Rapid Responses to:
|
|
Rapid Responses: Rapid Responses published:
-
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif)
Pre-injury Harris Hip Score Assessment, Blinding and Multiple Endpoints.
- Rudolf W. Poolman MD PhD (7 April 2008)
-
Re: Pre-injury Harris Hip Score Assessment, Blinding and Multiple Endpoints.
- Frede Frihagen, Lars Nordsletten, and Jan Erik Madsen (15 April 2008)
|
|
|||
|
Rudolf W. Poolman MD PhD, Consultant Orthopaedic Surgeon Onze Lieve Vrouwe Gasthuis, Amsterdam, Oosterpark 9, 1091 AC, The Netherlands
Send response to journal:
|
Sir,
With Interest I read the important article by Frede Frihagen, Lars Nordsletten and Jan Erik Madsen. The authors choose the Harris Hip Score as their primary outcome. The including surgeon evaluated the pre-injury Harris Hip Score. This is interesting since this instrument consist of four domains: 1) pain, 2) function (including gait based on physical examination), 3) absence of deformity based on physical examination, and 4) Range of Motion (ROM) based on physical examination. I wonder how the investigators were able to assess the pre-injury gait, deformity and ROM? Or did the investigators omit these domains in their analysis? Furthermore, I wonder how the outcome assessors where blinded during their assessment of the Harris Hip Score. Hemiarthroplasty requires a larger incision compared to the small incision used for cannulated screw fixation; could this have influenced blinding? Finally, the authors describe they used multiple comparisons. I wonder why they did not use statistical correction such as, the Bonferonni method(1), although this is a rather strict method. The three outcomes would set the level of significance at atleast 0.017 instead of 0.05. The sweeping results regarding re-operation rates and its effect on function would stand up to this test. Still, using statistical correction for multiple endpoints would represent more realistic findings. Kind regards, Rudolf W. Poolman MD PhD Reference (1) Bland JM, Altman DG. Statistics notes: Multiple significance tests: the Bonferroni method. BMJ 1995; 310(6973):170. Competing interests: None declared |
|||
|
|
|||
|
Frede Frihagen, MD Ulleval University Hospital, 0407 Oslo, Norway, Lars Nordsletten, and Jan Erik Madsen
Send response to journal:
|
We certainly appreciate the interest in, and thorough read of, our article. Some important methodological questions are raised. Firstly, there are two questions relating to the data collection and Harris Hip Score: 1) We wanted to include a pre-morbid functional score in the study to be able to ascertain well matched treatment groups, as recommended by among others the Cochrane collaboration.(1) As none of the patients - obviously - were available for a pre-facture interview or examination, we had to rely on a post injury score. In our pre-fracture Harris Hip Score we asked the patients and/or their caregivers about pain and function, including range of motion and absence of deformity. We believe that it was much better to have a baseline functional score based on recall than to have nothing at all, and if the patients over- or under- estimated their pre-fracture function, this should be equally distributed between the groups. 2) The next comment related to the postoperative scoring of the Harris Hip Score and raised the question of whether the size of incision unblinded the assessors. We solved that quite simply by not letting the outcome assessors undress the patient, as we felt that the potential problem of reducing the quality of assessment by far was outweighed by the benefit of protecting the blinding. 3) The third question raised concerned the data analysis and the lack of correction for multiple comparisons. In our discussion in the article this is mentioned. Our pre-defined main outcome measure, the Harris Hip Score after 12 months, however, showed a statistically and clinically significant result in favour of the treatment with hemiarthroplasty. The other pre-specified functional tests are probably not independent of Harris Hip Score, indeed they all pointed towards hemiarthroplasty being the better treatment. The benefit of using the Bonferroni correction is thus questionable because it may lead to a type II error,(2) especially in the presence of a plausible biological explanation: A displaced femoral neck fracture in an elderly patient is broken beyond repair and should be replaced instead of fixed. Best regards,
References (1) Parker MJ, Gurusamy K. Internal fixation versus arthroplasty for intracapsular proximal femoral fractures in adults. Cochrane Database Syst Rev 2006;(4):CD001708. (2) Perneger TV. What's wrong with Bonferroni adjustments BMJ 1998;316:1236-1238 Competing interests: None declared |
|||