Background As orthopaedic surgery becomes more evidence-based, the need for rigorous research has increased. This results in more complex studies that employ more sophisticated statistical analysis, often some form of regression. These statistical techniques require the data to meet certain assumptions for the findings to be considered valid. The purpose of this study is to determine the common regression techniques employed in the orthopaedic surgery literature, and demonstrate how often the assumptions of regression analyses are met and reported. Methods Studies published in the Journal of Bone & Joint Surgery (JBJS) in 2017 and 2018 were reviewed. Commentaries, editorials, and systematic reviews were excluded. The statistical analyses performed in each study were documented. When regression analyses were utilized, the article was reviewed for evidence that the necessary assumptions underlying the statistical methodology were assessed and met. Results From the 470 studies that were reviewed, the most common statistical test reported was the independent-samples t-test (n=215, 45.7%). Also, 201 studies (42.8%) implemented some form of regression analysis. The most common regression was a logistic regression (n= 106). None of the 201 studies using regression analysis reported meeting all of the necessary assumptions to appropriately use a regression test. Conclusion Many recent studies published in JBJS depended on regression analyses to reach their conclusions, but none fully reported the necessary assumptions of these tests. Orthopaedic surgery journals should be more transparent in reporting the methodology of statistical tests, and readers must beware of possible gaps in statistical methodology and critically evaluate the studies' findings.
Keywords: orthopaedic surgery; regression analysis; statistical analysis; statistical assumptions; study validity.
Copyright © 2021, Christiano et al.