Handling dropout and clustering in longitudinal multicentre clinical trials