Causal Effect Estimation in Longitudinal Study with Applications to Cardiovascular Risk Factors

• Activity Number: 21
• Type: Contributed
• Date/Time: Sunday, August 11, 2002 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section*
• Author(s): Wei Wu*+ and Cameron Chumlea and Shumei Guo
• Affiliation(s): Wright State University School of Medicine and Wright State University School of Medicine and Wright State University School of Medicine
• Keywords: causal effect ; longitudinal data ; overweight ; high risk for cardiovascular disease ; G-computation algorithm

Abstract

The present study estimates the causal effect of BMI on levels of low-density lipoprotein cholesterol (LDL) using longitudinal data. BMI, weight (kg)/height (m2), is a binary time-varying exposure (BMI >=25 kg/m2 equals overweight and BMI < 25 kg/m2 equals normal), and current measure of LDL is the dichotomous outcome measure (LDL>=130 mg/dl equals high risk for cardiovascular disease and LDL < 130 mg/dl equals normal). Longitudinal data from 296 men and 321 women 10 to 60 years of age enrolled in the Fels Longitudinal Study were examined. Logistic regression was used to estimate the conditional probability of the current outcome of LDL conditioned on previous LDL levels (high risk or normal), previous BMI (overweight or non-overweight), current BMI, and the current age. G-computation algorithm of Robins (Robins, J.M. 1986) was used for computation. The causal effect of overweight on high risk LDL level in probability increased with age and ranged from 0.25 to 0.32 for men and from 0.37 to 0.48 for women before 40 years of age. After age 40, causal effect of overweight was 0.35 for men and 0.50 for women.