The need for confidence intervals in the presentation of orthopaedic data.

I844 THE JOURNAL OF BONE AND JOINT SURGERY Reports in the orthopaedic literature and presentations at scientific meetings of such societies as The American Academy of Orthopaedic Surgeons are often criticized because of the large percentage of patients involved who were lost to follow-up, or because of the small number of patients in the study. Such criticism, although valid, does not give enough specific information regarding the shortcomings of these studies and the actual limitations of the data. Conversely. every study. no matter how small the sample size, does contain some information. Almost all orthopaedic publications require the presentation of a p value of a hypothesis test before a statement can be made concerning the superiority of one treatment modality over another. The p value is the probability that the chance selection of patients might suggest a difference in treatment that was not real. Only when the p value is small are we willing to believe that the observed difference in treatment is very likely real and not simply due to chance. However, when a simple estimate for a single group of patients (such as the tenyear results of a particular total knee prosthesis) is presented, it is often assumed that the estimate, which is usually based on a relatively small sample, is the same value that would have been obtained from a sample size of several thousand or more. The confidence interval for an estimated proportion (or other statistic) is much more informative than a p value when different treatments are compared. The confidence interval provides a measure of the magnitude of the possible difference between two groups of patients, regardless of whether or not the p value was small. The need to present confidence intervals has recently been discussed at meetings of The American Academy of Orthopaedic Surgeons5, in an editorial in The Journal of Bone and Joint Surgery’9, and in a recent survivorship analysis of the total condylar kneearthroplasty prosthesis’4.