Lifestyle, Coronary CA, Cardiovascular Events, and Mortality
Lifestyle, Coronary CA, Cardiovascular Events, and Mortality
The Multi-Ethnic Study of Atherosclerosis (MESA) is a prospective study of the risk factors for and prevalence and progression of cardiovascular disease. The MESA design methods have been published previously. Briefly, 6,814 participants who were 44–84 years of age and self-identified as white, African American, Hispanic, or Chinese were recruited at 6 academic centers (Columbia University, New York, New York; Johns Hopkins University, Baltimore, Maryland; Northwestern University, Chicago, Illinois; University of California-Los Angeles, Los Angeles, California; University of Minnesota, Minneapolis, Minnesota; and Wake Forest University, Winston-Salem, North Carolina) from 2000 to 2002. All participants were free from clinical cardiovascular disease at the time of enrollment. The study protocol was approved by the institutional review board at each site. All participants gave informed consent before enrollment. Anthropometric measurements, lifestyle behaviors, medical history, and laboratory data were all assessed, as described previously.
CAC was measured using electron beam or multidetector computed tomography. All participants were scanned twice consecutively at baseline, and scans were read by an expert physician-reader at Harbor-UCLA Medical Center, California. Results from each participant's 2 scans were averaged to provide a more accurate point estimate of his or her Agatston calcium score. More information on the methodology of CAC measurement and interpretation is available in the Web Appendix (available at http://aje.oxfordjournals.org/) and prior reports.
To quantify CAC progression, we randomly selected 2,954 participants to have a follow-up computed tomography examination between September 2002 and February 2004. There were 2,805 women selected between March 2004 and October 2005 and 1,406 selected between October 2005 and February 2008. The mean time between the baseline scan and the last follow-up scan was 3.1 (standard deviation (SD), 1.3) years.
CHD events consisted of nonfatal myocardial infarction, resuscitated cardiac arrest, angina, coronary revascularization, and death due to CHD. Event criteria are available in the Web Appendix and online (http://www.mesa-nhlbi.org). Participants were contacted at intervals of 9–12 months, and information was collected about interim hospitalizations, outpatient CHD diagnoses, and deaths. Death certificates and medical records were requested. Next-of-kin interviews were conducted for participants with out-of-hospital cardiovascular deaths. Medical records were obtained for 98% of participants with hospitalized CHD events and 95% of participants with outpatient cardiovascular diagnoses. Follow-up telephone interviews were completed for 92% of living participants. The last follow-up was in June of 2011.
We selected 4 lifestyle-related variables for the scoring system: diet, BMI, smoking status, and physical activity level. For each variable, a binary score was defined (Mediterranean-style diet vs. unhealthy diet, optimal BMI vs. suboptimal BMI, never smoker vs. ever smoker, and regular physical activity vs. sedentary lifestyle), with a score of 1 awarded for each healthy behavior.
Dietary health was categorized based on a previously published scale quantifying adherence to a Mediterranean diet. The scale has been associated with mortality in over 20,000 adults and is easily reproducible. Per this method, participants were awarded points for consuming more healthy foods (vegetables, legumes, fruits, nuts, cereal/grains, and fish) and fewer detrimental foods (full-fat dairy, meat, poultry, and saturated fat) than the median intake and for having an optimal alcohol intake. This amounted to a possible diet score of 11 points. Participants with a diet score above the median received 1 point toward their comprehensive lifestyle score. (For complete dietary methodology, see the Web Appendix.)
BMI was categorized as optimal (≥18.5 and ≤24.9) or suboptimal (≥25 or ≤18.4) based on current World Health Organization and National Institutes of Health guidelines. Participants with an optimal BMI received 1 point toward their lifestyle score. Smoking status was categorized as never smoker or previous/current smoker, with 1 point awarded for never smoking.
Physical activity data were obtained from the MESA Typical Week Physical Activity Survey. Participants who averaged more than 150 minutes/week of moderate-intensity physical activity or more than 75 minutes/week of vigorous-intensity activity were considered active based on current AHA guidelines and were awarded 1 point. The complete survey methodology is available in the Web Appendix and online. The addition of points for diet, physical activity, smoking status, and BMI amounted to a comprehensive lifestyle score ranging from 0 (least healthy) to 4 (healthiest) for each participant.
