“Has Evidence Based Medicine corrupted our understanding of causation?”

Professor James Isbister


This presentation will take an iconoclastic approach challenging many aspects of the current practice of EBM. The presenter feels comfortable that Archie Cochrane and Austin Bradford Hill may continue to rest in peace in the knowledge that there are defenders of the faith attempting to recapture EBM from fundamentalist attack!

To understand the underlying pathophysiology of disease, medical diagnosis and benefits/risks of therapy it is important to first define “causation” as a scientific term. Examples from transfusion medicine will be used to illustrate causation arguments. The general principles set forth in Koch’s postulates originally formulated by Robert Koch and Friedrich Loeffler in 1884 continue to provide medical science with a basis for determining the aetiology of many diseases, especially “simple diseases” for which necessary and sufficient conditions can be established. Although Koch’s postulates are of central importance in microbiology, they have limitations, some of which were acknowledged by Koch himself. The limitations have been leading to a redefinition of Koch’s postulates to make there more relevant to current understanding of infectious disease, but also applying the basic concepts to non-infectious disease. Establishing cause and effect is more problematic in the case of “complex diseases”. In statistical language, causation is expressed as a probability. Necessary and/or sufficient conditions are usually absent, leaving only contributing factors.

The randomised controlled trial (RCT) is considered the gold standard in evidence-based medicine (EBM), applying statistical probabilities to arrive at “the truth.” This current view of EBM often fails to acknowledge that most important advances in medicine usually start with clinical observations and the RCT is but one tool for acquiring knowledge about causation. The EBM classifications of levels of evidence, which invariably give the highest quality ranking to evidence derived from RCTs, is based in probabilistic causation. As a result, it offers only a limited perspective of evidence of causation and efficacy for many of modern medicine’s understanding of disease and therapeutic interventions. Indeed, cynics of EBM, as practiced today, accuse epidemiologists and statisticians, especially those from the frequentist school, that they have “stolen” and redefined the word causation. The word causation is not, and never was, a word in the statistical lexicon.