Estrogen replacement therapy and breast cancer risk.

Assessment of the effects of menopausal estrogens on the risk of breast cancer is obviously complex, with proper evaluations requiring attention to the effects of selection, recall, and surveillance biases; confounding factors; detailed exposure relations; and subgroup variations. Although much discussion has ensued over the extent to which variations between studies might be due to a variety of methodological differences, it is likely that the limited statistical power of many studies may be largely responsible for the divergent effects observed, particularly with regard to assessing the effect of long-term usage. Thus, in case-control studies in which population controls have been used and for which there have been a sufficient number of women using estrogens for approximately 10 or more years, it appears as though such an exposure may be associated with some elevated risk, possibly on the order of 30-80 percent (16, 34, 36-38)--a magnitude of risk that is obviously extremely difficult to detect in most epidemiologic studies. Evaluation of risks associated with short-term use is even more difficult, which has led most investigators to conclude that usage of less than 5 years is associated with no detectable excess risk. Problems with statistical power also plague evaluations of more detailed measures of use, including dosage and type of preparations, although there is some indication that nonconjugated preparations may be associated with a higher risk than other types of estrogens (16, 20, 36, 44, 51). In addition, there is some evidence that recent use may be associated with a modest elevation in risk (32, 33, 53), although effects of surveillance bias are difficult to resolve. Interpretation of the effect is complicated by several studies which have found estrogens to be associated with an increased risk of breast cancer incidence but a decreased risk of death, with the mortality ratios ranging from approximately 0.5 to 0.8 (33, 92, 93). Although one of these studies hypothesized that the relation with mortality may reflect a tendency for women with clinically recognized breast cancers and those at high risk not to be prescribed estrogens for long periods of time, studies showing that estrogen users have more favorable tumor characteristics at diagnosis than nonusers (20, 36) suggest that screening biases may be involved. The numerous contradictions regarding effects of menopausal estrogens from observational studies have heightened interest in the results of several ongoing controlled clinical trials.(ABSTRACT TRUNCATED AT 400 WORDS)