[Roaming through methodology: XXIX. P]

Abstract

The abundant use of p-values and statistical significance in medical literature often leads to invalid conclusions about the reported study results. A statistically significant result may be clinically irrelevant, while a clinically relevant effect may be ignored due to lack of statistical significance. These false conclusions can be attributed to the fact that p-values are often considered as a measure of relevance, validity and precision of a study outcome. A p-value, however, only conveys information about precision. In addition, a more informative means to describe the precision of a study exists, viz. confidence intervals. It is therefore important that both the editors of medical journals and the authors of scientific papers continue to use confidence intervals instead of p-values.