Sample size and statistical power of randomised, controlled trials in orthopaedics.

Abstract

We reviewed all 717 manuscripts published in the 1997 issues of the British and American volumes of the Journal of Bone and Joint Surgery and in Clinical Orthopaedics and Related Research, from which 33 randomised, controlled trials were identified. The results and sample sizes were used to calculate the statistical power of the study to distinguish small (0.2 of standard deviation), medium (0.5 of standard deviation), and large (0.8 of standard deviation) effect sizes. Of the 33 manuscripts analysed, only three studies (9%) described calculations of sample size. To perform post-hoc power assessments and estimations of deficiencies of sample size, the standard effect sizes of Cohen (small, medium and large) were calculated. Of the 25 studies which reported negative results, none had adequate power (beta < 0.2) to detect a small effect size and 12 (48%) lacked the power necessary to detect a large effect size. Of the 25 studies which did not have an adequate size of sample to detect small differences, the average used was only 10% of the required number Our findings suggest that randomised, controlled trials in clinical orthopaedic research utilise sample sizes which are too small to ensure statistical significance for what may be clinically important results.