Both drugs are associated with the same decreases in mean blood pressure in both scenarios. Given that we attach the same meaning to any blood pressure decrease of the same amount, does scenario B constitute equivalent evidence for the superiority of drug B as does scenario A?
Or might regression to the mean be an explanation for scenario B but not for scenario A, such that scenario B convinces us less of drug Bs effectiveness?
If Group B in the right-hand scenario is regressing to the mean, then that means they started out high not due to meaningful, stable characteristics but due to being selected based on random fluctuation. This selection bias would compromise the validity of the right-hand trial. But if you plotted individual changes, regression to the mean likely would look different from a systematic treatment-based change; the former would I think have a wider spread (would be less consistent). So I'd examine these individual changes. In fact it's almost always best to complement a plot of mean change with one of individual change; you can detect much more about the nature of any effects.
