[bjornsing]

Not saying they are wrong, but my feeling is that selection effects probably dwarf any causal effects here. You don’t need much skew in the health distribution of people who are drawn to an intervention like this to get this kind of correlation and effect size.