Common wrong beliefs about statistical testing: recent trends in biomedical sciences

Purpose of review Hypothesis testing is central to epidemiological research, yet its dominant implementation – null hypothesis significance testing (NHST) – is both conceptually flawed and prone to serious misinterpretation. NHST promotes a dichotomous framework that often obscures the real-world implications of biomedical findings and encourages ritualistic decision rules in medicine. This review aims to identify and address persistent misconceptions in statistical testing that distort evidence interpretation, undermining decision-making in public health and clinical research. Recent findings The most frequently reported misconceptions in recent literature include the conflation of statistical with clinical significance, the misuse of P-values and confidence intervals as tools for making dichotomous-finalistic statements, and the flawed application of power calculations and multiple testing corrections. Crucially, NHST fosters decision-making that is detached from the actual stakes of public health practice, disregarding the context-specific costs, risks, and benefits that demand critical reasoning and tailored analysis. Summary To improve scientific inference and support sound public health decision-making, this review advocates adopting statistical compatibility, defining effect relevance based on prior knowledge about the cost-benefit ratio rather than convention, and prioritizing precision over purely mathematical exercises aimed at power estimation. Moreover, it also warns against new forms of oversimplification, including equivalence testing and novel forms of dichotomization, which risk perpetuating the same conceptual errors and may reflect a deeper inclination toward cognitive ease over interpretive complexity. In this regard, a socio-cultural shift is needed, away from automatic procedures and toward thoughtful, context-aware interpretations of evidence.