Confounding in randomised controlled trialsBMJ 2010; 341 doi: https://doi.org/10.1136/bmj.c5403 (Published 06 October 2010) Cite this as: BMJ 2010;341:c5403
- Philip Sedgwick, senior lecturer in medical statistics
- 1Section of Medical and Healthcare Education, St George’s, University of London, Tooting, London, UK
A randomised controlled trial evaluated the effectiveness of specialised care for early psychosis in one London borough.1 Participants were 144 people aged between 16 and 40 years presenting to mental health services for the first or second time with non-organic, non-affective psychosis. Using simple random allocation, individuals were allocated to specialised or control care. Specialised care consisted of assertive outreach with evidence based biopsychosocial interventions, whereas control care comprised standard care. Both treatments were delivered by community health teams.
Primary outcome measures included the occurrence of relapse within 18 months of follow-up. The unadjusted odds ratio for relapse for patients in the specialised care group compared with those receiving standard care was 0.46 (95% CI 0.22 to 0.97). When adjusted for sex, previous psychotic episode, and ethnicity, the odds ratio was 0.55 (95% CI 0.24 to 1.26).
Which of the following statements, if any, are true?
a) Randomisation of trial participants reduced confounding between treatment groups
b) Odds ratios were calculated because the trial was retrospective