Determinants of patients' treatment preferences in a clinical trial

Several researchers have suggested that patients' preferences for a particu lar form of treatment should be taken into account in clinical trials. Pref erences may influence the outcome of treatment, especially in trials when p atients cannot be blinded to the type of treatment received and the outcome is based on patients' evaluations of therapy. Participants in this study w ere 136 edentulous patients who took part in a randomised controlled clinic al trial comparing two types of treatments for edentulism: conventional den tures and implant-supported prostheses. Prior to receiving treatment, subje cts were required to complete a questionnaire regarding their satisfaction with their present prostheses. In addition, they were asked to indicate whi ch treatment they would prefer if given a choice. The objective of this stu dy was to determine whether there are important differences among study par ticipants between patients who have a treatment preference and those who do not. The effects of satisfaction with pre-treatment prostheses, age, gende r and level of education on preferences were examined. Level of satisfactio n with the original dentures and level of education were significant predic tors of preference. Compared to subjects who rated their satisfaction with their current condition as 'low', the odds ratios associated with having a preference for implant treatment were 0.31 (95% CI: 0.09 to 0.96) for subje cts who rated their prostheses in the 'medium' range and 0.11 (95% CI: 0.03 to 0.41) for those who rated in the 'high' range. In addition, subjects wi th high levels of education were significantly less likely to have a prefer ence for either conventional or implant treatments (OR=0.18, 95% CI: 0.02 t o 0.77 and OR=0.20, 95% CI: 0.05 to 0.76, respectively) compared to those w ith low education. Neither age nor gender was a significant predictor of pr eference. We suggest that study designs which incorporate patients' prefere nces must take into account possible differences between preference groups that might confound the relationship between preference and the outcome of interest.