Abstract
Landmark results from the Women's Health Initiative trial showed that random assignment to menopausal hormone therapy (MHT) elevated risks of breast cancer and other adverse events. Recent analyses argue that MHT risks are small. These analyses report intention-to-treat (ITT) effects, ignoring the fact that many women assigned intervention were non-adherent, while many women assigned control initiated treatment. Instrumental variable (IV) methods and adherence data allow us to estimate effects of MHT on compliers who took MHT if and only if assigned. IV estimates show risks and benefits that are substantially larger than the ITT estimates used to inform MHT guidelines.