Electronic Letters to:

Trauma: J. P. Stannard, A. K. Singhania, R. R. Lopez-Ben, E. R. Anderson, R. C. Farris, D. A. Volgas, G. R. McGwin, Jr, and J. E. Alonso Deep-vein thrombosis in high-energy skeletal trauma despite thromboprophylaxis J Bone Joint Surg Br 2005; 87-B: 965-968 [Abstract] [Full text] [PDF]

eLetters: Submit a response to this article

Electronic letters published:

Response to Mr Cooper from the authors

James P Stannard   (14 October 2005)

Letter from Mr Cooper

Julian P Cooper   (6 September 2005)

Response to Mr Cooper from the authors 14 October 2005
James P Stannard, Orthopaedic Surgeon University of Alabama at Birmingham, USA Send letter to journal: Re: Response to Mr Cooper from the authors james.stannard {at}ortho.uab.edu James P Stannard	Sir, We thank Mr Cooper for his letter and understand his concern. The decision regarding the most appropriate test to use was made by Dr Gerald McGwin, who is an epidemiologist and statistician, in consultation with me. The decision to use the one sided test was made after careful consideration of our expectations in this study. We hypothesised prior to initiating the study that the pelvic fracture group was likely to have more frequent incidence of pelvic deep-vein thrombosis (DVT). Our statistical analysis was undertaken to reflect this hypothesis. The two sided test would have been appropriate if we did not have an expectation that the pelvic trauma patients would have a higher incidence of pelvic DVT. We acknowledge that the selection of the test is certainly open for debate, but believe the one sided test was appropriate in this case. J. P. STANNARD, MD University of Alabama, Birmingham, USA.
Letter from Mr Cooper 6 September 2005
Julian P Cooper, Consultant Orthopaedic Trauma Surgeon University Hospital Birmingham NHS Foundation Trust (Selly Oak Hospital) Send letter to journal: Re: Letter from Mr Cooper julian{at}cooperfrcs.fsnet.co.uk Julian P Cooper	Sir, I read this article with interest. While the rate of 11.5% of venous thromboembolism (VTE) is interesting, the conclusion that 'pelvic deep-vein thrombosis is significantly more frequent after pelvic than non-pelvic trauma' appears to be supported by a flawed statistical analysis. Repeating Fisher’s exact test with data from the paper reveals that the p-values are one-sided. One-sided tests appear to show greater significance, but conventionally tests are two-sided unless one direction of departure from the null hypothesis can be, with certainty, disregarded prior to analysis.1 There would seem to be no basis for such an assumption in this study. Using both the R2 and StatsDirect3 statistical packages, one-sided p-values for 'definite' and 'presumed + definite' clots are 0.053 and 0.0868 respectively, as opposed to the values of 0.056 and 0.087 stated in the paper. Two-sided values calculated by the same packages are 0.0753 and 0.1669 respectively. Armitage and Berry1 recommend that two-tailed p-values are more robustly obtained by doubling the one-tailed value, giving values of 0.106 and 0.1736 for the two comparisons. Whatever method is used for the calculation it is clear from the data in this study that there is no statistically significant difference in the frequency of pelvic deep-vein thrombosis. Furthermore, a meta-analysis, summarised in practice guidelines from the Eastern Association for the Surgery of Trauma,4 showed that the only established risk factors for VTE in trauma are spinal fractures, spinal cord injury and increasing age. The paper of Stannard et al includes no information on associated spinal injuries or the comparative ages of the groups—so a major source of confounding cannot be assessed. Given these concerns, I would respectfully suggest that the conclusion regarding frequency of pelvic deep-vein thrombosis cannot be sustained with these data. J. P. COOPER, BSc, FRCS(Tr & Orth) Selly Oak Hospital, Birmingham, UK. 1. Armitage P, Berry G. Statistical methods in medical research Third ed. Blackwell: Oxford, 1994:96-140. 2. R Development Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing: Vienna, 2004. 3. StatsDirect statistical software (version 2.4.5). URL http://www.statsdirect.com. 2005. 4. Rogers FB, Cipolle MD, Velmahos G, Rozycki G. EAST practice management guidelines for the management of venous thromboembolism in trauma patients. Eastern Association for the Surgery of Trauma URL. http://www.east.org/tpg/dvt.