If science-based medicine reflects the application of the best evidence, then we should expect practices to change when new data emerges. In the long run that’s generally true, and the progressive gains we’ve seen in the management of disease reflect this. But in the short run, change can be maddeningly slow, and there are many areas of medicine where we could be doing a better job of applying what we already know to improve outcomes and reduce harms. One area where this is obvious is drug treatments, which can provide remarkable benefits but are also sources of significant harms.
Somewhat problematically, the real world is often the setting where the full extent of harms from treatments are identified. Bringing new drugs to market means tradeoffs: Do you demand larger and longer clinical trials to get as much information as possible about a drug before it’s sold? Or do you approve based on more preliminary, potentially weaker evidence, to meet (potentially) important patient need? There is no set formula or right answer to this questions – it’s ultimately a value judgement exercised by regulators like the FDA, who decide which drugs are allowed for sale (the benefits are assumed, overall, to exceed the harms) or removed for sale (when the opposite is felt to be the case).