The variance of baseline descriptive variables was analyzed across lifestyle scores. Given the non-normality of the data, variance was measured using the Kruskal-Wallis equality-of-populations rank test.
CAC incidence was calculated for baseline zero CAC scores, and CAC progression was calculated for baseline nonzero scores. Annual CAC progression was calculated using the following equation: (mean phantom-adjusted CACfollow-up − mean phantom-adjusted CACbaseline)/(years between scans). Participants were excluded if coronary stents, which make scans uninterpretable, were placed during follow-up. For participants with multiple follow-up scans (n = 1,118), the final scan was used for the progression analysis.
The odds ratio for CAC incidence by lifestyle score was calculated using a generalized linear model with the logit link function. Annual CAC progression was calculated using median regression analysis because of the non-normality of CAC distribution. Hierarchical multivariable models were used: model 1 was unadjusted; model 2 was adjusted for age, sex, race, study site, and income; and model 3 was additionally adjusted for intermediate variables on the atherosclerosis pathway, including hypertension, use of medications to control hypertension, elevated fasting plasma glucose, use of medications for diabetes, elevated high-density lipoprotein cholesterol (HDL-C), elevated non–HDL-C, elevated triglycerides, use of lipid-lowering medications, and C-reactive protein level.
Time to CHD event or all-cause death was plotted using Kaplan-Meier statistics. Cox proportional hazards models were used to calculate the hazard ratio for suffering a CHD event or death based on lifestyle score in separate analyses. Models 1–3 were performed as above. An additional model (model 4) was then included to adjust for baseline CAC and determine whether lifestyle score added predicted value to baseline atherosclerosis. Hazard ratios and 95% confidence intervals were plotted by lifestyle score for each model. All statistical analyses were performed using STATA, version 12 (StataCorp LP, College Station, Texas). This data analysis plan was peer reviewed and approved by the MESA Publications and Presentations committee. A complete copy of the a priori statistical analysis plan is available at www.mesa-nhlbi.org.
Materials and Methods
The Multi-Ethnic Study of Atherosclerosis (MESA) is a prospective study of the risk factors for and prevalence and progression of cardiovascular disease. The MESA design methods have been published previously. Briefly, 6,814 participants who were 44–84 years of age and self-identified as white, African American, Hispanic, or Chinese were recruited at 6 academic centers (Columbia University, New York, New York; Johns Hopkins University, Baltimore, Maryland; Northwestern University, Chicago, Illinois; University of California-Los Angeles, Los Angeles, California; University of Minnesota, Minneapolis, Minnesota; and Wake Forest University, Winston-Salem, North Carolina) from 2000 to 2002. All participants were free from clinical cardiovascular disease at the time of enrollment. The study protocol was approved by the institutional review board at each site. All participants gave informed consent before enrollment. Anthropometric measurements, lifestyle behaviors, medical history, and laboratory data were all assessed, as described previously.
Coronary Artery Calcium
CAC was measured using electron beam or multidetector computed tomography. All participants were scanned twice consecutively at baseline, and scans were read by an expert physician-reader at Harbor-UCLA Medical Center, California. Results from each participant's 2 scans were averaged to provide a more accurate point estimate of his or her Agatston calcium score. More information on the methodology of CAC measurement and interpretation is available in the Web Appendix (available at http://aje.oxfordjournals.org/) and prior reports.
To quantify CAC progression, we randomly selected 2,954 participants to have a follow-up computed tomography examination between September 2002 and February 2004. There were 2,805 women selected between March 2004 and October 2005 and 1,406 selected between October 2005 and February 2008. The mean time between the baseline scan and the last follow-up scan was 3.1 (standard deviation (SD), 1.3) years.
