We describe a trial design of postmenopausal hormone therapy. Our goal was to design a trial that allows more patient management than usual, would include placebo effect in the therapy, would imitate the normal practices of health care as much as possible, and would be based on ordinary health services. We gave women a randomized recommendation to use or not use hormone therapy, invited them to two discussion groups, and sent them two questionnaires. They were asked to consult their own physician and to pay for the therapy themselves. The design worked well regarding recruitment, but compliance after 6 months was not satisfactory. The main problems were the women's aversion to randomization, the attitudes of the women's own physicians, and the difficulty in stopping the use of hormone therapy. The last-mentioned problem could be avoided by different inclusion criteria. Unless societal perception of research can be changed, most solutions to improve compliance would lead toward traditional trial designs.