Continuation Ratio and Generalized Estimating Equation Analysis of a Longitudinal Asthma Study

Abstract

Two randomized controlled trials were conducted to find out whether a new treatment for asthma has a significant effect on the patients. These were multi-center trials with a parallel design, the control arm receiving a Placebo. The data were collected over a period of about 20 days before administering the intervention and for almost 80 days after the intervention. Thus, each patient has many observations recorded, making the data longitudinal. The data are summarized using first descriptive statistics and graphical displays. Then, a continuation ratio model with a lagged covariate to account for the longitudinal aspect is used to model the data. Finally, Generalized Estimating Equations methods are used. These methods have acquired popularity in recent years to account for longitudinal correlation structures. To apply the continuation ratio, the data have to be appropriately restructured. Then, the logistic regression is used to model the symptoms. The results of this procedure show that the treatment is statistically significant. However, the goodness of fit tests show that the model is inadequate. This issue is explored in the last subsection of Chapter 3. Using Generalized Estimating Equations to analyze the number of times rescue medication was used, we concluded that there is no statistically significant difference between the Active and Control groups. However, we noticed that the use of rescue medication decreased with time from the start of treatment.