Using Area Under the Curve for Longitudinal Modeling of Blood Pressure in the Bogalusa Heart Study

• Activity Number: 21
• Type: Contributed
• Date/Time: Sunday, August 11, 2002 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section*
• Author(s): Nancy Cook*+ and Bernard Rosner
• Affiliation(s): Harvard Medical School and Harvard Medical School
• Keywords: repeated measures ; random effects ; growth curves ; tracking correlations ; longitudinal ; blood pressure

Abstract

The Bogalusa Heart Study included unequally spaced repeated measures of blood pressure (BP) from childhood to young adulthood. We used longitudinal growth curve models to predict adult BP from childhood measures. Using quadratic random effects models, we computed the area under the curve to estimate the average level for each participant within the age periods 5 to 14 and 20 to 34 years. Tracking correlations of these averages over time were computed by race and gender. These were approximately 0.6 for systolic BP, 0.5-0.6 for K4 diastolic BP, and 0.4-0.6 for K5 diastolic BP. We modeled young adult BP as a function of childhood BP and childhood and young adult anthropometric measures. Patterns of BMI, but not height, were predictive of young adult BP and development of hypertension within race and gender groups after control of childhood BP, and were somewhat different in whites and blacks. This method adjusts for the uneven age distribution of the repeated measures and captures the underlying or average BP over childhood or adulthood, leading to improved estimates of risk prediction.