- Email: [email protected]
Risk of endometrial cancer in relation to use of oestrogen combined with cyclic progestagen therapy in postmenopausal women
THE LANCET
Risk of endometrial cancer in relation to use of oestrogen combined with cyclic progestagen therapy in postmenopausal women Shirley A A Beresford, Noel S Weiss, Lynda F Voigt, Barbara McKnight
Summary Background Postmenopausal oestrogen therapy reduces the risk of osteoporosis and cardiovascular diseases but is associated with an increased risk of endometrial cancer. We have assessed the impact of a regimen of oestrogen with cyclic progestagen on risk of endometrial cancer for postmenopausal women. Methods We did a population-based case-control study of women aged 45–74 years in western Washington State, USA. Cases were identified from a regional cancer registry as having histologically confirmed endometrial cancer during 1985–91. 832 (72%) of 1154 eligible cases completed interviews. Controls were identified by random digit dialling, screened for intact uterus, frequency matched for age and county, and randomly assigned a reference date within 1985–91. Interviews with 1114 (73%) of 1526 eligible controls were done. The women provided information about use of hormone replacement therapy, and reproductive and medical history before diagnosis date (cases) or reference date (controls). Findings Relative to women who had never used hormones (for >6 months), women who had taken unopposed oestrogen had a four-fold increase (95% CI 3·1–5·1) in risk of endometrial cancer. Women who used a combined therapy of oestrogen with cyclic progestagen (eg, medroxyprogesterone acetate) had a relative risk of 1·4 (1·0–1·9). Among women with fewer than 10 days of added progestagen per month, the relative risk was 3·1 (1·7–5·7), whereas that for women with 10–21 days of added progestagen was 1·3 (0·8–2·2). The use of these combined regimens for 5 or more years was associated with risks of 3·7 (1·7–8·2) and 2·5 (1·1–5·5), respectively, relative to non-users of hormones. Interpretation Postmenopausal women who use combined therapy of oestrogen with cyclic progestagen on a longterm basis have an increased risk of endometrial cancer compared with those who are not on hormone replacement, even when progestagen is added for 10 or more days per month. This increase is much smaller than that associated with unopposed oestrogen, but needs to be confirmed.
Lancet 1997; 349: 458–61
Departments of Epidemiology (Prof S A A Beresford PhD, N S Weiss MD, L F Voigt PhD) and Biostatistics (B McKnight PhD), University of Washington, USA, and Programs in Epidemiology (S A A Beresford, N S Weiss, L F Voigt) and Biostatistics (B McKnight), Fred Hutchinson Cancer Research Center, Seattle, Washington Correspondence to: Prof Shirley A A Beresford, Department of Epidemiology, Box 357236, University of Washington, Seattle, WA 98195–7236, USA
458
Introduction Postmenopausal oestrogen therapy can benefit women in both the short term by reducing the frequency of hot flushes, and in the long term by reducing the risk of osteoporosis and cardiovascular disease.1 Although use of oestrogen alone is associated with an increased risk of endometrial cancer, addition of a progestagen reduces the incidence of endometrial hyperplasia2–4 and endometrial cancer.5–8 The addition of progestagen does not seem to interfere with the beneficial effects of oestrogen therapy,9,10 although the risk of cardiovascular disease may vary according to type of progestagen.4,11 In combination therapies, progestagen is commonly taken cyclically—oestrogen is taken for all or most days each month, but the progestagen for only part of the month. An alternative that has become popular is continuous combined therapy, in which both oestrogen and progestagen are taken every day of the month. Questions about the risk of endometrial cancer associated with oestrogen combined with cyclic progestagen therapy, compared with unopposed oestrogen therapy, have not yet been adequately answered. For example, how many days per month must progestagen be added to oestrogen therapy to achieve the reduced risk? There was a suggestion from Voigt and colleagues6 that 10 or more days of progestagen are needed, but the power of this study was limited. How safe is the long-term use of oestrogens combined with progestagens with respect to endometrial cancer risk? In our study we aimed to find out more about the relation between the risk of developing endometrial cancer and cyclic progestagen therapy.
Methods We identified female residents of King, Pierce, and Snohomish counties of Washington State, USA, aged 45–74 years, who developed a histologically confirmed endometrial carcinoma during 1985–91, from the Cancer Surveillance System, a population-based cancer registry that has served western Washington since 1974. For 1985 and 1986, recruitment was limited to King County, and to an upper age of 64 years. For 1991, recruitment was limited to an upper age of 69 years. During 1986–93, controls were identified by random-digitdialling telephone calls12 within the three counties. A short questionnaire was used to identify women aged 45–74 years as potential controls. If more than one eligible woman lived at the address, one was selected randomly. Controls were chosen with stratified sampling techniques, in such a way that the age and county distribution of controls was broadly similar to that of cases. Controls were randomly assigned a reference date, the distribution of which corresponded roughly to the date of diagnosis of the cases. The study was approved by the Institutional Review Board of the University of Washington, USA. All eligible participants were interviewed in person, except for 34 (3%) of the cases and 39 (5%) of the controls who were interviewed by telephone. On average there was a shorter interval between reference year and interview date for cases than for controls (median 21 months for cases and 26 months for controls). Adjustment for this difference did not change the results. At interview, the participants were asked about whether
Vol 349 • February 15, 1997
THE LANCET
Number of cases (394)
Number of controls (788)
Age (years) 45–49 50–54 55–59 60–64 65–69 70–74
65 (17%) 70 (18%) 78 (20%) 88 (22%) 63 (16%) 30 (8%)
122 (16%) 151 (19%) 172 (22%) 157 (20%) 118 (15%) 68 (9%)
County King Pierce Snohomish
249 (63%) 88 (22%) 57 (15%)
572 (73%) 123 (16%) 93 (12%)
Reference or diagnosis year 1985–87 1988–89 1990–91
112 (28%) 134 (34%) 148 (38%)
187 (24%) 370 (47%) 231 (29%)
Parity (livebirths plus stillbirths) 0 1 2 3 Ä4
50 (13%) 49 (12%) 99 (25%) 91 (23%) 105 (27%)
79 (10%) 78 (10%) 208 (26%) 199 (25%) 224 (28%)
Oral contraceptive use Never used 1 month–5 years Ä5 years
256 (65%) 93 (24%) 45 (11%)
459 (58%) 183 (23%) 146 (19%)
Ethnic origin White or hispanic Non-white
378 (96%) 16 (4%)
749 (95%) 39 (5%)
Marital status Never married Married or living as married Separated or divorced Widowed
18 (5%) 284 (72%) 48 (12%) 44 (11%)
25 (3%) 543 (69%) 123 (16%) 97 (12%)
Annual household income <$15 000 $15 000–$30 000 $30 000–$45 000 >$45 000 Refused or unknown
65 (17%) 125 (32%) 83 (21%) 107 (27%) 14 (4%)
150 (19%) 233 (30%) 180 (23%) 213 (27%) 12 (2%)
Education High school or less Some college or technical school College graduate or more
197 (50%) 109 (28%) 88 (22%)
348 (44%) 259 (33%) 181 (23%)
Body-mass index (quartiles of Quetlet’s index)* 13·01–21·49 21·50–23·96 23·97–27·42 27·43–63·21 Refused or unknown
47 (12%) 68 (17%) 68 (17%) 209 (53%) 2 (1%)
195 (25%) 191 (24%) 187 (24%) 212 (27%) 3 (0%)
*1 year before reference date.
