- © 1999 by American Society of Clinical Oncology
Risk of Menopause During the First Year After Breast Cancer Diagnosis
- From the Departments of Medicine and Clinical Epidemiology, Samuel Lunenfeld Research Institute, Mount Sinai Hospital; Department of Medicine, Toronto-Sunnybrook Regional Cancer Centre; Department of Medicine, Women's College Hospital; and Department of Public Health Science, University of Toronto, Toronto, Ontario, Canada.
- Address reprint requests to Pamela J. Goodwin, MD, Marvelle Koffler Breast Centre, Mount Sinai Hospital, 1284-600 University Ave, Toronto, Ontario M5G 1X5, Canada; email pgoodwin{at}mtsinai.on.ca
Abstract
PURPOSE: Premenopausal women with breast cancer often enter a premature menopause during initial treatment of their malignancy, with resulting loss of childbearing capacity, onset of menopausal symptoms, and subsequent prolonged exposure to long-term risks of menopause. Adjuvant therapy is believed to contribute to this early menopause.
PATIENTS AND METHODS: One hundred eighty-three premenopausal women with locoregional breast cancer (tumor-node-metastasis staging system classification, T1-3 N0-1 M0) who had undergone surgical treatment and provided information on menopausal status at diagnosis and 1 year later were enrolled. Systemic adjuvant therapy was recorded. Univariate and multivariate predictors of menopause were examined.
RESULTS: Age, weight gain, tumor stage, nodal stage, and systemic adjuvant therapy (chemotherapy, tamoxifen) were all significant univariate correlates of menopause. In multivariate analysis, age, chemotherapy, and hormone therapy (tamoxifen) made significant independent contributions to the onset of menopause.
CONCLUSION: Age and systemic chemotherapy are the strongest predictors of menopause in women with locoregional breast cancer. They independently contribute to menopause. A graphic representation of our multivariate model allows an estimation of risk of menopause according to patient age and planned adjuvant treatment, and it may facilitate clinical decision-making.
BREAST CANCER IS ONE of the most commonly diagnosed malignancies among premenopausal women. An estimated 180,000 women were diagnosed with breast cancer in the United States1 in 1995; approximately 25% of these women were premenopausal.2-5 Screening programs and increased awareness of breast problems have led to diagnosis at an earlier stage of disease.6,7 This, combined with the widespread use of adjuvant drug treatment,8-10 notably chemotherapy in premenopausal women, has led to ever-increasing numbers of young women who are surviving long periods of time after breast cancer diagnosis. As a result, long-term effects of treatment are becoming increasingly important. One of these long-term sequelae—premature menopause4,11,12 with resulting loss of childbearing capacity and prolonged exposure to risks of menopause, including heart disease, osteoporosis, and symptoms such as hot flushes, genitourinary problems, and psychologic distress—has been receiving increasing attention in recent years.13,14 Hormone replacement therapy is commonly used in women without breast cancer to treat, or prevent, these menopausal sequelae.15,16 However, its use in women with breast cancer is not widely accepted because of concerns that it may increase the risk of new breast primary tumors and recurrence of previously diagnosed cancer.17-20
As adjuvant treatment is used more often in women with early-stage disease who are at low risk of breast cancer recurrence, tradeoffs between the beneficial effects of this treatment on recurrence and detrimental effects, including loss of childbearing capacity and menopausal complications, are becoming increasingly salient. The absolute size of the benefits of treatment decreases with decreasing risk, but the risks associated with menopause are constant. An enhanced ability to predict the likelihood of menopause onset would facilitate decision-making by women and physicians facing these tradeoffs, leading to a more informed choice of treatment.
In the past, studies examining menopause during adjuvant therapy for breast cancer have focused on two main issues: the risk of menopause with specific types of treatment and the impact of menopause on prognosis. Excellent recent reviews of these studies have been published.4,11,12 In this report, we focus on the risk of menopause arising from adjuvant treatment.
