Professor Evans, you're an expert in pharmacovigilance - continuous surveillance of adverse drug effects after the medication is approved, when doctors start prescribing it to patients. What are the best practices to detect adverse drug reactions to medicines that are already on the market? And what are the common problems today with adverse drug effects surveillance? To me, there are two things that we can do in large measure. The first is that assuming that all the trials that can be funded have been carried out, we need to do better surveillance in large electronic health record databases. We need to make sure that the data in those are able to be used for research and is having continuous monitoring. Particularly of new drugs, but also of all drugs, so that where harms are occurring, we can detect them. And in some areas, we do not do that very well, based on records. The second thing that we can do is make it easier for physicians to report suspected adverse reactions. Human beings, whether patients or health professionals, are very good at being suspicious that medicine has caused an adverse effect. Sometimes, indeed, a lot of the time, that suspicion is incorrect.

Nevertheless, having that suspicion is worthwhile. And making the process of reporting that to the FDA in the US, or regulatory authorities around the world, ought to be Be a minimal effort for the physician or health professional who sees that particular effect. In the UK, and indeed in Europe, we have been allowing patients to report such things themselves. But they often report the trivial ones, which can be important to them and may be important for regulators to know. But we really need very good reporting from physicians. And making that process as simple as possible is, I think, the thing we can do best, and I have done both of those things, we have analysts who are able to analyze the data that comes in, and we can use perhaps very big data to use the buzzword in the era of computing Big Data approaches to electronic health records to be able to detect potential adverse effects. The problem is, of course, that bit like diagnostic testing for size golf too, you will have false positives and false negatives. You will find harms where there isn't really any; and you will fail to find harms where there are some, but doing all of this better is possible.