Table 1: Demographic and reproductive characteristics in cases and controls using combined therapy or no therapy or not they had had a hysterectomy, about their use of hormone replacement therapy, and about their reproductive and medical history up to the time of diagnosis (cases) or to the reference date (controls). To enhance recall of hormone therapy used, we provided pictures of commonly used preparations. The usefulness of such memory prompts has been reported. 13 Individuals interviewed by telephone received the pictures by post before the interview. 72 of the 1254 potential cases identified did not have an epithelial endometrial cancer, 19 had difficulty in communicating, and nine were not interviewed because they had no telephone or were not living at a private address in one of the three study counties at the time of diagnosis. The remaining 1154 cases were eligible. Of these, complete interviews were obtained for 832 (72·2%). 100 (9%) women died before interview, the physicians of 63 (5%) women refused to allow us to contact their patient, and 158 (14%) cases themselves refused to participate. 454 potential cases had an independent pathology review done by two pathologists who were unaware of the Cancer Surveillance System classification. 402 of these cases were initially classified by the Cancer Surveillance System as invasive
Vol 349 • February 15, 1997
Progestagen use
Number of cases
Number of controls
Odds ratio*
Never used hormones <10 days per month 10–21 days per month
270 (84·4%) 25 (7·8%) 25 (7·8%)
593 (86·8%) 26 (3·8%) 64 (9·4%)
1·0 3·1 (1·7–5·7) 1·3 (0·8–2·2)
Analysis restricted to participants whose experience had not been previously reported. *Adjusted for age (continuous), body-mass index (continuous), and county lived in (three groups).
Table 2: Endometrial cancer in users of oestrogen combined with cyclic-progestagen therapy: influence of number of days per month progestagen is added epithelial endometrial cancer, and 52 had other histological classifications, including carcinoma in situ. Of the former cases, 367 (91%) were confirmed by independent pathology review. To recruit controls, 52 045 random-digit-dialling telephone calls were made, of which 23 527 were found to be residential numbers. There were 2113 numbers for which no answer was obtained; we assumed about 20% of these telephone numbers would have been for residential addresses. 14,15 Household screening was completed for 22 650 of the residential numbers, giving a screening response of 96·3%. 2620 women were found to be eligible for the study by both age and county; 1975 (75·4%) women agreed to be interviewed (overall response for controls was 73·0%). 860 (44%) were immediately excluded because they had had a hysterectomy, leaving 1114 women. Controls were also excluded from our analysis if they had been diagnosed with cancer of the endometrium or had had a hysterectomy more than 6 months before their reference date. Only hormone use that was probably postmenopausal was included in the analysis. Therefore, if any participant used hormones before age 45 years, for regulation of menstruation or for depression, anxiety, or emotional distress, or for premenstrual syndrome. Responses were reviewed independently by two of the investigators who were unaware of case-control status. Hormone therapy was not judged to be postmenopausal-hormone therapy unless there was additional evidence that it was used after menopause. Hormone use for infertility or to prevent miscarriage was also not counted as postmenopausal hormone use. Because use of unopposed oestrogen for less than 6 months has little or no effect on the incidence of endometrial cancer, we focused our analysis on women who used a combined oestrogenprogestagen regimen for at least 6 months. Durations of use of at least 6 months are also associated with better recall of dose and type of therapy.13 Women who had used hormone therapy for less than 6 months were grouped with non-users in our analysis. Because their numbers were so small, we excluded women with a history of continuous combined therapy (six cases, 17 controls) from this analysis. In addition, we excluded combined therapy users who had ever used unopposed oestrogen replacement therapy (76 cases and 74 controls) from most of our analyses because we wanted to assess the effect of oestrogen combined with cyclic progestagen in women whose risk of endometrial cancer was not already increased by use of unopposed oestrogen. Variations in regimens of combined therapy were counted as different regimens only if one had added progestagen for less than 10 days per month and the other had added progestagen for 10 days per month or more. There were four cases and 14 controls who had used more than one regimen of combined therapy. To be able to interpret the findings in a straightforward way, we excluded these 18 women. We classified the participants in four groups: those who had never used hormones for as long as 6 months; those who had used only unopposed oestrogens; those who had used only oestrogen combined with cyclic progestagen; and those who had at some time used unopposed oestrogens and at other times had used combined oestrogen-progestagen therapy.
Statistical analysis The analyses were by logistic regression techniques, with adjustment for potential confounders listed in table 1. Models were fitted via the method of maximum likelihood. We calculated odds ratios and CIs, for exposure relative to non-exposure.
459
THE LANCET
Duration of hormone therapy
Number of cases
Number of controls
Odds ratio*
Duration of hormone therapy
Number of cases
Number of controls
Odds ratio*
Never used hormones
337 (85·5%)
685 (86·9%)
1·0
Never used hormones
337 (87·3%)
685 (89·0%)
1·0
Progestagen added <10 days per month Duration of combined HRT 6–35 months 12 (3·1%) 36–59 months 3 (0·8%) Ä60 months 15 (3·8%)
14 (1·8%) 7 (0·9%) 12 (1·6%)
2·1 (0·9–4·7) 1·4 (0·3–5·4) 3·7 (1·7–8·2)
Progestagen added 10–21 days per month Duration of combined HRT 6–35 months 10 (2·5%) 36–59 months 5 (1·3%) Ä60 months 12 (3·1%)
31 (3·9%) 23 (2·9%) 16 (2·0%)
0·8 (0·4–1·8) 0·6 (0·2–1·6) 2·5 (1·1–5·5)
*Adjusted for age (continuous), body-mass index (continuous), and county lived in (three groups). HRT=hormone replacement therapy.