The risk of menopause with polyagent adjuvant chemotherapy has been reported to range from 53% to 89%.11 Some of the variability reflects differing definitions of menopause, some reflects patient and treatment-related characteristics, and some reflects follow-up duration. In most reports, age has been shown to be a strong predictor of menopause onset. Unfortunately, most studies have grouped patients into old (> 40 years) and young (< 40 years), thereby providing only crude estimates of risk at specific ages. Bines et al4 reported 40% of women under 40 years of age and 76% of those over 40 years became menopausal during adjuvant chemotherapy with cyclophosphamide, methotrexate, and fluorouracil (CMF). Bianco et al21 reported an increasing risk of menopause with increasing age. Cumulative dose of chemotherapy has been associated with risk of menopause.4 It is unclear whether duration of treatment or dose-intensity exerts effects that are independent of cumulative dose. One cycle of perioperative CMF has been reported to result in menopause in 31% of women,22 a rate that is substantially lower than that reported with multiple cycles. There is sparse evidence that different types of chemotherapy are associated with differing risks of menopause; however, almost all have used alkylating agent–based chemotherapy, most frequently CMF. Existing evidence suggests that risk of menopause is lower with single-agent melphalan (26%)23 or with doxorubicin and cyclophosphamide (AC; 34%), but the latter usually involves four cycles given over 12 weeks with a lower cumulative dose of chemotherapy.24 Little information is available regarding the age-specific risk of menopause in women receiving tamoxifen only.
There is some evidence to suggest that chemotherapy-induced amenorrhea is associated with an improved prognosis in locoregional breast cancer, although reports are not consistent. Some reports have found a significant protective effect of treatment-induced menopause, whereas others have failed to find any difference in outcome in those who become menopausal and those who do not. Two recent reviews11,12 have come to different conclusions regarding the existence of a beneficial prognostic effect of treatment-induced amenorrhea.
In this report, we will examine factors predicting onset of menopause in a cohort of premenopausal women with newly diagnosed breast cancer receiving either adjuvant CMF, cyclophosphamide, epirubicin, and fluorouracil (CEF), tamoxifen, or no treatment. Our goal is to generate a model that will provide clinically useful estimates of the risk of menopause occurring in women receiving adjuvant treatment.
PATIENTS AND METHODS
Population Assembly
This study included an inception cohort of 183 premenopausal women diagnosed with locoregional breast cancer (tumor-node-metastasis staging system classification, T1-3 N0-1 M0) at three participating institutions (Mount Sinai Hospital, St Michael's Hospital, and Women's College Hospital) between July 1992 and June 1996 at the University of Toronto. All had undergone surgical resection of their breast primary tumor (mastectomy or lumpectomy with margins clear of invasive disease) with axillary node dissection. Women were enrolled after surgery and before (or during the first month of) administration of systemic adjuvant therapy. These women formed part of a larger inception cohort27 that also included peri- and postmenopausal women. Women enrolled during the first 3 years of the larger study who were not asked to provide information on follow-up menopausal status and women who had had a hysterectomy (with or without oophorectomy) are not included in this report. The overall inception cohort was assembled to examine prognostic effects of body size and nutrition-related factors. Those effects will be reported elsewhere. All women provided written informed consent to participate in this study, in keeping with guidelines of the Human Subjects Committee of the University of Toronto.
Data Collection
Information was collected on age, body size (fasting weight, height, and body mass index [BMI], expressed as weight [kg]/height [m2]), tumor stage and size, nodal stage, number of nodes involved and removed, hormone receptor status (estrogen, progesterone), surgical treatment (mastectomy, lumpectomy), radiation treatment (yes v no), adjuvant treatment (CMF, CEF, tamoxifen alone, tamoxifen combined with CMF or CEF, or no adjuvant treatment), and duration of adjuvant treatment (months).
