Were all clinically important outcomes considered?
This is very important. Many studies only look at "surrogate endpoints".
For example, a study may only examine effects of a drug on cholesterol
level without reporting death rates, only to find that a drug that seems
to lower cholesterol effectively results in a significant risk of death.
Application to Pollock et al's
thyroxine paper
In this study, several different endpoints were considered. Death would
obviously be a significant endpoint but one must assume there were no deaths
during the study period. The subjective measures of physical and psychological
well-being are certainly valid and would be valued by patients.