For healthcare professionals only

Endgames Statistical Question

Cluster randomised controlled trials: sample size calculations

BMJ 2013; 346 doi: (Published 03 May 2013) Cite this as: BMJ 2013;346:f2839
  1. Philip Sedgwick, reader in medical statistics and medical education
  1. 1Centre for Medical and Healthcare Education, St George’s, University of London, Tooting, London, UK
  1. p.sedgwick{at}

Researchers evaluated the effects of training primary care health professionals in counselling patients about health related behaviour change. A cluster randomised trial with general practices as the unit of randomisation was performed. The intervention consisted of behaviour change counselling in four risk behaviours (smoking, alcohol use, exercise, and healthy eating). Control treatment consisted of standard care.1

The primary outcome measure was the proportion of patients who self reported making beneficial changes in at least one of the four risky behaviours at three months. Sample size was based on the ability to demonstrate an increase in the proportion of patients reporting beneficial change on one or more of the four health behaviours from 50% in the control group to 65% for the intervention group at three months. To do so a randomised controlled trial using simple random allocation with 80% power and a critical level of significance of 0.05 would require 340 patients. To account for clustering effects from randomised practices, an intracluster correlation coefficient of 0.05 was applied; this increased the sample size to 1104 patients, with 24 practices recruiting 46 patients each.

The researchers reported that training primary care health professionals in behaviour change counselling did not increase patients’ self reported beneficial behaviour change at three months.

Which of the following statements, if any, are true?

  • a) Measurements of the primary outcome for patients in the same general practice will tend to be correlated

  • b) If measurements of the primary outcome for patients within a general practice are correlated, sample size will need to be increased to demonstrate the smallest …

View Full Text

Log in

Log in through your institution


* For online subscription