Table 3: Duration of therapy and endometrial cancer in users of oestrogen combined with cyclic progestagen Variables were retained in the final model if they changed the odds ratio by more than 10%. We use the term relative risk interchangeably with odds ratio in this manuscript, since these two terms are numerically almost identical in case-control studies of uncommon diseases.16 Data that had been included in an earlier paper (women at ages 45–64 years in King County with a reference date in years 1985, 1986, or 1987)6 were excluded only from the analysis of endometrial cancer risk according to whether progestagen was taken for less than 10 days or more than 10 days per month.
Results There were 337 cases and 685 controls who had never used hormones for as long as 6 months. Unopposed oestrogen was taken for at least 6 months by 324 cases and 180 controls, and was associated with a relative risk of endometrial cancer, adjusted for age, body-mass index 1 year before the reference date, and county of residence, of 4·0 (95% CI 3·1–5·1), with non-users of hormones as the reference group. The relative risks of endometrial cancer among women who had used unopposed oestrogens for at least 6 months and combined oestrogenprogestagen therapy for at least 6 months (76 cases and 74 controls) and among women who used only combined oestrogen-progestagen therapy (67 cases and 134 controls) were 2·7 (1·9–4·0) and 1·4 (1·0–1·9), respectively. Analysis of data from cases and controls in the early part of this study6 showed that only women who took progestagen for less than 10 days during the cycle were at an increased risk of endometrial cancer relative to women who had not taken hormones. Table 2 presents comparable data for all women in our study excluding those who were included in the earlier analysis. Compared with women who had not used hormones for more than 6 months, women who used oestrogen combined with cyclic progestagen therapy, with progestagen added for less than 10 days per month, had a relative risk of endometrial cancer of 3·1 (1·7–5·7). In contrast, the relative risk of endometrial cancer in women with progestagen added for 10 to 21 days per month was 1·3 (0·8–2·2). In the main analyses, only women who used combined therapy or had never used hormone therapy were included. Table 1 shows the distribution of demographic and reproductive characteristics of these 394 cases and 788 controls. The age distribution of cases and controls was similar. Consistent with known risk factors for endometrial cancer, slightly higher proportions of cases than controls had never used oral contraceptive pills, were nulliparous, had never been married, and had a high bodymass index 1 year before the reference date. 460
Current users only: progestagen added <10 days per month 6–59 months 11 (2·8%) 13 (1·7%) Ä60 months 14 (3·6%) 9 (1·2%)
2·2 (0·9–5·2) 4·8 (2·0–11·4)
Current users only: progestagen added 10–21 days per month 6–59 months 12 (3·1%) 48 (6·2%) Ä60 months 12 (3·1%) 15 (1·9%)
0·7 (0·4–1·4) 2·7 (1·2–6·0)
*Adjusted for age (continuous), body-mass index (continuous), and county lived in (three groups). HRT=hormone replacement therapy.
Table 4: Duration of therapy and endometrial cancer in current users of oestrogen combined with cyclic progestagen
Of those who knew the progestagen in their combined therapy, 89 of 93 controls and 47 of 48 cases identified it as medroxyprogesterone acetate. Women who used a progestagen for less than 10 days per month and women who used it for 10 or more days, there was an increased risk of endometrial cancer with 5 or more years of use of oestrogen and cyclic progestagen. Odds ratios were 3·7 (1·7–8·2) and 2·5 (1·1–5·5), respectively (table 3). Our finding of excess risk of endometrial cancer associated with long duration (5 or more years) of use of oestrogen combined with cyclic progestagen therapy was essentially unchanged when the dose of oestrogen was taken into account. For women taking oestrogen at daily doses of 0·9 mg or less conjugated oestrogen (36 cases and 73 controls) combined with cyclic progestagen, the odds ratios associated with durations of use of 5 years or more were 3·1 (1·2–8·1) for less than 10 days and 2·1 (0·8–5·4) for 10 days per month or more of progestagen. The increased risk associated with 5 years or more of use of a progestagen regimen of 10 to 21 days per month was present even for the highest dose of medroxyprogesterone acetate used, 10 mg per day (relative risk 2·7 [1·0–6·8], based on eight cases and 11 controls). Several respondents could not be included in the dose analysis: 26 cases and 34 controls: 26 cases and 34 controls for whom dose of progestagen was not shown; one case and four controls who took a progestagen other than medroxyprogesterone acetate; and four cases and two controls who took two regimens at different doses (one at a ¶5 mg dose and one at a 10 mg dose). Because the optimum number of days progestagen should be taken each cycle to minimise the risk of hyperplasia,2 and thus endometrial cancer, is not known, we subdivided the analysis according to number of days of progestagen used per month: between 10 and 12 days and 13 days or more. There were not enough data in the latter group for analysis (only 13 women, or 1% of the sample analysed here). Nevertheless, the finding of an increased risk of endometrial cancer associated with duration of oestrogen combined with cyclic progestagen use of 5 years or more persisted when the analysis was restricted to women who took progestagen for 10 to 12 days each month. The odds ratio associated with such long-term use of this regimen was 3·1 (1·3–7·4). To improve the homogeneity of the study group, we repeated the analyses for current users of oestrogen with cyclic progestagen therapy compared with women who had never used hormone replacement therapy. There was an increased risk of endometrial cancer with use of combined therapy for 5 years or more, both in women who used a progestagen for less than 10 days per month (relative risk 4·8 [2·0–11·4] and in women who used a progestagen for 10 or more days per month (relative risk 2·7 [1·2–6·0]).
