fbpx

GenstatMixed model analysis for clinical trials 

 C. Supakorn

 

Mixed models are a powerful tool for data analysis in agriculture, ecology, forestry, medicine etc. In clinical studies, trials are usually conducted across multiple centres. This is often because there’s a limited number of suitable experimental subjects in a single centre and/or to include subjects from a wide demographic range in the study. Subsequent analysis of data from a multi-centre trial must consider the possibility of differences between the individual centres and their effect on treatment. Ignoring centre in the model will typically result in incorrect conclusions being made! 

When a mixed model is implemented to analyse data from a multi-centre trial, assessment of treatment effects across centres is achievable. However, we must decide whether centre and the centre by treatment interaction are fitted as fixed or random termssince this will affect the interpretation of our results: when fitted as fixed, inference can only be made about the centres observed, but if fitted as random, inference can be applied to the broader population of centres. 

 

Let’s see an example. This is described in the book Applied Mixed Models in Medicine by Brown and Prescott (2015).  

The data is from a study to compare three drug treatments (A, B, and C) for controlling hypertension. Twenty-nine centres participated in the trial and patients were randomly assigned to the drug treatment groups. Diastolic blood pressure was recorded for each patient pre-treatment (dbp1) and post (dbptreatment. The main objective of this trial was to assess the effect of the three treatments on diastolic blood pressure. 

The full model for analysing this data includes dbp1 as a baseline covariate, treatment (treat) as a fixed term, and centre and the centre by treatment interaction (centre.treat) as either fixed or random terms.  

 

where  

 

The REML function in Genstat can be used to analyse this data. Watch this video!  

 

 

Reference:  

Brown, H. and Prescott, R. (2015). Applied Mixed Models in Medicine. Third Edition. John Wiley & Sons Ltd, England.