A Statistics Review for the Arthroplasty Community

When you think about it, the fact that research articles can condense months’ worth of work (and sometimes years) and analysis into a tidy 3000 word paper is pretty incredible. This compression also makes every detail about the study, from its methods to how the results are presented as well as the statistical tests performed, of utmost importance. It is these statistical analyses that allow for the most fundamental question of each research paper to be answered: Is it likely that a true difference exists between the cohorts being studied? Yet, it is also these statistical assessments that readers have the hardest time interpreting on their own. As Mark Twain simplified for all of us: “There are three kinds of lies: lies, damned lies, and statistics.” And when it comes to statistics, very few of us have enough in-depth knowledge and experience to properly uncover what is real and what is a mistake. We can all review literature and impart our own knowledge of a topic to better understand a research study. Similarly, we can look at two groups of data, and by and large be able to tell if they are similar or different. But what a reader cannot do (at least not easily) is calculate the statistical differences between the groups being studied or to fully grasp if the proper statistical tests were performed. We see the magical “P < .05” and are too often naïve enough to assume that the statistical difference calculated was done without error and has clinical relevance. We also misconceive that association implies causation. These are potential problems for all of us.