Vol 349 • February 15, 1997
THE LANCET
Discussion Our study has confirmed the lower risk of endometrial cancer associated with use of a combined oestrogenprogestagen regimen than with unopposed oestrogens. The addition of cyclic progestagens to oestrogen therapy leads to differentiation of the endometrium, and thus to sloughing of the cell lining each month.17 This regular replenishment of the cells on the wall of the uterus has been hypothesised18 to discourage the formation of hyperplasia and thus cancer. However, our results do not support the optimistic view of an overall lower risk of endometrial cancer being found with combined therapy than with no therapy.18,19 Our study provides evidence for the risk of endometrial cancer associated with oestrogen combined with cyclic progestagen therapy being dependent on the number of days the progestagen is taken each month. In a previous study we had found a higher risk associated with a shorter monthly duration of progestagen.6 Our results are consistent with those obtained in studies of hyperplasia, which have found a relatively low incidence in women who received added progestagen for at least 10 days per month.2,3 However, in one study of combined therapy with 11 cases and nine controls,5 no variation in cancer risk was found with different number of days per month of progestagen. In our study, relative to women who never had postmenopausal hormone therapy, women who had taken oestrogen combined with cyclic progestagen therapy for 5 years or longer were at relative risk of at least two of endometrial cancer. Our study has limitations in common with most casecontrol studies. Differential recall between cases and controls of regimens of hormone replacement therapy is possible. No independent checks on women’s recall were made in this study. However, other studies20,21 have shown that verification from medical records correlates well with women’s recall of hormone use. Women can also provide accurate information on the specific type, scheduling, duration, and when they started hormone therapy.13,21 A study of older women found significantly greater recall of proper names by those taking oestrogen replacement therapy than by women not taking oestrogen.23 There may have been unmeasured confounders biasing the odds ratio estimates we calculated. The estimates were not substantially changed by adjustment for different measured confounders, and by repeated analyses on different subsets of data; thus, an influence of bias is unlikely. However, the non-response was about 27% for both cases and controls. If reasons for non-response were different for cases and controls, a bias could have been introduced. In the cases, the number of deaths between diagnosis and interview date counterbalanced the additional number of refusals in the control group. Although this is a comparatively large study, the numbers of women who had used one regimen of combined therapy for more than 5 years and for less than 5 years were modest. Thus, the CIs of the estimates of short-duration and long-duration use overlapped. If a woman takes postmenopausal hormone therapy of oestrogen combined with cyclic progestagen, what is her risk of endometrial cancer? Our study shows that compared with unopposed oestrogen therapy, the risk of endometrial cancer associated with combined hormone therapy is greatly reduced. Compared with no hormone use, use of combined therapy for less than 5 years is not
Vol 349 • February 15, 1997
associated with an increased risk unless the progestagen is added for fewer than 10 days each month. Our data suggest that long-term (5 years or more) use of combined therapy, even when the progestagen is added for more than 10 days per month, is associated with an increased risk of endometrial cancer. Further investigation is needed because of the high frequency of combined oestrogenprogestagen use, and the recommendations that such use be continued for long durations to maintain beneficial effects on bone and on the cardiovascular system. This study was funded by the National Cancer Institute, USA (grant numbers R01 CA47749 and R35CA39779). We thank the interviewers; the independent pathologists who rereviewed the histology; Barbara Hulka for guidance; and all the participants.
References 1 2
3
4
5
6
7 8 9
10
11
12 13 14
15
16 17 18
19 20
21 22
23
Whitehead MI, Hillard TC, Crook D. The role and use of progestogens. J Obstet Gynaecol 1990; 75: 59–76 (suppl). Whitehead MI, Townsend PT, Pryse-Davis J, et al. Effects of estrogens and progestagen on the biochemistry and morphology of the postmenopausal endometrium. N Engl J Med 1981; 305: 1599–605. Sturdee DN, Wade-Evans T, Peterson MEL, et al. Relations between bleeding pattern, endometrial histology, and estrogen treatment in menopausal women. BMJ 1978: i: 1575–77. The Writing Group for the PEPI Trial. Effects of hormone replacement therapy on endometrial histology in postmenopausal women. JAMA 1996; 275: 370–75. Brinton LA, Hoover RN. The endometrial cancer collaborative group. Estrogen replacement therapy and endometrial cancer risk: unresolved issues. J Obstet Gynecol 1993; 81: 265–71. Voigt LF, Weiss NS, Chu J, Daling JR, McKnight B, van Belle G. Progestagen supplementation of exogenous oestrogens and risk of endometrial cancer. Lancet 1991; 338: 274–77. Jick SS, Walker AM, Jick H. Estrogens, progesterone, and endometrial cancer. Epidemiology 1993; 4: 20–24. Jick SS. Combined estrogen and profesterone use and endometrial cancer. Epidemiology 1993; 4: 384. Psaty BM, Heckbert SR, Atkins D, et al. The risk of myocardial infarction associated with the combined use of estrogens and progestagens in postmenopausal women. Arch Intern Med 1994; 154: 1333–39. Ernster VL, Bush TL, Huggins GR, Hulka BS, Kelsey JL, Schottenffeld D. Benefits and risks of menopausal estrogen and/or progestagen hormone use. Prev Med 1988; 17: 201–23. The Writing Group for the PEPI Trial. Effects of estrogen or estrogen/progestagen regimens on heart disease risk factors in postmenopausal women. The Postmenopausal Estrogen/Progestagen Trial. JAMA 1995; 273: 199–208. Waksberg J. Sampling methods for random digit dialing. J Am Stat Assoc 1978; 73: 40–46. Beresford SAA, Coker AL. Pictorially assisted recall of past hormone use in case-control studies. Am J Epidemiol 1989; 130: 202–05. Voigt LF, Davis S, Heuser L. Random digit dialing: the potential effect on sample characteristics of the conversion of non-residential phone numbers. Am J Epidemiol 1992; 136: 1393–99. Genalo MA, Hickman RD. A study of ring/no answer numbers in random digit dialing telephone surveys. Proceedings of the Section on Survey Research Methods of the American Statistical Association. Alexandria: American Statistical Association, 1988: 552–57. Breslow NE, Day NE. The analysis of case control studies, volume 1. Lyons: International Agency for Research on Cancer, 1980. Whitehead MI, Fraser D. The effects of estrogens and progestogens on the endometrium. Obstet Gynecol Clin North Am 1987; 14: 299–320. Gambrell RD, Massey FM, Castaneda TA, Ugenas AJ, Ricci CA, Wright JM. Use of the progestogen challenge test to reduce the risk of endometrial cancer. J Obstet Gynaecol 1980; 55: 732–38. Hammond CB, Jelovsek FR, Lee KL, et al. Effect of long-term estrogen replacement therapy. Am J Obstet Gynecol 1979; 133: 537–47. Stadel BV, Weiss N. Characteristics of menopausal women: a survey of King and Pierce counties in Washington, 1973–1974. Am J Epidemiol 1975; 102: 209–16. Paganini-Hill A, Ross RK. Reliability of recall of drug usage and other health-related information. Am J Epidemiol 1982; 116: 114–22. Persson I, Bergkvist L, Adami H. Reliability of women’s histories of climacteric oestrogen treatment assessed by prescription forms. Int J Epidemiol 1987; 16: 222–28. Robinson D, Friedman L, Marcus R, Tinkleberg J, Yesavage J. Estrogen replacement therapy and memory in older women. J Am Geriatr Soc 1994; 42: 919–22.
