No, you rake together all the data you can grab on how the drug's used in the real world and look for adverse events

Yes, but I was mainly thinking about the trials before it goes into general use when I was talking about testing for interactions with other drugs. At that stage, you are primarily concentrating on a (relatively) small set of other drugs where you suspect possible interactions surely? Once it goes into wider use, yes, you are gathering what real world data you can. Which is exactly why the NHS England data is so valuable.

Thank you for your comments so far. It is always useful to hear from someone with more experience of a subject than I have!