FWIW, in the Medium app at least (on a mobile device this AM) the link to the phenol explorer database seems not to work. (it may or may not in the web version). Simple enough to search for it once I knew something like that existed. http://phenol-explorer.eu
That mentioned, a few decades ago I worked with an earlier version of the biobank. Adjusting for smoking habits would have been fairly simple - but the coffee consumption data is still self-reported nutritional data - which I'd consider only accurate enough to maybe form a hypothesis. But it's what we currently have. Maybe if those GWAS studies found traits that do something similar to what those components of coffee do, we'd be able to use genetic randomization*. For instance, I'm a 'fast' caffeine metabolizer and do drink coffee (AM only, typically 300-450ml). I'd be willing to trust a nutritional questionaire that asked if I drank *any* coffee. So, are fast metabolizers who drink coffee more or less susceptible to dermentia? Gotta wonder if there are other genetic traits that affect, say NRF2...(?)

*if you haven't covered genetic randomization, might I suggest it?
Thanks for writing this!