CHD Events
CHD events consisted of nonfatal myocardial infarction, resuscitated cardiac arrest, angina, coronary revascularization, and death due to CHD. Event criteria are available in the Web Appendix and online (http://www.mesa-nhlbi.org). Participants were contacted at intervals of 9–12 months, and information was collected about interim hospitalizations, outpatient CHD diagnoses, and deaths. Death certificates and medical records were requested. Next-of-kin interviews were conducted for participants with out-of-hospital cardiovascular deaths. Medical records were obtained for 98% of participants with hospitalized CHD events and 95% of participants with outpatient cardiovascular diagnoses. Follow-up telephone interviews were completed for 92% of living participants. The last follow-up was in June of 2011.
Lifestyle Score
We selected 4 lifestyle-related variables for the scoring system: diet, BMI, smoking status, and physical activity level. For each variable, a binary score was defined (Mediterranean-style diet vs. unhealthy diet, optimal BMI vs. suboptimal BMI, never smoker vs. ever smoker, and regular physical activity vs. sedentary lifestyle), with a score of 1 awarded for each healthy behavior.
Dietary health was categorized based on a previously published scale quantifying adherence to a Mediterranean diet. The scale has been associated with mortality in over 20,000 adults and is easily reproducible. Per this method, participants were awarded points for consuming more healthy foods (vegetables, legumes, fruits, nuts, cereal/grains, and fish) and fewer detrimental foods (full-fat dairy, meat, poultry, and saturated fat) than the median intake and for having an optimal alcohol intake. This amounted to a possible diet score of 11 points. Participants with a diet score above the median received 1 point toward their comprehensive lifestyle score. (For complete dietary methodology, see the Web Appendix.)
BMI was categorized as optimal (≥18.5 and ≤24.9) or suboptimal (≥25 or ≤18.4) based on current World Health Organization and National Institutes of Health guidelines. Participants with an optimal BMI received 1 point toward their lifestyle score. Smoking status was categorized as never smoker or previous/current smoker, with 1 point awarded for never smoking.
Physical activity data were obtained from the MESA Typical Week Physical Activity Survey. Participants who averaged more than 150 minutes/week of moderate-intensity physical activity or more than 75 minutes/week of vigorous-intensity activity were considered active based on current AHA guidelines and were awarded 1 point. The complete survey methodology is available in the Web Appendix and online. The addition of points for diet, physical activity, smoking status, and BMI amounted to a comprehensive lifestyle score ranging from 0 (least healthy) to 4 (healthiest) for each participant.
Statistical Analysis
The variance of baseline descriptive variables was analyzed across lifestyle scores. Given the non-normality of the data, variance was measured using the Kruskal-Wallis equality-of-populations rank test.
CAC incidence was calculated for baseline zero CAC scores, and CAC progression was calculated for baseline nonzero scores. Annual CAC progression was calculated using the following equation: (mean phantom-adjusted CACfollow-up − mean phantom-adjusted CACbaseline)/(years between scans). Participants were excluded if coronary stents, which make scans uninterpretable, were placed during follow-up. For participants with multiple follow-up scans (n = 1,118), the final scan was used for the progression analysis.
The odds ratio for CAC incidence by lifestyle score was calculated using a generalized linear model with the logit link function. Annual CAC progression was calculated using median regression analysis because of the non-normality of CAC distribution. Hierarchical multivariable models were used: model 1 was unadjusted; model 2 was adjusted for age, sex, race, study site, and income; and model 3 was additionally adjusted for intermediate variables on the atherosclerosis pathway, including hypertension, use of medications to control hypertension, elevated fasting plasma glucose, use of medications for diabetes, elevated high-density lipoprotein cholesterol (HDL-C), elevated non–HDL-C, elevated triglycerides, use of lipid-lowering medications, and C-reactive protein level.
Time to CHD event or all-cause death was plotted using Kaplan-Meier statistics. Cox proportional hazards models were used to calculate the hazard ratio for suffering a CHD event or death based on lifestyle score in separate analyses. Models 1–3 were performed as above. An additional model (model 4) was then included to adjust for baseline CAC and determine whether lifestyle score added predicted value to baseline atherosclerosis. Hazard ratios and 95% confidence intervals were plotted by lifestyle score for each model. All statistical analyses were performed using STATA, version 12 (StataCorp LP, College Station, Texas). This data analysis plan was peer reviewed and approved by the MESA Publications and Presentations committee. A complete copy of the a priori statistical analysis plan is available at www.mesa-nhlbi.org.