Choice of Time Scale and Its Effect on Significance of Predictors in Longitudinal Studies

• Activity Number: 72
• Type: Contributed
• Date/Time: Sunday, August 7, 2005 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Michael Pencina*+ and Martin G. Larson and Ralph B. D'Agostino
• Companies: Boston University and Boston University and Boston University
• Keywords: survival models ; time on study ; age to event ; proportional hazards

Abstract

Time-to-event regression analyses are a frequent tool in biomedical research. In clinical trials, this time is usually measured from the beginning of the study. The same approach was adopted by longitudinal observational studies. However, in recent years, a strong argument has been made to use the date of birth as a starting point and thus use age as the time-to-event. In this paper, we consider five proportional hazards regression models that differ in the choice of time scale and in the method of adjusting for the years before the study: two models with time on study as a scale, with linear and quadratic terms for baseline age, and three models with age as a scale, adjusted for entry age, stratified on birth cohort, and unadjusted. Using mathematical formulations of the models considered and numerical simulations, we explore the differences in the estimated regression coefficients reported in the literature and make recommendations about the proper choice of scale. We conclude that correct adjustment for the age at entry is crucial in reducing the bias of the estimated coefficients. Our findings are supported by four practical examples from the Framingham Heart Study.