Menopausal status at diagnosis was as follows: premenopausal, defined as regular menses; perimenopausal, defined as irregular menses with at least one period during the previous year; postmenopausal, defined as no menses for at least 1 year. Women who were premenopausal at diagnosis and whose menses stopped (and did not return) during the year after diagnosis were classified as postmenopausal at 1-year follow-up.
Statistical Analysis
Descriptive statistics were generated by examining distributions, means, and SDs for continuous variables. BMI was not normally distributed; an inverse transformation was used to obtain a normal distribution. For categorical variables, frequency tables were generated.
The candidate explanatory variables in the univariate analyses of menopause onset were age at diagnosis (continuous), BMI at diagnosis and change in BMI after diagnosis (continuous), type of surgery (mastectomy or lumpectomy), tumor stage (TX, T1, or T2/3), nodal stage (N0 or N1), adjuvant chemotherapy (yes or no), adjuvant tamoxifen (yes or no), and duration of follow-up (number of days or continuous). For the univariate analysis, t tests were performed using age, BMI, and duration of follow-up to find significant predictors for onset of menopause. Pearson's χ2 test was performed for the categorical variables tumor stage, nodal stage, adjuvant chemotherapy, and adjuvant hormone therapy. No difference was seen between CMF or CEF chemotherapy. As a result, they were combined in the multivariate analysis.
A multivariate logistic regression analysis was then performed using onset of menopause as a binary outcome variable. The candidate explanatory variables in the initial model were all those that were univariately significant at the .1 level. The purpose of this analysis was to identify those clinical factors that jointly provided the best prediction of the onset of menopause. In a stepwise fashion, the explanatory variables with the largest P values were removed from the model after it was confirmed that they were not involved in a significant two-way interaction with any one of the remaining variables. Significance was tested using χ2 tests based on the change in deviance. This was repeated until no variable could be removed at a threshold p value of .05. The resulting model contained age at diagnosis, adjuvant chemotherapy, and adjuvant hormone therapy. Interactions between these remaining terms were examined. The probability of the onset of menopause for various levels of the explanatory variables was calculated from this model and depicted graphically. Pointwise two-SE bars were calculated at a number of points; they can be loosely interpreted as pointwise 95% confidence intervals. To investigate the model's fit, deviance residuals obtained from the final model were inspected; no unusually large residuals were found.
RESULTS
Description of the Study Population
The mean age was 43.7 ± 5.2 years. Other population characteristics are listed in Table 1. The majority of women had undergone lumpectomy with postoperative radiation. Almost 50% had T1 tumors and almost two thirds had no axillary node involvement. The majority of women received adjuvant chemotherapy (45.4%, CMF; 13.7%, CEF). Just over 25% (47 women) received adjuvant tamoxifen; of these, 25 (53.2%) received combined chemotherapy and tamoxifen. None received AC chemotherapy.
Study Population
Univariate Predictors of Onset of Menopause
As shown in Table 2, age was a significant univariate predictor of menopause (P = .0001). The mean age in those who became menopausal was 3 years greater than it was in those who remained premenopausal (45.5 v 42.3 years); however, there was a steadily increasing risk of menopause with increasing age (Fig 1). Baseline body size, measured as BMI or weight (data not shown), did not predict menopause. In contrast, weight gain during the year after diagnosis, measured as BMI or as weight, was significantly associated with menopause, with women gaining larger amounts of weight having the greatest likelihood of becoming menopausal (P = .01).
Univariate Predictors of Menopause
Probability of menopause during the first year after diagnosis (from model shown in Table 3).
Of the tumor- and treatment-related variables, tumor stage, nodal stage, and type of systemic adjuvant therapy administered (chemotherapy v tamoxifen only v no chemotherapy) were all significantly associated with menopause. Women who became menopausal were more likely to have larger tumors and involved axillary nodes and to have received systemic adjuvant therapy (notably chemotherapy) than those who remained premenopausal.
