Misunderstood as usual

Randomised trials and other clinical studies are performed to learn about the effects of a drug or treatment among all patients, not least future ones and not just the ones included in the study. Generalisation from a small sample to an infinite population can, however, not be made without uncertainty, and the magnitude of this uncertainty is often presented in terms of p-values. These values depend on sample size and variability and have, in themselves, nothing to do with clinical relevance or scientific importance. Statistically significant findings are therefore not necessarily clinically important. The evidence for a statistically significant finding's clinical importance remains to be shown, and the interpretation of statistical nonsignificance as evidence of "no difference" is a common mistake. Furthermore, comparing the statistical significance of a factor studied by several researchers, some finding significance, others not, is not meaningful without considerations regarding the effects of sample size and heterogeneity of the studied patients.

More from Where the rubber meets the road
All posts