Covariates included sociodemographic variables (age, sex, race, and study site), clinical conditions (presence of coronary heart disease, diabetes, hypertension, osteoarthritis, peripheral artery disease, cerebrovascular disease, depression, and pulmonary disease as defined by algorithms identified by the clinical investigators in the study18), and physical and biological parameters, including smoking status, body mass index (BMI, calculated as weight in kilograms divided by the square of height in meters), physical activity (calculated using the Harvard Alumni study19- 22 algorithm based on walking and exercise expenditure in kilocalories per week), and measurements for total cholesterol, creatinine, albumin, fasting glucose, and glycated hemoglobin (HbA1c). Total cholesterol, creatinine, and albumin values were determined by a colorimetric technique on a Vitros 950 analyzer (Johnson & Johnson, New Brunswick, NJ). Levels of LDL-C were calculated by the equation described by Friedewald et al.23 Fasting glucose levels were measured using an automated glucose oxidase reaction (YSI 2300 Glucose Analyzer; YSI, Yellow Springs, Ohio). Levels of HbA1c were measured by a fully automated analyzer (Variant; Bio-Rad Laboratories Inc, Hercules, Calif) using the principle of ion exchange high-performance liquid chromatography. Medications taken in the past 2 weeks were brought in, recorded, and coded according to the Iowa Drug Information System. Using this system, participants using nonsteroidal anti-inflammatory drugs, angiotensin-converting enzyme inhibitors, and statins were identified.