JBJS welcomes reader comments on published articles. Letters to the Editor are reviewed by JBJS editors but are not peer-reviewed. To submit your letter, please follow the "submit a response" link that appears in the content box at the upper right of the full text of the article.

Letters to the Editor to:

Scientific Articles:
E. Louis Peak, Javad Parvizi, Michael Ciminiello, James J. Purtill, Peter F. Sharkey, William J. Hozack, and Richard H. Rothman
The Role of Patient Restrictions in Reducing the Prevalence of Early Dislocation Following Total Hip Arthroplasty. A Randomized, Prospective Study
J Bone Joint Surg Am 2005; 87: 247-253 [Abstract] [Full text] [PDF]
*Letters to the Editor: Submit a response to this article

Electronic letters published:

[Read Letter to the Editor] Total Hip Arthroplasty Dislocations and Power Analysis for Rare Events
Justin S. Cummins, James N. Weinstein   (24 May 2005)
[Read Letter to the Editor] Dr Parvizi and colleagues respond to Drs Cummins and Weinstein
Javad Parvizi, M.D., Louis Peak, M.D., Richard H. Rothman, M.D., Ph.D.   (24 May 2005)

Total Hip Arthroplasty Dislocations and Power Analysis for Rare Events 24 May 2005
Next Letter to the Editor Top
Justin S. Cummins,
Orthopedic Surgery Resident
Dartmouth Hitchcock Medical Center,
James N. Weinstein

Send letter to journal:
Re: Total Hip Arthroplasty Dislocations and Power Analysis for Rare Events

justin.s.cummins{at}hitchcock.org Justin S. Cummins, et al.

To the editor:

In the article by Peak et al, entitled “The Role of Patient Restrictions in Reducing the Prevalence of Early Dislocation Following Total Hip Arthroplasty” (Peak et al, J Bone Joint Surg Am. 2005; 87: 247- 253), the authors bring high quality data to an area where few studies exist.

After reviewing the methods section of the paper, there are two specific concerns and questions for the authors: First, why is a threefold difference in dislocation rates chosen as the threshold for detecting a difference between the two groups? As noted by the authors, multiple factors may contribute to dislocations in the early postoperative period, including surgical approach and implant position. Thus, it is difficult to imagine that discontinuing the use of an abduction pillow and a high toilet seat, as well as not riding in a car would result in a threefold increase in dislocation rates. This implies that if restrictions resulted in a 50% decrease in dislocations, they would still not be worthwhile to implement.

A second concern relates to the power calculation. With the numbers given in the article (alpha – 0.05, beta – 0.20, baseline dislocation rate – 1.0%), and attempting to power the study to detect a threefold difference, a much larger study group is needed. Using the method illustrated in a recent paper by Lochner et al.(1) for estimating the sample size needed to detect a difference in proportions, a minimum of 760 patients in each group would be needed to detect a threefold difference in dislocation rates. A more precise estimate using STATA (2) statistical software indicates that 866 patients would be needed for each group in order to appropriately minimize the risk of a type-II error. With the number of patients enrolled in this trial, the power is approximately 13% (beta = .87), which makes the risk of a type-II error very high.

Given the high threshold for detecting a difference and the apparent lack of sufficient statistical power, I would be hesitant to conclude that no difference exists between the restricted and unrestricted groups in regard to dislocation rate. Thank you for considering these comments and congratulations on your work.

References

1. Lochner H, Bhandari M, Tornetta P. Type-II error rates (beta errors) of randomized trials in orthopedic trauma. J Bone Joint Surg Am. 2001; 83: 1650-1655.

2. STATA version 8.2, 2004. StataCorp, College Station, TX 77845.

Dr Parvizi and colleagues respond to Drs Cummins and Weinstein 24 May 2005
Previous Letter to the Editor Top
Javad Parvizi, M.D.,
Surgeon
Rothman Institute,
Louis Peak, M.D., Richard H. Rothman, M.D., Ph.D.

Send letter to journal:
Re: Dr Parvizi and colleagues respond to Drs Cummins and Weinstein

parvj{at}aol.com Javad Parvizi, M.D., et al.

We thank Drs. Cummins and Weinstein for their interest in our article and for their shrewd comments.

The ‘expected’ three fold increase in dislocation rate was determined after extensive review of the literature and discussions with various arthroplasty surgeons. However, because of lack of any report in the literature pertinent to this subject, we could not select the difference (delta) based on historic publications. We chose the smallest possible difference between the groups in order to ‘overpower’ the study. We agree with the authors that one would expect to observe a substantial, and hopefully much more than a three fold, reduction in the incidence of dislocation by implementation of these restrictions.

We were cognizant of the importance of power analysis for this study. Statistical advice had been sought prior to initiation of the study and the size of the patient population was determined based on that advice. The power analysis was performed using the method described by Joseph L. Fleiss in the book Statistical Methods for Rates and Proportions.(1) This book is the authoritative text in the statistical field on comparative studies using rates and proportions. Using the methods from Chapter 3, "Determining Sample Sizes Needed to Detect a Difference Between Two Proportions", we set significance level á to be 0.05 and the power to be 0.80 for a one-tailed test of dislocation rates between the two groups. The rate of dislocation was determined to be low at approximately 1 percent. Since the dislocation rate was set to be so low, the lowest incidence table (TableA.3, p.260) was utilized. For a power of 0.80, and to detect a three-fold difference, a minimum sample size of 130 was needed in each group for this study. The method elected to determine the power for this study is well accepted, validated, and previously utilized by numerous investigators. However, despite our initial power analysis that determined that a population of 130 patients in each arm of the study would be sufficient, we decided to recruit a higher number of patients to account for attrition. There were no patients lost to follow-up. Based on our initial statistical evaluations and the consequent detailed examination of our protocol, we are confident that the patient population recruited into this study is markedly higher than one needed to avoid a type II error even with such a small rate of dislocation.

Although the dislocation rate for the ‘experimental’ patient population at the conclusion of the study did not happen to be two or three times higher than the ‘control’ group, this does not detract from the validity of the statistical methods used to determine the sample size.

The main objective of the study was to evaluate the dislocation rate. Furthermore, the validity of the conclusions of our study has been confirmed in a recent analysis of the dislocation rate in 1000 patients at our institution. The dislocation rate for these patients who were not subjected to the restrictions mentioned in the study was 0.6 (2).

Respectfully:

Louis Peak MD

Javad Parvizi MD

Richard H Rothman MD, PhD

(1)Joseph L. Fleiss in the book Statistical Methods for Rates and Proportions , 2nd edition, 1981, John Wiley & Sons Inc., New York, pp. 42-44, 260.

(2)Sharkey PF, Parvizi J, Hozack WJ, Rothman RH. Ultra-High Volume and Early Outcomes of Primary and Revision Total Hip Arthroplasty. J Arthroplasty 19(6): 694-9, 2004