What influences participants' treatment preference and can it influence outcome? Results from a primary care-based randomised trial for shoulder pain.

Abstract

BACKGROUND: In randomised clinical trials (RCTs), outcome may be influenced by the opinions of the participants about the efficacy of treatments. AIM: To examine how initial treatment preferences of participants in a shoulder pain trial affected functional outcome and future treatment preferences. DESIGN OF STUDY: Observational cohort study nested within a multicentre, pragmatic RCT of steroid injection versus physiotherapy for unilateral shoulder pain. SETTING: Nine general practices in north Staffordshire. METHOD: Two hundred and seven adults were randomised in the trial. Disability scores and preferences of the participants for the trial treatments were elicited at two points: prior to randomisation and 6 months post-randomisation. A good functional outcome was defined as at least a halving in the disability score at the 6 months follow-up point. RESULTS: Pre-randomisation preferences were: 40% for injection and 20% for physiotherapy, and 40% gave no preference. A good outcome was achieved in a higher percentage of participants who gave a pre-randomisation treatment preference compared with those who did not (62% compared with 48% percentage difference = 14%; 95% confidence interval [CI] = -1 to 27%) with similar percentages in each preferred treatment group. However, receiving the preferred treatment did not confer any additional benefit in those who expressed a preference (receiving preferred treatment = 56%; not receiving preferred treatment = 69%). At 6 months post-randomisation, participants with a good, as opposed to poor, outcome were more likely to report as their preferred treatment the one to which they had been randomised, irrespective of pre-randomisation preference and whether the preferred treatment was received. CONCLUSION: This analysis suggests that preferences prior to treatment can affect outcome, but that treatment outcome is a stronger influence on post-treatment preferences. We present some empirical evidence to support the statement that treatment preferences can have important effects on the results of RCTs.