461
Risk of endometrial cancer in relation to use of oestrogen combined with cyclic progestagen therapy in postmenopausal women Shirley A A Beresford, Noel S Weiss, Lynda F Voigt, Barbara McKnight
Summary Background Postmenopausal oestrogen therapy reduces the risk of osteoporosis and cardiovascular diseases but is associated with an increased risk of endometrial cancer. We have assessed the impact of a regimen of oestrogen with cyclic progestagen on risk of endometrial cancer for postmenopausal women. Methods We did a population-based case-control study of women aged 45–74 years in western Washington State, USA. Cases were identified from a regional cancer registry as having histologically confirmed endometrial cancer during 1985–91. 832 (72%) of 1154 eligible cases completed interviews. Controls were identified by random digit dialling, screened for intact uterus, frequency matched for age and county, and randomly assigned a reference date within 1985–91. Interviews with 1114 (73%) of 1526 eligible controls were done. The women provided information about use of hormone replacement therapy, and reproductive and medical history before diagnosis date (cases) or reference date (controls). Findings Relative to women who had never used hormones (for >6 months), women who had taken unopposed oestrogen had a four-fold increase (95% CI 3·1–5·1) in risk of endometrial cancer. Women who used a combined therapy of oestrogen with cyclic progestagen (eg, medroxyprogesterone acetate) had a relative risk of 1·4 (1·0–1·9). Among women with fewer than 10 days of added progestagen per month, the relative risk was 3·1 (1·7–5·7), whereas that for women with 10–21 days of added progestagen was 1·3 (0·8–2·2). The use of these combined regimens for 5 or more years was associated with risks of 3·7 (1·7–8·2) and 2·5 (1·1–5·5), respectively, relative to non-users of hormones. Interpretation Postmenopausal women who use combined therapy of oestrogen with cyclic progestagen on a longterm basis have an increased risk of endometrial cancer compared with those who are not on hormone replacement, even when progestagen is added for 10 or more days per month. This increase is much smaller than that associated with unopposed oestrogen, but needs to be confirmed.
Lancet 1997; 349: 458–61
Departments of Epidemiology (Prof S A A Beresford PhD, N S Weiss MD, L F Voigt PhD) and Biostatistics (B McKnight PhD), University of Washington, USA, and Programs in Epidemiology (S A A Beresford, N S Weiss, L F Voigt) and Biostatistics (B McKnight), Fred Hutchinson Cancer Research Center, Seattle, Washington Correspondence to: Prof Shirley A A Beresford, Department of Epidemiology, Box 357236, University of Washington, Seattle, WA 98195–7236, USA
458
Introduction Postmenopausal oestrogen therapy can benefit women in both the short term by reducing the frequency of hot flushes, and in the long term by reducing the risk of osteoporosis and cardiovascular disease.1 Although use of oestrogen alone is associated with an increased risk of endometrial cancer, addition of a progestagen reduces the incidence of endometrial hyperplasia2–4 and endometrial cancer.5–8 The addition of progestagen does not seem to interfere with the beneficial effects of oestrogen therapy,9,10 although the risk of cardiovascular disease may vary according to type of progestagen.4,11 In combination therapies, progestagen is commonly taken cyclically—oestrogen is taken for all or most days each month, but the progestagen for only part of the month. An alternative that has become popular is continuous combined therapy, in which both oestrogen and progestagen are taken every day of the month. Questions about the risk of endometrial cancer associated with oestrogen combined with cyclic progestagen therapy, compared with unopposed oestrogen therapy, have not yet been adequately answered. For example, how many days per month must progestagen be added to oestrogen therapy to achieve the reduced risk? There was a suggestion from Voigt and colleagues6 that 10 or more days of progestagen are needed, but the power of this study was limited. How safe is the long-term use of oestrogens combined with progestagens with respect to endometrial cancer risk? In our study we aimed to find out more about the relation between the risk of developing endometrial cancer and cyclic progestagen therapy.
Methods We identified female residents of King, Pierce, and Snohomish counties of Washington State, USA, aged 45–74 years, who developed a histologically confirmed endometrial carcinoma during 1985–91, from the Cancer Surveillance System, a population-based cancer registry that has served western Washington since 1974. For 1985 and 1986, recruitment was limited to King County, and to an upper age of 64 years. For 1991, recruitment was limited to an upper age of 69 years. During 1986–93, controls were identified by random-digitdialling telephone calls12 within the three counties. A short questionnaire was used to identify women aged 45–74 years as potential controls. If more than one eligible woman lived at the address, one was selected randomly. Controls were chosen with stratified sampling techniques, in such a way that the age and county distribution of controls was broadly similar to that of cases. Controls were randomly assigned a reference date, the distribution of which corresponded roughly to the date of diagnosis of the cases. The study was approved by the Institutional Review Board of the University of Washington, USA. All eligible participants were interviewed in person, except for 34 (3%) of the cases and 39 (5%) of the controls who were interviewed by telephone. On average there was a shorter interval between reference year and interview date for cases than for controls (median 21 months for cases and 26 months for controls). Adjustment for this difference did not change the results. At interview, the participants were asked about whether
Vol 349 • February 15, 1997
THE LANCET
Number of cases (394)
Number of controls (788)
Age (years) 45–49 50–54 55–59 60–64 65–69 70–74
65 (17%) 70 (18%) 78 (20%) 88 (22%) 63 (16%) 30 (8%)
122 (16%) 151 (19%) 172 (22%) 157 (20%) 118 (15%) 68 (9%)
County King Pierce Snohomish
249 (63%) 88 (22%) 57 (15%)
572 (73%) 123 (16%) 93 (12%)
Reference or diagnosis year 1985–87 1988–89 1990–91
112 (28%) 134 (34%) 148 (38%)
187 (24%) 370 (47%) 231 (29%)
Parity (livebirths plus stillbirths) 0 1 2 3 Ä4
50 (13%) 49 (12%) 99 (25%) 91 (23%) 105 (27%)
79 (10%) 78 (10%) 208 (26%) 199 (25%) 224 (28%)
Oral contraceptive use Never used 1 month–5 years Ä5 years
256 (65%) 93 (24%) 45 (11%)
459 (58%) 183 (23%) 146 (19%)
Ethnic origin White or hispanic Non-white
378 (96%) 16 (4%)
749 (95%) 39 (5%)
Marital status Never married Married or living as married Separated or divorced Widowed
18 (5%) 284 (72%) 48 (12%) 44 (11%)
25 (3%) 543 (69%) 123 (16%) 97 (12%)
Annual household income <$15 000 $15 000–$30 000 $30 000–$45 000 >$45 000 Refused or unknown
65 (17%) 125 (32%) 83 (21%) 107 (27%) 14 (4%)
150 (19%) 233 (30%) 180 (23%) 213 (27%) 12 (2%)
Education High school or less Some college or technical school College graduate or more
197 (50%) 109 (28%) 88 (22%)
348 (44%) 259 (33%) 181 (23%)
Body-mass index (quartiles of Quetlet’s index)* 13·01–21·49 21·50–23·96 23·97–27·42 27·43–63·21 Refused or unknown
47 (12%) 68 (17%) 68 (17%) 209 (53%) 2 (1%)
195 (25%) 191 (24%) 187 (24%) 212 (27%) 3 (0%)
*1 year before reference date.