The impact of duration of chemotherapy on menopause onset could not be examined because 99 (91.6%) of the 108 women who received chemotherapy received six cycles of treatment. A similar duration of treatment was planned in the nine remaining women, but treatment was stopped early for a variety of reasons (treatment was stopped after one cycle in two women, four cycles in four women, and five cycles in three women).
Multivariate Analysis of Onset of Menopause
Variables included in the multivariate model included age, tumor size, nodal status, chemotherapy, hormone therapy, and all relevant interaction terms. Weight gain was not included in the model because it was believed to occur as a result of menopause, rather than being a causal factor for menopausal onset.27
The final model is shown in Table 3. Age at diagnosis, use of chemotherapy, and use of hormone therapy were each significantly, and independently, associated with menopause onset. None of the interaction terms made a significant contribution to the model. There was no evidence that risk of menopause differed with CEF versus CMF, and these factors were combined in the graphic representation of the model. The use of tamoxifen in addition to either type of chemotherapy resulted in a small, but statistically significant, increase in the risk of menopause.
Final Multivariate Model
Figure 1 shows the probability of menopause according to age at diagnosis and type of systemic adjuvant therapy administered, as predicted by this model. Approximate 95% confidence intervals for these curves overlap for the two upper curves (chemotherapy only, combined chemotherapy and tamoxifen); they also overlap for the two lower curves (tamoxifen only, no systemic therapy). Beyond the age of 35, there was no overlap of 95% confidence intervals for the upper two curves with those for the lower two curves. That is, the risk of menopause was significantly increased when chemotherapy (as opposed to tamoxifen only or no adjuvant therapy) was administered to women over the age of 35.
DISCUSSION
We have demonstrated that two factors, age and use of systemic chemotherapy, are important predictors of menopause onset in premenopausal women with newly diagnosed breast cancer. Use of either CMF or CEF, whether in combination with tamoxifen or not, increased the risk of menopause in 40-year-old women from less than 5% to more than 40%. In 50-year-old women, this risk was increased from approximately 20% to close to 100%. Furthermore, although risk of menopause was low in many of the treatment groups before the age of 35, beyond that age, there was a clear separation of risk in those receiving chemotherapy and those not receiving chemotherapy; in women over the age of 35, 95% confidence intervals for those receiving chemotherapy did not overlap the 95% confidence intervals for those not receiving chemotherapy. The absence of a difference in risk of menopause in those receiving CMF and CEF was not unexpected given the similarity of the protocols of these treatments.25,26 Both treatments are given for 6 months and both use 14 days of oral cyclophosphamide, with fluorouracil and either methotrexate or epirubicin given intravenously on days 1 and 8.
Earlier reports have identified a higher risk of menopause associated with larger cumulative dose or longer duration of treatment (1 year or longer). Unfortunately, we were unable to examine the effect of cumulative dose or duration of chemotherapy on risk of menopause because all women received similar doses25,26 and the vast majority of study subjects (91.6%) received 6 months of chemotherapy. None received more than this; nine women (8.4%) embarked on a planned 6 months of treatment but stopped early. The lower cumulative dose of cyclophosphamide in the CEF regimen (75 mg/m2) as opposed to the CMF regimen (100 mg/m2) may have contributed to the slightly lower rate of menopause onset seen with the CEF regimen (55.6% v 64.6%); however, other differences between the regimens may also have been important. Because AC chemotherapy was not used as adjuvant therapy for the premenopausal women in our inception cohort, we were not able to examine risk of menopause after this treatment.
Age was a strong predictor of risk of menopause in our cohort. In those receiving chemotherapy, the risk began to increase at age 35, and by age 45, a large majority became menopausal during chemotherapy. In those not receiving chemotherapy, the risk of menopause onset was low before the age of 45, regardless of whether tamoxifen was used, and it began to increase steadily after the age of 45.
