The importance of clear language

Open Access-This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the source is credited. Most manuscripts submitted to Acta Orthopaedica, as well as to other medical scientific journals, describe treatment effects, complications, side effects, quality of life, health economy, etc. in a limited group of patients. The aim of the studies, however, is to learn something of importance for all patients, including future ones, not only for those who participated in the studies. Statistical analysis is performed to quantify the empirical support for the findings. The evaluation is based on sample-to-population inference: an observable sample is studied in order to learn about the unobservable population it represents. A consequence of the population not being fully observable is that the findings will always be uncertain to some degree. Statistical methods can, however, be used to quantify this uncertainty. It is often presented in terms of confidence intervals and p-values. Unfortunately, instead of recognizing that p-values and confidence intervals describe generalization uncertainty, many authors seem to believe that these measures represent some obscure property of the observed sample itself. For example, when two groups have different mean values, but the difference is not statistically significant, the observed difference is often described as " there was no difference ". Another common phrase is that " there was no statistical difference ". The first description is wrong, because it is a misrep-resentation of what has actually been observed. The second characterization is ridiculous, because a statistically insignificant difference in mean values is as much " statistical " as a statistically significant one. Observed data should be described correctly: for example, " the two groups had similar mean values " or " differed in mean value, but the difference was not statistically significant ". Similar confusion is common for statistically significant findings. Statistical significance can, for example, be considered to be a token of practical significance or importance. However, uncertainty and practical significance are two fundamentally different things. Practical significance cannot be derived from a p-value. Note that the p-value is not an effect measure; a lower p-value does not necessarily correspond to a greater difference. The same difference in revision rates between two specific knee arthroplasties is, for example, likely to get a much lower p-value when tested in the large Swedish Knee Arthro-plasty Register than in …