Current issues of ACP Journal Club are published in Annals of Internal Medicine


Meta-analysis: Estrogen replacement therapy increases the risk for breast cancer

ACP J Club. 1991 July-Aug;115:18. doi:10.7326/ACPJC-1991-115-1-018

Source Citation

Steinberg KK, Thacker SB, Smith SJ, et al. A meta-analysis of the effect of estrogen replacement therapy on the risk of breast cancer. JAMA. 1991 Apr 17;265:1985-90.



To investigate the effect of duration of estrogen replacement therapy on risk for breast cancer.

Data sources

MEDLINE, CANCERLIT, Current Contents, and EXCERPTA MEDICA databases from 1966 through September 1989 were used to identify published material on noncontraceptive estrogen use and the risk for breast cancer.

Study selection

Of 240 citations, studies were selected if the effect of estrogen replacement therapy on the risk for breast cancer was determined in women who had natural menopause or who had a premenopausal hysterectomy. 34 studied the relation between noncontraceptive, menopausal estrogen use and breast cancer. 25 studies excluded women with pre-existing cancer and presented comparable data.

Data extraction

Studies (with identifiers masked) were assigned quality scores of 0 to 100 by 3 independent raters, based on sampling, data-collection, and follow-up methods; histologic confirmation of diagnosis; quantification of estrogen dosage; and adjustment for confounding risk factors. Data were pooled for the 16 case-control studies that described duration of estrogen use. Proportional increase in risk for breast cancer per year of use was calculated.

Main results

Quality scores ranged from 38 to 83 (median 64) for 10 case-control studies that found increased risk for cancer with increasing duration. For 9 studies that found no increased risk, quality scores were 12 to 62 (median 25). 2 follow-up studies found increased risk (scores 48 and 82), and 5 found no increased risk (range 13 to 72, median 57). 15 case-control studies (including women with all types of menopause) found mean proportional risk for breast cancer increased significantly for each year of estrogen use (0.018, 95% CI 0.010 to 0.025); after 15 years, the relative risk (RR) was 1.3 (CI 1.2 to 1.6). There was, however, no effect of use of estrogen for < 5 years. 10 studies of postmenopausal women gave a lower increase in risk with duration (0.015) than 4 studies that included premenopausal women (0.055). Estrogen use increased risk for women who had had oophorectomy (0.028) and for those who had experienced natural menopause (0.019). Increased risk for breast cancer in women with a family history of breast cancer was augmented in those who had taken estrogen (RR 3.4, CI 2.0 to 6.0) compared with women who had not (RR 1.5 CI 1.2 to 1.7).


Meta-analysis of case-control studies indicate an increasing risk for breast cancer with > 5 years duration of estrogen replacement therapy. This effect was most evident among the studies that used sounder methods.

Source of funding: Not stated.

Address for article reprint: Dr. K.K. Steinberg, Center for Environmental Health and Injury Control, Centers for Disease Control, MS F-18, 1600 Clifton Road, Atlanta, GA 30333, USA.


For all known carcinogens, there is a latent interval between the start of exposure and the appearance of tumors. This interval may be several years in length. Several risk factors for breast cancer suggest that ovarian hormones are involved in the cause of the disease; however, it has not been convincingly shown that exogenous estrogens increase risk for breast cancer. Individual studies may lack the power to show an association if there are small numbers of cases with a duration of use greater than the latent interval.

The study by Steinberg and colleagues combines the results of 16 case-control studies in a meta-analysis and suggests no increase in risk for breast cancer when estrogens are used for > 5 years; but after 5 years, there is a small linear increase in risk with increasing duration of use. The study was carefully done, and although the authors have had to make some assumptions, these appear to be reasonable.

The study result may be influenced, as the authors note, by publication bias. Further, because it involves only case-control studies, differential recall between cases and controls about exposure to estrogen in the past could bias the results. Large observational studies, in which estrogen use is determined in an inception cohort that is then followed and observed for the development of breast cancer, can be expected to provide better evidence concerning the association of estrogens with cancer risk. 2 studies included in this article, but not in the meta-analysis, do show an increased risk with long duration of use. 4 other cohort studies do not show this association, but may have failed to do so because of too few cases with long exposure.

Norman F. Boyd, MD
The Ontario Cancer InstituteToronto, Ontario, Canada