Giving medicine a fair trialBMJ 2000; 321 doi: https://doi.org/10.1136/bmj.321.7275.1529/a (Published 16 December 2000) Cite this as: BMJ 2000;321:1529
Patients' preferences should be assessed
- Laurie Allan, director chronic pain services,
- Lance Tooke, clinical research fellow, pain management
- Northwick Park Hospital, Harrow, Middlesex HA1 3UJ
- Sandwell Health Authority, West Bromwich B70 9LD
- Unit of Clinical Epidemiology and Trials, National Institute for Cancer Research, 16132 Genoa, Italy
- Clinical Trials Research Group, Biomedical Ethics Unit, McGill University, Montreal, Quebec, Canada H3A1 W9
- Department of Radiation Oncology, Sir Charles Gairdner Hospital, Nedlands, Western Australia 6008, Australia
- Cancer Foundation of Western Australia
- Western Australian Clinical Oncology Group, Subiaco, Western Australia 6008, Australia
EDITOR—Ashcroft in his editorial calls for trials that do not second guess what patients want.1 We agree that sometimes the search for the perfect design of a clinical trial is impractical. Ashcroft argues that if we are uncertain about treatments then the best treatment for the patient is the trial.
Although trials should be simple, timely, and well designed to answer well posed questions, we cannot agree that they should not assess what patients prefer. The preference of a patient deserves special emphasis when diseases or treatments affect quality of life, when the treatment entails risks or side effects, or when the choice between treatments is a “close call.”2 This is particularly relevant in chronic diseases where a particular symptom such as pain is the problem. The balance between the efficacy and the profile of side effects of the treatment in relation to the overall pain experience is one that only patients can judge and may be more accurate than pain measures alone. Preference for a particular treatment may promote compliance and contribute to its success. We understand Ashcroft's difficulty that endpoints relevant to patients—for example, quality of life—make trials harder to run and take longer to implement, but we cannot agree that the results are harder to generalise and apply.
Those who apply a true evidence based approach include only studies that are randomised, double blind, and placebo controlled in systematic reviews. The goal of good clinical trial design is to eliminate chance and bias. Without randomisation, treatment effects are exaggerated up to 40%; without effective blinding, exaggeration may reach 20%.3 The larger the sample population the more likely the results are to be credible; small trials overestimate treatment effect by as much as 30%.4 These facts constitute a hierarchy that has not yet been …