Sample size and statistical power of randomised, controlled trials in orthopaedics / Author’s reply
I cannot dispute the findings in the article by Freedman, Back and Bernstein,’ entitled `Sample size and statistical power of randomised, controlled trials in orthopaedics’ in the April 2001 issue, the conclusions of which were that most randomised trials in orthopaedics do not have adequate patient numbers to be able to detect statistically significant results, but question the authors’ comments that randomised trials undertaken without a power calculation to determine the number of patients to be studied are unethical.
In the ideal situation dilemmas over treatment are best resolved by a well-conducted randomised trial. To avoid type-II errors, however, trials in excess of 1000 patients are often required, thereby necessitating a multicentre study. While this is to be encouraged, practical and financial restrictions mean that it is generally not possible. An alternative is to encourage smaller randomised trials with appropriate methods and standardised reporting of results. The results of these smaller studies can then be combined in meta-analysis, thereby achieving a study of appropriate numbers and statistical significance. Such a practice is now routine in many branches of medicine. One example of such a meta-analysis in orthopaedics was to combine the 15 randomised trials involving 2023 cases comparing the Gamma nail with the sliding hip screw.2 Individually, no study was large enough to give valid results, but collectively clear conclusions could be made. Small trials also enable individual clinicians to have their own research projects, rather than being a small part of a large multicentre study, and such studies are easier to organise and finance. I have to disagree therefore with the conclusions of Freedman et al. Small randomised trials are not unethical, but can have a useful role in resolving dilemmas of treatment.
M. PARKER, MD FRCS Ed Peterborough District Hospital Peterborough, UK.
1. Freedman KB, Back S, Bernstein J. Sample size and statistical power of randomised controlled trials in orthopaedic surgery. J Bone Joint Surg [Br] 2001;83-B:397-402.
2. Parker MJ, Handoll HHG. Gamma and other cephalocondylic intramedullary nails versus extramedullary implants for extracapsular hip fractures (Cochrane Review). In: The Cochrane Library, Issue 2, 2001. Oxford: Update Software.
We thank Mr Parker for his letter in which he takes issue with our claim that it is unethical to perform randomised controlled studies with inadequate samples. We believe that it is wrong to conduct any study if it is known that it will not yield meaningful results.
Mr Parker suggests that we should “encourage small randomised trials”, in the hope that they can be combined by a metaanalysis to one meaningful project. We disagree. It is much better that the investigators collaborate at the outset. Why leave it to chance that additional studies will be done and that they will be combined in a meta-analysis? There is no doubt that a metaanalysis of studies comprising 2023 cases is inferior to and makes weaker conclusions than one unified trial studying a similar number of patients. Although it is laudable that individual clinicians can “have their own research projects” it is not a sound policy to encourage them in the hope that a critical mass will be reached.
The word ‘unethical’ is perhaps too harsh, as the individual investigator has only the best intentions, but good intentions are often not enough. We believe that we owe our patients the best possible advice and counsel and that, in turn, demands that our studies are of appropriate power.
It may be easier to finance small studies as Mr Parker suggests, but “you get what you pay for”.
K. B. FREEDMAN, MD S. BACK, BA J. BERNSTEIN, MD, MS University of Pennsylvania Philadelphia, USA.
Copyright British Editorial Society of Bone & Joint Surgery Nov 2001
Provided by ProQuest Information and Learning Company. All rights Reserved