Link. We desperately need observational studies but the RCT critique is vintage. I’d hoped for more on models and Bayesian approaches to confidence.