The Heart and Soul Study is an observational study designed to investigate the influence of psychosocial factors on the progression of CHD. Methods have been described previously.5
Briefly, participants were recruited from outpatient clinics in the San Francisco Bay Area if they met one of the following inclusion criteria: history of myocardial infarction or coronary revascularization, angiographic evidence of > 50% stenosis in ≥ 1 coronary vessels, exercise-induced ischemia by treadmill or nuclear testing, or documented diagnosis of CHD. Participants were excluded if they were unable to walk 1 block, experienced myocardial infarction in the past 6 months, or were likely to move out of the area within 3 years. The study protocol was approved by the Institutional Review Boards of participating institutions, and all participants provided written informed consent. Between September 2000 and December 2002, 1024 participants enrolled and underwent a baseline study appointment that included a medical history, physical examination, and health status questionnaire. Outpatient 24-hour timed urine collections and fasting (12-hour) morning venous blood samples were obtained. Participants were followed through June 30, 2009. For the present analysis, we excluded participants with missing urine collections (N=57) or missing covariate data (N=81), providing a final analytic sample of 886 (87%) participants.
The protocol for timed urine collection has been described previously.6
In brief, participants received instructions on urine collection and specimen refrigeration. They were asked to void at the end of their study appointment and to begin the collection from that point forward. Research personnel arrived at the patient’s home 24 hours after the timed collection was initiated. If patients reported missing any urine or collections were < 1 or > 3 liters, collections were repeated. If participants were unable to collect all urine, no data were recorded. Urine volume was recorded (mL), and creatinine was measured by the rate Jaffe method. Urinary creatinine excretion was calculated in milligrams per day (urine volume [mL] times urine creatinine [mg/dL] divided by 100).
Between the baseline examination and May 1, 2009 annual telephone interviews were conducted with study participants (or their proxy) for vital status. For any reported event, medical records, death certificates, and coroner’s reports were retrieved. Date of death was recorded to provide time-to-event data from the baseline examination.
Patient demographics and co-morbid diseases were determined by questionnaire. Systolic and diastolic blood pressures were measured after 5 minutes of rest in subjects in the supine position by trained research personnel. Participants were instructed to take their blood pressure medications on the morning of the intake appointment and not to smoke or consume caffeine 5 hours before the visit. Weight and height were measured in participants wearing light clothing and no shoes. BMI was calculated as weight in kilograms divided by height in meters squared.
Serum cystatin C concentrations were measured with a particle-enhanced immunonephelometric assay7
(N Latex Cystatin-C, Dade Behring, Inc, Deerfield, Ill) and used to calculate estimated GFR (eGFRcys) with the following formula: eGFRcys=76.7× cystatin C−1.19
. This formula, which has been validated with comparison with iothalamate-measured GFR in a pooled cohort of kidney disease studies, showed little bias and provided a non–creatinine based method to adjust for kidney function for this study.8
Total cholesterol and high-density lipoprotein (HDL) cholesterol were measured with standard clinical chemistry analyzers. High sensitivity C-reactive protein (CRP) was measured with the Roche (Indianapolis, Ind) and the Beckman Extended Range (Galway,Ireland) assays.9
Fasting glucose was measured by a standard clinical analyzer, and fasting insulin was measured by ELISA (Linco Research, St Charles, Mo). Participants provided rest transthoracic echocardiograms that were read by a single expert cardiologist blinded to all clinical data as described previously.10
Left ventricular mass was calculated with the truncated ellipsoid method11
and indexed to body surface area. Left ventricular ejection fraction was determined by biplane method of disks.12
Exercise treadmill test by the modified Bruce protocol was performed and the maximum number of metabolic equivalents (METs) and heart rate achieved were recorded.
We began by exploring the association of sex-specific BMI quintiles in men (24.0 kg/m2
, 26.3 kg/m2
, 28.7 kg/m2
, 31.6 kg/m2
) and women (24.6 kg/m2
, 27.4 kg/m2
, 30.0 kg/m2
, 34.6 kg/m2
) with mortality. We verified the linearity of the association between BMI and mortality to maximize statistical power. The distribution of demographic variables and standard CHD risk factors were compared across BMI quintiles using ANOVA tests for continuous variables and the chi square tests for categorical variables, as appropriate. We performed Cox proportional hazards regression to examine the association between BMI and all-cause mortality, using the lowest sex-specific BMI quintile as the referent category. Patient demographics, co-morbid diseases, physical measurements and laboratory values were added to the Cox proportional hazards regression models. Prior analyses have demonstrated a linear relationship of urine creatinine excretion rate with mortality in this cohort.1
Urinary creatinine excretion rate was included in the final model as a continuous covariate to evaluate whether it attenuated the association of BMI with all-cause mortality.
We performed a sensitivity analyses to investigate whether over- or under- urine collections may have introduced bias. We excluded participants whose measured 24-hour urinary creatinine clearance was > 30% different from their eGFRcys. This takes advantage of the fact that urinary creatinine excretion rate divided by serum creatinine equals creatinine clearance (in mL/min), an estimate of GFR. eGFRcys provided another estimate of GFR that was independent of the quality of timed urine collections and of creatinine kinetics. Because the eGFRcys estimate is given as GFR normalized to body surface area, to compare the two renal function estimates, measured creatinine clearance was divided by body surface area and multiplied by 1.73.
We also explored whether measures of subclinical cardiovascular disease might attenuate the associations of BMI with mortality in our population. We stratified the outcome of mortality into early (< 3 years from enrollment) versus late mortality (≥ 3 years from enrollment). This cut-point was chosen as the half-way point of the median follow-up in our study (6 years). Additional analyses evaluated whether adjustment for self-reported overall health (on a 5-point Likert scale), total METS achieved on the modified Bruce protocol treadmill test, and left ventricular mass index, attenuated the association of BMI with mortality in Cox models.
All analyses were conducted using SAS (version 9.2, SAS Institute Inc., Cary, NC, USA). P-values < 0.05 were considered statistically significant.