A very good point. These studies should be comparing QALYs (quality-adjusted life years, a measure of disease burden) instead of relative prevalence of a handful of negative outcomes, the latter of which is much more vulnerable to p-hacking.