Full Program »
Evidence and Mechanistic Reasoning: A Matter of Inductive RiskEvidence-Based Medicine is not consistent in its treatment of causal knowledge. On the one hand, various portions of “the gold standard” method for EBM seem to require causal knowledge. For example, it is difficult to avoid making use of causal knowledge in hypothesis creation, the selection of a relevant placebo and in the application of data from randomized controlled trials to individual patients.
On the other hand, EBM has a history of rejecting the kind of causal knowledge derived from mechanistic reasoning (also called pathophysiological rationale) as dangerous. In the “debut” article of the EBM movement, the Evidence Based Working Group writes, “Evidence-based medicine de-emphasizes intuition, unsystematic clinical experience, and pathophysiologic rationale [another term for mechanistic reasoning] as sufficient grounds for clinical decision making and stresses the examination of evidence from clinical research.” The reasoning for this exclusion of knowledge based on mechanistic reasoning largely rests on a series of standard examples from the history of medicine in which insufficient knowledge about mechanisms led to poor, sometimes horrific, patient outcomes. One such example is the Cardiac Arrhythmia Suppression Trial (CAST) which discovered that the antiarrhythmic drugs prescribed to reduce mortality in patients who had had a myocardial infarction based on causal assumptions about the role of cardiac arrhythmia in MIs were actually harming rather than helping patients, contributing to thousands of deaths. Other examples include bloodletting to treat a wide array of illnesses, and back sleeping for infants to prevent SIDS.
Critics of EBM claim that these historical examples of failed mechanistic reasoning do not indict mechanistic reasoning as a whole. They only provide cases in which the relevant mechanisms were not properly understood. The problem, they claim, is not that mechanistic reasoning was used, but that it was not used correctly or taken far enough. Under this argument, there is no strong difference between the cases of causal reasoning EBM accepts and those it rejects. If EBM is interested both in internal consistency and continuing to use causal knowledge in hypothesis creation, placebo selection, individual patient application and in other situations, EBM ought to accept mechanistic reasoning as just as legitimate as the other causal knowledge it helps itself to.
I am sympathetic to this claim, but argue that it does not capture what EBM supporters are implicitly worried about when talking about causal knowledge: the amount of risk of error (because of confounders) that is acceptable in causal stories differs from situation to situation. I argue that it is this tolerance of inductive risk which separates the list of situations in which supporters of EBM are perfectly happy to make use of causal knowledge from the situations in which use of causal knowledge is considered dangerous.
University of Durham