How To Lie with Statistics and Figures

Medical practitioners have had to learn more and more statistics not only to perform studies but more importantly, to interpret the medical literature and apply new findings to practice. We believe that because of a lack of formal training in statistics, the prevalence of some common errors is simply because of a lack of knowledge and awareness about these errors, which frequently leads to the misinterpretation of study results. This article reviews some of the common pitfalls and tricks that are prevalent in the reporting of results in the medical literature. Common errors include the use of the wrong average, misinterpretation of statistical significance as practical significance, reaching false conclusions because of errors in statistical power interpretation, and false assumptions about causation caused by correlation. Additionally, we review some design and reporting practices that are misguided, the use of post hoc analysis in study design, and the pervasiveness of "spin" in scientific writing. Last, we review and demonstrate common pitfalls with the presentation of data in graphs, which adds another potential opportunity to introduce bias. The tests used in the medical literature continue to change and evolve, usually for the better. With these changes, there will certainly be opportunities to introduce unintentional bias. The more aware we are of this, the more likely we are to find it and correct it.