Table 1: Demographic and reproductive characteristics in cases and controls using combined therapy or no therapy or not they had had a hysterectomy, about their use of hormone replacement therapy, and about their reproductive and medical history up to the time of diagnosis (cases) or to the reference date (controls). To enhance recall of hormone therapy used, we provided pictures of commonly used preparations. The usefulness of such memory prompts has been reported. 13 Individuals interviewed by telephone received the pictures by post before the interview. 72 of the 1254 potential cases identified did not have an epithelial endometrial cancer, 19 had difficulty in communicating, and nine were not interviewed because they had no telephone or were not living at a private address in one of the three study counties at the time of diagnosis. The remaining 1154 cases were eligible. Of these, complete interviews were obtained for 832 (72·2%). 100 (9%) women died before interview, the physicians of 63 (5%) women refused to allow us to contact their patient, and 158 (14%) cases themselves refused to participate. 454 potential cases had an independent pathology review done by two pathologists who were unaware of the Cancer Surveillance System classification. 402 of these cases were initially classified by the Cancer Surveillance System as invasive
Vol 349 • February 15, 1997
Progestagen use
Number of cases
Number of controls
Odds ratio*
Never used hormones <10 days per month 10–21 days per month
270 (84·4%) 25 (7·8%) 25 (7·8%)
593 (86·8%) 26 (3·8%) 64 (9·4%)
1·0 3·1 (1·7–5·7) 1·3 (0·8–2·2)
Analysis restricted to participants whose experience had not been previously reported. *Adjusted for age (continuous), body-mass index (continuous), and county lived in (three groups).
Table 2: Endometrial cancer in users of oestrogen combined with cyclic-progestagen therapy: influence of number of days per month progestagen is added epithelial endometrial cancer, and 52 had other histological classifications, including carcinoma in situ. Of the former cases, 367 (91%) were confirmed by independent pathology review. To recruit controls, 52 045 random-digit-dialling telephone calls were made, of which 23 527 were found to be residential numbers. There were 2113 numbers for which no answer was obtained; we assumed about 20% of these telephone numbers would have been for residential addresses. 14,15 Household screening was completed for 22 650 of the residential numbers, giving a screening response of 96·3%. 2620 women were found to be eligible for the study by both age and county; 1975 (75·4%) women agreed to be interviewed (overall response for controls was 73·0%). 860 (44%) were immediately excluded because they had had a hysterectomy, leaving 1114 women. Controls were also excluded from our analysis if they had been diagnosed with cancer of the endometrium or had had a hysterectomy more than 6 months before their reference date. Only hormone use that was probably postmenopausal was included in the analysis. Therefore, if any participant used hormones before age 45 years, for regulation of menstruation or for depression, anxiety, or emotional distress, or for premenstrual syndrome. Responses were reviewed independently by two of the investigators who were unaware of case-control status. Hormone therapy was not judged to be postmenopausal-hormone therapy unless there was additional evidence that it was used after menopause. Hormone use for infertility or to prevent miscarriage was also not counted as postmenopausal hormone use. Because use of unopposed oestrogen for less than 6 months has little or no effect on the incidence of endometrial cancer, we focused our analysis on women who used a combined oestrogenprogestagen regimen for at least 6 months. Durations of use of at least 6 months are also associated with better recall of dose and type of therapy.13 Women who had used hormone therapy for less than 6 months were grouped with non-users in our analysis. Because their numbers were so small, we excluded women with a history of continuous combined therapy (six cases, 17 controls) from this analysis. In addition, we excluded combined therapy users who had ever used unopposed oestrogen replacement therapy (76 cases and 74 controls) from most of our analyses because we wanted to assess the effect of oestrogen combined with cyclic progestagen in women whose risk of endometrial cancer was not already increased by use of unopposed oestrogen. Variations in regimens of combined therapy were counted as different regimens only if one had added progestagen for less than 10 days per month and the other had added progestagen for 10 days per month or more. There were four cases and 14 controls who had used more than one regimen of combined therapy. To be able to interpret the findings in a straightforward way, we excluded these 18 women. We classified the participants in four groups: those who had never used hormones for as long as 6 months; those who had used only unopposed oestrogens; those who had used only oestrogen combined with cyclic progestagen; and those who had at some time used unopposed oestrogens and at other times had used combined oestrogen-progestagen therapy.
Statistical analysis The analyses were by logistic regression techniques, with adjustment for potential confounders listed in table 1. Models were fitted via the method of maximum likelihood. We calculated odds ratios and CIs, for exposure relative to non-exposure.
459
THE LANCET
Duration of hormone therapy
Number of cases
Number of controls
Odds ratio*
Duration of hormone therapy
Number of cases
Number of controls
Odds ratio*
Never used hormones
337 (85·5%)
685 (86·9%)
1·0
Never used hormones
337 (87·3%)
685 (89·0%)
1·0
Progestagen added <10 days per month Duration of combined HRT 6–35 months 12 (3·1%) 36–59 months 3 (0·8%) Ä60 months 15 (3·8%)
14 (1·8%) 7 (0·9%) 12 (1·6%)
2·1 (0·9–4·7) 1·4 (0·3–5·4) 3·7 (1·7–8·2)
Progestagen added 10–21 days per month Duration of combined HRT 6–35 months 10 (2·5%) 36–59 months 5 (1·3%) Ä60 months 12 (3·1%)
31 (3·9%) 23 (2·9%) 16 (2·0%)
0·8 (0·4–1·8) 0·6 (0·2–1·6) 2·5 (1·1–5·5)
*Adjusted for age (continuous), body-mass index (continuous), and county lived in (three groups). HRT=hormone replacement therapy.
