The probability of probability and research truths

Emerg Med Australas. 2017 Apr;29(2):242-244. doi: 10.1111/1742-6723.12740. Epub 2017 Feb 15.

Abstract

The foundation of much medical research rests on the statistical significance of the P-value, but we have fallen prey to the seductive certainty of significance. Other scientific disciplines work to a different standard. This may partly explain why medical reversal is an increasing phenomenon, whereby new studies (based on the 0.05 standard) overturn previous significant findings. This has generated a crisis in the rigour of evidence-based medicine, as many people erroneously believe that a P < 0.05 means the treatment effect is clinically important. However, statistics are not facts about the world. Nor should they be based on an arbitrary threshold that arose for historical reasons. This arbitrary threshold encourages an unthinking automatic response that contributes to industry's influence on medical research. Examples from emergency medicine practice illustrate these themes. Study replication needs to be valued as much as discovery. Careful and thoughtful unbiased thinking about the results we do have is undervalued.

Keywords: P-value; medical reversal; probability; replication; statistical significance.

MeSH terms

  • Biomedical Research / methods
  • Biomedical Research / standards*
  • Biomedical Research / trends
  • Evidence-Based Medicine / standards
  • Evidence-Based Medicine / statistics & numerical data
  • Humans
  • Probability*
  • Research Design / standards
  • Research Design / statistics & numerical data