Statistics in experimental studies on the human spine: Theoretical basics and review of applications

Proper statistical analysis is essential in the research studies. In particular, as regards the in vitro testing of the lumbar spine, the criteria for the standardization have been extensively discussed but the use of statistics has not been reviewed. Unfortunately, cadaveric testing is a very difficult and complex experimental field, in which many factors such as collection times, costs, funding, personnel and logistic issues, availability of implant components, can determinate low sample sizes, with impacts on the statistical evaluation. Moreover, as in any other field, some errors can be commonly made in the choice of the most appropriate statistical tests. The present tutorial article provides a comprehensive overview of the theoretical basics of the statistical approach, focusing on the experimental testing of the lumbar spine. Reference values about adequate sample size and statistical power were provided. The use of statistics was reviewed in a selection of 20 papers, chosen among the most cited and representative contributions for the in vitro study of the lumbar spine. Overall, only 7 in 20 papers applied fully corrected procedures. The achieved power, calculated for the medium conventional level of the effect size, ranged from 0.13 to 0.99 but overall was found rather poor, below 0.6. The study also revealed that applying incorrect statistical tests and having low sample size can impact p-value and power, respectively, with consequences on the identification of the significant outcomes. We emphasize that this review is not intended as a mere critical analysis from a statistical perspective, but as a constructive investigation aimed to support the researchers in the challenging field of cadaveric testing. It is worth noting that the provided recommendations and reference values can be also accounted for experimental scenarios testing the other spine regions, either in human or in animal models.