Grodstein (1) suggests that the useful questions raised by our paper (2) are how and to what extent the differences in cardiovascular health of women who do and do not use postmenopausal estrogen replacement therapy (ERT) affect the interpretation of observational studies on the relation between ERT use and risk for cardiovascular disease morbidity and mortality. She argues that the relation is likely to be causal, but we are not ready to accept that interpretation without further clinical trial data that would document the effect of long-term use of ERT according to the women's characteristics. Although it is true that many observational studies show the benefit of ERT on risk for cardiovascular disease, they also show that the benefit of ERT is apparent for diseases not thought to be related to estrogen exposure. For example, among women enrolled in the Study of Osteoporotic Fractures, longterm current ERT users had a lower risk of all-cause mortality after removal of death due to breast, endometrial, and colon cancer, hip fractures, stroke, and atherosclerotic heart disease (3). Similarly, Sturgeon et al. reported lowered relative risk among current users (defined as use within 1.9 years or less) for almost all causes of death, including cancers (4). Grodstein suggests that many physicians stop ERT when women become ill with cancer, leaving healthy women remaining as current users. However, this explanation is not easily applied to mortality from injuries, which is also reduced among ERT users (4). The ubiquitous benefit of ERT suggests that selection factors for ERT use are important to consider in interpreting ERT-heart disease associations.