Sample size calculations in orthopaedics randomised controlled trials: revisiting research practices.

The purpose of this study was to identify how often sample size calculations were reported in recent orthopaedic randomized controlled trials (RCTs) and to determine what proportion of studies that failed to find a significant treatment effect were at risk of type II error. A pre-defined computerized search was performed in MEDLINE to identify RCTs published in 2012 in the 20 highest ranked orthopaedic journals based on impact factor. Data from these studies was used to perform post hoc analysis to determine whether each study was sufficiently powered to detect a small (0.2), medium (0.5) and large (0.8) effect size as defined by Cohen. Sufficient power (1-β) was considered to be 80% and a two-tailed test was performed with an alpha value of 0.05. 120 RCTs were identified using our stated search protocol and just 73 studies (60.80%) described an appropriate sample size calculation. Examination of studies with negative primary outcome revealed that 68 (93.15%) were at risk of type II error for a small treatment effect and only 4 (5.48%) were at risk of type II error for a medium sized treatment effect. Although comparison of the results with existing data from over 10 years ago infers improved practice in sample size calculations within orthopaedic surgery, there remains an ongoing need for improvement of practice. Orthopaedic researchers, as well as journal reviewers and editors have a responsibility to ensure that RCTs conform to standardized methodological guidelines and perform appropriate sample size calculations.