It is problematic to apply average study results to target populations. For example in the European Atrial Fibrillation Trial (EAFT) the risk of intracranial hemorrhage in patients with atrial fibrillation taking warfarin was minimal. However, when warfarin was given to patients in sinus rhythm it increased the risk of hemorrhage. In another example carotid endarterectomy appeared in one study to increase overall mortality by 0.5%. But in a subgroup analysis of patients with severe carotid stenosis the procedure had a clear benefit. This example illustrates why the terms ‘external validity’ and ‘generalizability’ can be misleading: the problem of extrapolation also arises when applying average study results to individuals or subpopulations within the study (particularizing). One strategy for dealing with the problem of extrapolation is simple induction. The CONSORT checklist includes an item about ‘generalizability’ that involves asking whether target populations would have met the inclusion criteria for a trial. But even if a patient would not have met the inclusion criteria it is nevertheless sometimes safe to implement the study results, while in other cases although the patient would have met the inclusion criteria it is unsafe to do so. Critics and proponents of Evidence-Based Medicine (EBM) suggest that knowledge of pathophysiological mechanisms can solve the problem of extrapolation. However, no explicit guidelines for implementing this suggestion are provided. In this programme of research I study what mechanisms are, how to characterize them, and how they can be used to support efficacy and extrapolation.
The work on mechanisms has influenced the types of evidence accepted as relevant for the purpose of making healthcare decisions.