STUDY SIZES AND STATISTICAL ANALYSIS IN ORTHOPAEDIC RESEARCH

Abstract

Aim: To review the study size and requirements of studies looking at factors affecting outcome following total hip arthroplasty.

Background: The orthopaedic literature is full of claims that new products out-perform older ones, cemented and un-cemented components are equal, cementing grades and mantles are all important and that component orientation is vital to longevity. We are also aware of patients who have performed well despite having numerous adverse features to their joint replacements.

We have searched the available literature for factors that have been implicated in the survival of hip replacements. We have used these to determine the likely study sizes required to provide meaningful data.

Method: We identified over 50 variables that have been implicated in the survival of hip replacements. Assuming all of them to be of equal relevance the study size required for multivariate analysis to be possible would be in excess of 50000. Some of these variables are less likely to be of great significance and this number could potentially be reduced to 25–30000.

Discussion: Because total hip replacement has a greater than 90% success rate at 10 years in nearly all series this makes the numbers required for multivariate analysis much larger. Individual factors affecting survival will be difficult to identify unless large series are considered.

Conclusion: The only way to reduce the numbers required for meaningful studies is to increase the matching of as many variables as possible to ensure that investigators conclusions are valid in the majority of small series orthopaedic studies.

The use of Joint Registers is likely to be the only way of obtaining the volume of data required to detect individual factors affecting survival. Care will still need to be taken interpreting this data as there are still numerous variables which are not accounted for in the Joint Register.