10-year CHD risk as a surrogate outcome.
In summary, in the PREMIER trial, two multi-component behavioral interventions incorporating diet and physical activity recommendations significantly lowered estimated 10-year CHD risk by 12-14% relative to a control condition. These estimated reductions in CHD risk are substantial and support research and translational efforts to implement counseling on lifestyle change as part of routine medical care. Given that heart disease remains the leading cause of death in the United States,37 translation of these findings into clinical practice should have a substantial public health impact.
FUNDING SOURCES There was no project-specific funding for this study. Dr. Maruthur was supported by a training grant (5 T32 HL007024-31) from the National Heart, Lung, and Blood Institute (NHLBI), NIH. Dr. Wang’s effort was supported by a grant UL1 RR 025005 from the National Center for Research Resources (NCRR), a component of the NIH and the NIH Roadmap for Medical Research. The PREMIER trial was sponsored by NHLBI, NIH grants UO1 HL60570, UO1 HL60571, UO1 HL 60573, UO1 HL60574, UO1 HL62828, and MO1 RR00052. The contents of the manuscript are solely the responsibility of the authors and do not necessarily represent the official view of NCRR or NIH.
Publisher's Disclaimer: This is an un-copyedited author manuscript that was accepted for publication in Circulation, copyright The American Heart Association. This may not be duplicated or reproduced, other than for personal use or within the “Fair Use of Copyrighted Materials” (section 107, title 17, U.S. Code) without prior permission of the copyright owner, The American Heart Association. The final copyedited article, which is the version of record, can be found at http://www.ahajournals.org. The American Heart Association disclaims any responsibility or liability for errors or omissions in this version of the manuscript or in any version derived from it by the National Institutes of Health or other parties.
Clinical Trial Registration Information: Not applicable as trial began enrollment before 2005