We have considered menstrual function only during the first year after breast cancer diagnosis. Some women whose menses stopped during this time may have had a subsequent return of menses, but it is likely that this was an infrequent event. In previous reports, less than 11% of women over 40 who became menopausal during treatment experienced a return of menses; 12% to 15% of younger women experienced a return of menses after a period of amenorrhea.4 Because many of these women had a brief period of amenorrhea, with a return of menses during the first year after diagnosis, these observations are not directly applicable to our cohort. It is probable the frequency of return of menses after our 1-year follow-up was complete was lower than these published estimates. It is also possible that women who continue to have regular menses 1 year after diagnosis remain at increased risk for an early menopause as a result of their adjuvant therapy. Longer follow-up is needed to address this issue.
In this report, we have focused only on women who were premenopausal with regular menses at diagnosis. An additional 23 women (mean age, 50.1 years) who participated in our cohort study were perimenopausal at study entry. That is, they had irregular menses, with at least one period in the preceding 12 months. Menses stopped in all but four of these women during the subsequent year (three continued to have irregular menses and one began menstruating regularly). Because of this high rate of menopause onset (82.6%), we were unable to examine predictors of menopause in a statistically meaningful fashion in these perimenopausal women.
Our observations of the overall risk of menopause during adjuvant treatment for breast cancer are similar to those in the published literature; however, we have extended previous observations by providing estimates of the risk of menopause onset according to age at diagnosis and the specific type of systemic adjuvant treatment administered. This information should be useful in counseling patients about the side effects of treatment and, in patients concerned about fertility or menopausal complications, in assisting them in making tradeoffs between possible benefits of adjuvant treatment and increased risk of infertility or menopausal complications.
Our predictive model will likely be useful in a variety of clinical situations. For example, a woman with a low-risk, hormone receptor–positive breast cancer may be offered a choice of adjuvant chemotherapy or tamoxifen. If pregnancy is a priority for her, and she is under 35 years of age, chemotherapy might be the preferred option. Chemotherapy would be associated with a low risk of amenorrhea (< 15%) and would be completed in a short time, whereas tamoxifen, although associated with a negligible risk of menopause, would result in a 5-year deferral of pregnancy. A 40-year-old woman in the same situation would have a more difficult decision—chemotherapy would be associated with a 40% risk of amenorrhea. This woman must balance the not insubstantial risk of infertility resulting from chemotherapy against the prolonged duration of tamoxifen administration (which would require a delay in pregnancy) and the absolute reduction in risk of recurrence of her breast cancer resulting from either treatment. If the latter is small, she may choose to forego adjuvant therapy.
When childbearing is not a concern, the risk of premature menopause with its associated sequelae (vasomotor, genitourinary and psychological symptoms; increased risk of osteoporosis, heart disease and possibly other conditions) will often play a greater role in decision-making. A woman under 35 would likely be reassured to know her risk of menopause during chemotherapy is relatively low (and close to zero with tamoxifen). In contrast, a 45-year-old woman at risk for heart disease might conclude that the risks of early menopause resulting from chemotherapy outweigh any potential benefits of chemotherapy over tamoxifen and opt for tamoxifen or, when risk of recurrence is low, no treatment.
Estimates of the risk of menopause according to age and treatment facilitates the decision-making process regarding adjuvant therapy in breast cancer. This process requires that each woman balance potential benefits of treatment against potential adverse effects and future risks. When absolute benefits are small, as is the case in low-risk, node-negative breast cancer, risk of menopause and its sequelae may tip the balance toward or away from specific treatments. When absolute benefits are larger, as is the case in node-positive and high-risk, node-negative disease, effects of adjuvant therapy on disease recurrence and survival will usually outweigh menopausal considerations.
Acknowledgments
Supported by the Canadian Breast Cancer Research Initiative and the Medical Research Council of Canada.
- Received December 28, 1998.
- Accepted April 12, 1999.
