The current pyramid of evidence for interventions
This is what the actual pyramid of evidence should look like: its a lot more complex.
Just because something is a "meta-analysis" doesn't mean it suddenly trumps every other study. You can have meta-analyses of anything... meta-analyses of worm studies, meta-analysis of majorly flawed observational studies, meta-analyses of mixes of RCTs and observational studies... or totally irrelevant MA PICOTS.
There are three main problems with meta-analyses (except irrelevant PICOTS).
1) Inclusion of majorly flawed studies makes the meta analysis useless
2) Inexplicable variation in results in the studies with studies crossing the lines of no or opposite effect in the forest plot (measured by I^2 over 50% typically).. and no a-priori heterogeneity plan
3) Not at least discussing using peto and fixed effects in sensitivity testing with your statistician... as results can change from MH random effects, and often MH or random can be inappropriate
Similarly just as something is a clinical guideline doesn't mean its good, even AGREE-II doesn't ensure a clinical guideline isn't flawed.
Clinical guidelines can be majorly flawed in many ways, such as being out of date, poor search strategy, mixing expert opinion with high quality evidence, not making it clear on which statements are backed by which evidence in which PICOTS etc.