A study in a nuclear medicine journal investigated how often a statistically significant cutoff between normal and abnormal values would misclassify healthy patients. Analyzing 32 test results, the typical cutoff fell just 0.66 SDs from the mean. At that point, one SD encompasses 34% of normally distributed data on both sides of the mean. Consequently, 34% of healthy patients fall beyond 0.66 SDs and would be erroneously deemed abnormal. The authors conclude statistical significance frequently fails to ensure clinical usefulness on an individual level.
Citation: Heston TF, Wahl RL. How often are statistically significant results clinically relevant? Not often. J Nucl Med. 2009;50(Suppl 2):1370. DOI: 10.5281/zenodo.10198933