Table 3: Duration of therapy and endometrial cancer in users of oestrogen combined with cyclic progestagen Variables were retained in the final model if they changed the odds ratio by more than 10%. We use the term relative risk interchangeably with odds ratio in this manuscript, since these two terms are numerically almost identical in case-control studies of uncommon diseases.16 Data that had been included in an earlier paper (women at ages 45–64 years in King County with a reference date in years 1985, 1986, or 1987)6 were excluded only from the analysis of endometrial cancer risk according to whether progestagen was taken for less than 10 days or more than 10 days per month.
Results There were 337 cases and 685 controls who had never used hormones for as long as 6 months. Unopposed oestrogen was taken for at least 6 months by 324 cases and 180 controls, and was associated with a relative risk of endometrial cancer, adjusted for age, body-mass index 1 year before the reference date, and county of residence, of 4·0 (95% CI 3·1–5·1), with non-users of hormones as the reference group. The relative risks of endometrial cancer among women who had used unopposed oestrogens for at least 6 months and combined oestrogenprogestagen therapy for at least 6 months (76 cases and 74 controls) and among women who used only combined oestrogen-progestagen therapy (67 cases and 134 controls) were 2·7 (1·9–4·0) and 1·4 (1·0–1·9), respectively. Analysis of data from cases and controls in the early part of this study6 showed that only women who took progestagen for less than 10 days during the cycle were at an increased risk of endometrial cancer relative to women who had not taken hormones. Table 2 presents comparable data for all women in our study excluding those who were included in the earlier analysis. Compared with women who had not used hormones for more than 6 months, women who used oestrogen combined with cyclic progestagen therapy, with progestagen added for less than 10 days per month, had a relative risk of endometrial cancer of 3·1 (1·7–5·7). In contrast, the relative risk of endometrial cancer in women with progestagen added for 10 to 21 days per month was 1·3 (0·8–2·2). In the main analyses, only women who used combined therapy or had never used hormone therapy were included. Table 1 shows the distribution of demographic and reproductive characteristics of these 394 cases and 788 controls. The age distribution of cases and controls was similar. Consistent with known risk factors for endometrial cancer, slightly higher proportions of cases than controls had never used oral contraceptive pills, were nulliparous, had never been married, and had a high bodymass index 1 year before the reference date. 460
Current users only: progestagen added <10 days per month 6–59 months 11 (2·8%) 13 (1·7%) Ä60 months 14 (3·6%) 9 (1·2%)
2·2 (0·9–5·2) 4·8 (2·0–11·4)
Current users only: progestagen added 10–21 days per month 6–59 months 12 (3·1%) 48 (6·2%) Ä60 months 12 (3·1%) 15 (1·9%)
0·7 (0·4–1·4) 2·7 (1·2–6·0)
*Adjusted for age (continuous), body-mass index (continuous), and county lived in (three groups). HRT=hormone replacement therapy.
Table 4: Duration of therapy and endometrial cancer in current users of oestrogen combined with cyclic progestagen
Of those who knew the progestagen in their combined therapy, 89 of 93 controls and 47 of 48 cases identified it as medroxyprogesterone acetate. Women who used a progestagen for less than 10 days per month and women who used it for 10 or more days, there was an increased risk of endometrial cancer with 5 or more years of use of oestrogen and cyclic progestagen. Odds ratios were 3·7 (1·7–8·2) and 2·5 (1·1–5·5), respectively (table 3). Our finding of excess risk of endometrial cancer associated with long duration (5 or more years) of use of oestrogen combined with cyclic progestagen therapy was essentially unchanged when the dose of oestrogen was taken into account. For women taking oestrogen at daily doses of 0·9 mg or less conjugated oestrogen (36 cases and 73 controls) combined with cyclic progestagen, the odds ratios associated with durations of use of 5 years or more were 3·1 (1·2–8·1) for less than 10 days and 2·1 (0·8–5·4) for 10 days per month or more of progestagen. The increased risk associated with 5 years or more of use of a progestagen regimen of 10 to 21 days per month was present even for the highest dose of medroxyprogesterone acetate used, 10 mg per day (relative risk 2·7 [1·0–6·8], based on eight cases and 11 controls). Several respondents could not be included in the dose analysis: 26 cases and 34 controls: 26 cases and 34 controls for whom dose of progestagen was not shown; one case and four controls who took a progestagen other than medroxyprogesterone acetate; and four cases and two controls who took two regimens at different doses (one at a ¶5 mg dose and one at a 10 mg dose). Because the optimum number of days progestagen should be taken each cycle to minimise the risk of hyperplasia,2 and thus endometrial cancer, is not known, we subdivided the analysis according to number of days of progestagen used per month: between 10 and 12 days and 13 days or more. There were not enough data in the latter group for analysis (only 13 women, or 1% of the sample analysed here). Nevertheless, the finding of an increased risk of endometrial cancer associated with duration of oestrogen combined with cyclic progestagen use of 5 years or more persisted when the analysis was restricted to women who took progestagen for 10 to 12 days each month. The odds ratio associated with such long-term use of this regimen was 3·1 (1·3–7·4). To improve the homogeneity of the study group, we repeated the analyses for current users of oestrogen with cyclic progestagen therapy compared with women who had never used hormone replacement therapy. There was an increased risk of endometrial cancer with use of combined therapy for 5 years or more, both in women who used a progestagen for less than 10 days per month (relative risk 4·8 [2·0–11·4] and in women who used a progestagen for 10 or more days per month (relative risk 2·7 [1·2–6·0]).
