It's refreshing to meet someone as cynical as myself obsgraham. I'd agree there is never "the final answer," provided by research, and of course, we can never "prove" or "disprove" our interventions absolutely, but I think where we have moved on, is that we are using more robust evidence, relying on realistic outcome measures, (such as the Grade 1 RCT's in statin intervention), and are less likely to rely on studies of poor design, using proxy outcome measures (which don't translate to clinical benefit), or even worse "expert opinion."
A meta analysis or systematic review shouldn't automatically refute the past 25 studies, it should add collective weight to the accuracy (or innacuracy) of the previous studies.
The RCGP stuff I cited does rely on high quality studies.
The decision to offer statins to those of CVD risk of 20% or above (in 10yrs), was made by a committee- I suspect this was made on the basis of cost, rather than clinical effectiveness, and I accept,the method we use to establish the risk calculation is far from perfect. (Although better than what we had before.)