Endgames Statistical Question

Randomised controlled trials: tests of interaction

BMJ 2014; 349 doi: https://doi.org/10.1136/bmj.g6820 (Published 14 November 2014) Cite this as: BMJ 2014;349:g6820
  1. Philip Sedgwick, reader in medical statistics and medical education1
  1. 1Institute for Medical and Biomedical Education, St George’s, University of London, London, UK
  1. p.sedgwick{at}sgul.ac.uk

Researchers evaluated the efficacy of an intervention based in general practice on the occurrence of repeat episodes of deliberate self harm.1 A cluster randomised controlled trial study was performed, involving 98 general practices. The intervention consisted of a letter from the general practitioner inviting the patient to consult, with guidelines on the assessment and management of deliberate self harm for the GP to use in consultations. The control treatment was standard care. Participants were patients registered with one of the study practices who had attended an emergency department after an episode of deliberate self harm (referred to as the index episode). A total of 1932 patients were recruited; 964 were allocated to the intervention and 968 to the control.

The primary outcome was a repeat episode of deliberate self harm in the 12 months after the index episode (categorised as yes or no). Although a higher proportion of patients had a repeat episode of deliberate self harm in the intervention group than in the control group, the difference was not significant (21.9% (n=211) v 19.5% (n=189); odds ratio 1.2, 95% confidence interval 0.9 to 1.5; P=0.16). No significant interaction existed between the treatment group and the patient’s sex (P=0.51) or the method used to cause self harm (categorised as poisoning, laceration, or other) (P=0.64). A significant interaction existed between the treatment group and a history of deliberate self harm before the index episode (categorised as yes or no) (P=0.017). A subsequent subgroup analysis showed an apparent benefit of the intervention for patients with a history of deliberate self harm (0.57, 0.33 to 0.98), but an adverse effect for patients …

View Full Text

Sign in

Log in through your institution

Subscribe