Article type
Year
Abstract
Background: Patients usually have treatment preferences which may conflict with the principle of randomization. Respecting preferences in a trial format could enhance recruitment rates and statistical efficiency. Also, one might assume beneficial effects for trial participants if they are invited to actively take part in management decisions rather than receiving a treatment by chance. It is unclear if this assumption is substantiated by the available data.
Objectives: In a systematic review, we studied whether risk profiles and effect sizes differ between experimental and control cohorts in clinical trials with parallel randomized and observational arms.
Methods: Trial formats incorporating patient's preferences (that is, Brewins partially-randomized patient preference trials [in PRPPT patients issuing a clear preference receive their treatment of choice, and, if undecided, are offered randomization], comprehensive cohort studies, two-stage trial designs) were eligible for this review. We scanned Medline, Embase, Cancerlit, Scisearch, the Cochrane Central database and the internet for studies published between 1966 and August 2004. We made no restrictions for language. Two authors independently abstracted data from original studies and solved discrepancies by consensus. Quantitative analysis was restricted to primary endpoints. We calculated event rates, risk differences or differences in means (d) with 95% confidence intervals (CI) between randomly generated and preferred experimental and control groups.
Results: Of 721 citations originally identified, 22 studies with 8829 patients were included. Trials covered a broad spectrum of conditions and interventions (e.g., counseling versus antidepressants, surgical versus medical management of unstable angina). Four trials met the original PRPPT format.
We found no obvious differences in baseline profiles or absolute and relative effect sizes between randomized and preference arms. One trial showed favorable results with local over general anesthesia for endometrial microwave ablation only in the randomized group (semantic differential scale d 0.42, 95% CI 0.19 to 0.64). In contrast, women preferring waterbirth over landbirth had a shortened first stage of labor (d -221 minutes, 95% CI -443 to 2 minutes), which was not reproduced in the randomized cohort.
Conclusions: Patients preferences have minor influence on effect sizes with different therapeutic interventions.
Objectives: In a systematic review, we studied whether risk profiles and effect sizes differ between experimental and control cohorts in clinical trials with parallel randomized and observational arms.
Methods: Trial formats incorporating patient's preferences (that is, Brewins partially-randomized patient preference trials [in PRPPT patients issuing a clear preference receive their treatment of choice, and, if undecided, are offered randomization], comprehensive cohort studies, two-stage trial designs) were eligible for this review. We scanned Medline, Embase, Cancerlit, Scisearch, the Cochrane Central database and the internet for studies published between 1966 and August 2004. We made no restrictions for language. Two authors independently abstracted data from original studies and solved discrepancies by consensus. Quantitative analysis was restricted to primary endpoints. We calculated event rates, risk differences or differences in means (d) with 95% confidence intervals (CI) between randomly generated and preferred experimental and control groups.
Results: Of 721 citations originally identified, 22 studies with 8829 patients were included. Trials covered a broad spectrum of conditions and interventions (e.g., counseling versus antidepressants, surgical versus medical management of unstable angina). Four trials met the original PRPPT format.
We found no obvious differences in baseline profiles or absolute and relative effect sizes between randomized and preference arms. One trial showed favorable results with local over general anesthesia for endometrial microwave ablation only in the randomized group (semantic differential scale d 0.42, 95% CI 0.19 to 0.64). In contrast, women preferring waterbirth over landbirth had a shortened first stage of labor (d -221 minutes, 95% CI -443 to 2 minutes), which was not reproduced in the randomized cohort.
Conclusions: Patients preferences have minor influence on effect sizes with different therapeutic interventions.