Research projects

Methodology projects

The Design and Analysis of Trials of Group Treatments


Group administration of treatment, i.e. exercise classes for the treatment of musculo-skeletal disorders and self-help groups for smoking cessation or alcohol problems, is becoming an increasingly attractive option, mainly due to reduction of costs and effectiveness of therapy through between-patients interaction. However, when trials compare group with individual therapies, group clustering effects are too often ignored, leading likely to incorrect sample size estimation and trial results, and exposing patients to ineffective treatments.

This is a project for exploring the design and analysis of trials where clustering occurs in the group therapy arm but not in the individual (control) arm, taking into account clustering effects (i.e. the between-group therapy variation), as well as the between-therapist variation (‘therapist effect’) in patient outcomes.

Research Questions

For trials with only partially nested data, this project aims to answer the following questions via simulation and analytical work:

  • What sample size is required for trials of group therapy versus individual therapy?
  • How large is the intra-cluster correlation due to group therapy?
  • What are the best methods of statistical analysis to estimate the treatment effect in trials comparing group administered treatment with individual therapy?
  • What are the best methods of estimating the causal effect of group administered interventions where there is non-compliance with treatment?

Duration of the project

April (2009) - March (2012)

Funding body

Medical Research Council - Methodology Research Programme

Members of the project

Chris RobertsPrincipal investigator
Dr Steve RobertsCo-investigator
Professor Graham DunnCo-investigator
Dr Eva BatistatouResearch associate

Links to other resources

  • Please follow this link to the MRC website
  • Please follow this link to the clsampsi STATA program, developed by Chris Roberts and Eva Batistatou to calculate power and number of clusters/cluster sizes for trials with clustering effects

Contact us

If you would like further information about this work please contact either the PI, Dr Chris Roberts (, or the project RA, Dr Evridiki Batistatou (