The impact of using a partially randomised patient preference design when evaluating alternative managements for heavy menstrual bleeding

Abstract

Objective To identify the advantages and disadvantages of using a partially randomised patient preference design rather than a conventional randomised controlled design when evaluating alternative managements for heavy menstrual bleeding.

Design Randomised controlled comparison of two clinical trial designs with subsequent follow up of the cohorts of women generated.

Participants Women attending a general gynaecology clinic for the first time because of heavy menstrual bleeding.

Interventions Partially randomised patient preference clinical trial design and conventional randomised controlled design.

Main outcome measures Overall participation; participation in randomised clinical trial of medical management compared with transcervical surgical resection of the endometrium; prognostic characteristics (socio–demographic and Short Form 36) of clinical trial groups; outcomes (clinical and Short Form 36) of clinical trial groups.

Results Overall, more women participated in the partially randomised patient preference design (130/135 vs97/138; difference 27%, 95% CI 18% to 34%) but there was no difference in the numbers who agreed to be randomised (90/135 vs97/138; difference −3%, 95% CI −15% to 7%). Women who chose medical management tended to have better general health, to be less restricted by their menstrual problems, with fewer having been previously treated by their general practitioner. Those who chose transcervical resection of the endometrium had all tried medical management and had higher bleeding scores. Follow up satisfactions and acceptability rates, and Short Form 36 scores were highest after transcervical resection of the endometrium, whether chosen or randomised. Acceptability and a desire to continue the same treatment was greater among those who chose medical management than those randomly allocated it.

Conclusions Use of the partially randomised patient preference design did not affect recruitment to the randomised controlled trial suggesting that a conventionally designed trial would not be biased by motivational factors in this context. Data from the preference groups informed the generalisability of the results but did tend to confirm conclusions that anyway reasonably followed from the randomised controlled trial. The extra resource implications of using the partially randomised patient preference design were significant reflecting the additional 40% who participated and the extra analyses entailed.