Vol 349 • February 15, 1997
THE LANCET
Discussion Our study has confirmed the lower risk of endometrial cancer associated with use of a combined oestrogenprogestagen regimen than with unopposed oestrogens. The addition of cyclic progestagens to oestrogen therapy leads to differentiation of the endometrium, and thus to sloughing of the cell lining each month.17 This regular replenishment of the cells on the wall of the uterus has been hypothesised18 to discourage the formation of hyperplasia and thus cancer. However, our results do not support the optimistic view of an overall lower risk of endometrial cancer being found with combined therapy than with no therapy.18,19 Our study provides evidence for the risk of endometrial cancer associated with oestrogen combined with cyclic progestagen therapy being dependent on the number of days the progestagen is taken each month. In a previous study we had found a higher risk associated with a shorter monthly duration of progestagen.6 Our results are consistent with those obtained in studies of hyperplasia, which have found a relatively low incidence in women who received added progestagen for at least 10 days per month.2,3 However, in one study of combined therapy with 11 cases and nine controls,5 no variation in cancer risk was found with different number of days per month of progestagen. In our study, relative to women who never had postmenopausal hormone therapy, women who had taken oestrogen combined with cyclic progestagen therapy for 5 years or longer were at relative risk of at least two of endometrial cancer. Our study has limitations in common with most casecontrol studies. Differential recall between cases and controls of regimens of hormone replacement therapy is possible. No independent checks on women’s recall were made in this study. However, other studies20,21 have shown that verification from medical records correlates well with women’s recall of hormone use. Women can also provide accurate information on the specific type, scheduling, duration, and when they started hormone therapy.13,21 A study of older women found significantly greater recall of proper names by those taking oestrogen replacement therapy than by women not taking oestrogen.23 There may have been unmeasured confounders biasing the odds ratio estimates we calculated. The estimates were not substantially changed by adjustment for different measured confounders, and by repeated analyses on different subsets of data; thus, an influence of bias is unlikely. However, the non-response was about 27% for both cases and controls. If reasons for non-response were different for cases and controls, a bias could have been introduced. In the cases, the number of deaths between diagnosis and interview date counterbalanced the additional number of refusals in the control group. Although this is a comparatively large study, the numbers of women who had used one regimen of combined therapy for more than 5 years and for less than 5 years were modest. Thus, the CIs of the estimates of short-duration and long-duration use overlapped. If a woman takes postmenopausal hormone therapy of oestrogen combined with cyclic progestagen, what is her risk of endometrial cancer? Our study shows that compared with unopposed oestrogen therapy, the risk of endometrial cancer associated with combined hormone therapy is greatly reduced. Compared with no hormone use, use of combined therapy for less than 5 years is not
Vol 349 • February 15, 1997
associated with an increased risk unless the progestagen is added for fewer than 10 days each month. Our data suggest that long-term (5 years or more) use of combined therapy, even when the progestagen is added for more than 10 days per month, is associated with an increased risk of endometrial cancer. Further investigation is needed because of the high frequency of combined oestrogenprogestagen use, and the recommendations that such use be continued for long durations to maintain beneficial effects on bone and on the cardiovascular system. This study was funded by the National Cancer Institute, USA (grant numbers R01 CA47749 and R35CA39779). We thank the interviewers; the independent pathologists who rereviewed the histology; Barbara Hulka for guidance; and all the participants.
References 1 2
3
4
5
6
7 8 9
10
11
12 13 14
15
16 17 18
19 20
21 22
23
Whitehead MI, Hillard TC, Crook D. The role and use of progestogens. J Obstet Gynaecol 1990; 75: 59–76 (suppl). Whitehead MI, Townsend PT, Pryse-Davis J, et al. Effects of estrogens and progestagen on the biochemistry and morphology of the postmenopausal endometrium. N Engl J Med 1981; 305: 1599–605. Sturdee DN, Wade-Evans T, Peterson MEL, et al. Relations between bleeding pattern, endometrial histology, and estrogen treatment in menopausal women. BMJ 1978: i: 1575–77. The Writing Group for the PEPI Trial. Effects of hormone replacement therapy on endometrial histology in postmenopausal women. JAMA 1996; 275: 370–75. Brinton LA, Hoover RN. The endometrial cancer collaborative group. Estrogen replacement therapy and endometrial cancer risk: unresolved issues. J Obstet Gynecol 1993; 81: 265–71. Voigt LF, Weiss NS, Chu J, Daling JR, McKnight B, van Belle G. Progestagen supplementation of exogenous oestrogens and risk of endometrial cancer. Lancet 1991; 338: 274–77. Jick SS, Walker AM, Jick H. Estrogens, progesterone, and endometrial cancer. Epidemiology 1993; 4: 20–24. Jick SS. Combined estrogen and profesterone use and endometrial cancer. Epidemiology 1993; 4: 384. Psaty BM, Heckbert SR, Atkins D, et al. The risk of myocardial infarction associated with the combined use of estrogens and progestagens in postmenopausal women. Arch Intern Med 1994; 154: 1333–39. Ernster VL, Bush TL, Huggins GR, Hulka BS, Kelsey JL, Schottenffeld D. Benefits and risks of menopausal estrogen and/or progestagen hormone use. Prev Med 1988; 17: 201–23. The Writing Group for the PEPI Trial. Effects of estrogen or estrogen/progestagen regimens on heart disease risk factors in postmenopausal women. The Postmenopausal Estrogen/Progestagen Trial. JAMA 1995; 273: 199–208. Waksberg J. Sampling methods for random digit dialing. J Am Stat Assoc 1978; 73: 40–46. Beresford SAA, Coker AL. Pictorially assisted recall of past hormone use in case-control studies. Am J Epidemiol 1989; 130: 202–05. Voigt LF, Davis S, Heuser L. Random digit dialing: the potential effect on sample characteristics of the conversion of non-residential phone numbers. Am J Epidemiol 1992; 136: 1393–99. Genalo MA, Hickman RD. A study of ring/no answer numbers in random digit dialing telephone surveys. Proceedings of the Section on Survey Research Methods of the American Statistical Association. Alexandria: American Statistical Association, 1988: 552–57. Breslow NE, Day NE. The analysis of case control studies, volume 1. Lyons: International Agency for Research on Cancer, 1980. Whitehead MI, Fraser D. The effects of estrogens and progestogens on the endometrium. Obstet Gynecol Clin North Am 1987; 14: 299–320. Gambrell RD, Massey FM, Castaneda TA, Ugenas AJ, Ricci CA, Wright JM. Use of the progestogen challenge test to reduce the risk of endometrial cancer. J Obstet Gynaecol 1980; 55: 732–38. Hammond CB, Jelovsek FR, Lee KL, et al. Effect of long-term estrogen replacement therapy. Am J Obstet Gynecol 1979; 133: 537–47. Stadel BV, Weiss N. Characteristics of menopausal women: a survey of King and Pierce counties in Washington, 1973–1974. Am J Epidemiol 1975; 102: 209–16. Paganini-Hill A, Ross RK. Reliability of recall of drug usage and other health-related information. Am J Epidemiol 1982; 116: 114–22. Persson I, Bergkvist L, Adami H. Reliability of women’s histories of climacteric oestrogen treatment assessed by prescription forms. Int J Epidemiol 1987; 16: 222–28. Robinson D, Friedman L, Marcus R, Tinkleberg J, Yesavage J. Estrogen replacement therapy and memory in older women. J Am Geriatr Soc 1994; 42: 919–22.
461