So at what point do you have enough information to do something.
Dan Strickland
11

Well it’s certainly possible to infer causality from large observational trials, but the point is that you can’t do it from a single trial and even then you’d ideally lead with a host of caveats. For example, the correlation between diet drink consumption and obesity has remained strong in virtually every large observational trial done, but the few RCTs on the subject have found no such effect.

I’d say that if you’ve got a number of large observational trials that all point in the same direction, and have been controlled for the usual suspects (BMI, income, etc), you can say with a bit of confidence that something is likely causal if you’ve got a plausible biological pathway. The problem is that most of these news stories spring up about a single study that didn’t control for something obvious like BMI and propose an association that goes against established literature.