Methods

MESA is a prospective cohort study of risk factors for subclinical atherosclerosis conducted at 6 field centers in the United States (Baltimore, Maryland; Chicago, Illinois; Saint Paul, Minnesota; Los Angeles, California; New York, New York; and Forsyth County, North Carolina). At study entry, participants were aged 45 to 84 years and self-reported no history of clinical cardiovascular disease. Recruitment and baseline examination of the original 6,814 MESA participants occurred during July 2000 to August 2002. A subsample of consenting participants with no contraindications underwent a cardiac MRI, with 3,541 of the MRIs including an assessment of the ascending aorta. Four additional examinations have been completed over the follow-up (most recently in 2010 to 2012). The study was approved by the institutional review boards of all participating institutions, and all participants provided written informed consent.

AF was ascertained through study electrocardiography, hospital discharge codes, and for participants aged ≥65 years enrolled in fee-for-service Medicare (55% of the cohort), from Medicare claims data obtained from the Centers for Medicare & Medicaid Services. Annual follow-up telephone calls to study participants through February 2012 were used to identify hospitalizations, and Medicare claims data were used to ascertain inpatient AF events through December 31, 2009. Discharges showing the International Classification of Diseases , ninth revision (ICD-9), codes 427.31 or 427.32 were classified as AF events. The date of AF incidence was defined as the date of the first record showing a diagnosis of AF. A review of 16 validation studies determined that the use of the ICD-9 codes to identify AF events has relatively good performance. At baseline, participants who self-reported AF or who had AF in the baseline electrocardiography or in a Medicare claim before study enrollment were excluded.

Measurements from physical examination and questionnaires were made at MESA baseline. Seated systolic and diastolic BPs were defined as the average of the last 2 of 3 BP measurements taken after a 5-minute seated rest using an automated oscillometric sphygmomanometer (Dinamap Pro 100; Critikon, Tampa, Florida). MAP is defined as the sum of diastolic BP and [(1/3) × systolic BP]. PP is defined as the difference between systolic and diastolic BPs. Aortic distensibility was evaluated using 1.5-T whole-body MRI systems, Signa CV/I or Signa LX (General Electric Medical Systems, Waukesha, Wisconsin), as previously described. Covariate variables were measured using standard protocols, as described in Supplementary Methods .

Among the original 6,814 MESA participants, for our primary analysis, we made the following exclusions: those who were ineligible (n = 5), those with prevalent AF (n = 58), those with no follow-up information (n = 20), and those missing information on covariates (n = 101). The subgroup analysis additionally excluded MESA participants who did not receive an aortic MRI at baseline or who had invalid MRI parameters (n = 3,189).

Using restricted cubic splines, we determined that the shape of associations of BP measurements (systolic BP, diastolic BP, MAP, and PP) and aortic distensibility with incident AF was approximately linear. Thus, we decided to model each measure as a continuous variable divided by its respective SD. Using Cox proportional hazards regression, we calculated adjusted hazard ratios and 95% confidence intervals for the associations of the BP measurements and aortic distensibility with incident AF. Follow-up time was defined as time in days between the baseline visit and incidence of AF, death, or last follow-up contact, whichever occurred first. Initial models were adjusted for age, gender, race/ethnicity, and MESA field center site, and fully adjusted models were additionally adjusted for education level, height, body mass index, smoking status, antihypertensive medication use, diabetes, electrocardiography-based left ventricular hypertrophy, PR interval (<120, 120 to 199, or >199 ms), and heart rate at rest. Other than gender, site, education level, and heart rate, the other variables used for model adjustment were chosen in accordance with the CHARGE-AF risk score augmented model. In addition, we also ran analyses adjusting for MRI-based left ventricular mass in place of electrocardiography-based left ventricular hypertrophy (MRI-based left ventricular mass was used instead of electrocardiography-based left ventricular hypertrophy in all subgroup analyses). In further analyses, we also adjusted for incident heart failure and myocardial infarction as time-dependent covariates, in which events diagnosed on the same date as AF diagnosis were considered interim events. Finally, we ran a sensitivity analysis adjusting for different types of antihypertensive medications (angiotensin-converting enzyme inhibitors or angiotensin receptor blockers, β blockers, diuretics, and other types). We evaluated effect modification by age, gender, and race/ethnicity by adding multiplicative terms into regression models. The proportional hazards assumption was tested by including time-covariate interactions for the BP measurements and aortic distensibility, and no violation of the proportional hazards assumption was detected. All statistical analysis was performed using SAS, version 9.3 (SAS Institute, Cary, North Carolina).

Results

At baseline, there were 6,630 participants (3,130 men and 3,500 women) free of AF who met the inclusion criteria for the primary analysis (n = 4,885 available for left ventricular mass–adjusted models). PP was strongly correlated with systolic BP (r = 0.88, p <0.001), weakly correlated with diastolic BP (r = 0.17, p <0.001), and moderately correlated with MAP (r = 0.59, p <0.001). Table 1 presents the baseline characteristics of MESA participants included in the primary analysis by incident AF status. Those who experienced incident AF during follow-up were older, taller, and more likely to be men, ever smokers, and antihypertensive medication users. They also had higher levels of BP and left ventricular mass, lower levels of aortic distensibility, and longer PR intervals. There were 3,441 participants (1,571 men and 1,870 women) meeting the inclusion criteria for the aortic distensibility subgroup analysis, and their overall characteristics were similar to those included in the primary analysis (see Supplementary Table 1 ).

There were 307 newly diagnosed cases of AF (4.6%) over a mean follow-up period of 7.8 ± 1.7 years (incidence 5.9 cases per 1,000 person-years). Table 2 presents the association between BP measurements, modeled continuously per 1-SD increase, and incident AF. After adjusting for age, gender, race/ethnicity, and site (model 1), increases in systolic BP, MAP, and PP, but not diastolic BP, were significantly associated with greater risk for AF. Additional adjustment for education, heart rate at rest, and the remaining CHARGE-AF risk factors (model 2) resulted in slightly attenuated hazard ratios, with only systolic BP and PP remaining significant. Models further adjusted for MRI-based left ventricular mass in place of electrocardiography-based left ventricular hypertrophy (model 3) and interim heart failure and myocardial infarction events (model 4) further attenuated the association, with none of the individual BP measurements remaining significantly associated with AF risk. Sensitivity analyses further adjusting model 3 for specific antihypertensive medication types also slightly attenuated the association (see Supplementary Table 2 ). Diastolic BP showed an inverse association with incident AF in all models when modeled simultaneously with systolic BP. When modeled simultaneously with MAP, PP remained significantly positively associated with AF in all models ( Table 2 and Figure 1 ). In model 3, systolic BP and MAP were marginally negatively associated with incident AF when modeled simultaneously with PP. In all models, PP was more strongly associated with AF risk after adjustment for MAP or systolic BP compared with models including PP alone. We found no significant interactions between BP measurements and age, gender, or race/ethnicity (results not shown).

Stay updated, free articles. Join our Telegram channel

Join