Different Methods to Analyse the Results of a Randomized Controlled Trial with More Than One Follow-Up Measurement

In this chapter, an overview is given of different methods to analyse data from a randomised controlled trial (RCT) with more than one follow-up measurement. For a continuous outcome variable, a classical GLM for repeated measurements can be used to analyse the difference in development over time between the intervention and control group. However, because GLM for repeated measurements has some major disadvantages (e.g., only suitable for complete cases), it is advised to use more advanced statistical techniques such as mixed model analysis or Generalised Estimating Equations (GEE). The biggest problem with the analysis of data from an RCT with more than one follow-up measurement is the possible need for an adjustment for baseline differences. To take these differences into account a longitudinal analysis of covariance, an autoregressive analysis or a ‘combination’ approach can be used. The choice for a particular method depends on the characteristics of the data. For dichotomous outcome variables, an adjustment for baseline differences between the groups is mostly not necessary. Regarding the more advanced statistical techniques it was shown that the effect measures (i.e. odds ratios) differ between a logistic mixed model analysis and a logistic GEE analysis. This difference between these two methods was not observed in the analysis of a continuous outcome variable. Based on several arguments (e.g., mathematical complexity, unstable results, etc.), it was suggested that a logistic GEE analysis has to be preferred above a logistic mixed model analysis.