We appreciate the opportunity to again discuss the limitations of the observational literature evaluating hormone therapy and CHD in women. Before the publication of the Heart and Estrogen/progestin Replacement Study in 1998, postmenopausal hormone therapy was widely recommended to women to prevent CHD, largely on the basis of data from observational studies. However, ascertainment of hormone therapy use in these studies was limited, as we discussed and described in Appendix Tables 1 and 2 of our paper. Very few studies evaluated the type of estrogen used, whether progestins were used, or the dose of either formulation. Despite limitations, these data were used to make widespread recommendations to women about the likelihood of CHD protection associated with hormone therapy. Our meta-analysis was an attempt to evaluate the observational studies judged most likely to provide valid findings and to determine whether, in the aggregate, the data were sufficient and of high enough quality to support this widespread recommendation. Our findings suggested that they were not. We agree with Drs. Meyerson and Chausmer that even among the best studies, very few data are available on the type of estrogen used and that generalizations about all estrogen or progestin formulations are inappropriate. However, we strongly believe that any recommendations of hormone therapy in women for the prevention of CHD should be made only on the basis of randomized, controlled trials.

The concerns of Drs. Grodstein, Manson, and Stampfer are addressed in another publication (1). The important points of that response are as follows. First, the level of education sufficient to become a registered nurse can vary from nursing school diploma to bachelor's degree to advanced degree, a range that is very likely to be associated with important variation in socioeconomic status. Thus, we did not include the Nurses' Health Study in our original meta-analysis of the articles that accounted for socioeconomic status. We would have done so if we had assumed that including only registered nurses is equivalent to statistical adjustment for socioeconomic status. When we did include the Nurses' Health Study, the summary figures changed only minimally, as described in more detail elsewhere (1).

Second, when we included studies that adjusted for either alcohol use or physical activity in our meta-analysis (including the Nurses' Health Study), the pooled estimates did not suggest CHD benefit. The pooled estimates are reported elsewhere (1). Third, animal data and intermediate outcomes can suggest hypotheses or potential biological mechanisms of action, but only randomized, controlled trials of estrogens with or without progestins can prove that hormone therapy is cardioprotective.