A review of postmenopausal hormone therapy recommendations: potential for selection bias.

Objective: To survey the advice given to physicians in regard to postmenopausal hormone therapy, and thus to find out indirectly whether selective prescribing may have contributed to the findings of health effects.

Data Sources: Surveys on the health benefits of postmenopausal therapy were identified from previous reviews. Experts in various fields were consulted regarding relevant textbooks, and recommended booklists were used.

Methods Of Study Selection: Most of the surveys were conducted in the United States and concerned only estrogens; therefore, we focused on these data. Advice given to United States physicians on estrogen prescribing was abstracted from the Physicians' Desk Reference, five textbooks, and other sources written between 1945-1990.

Data Extraction And Synthesis: In advice given to physicians, there was concern that estrogens could cause cancer. In addition, it was not recommended to give estrogens to women in poor general health. Advice concerning cardiovascular diseases and estrogens was contradictory and variable.

Conclusions: If the advice observed reflects the prescribing practices of physicians, then surveys on the health impact of estrogen therapy may have underestimated the risk of breast cancer and overestimated the prevention of fractures. The impact on studies of cardiovascular diseases is less clear, and other mechanisms of selection bias may have been more important.