Nature recently published an excellent list of 20 things policymakers should understand about interpreting scientific claims, by William J Sutherland, David Spiegelhalter and Mark A Burgman, academics from Cambridge and Melbourne. Included are reminders that “scientists are human” and that “correlation does not imply causation”, as well as practical examples explaining why “regression to the mean can mislead” or how to “beware the base rate fallacy”. Continue reading
To assess the evidence, you need to know it is there.
Properly conducted clinical trials provide the best evidence for whether drugs work and are safe. But about half have never been published, and trials with positive results – where the drug concerned is shown to be safe and effective – are more likely to be published than negative ones. Patients can be harmed – for example if a treatment found to be harmful in an invisible trial is then given to patients in a new one – or medicines used ineffectively or wastefully, as a result. Researchers can’t plan research properly because they don’t know what has gone before. The problems with this situation have been well documented for years, but things may now be going to change.