Introduction

NRG Oncology/NSABP B‐47 is a phase III, multicenter, randomized adjuvant therapy trial designed to evaluate the addition of trastuzumab to adjuvant chemotherapy in patients with HER2-low breast cancer. The primary endpoint was invasive disease‐free survival (IDFS) and clinical results showed no benefit from the addition of trastuzumab to chemotherapy in this population1. Because the impact of trastuzumab on menstrual function was unknown at the time of protocol design, a secondary endpoint focused on menstrual health. The primary aims of the menstrual history (MH) sub-study were to assemble an observational cohort of pre- and perimenopausal women to evaluate the effect of chemotherapy with or without trastuzumab on treatment-related amenorrhea (TRA) and the associations between TRA and circulating reproductive hormone levels. Secondary exploratory aims examined associations between chemotherapy regimen, TRA, and IDFS benefit in premenopausal women. Here we report on these outcomes, as well as medical and demographic predictors of TRA.

Results

Patients

The B-47 protocol entered 3270 patients between February 8, 2011 and February 10, 2015. The MH sub-study enrolled 1458 eligible women, and 1231 consented to blood sample collection. This report uses data collected through June 30, 2019. Figure 1 shows the study CONSORT diagram. Both baseline and at least one follow-up MH form were available for 1428 patients. There was high compliance with form submission across the 36 months. Both baseline and at least one follow‐up blood hormone measure were available for 1123 patients.

Fig. 1
figure 1

CONSORT Diagram. NSABP B-47 Menstrual History Sub-study.

The mean age of participants was 44.1 years. Table 1 shows the distribution and tumor characteristics by randomized treatment groups and by chemotherapy regimens for all analyzed MH patients. There were no differences in characteristics between those assigned to trastuzumab vs no trastuzumab. Anthracycline regimen recipients were statistically significantly younger, had more positive nodes, were more likely to have hormone receptor (HR)-negative cancer, a higher histologic grade, and to have had a mastectomy, compared to those treated with the non-anthracycline regimen. Participants in the MH sub-study were significantly younger (P < 0.0001); more likely to be non-White (P = 0.003); Hispanic (P = 0.01); hormone-receptor positive (<0.0001); and to have received doxorubicin and cyclophosphamide (AC) followed by weekly paclitaxel (WP) for 12 weeks (P < 0.0001) as compared to the B-47 participants not in the MH sub-study (data not shown).

Table 1 Patient and tumor characteristics by treatment group and intended chemotherapy regimen for analyzed patients: NSABP B-47.

Amenorrhea

We hypothesized that the cessation of menses for at least 6 months’ duration would be associated with postmenopausal estradiol levels and elevated follicle-stimulating hormone (FSH) levels. However, the rate of agreement between amenorrhea status and estradiol and FSH levels at both 12 and 24 months was poor (Supplementary Table 1). At 12 months, the kappa statistics were 0.36, 0.11, and 0.09, for estradiol, FSH, and the combined indicator for estradiol and FSH, respectively. At 24 months, the kappa statistics were 0.44, 0.08, and 0.08, respectively. Of the 648 women who had data available for MH, estradiol, and FSH at 24 months, there were 410 with TRA, however, hormone levels were in the premenopausal range (estradiol ≥ 20 pg/mL and FSH < 50 mIU/mL). Of those, 360 (88%) reported the use of tamoxifen.

We also hypothesized that trastuzumab would have no effect on the rate of cessation of menses for ≥6 months. As shown in Supplementary Fig. 1, there were no differences in TRA rates between the trastuzumab and non-trastuzumab intention‐to‐treat groups. Furthermore, there were no statistically significant differences between treatment groups after adjusting for age at any of the three time points. At 12 months, 84.0% of patients in the trastuzumab group were amenorrheic vs 86.3% in the non-trastuzumab group (P = 0.20). Age-adjusted P values at 24 and 36 months were 0.81 and 0.83, respectively (Table 2).

Table 2 Results of logistic regressiona predicting amenorrhea status at 12, 24, and 36 months: NSABP B-47.

Table 2 also presents the results for other potential explanatory factors in predicting TRA at each time point. We first looked at age, confirming that it was a statistically significant predictor of TRA at 12, 24, and 36 months. After adjusting for age, there was no difference in amenorrhea status between groups defined by ethnicity, chemotherapy regimen, IHC score, number of positive nodes, or histologic grade. There was a statistically significant difference in race at 24 months with Black women having a decreased likelihood of TRA compared to White women. Body mass index (BMI) was a statistically significant predictor of TRA at 12 and 24 months, in that obese women were less likely to become amenorrheic. HR status/ET use was also a statistically significant predictor for TRA at each time point. Specifically, compared to HR‐negative patients, those who were HR‐positive and had reported use of tamoxifen or other endocrine therapy (ET) were more likely to be amenorrheic.

At each time point, we investigated a multivariable model with potential explanatory variables using backward elimination. For those models, the HR status/ET use was examined instead of HR status alone. At 12 and 24 months, all variables dropped out of the model except age, BMI, and HR status/ET use. However, at 36 months, chemotherapy regimen was a statistically significant predictor with age and HR status/ET. Supplementary Table 2 provides results of this multivariable model, showing that the non-anthracycline regimen had lower odds of TRA (OR = 0.60, 95% CI = 0.39–0.94; P = 0.02).

Estradiol and FSH

The distribution of estradiol and FSH levels categorized into estradiol <20 pg/mL and FSH ≥ 50 mIU/mL versus estradiol ≥20 pg/mL and/or FSH < 50 mIU/mL across all time points is presented in Fig. 2. The baseline percent of patients with estradiol and FSH levels in the postmenopausal range (estradiol < 20 pg/mL and FSH ≥ 50 mIU/mL) was 4.2%. This increased to 49.6% (3 months) and 52.6% (6 months), then dropped to 27.8% (12 months) and 22% (24 months). Supplementary Table 3 presents logistic regression results for predicting low estradiol and high FSH. At 6 months, there were statistically significant differences between chemotherapy regimens (P = 0.008), IHC scores (P = 0.004), and BMI groups (P < 0.0001). Obese women were less likely to have estradiol and FSH levels in the postmenopausal range as compared to women in the normal BMI range. BMI remained a statistically significant predictor at 12 months, but not at 24 months. Histologic grade was a statistically significant predictor at both 12 and 24 months (P = 0.03 at both time points), with increased odds of postmenopausal hormone levels in those with high vs low grade. ET use during the first 6 months after randomization was not available, but HR status/ET use was a statistically significant predictor at 12 and 24 months, with overall P values <0.0001 at both time points. When considering a multivariable model at 24 months, only age and HR status/ET use (controlling for baseline estradiol and FSH level), remained statistically significant. Patients with HR‐positive cancer with reported tamoxifen use were less likely to have estradiol and FSH levels in the postmenopausal range as compared to HR‐negative patients (OR = 0.24, 95% CI = 0.15–0.39).

Fig. 2
figure 2

Distribution of Combined Estradiol and FSH levels over Time.

Figure 3 shows the mean FSH levels respectively over time by treatment group (a), chemotherapy regimen (b), HR status/ET use (c), age group (d), and BMI (e). Treatment groups did not differ over time. Overall, chemotherapy regimen did not reach statistical significance (P = 0.08); however, the plot shows a qualitative interaction over time (P < 0.0001). HR status/ET use (defined as ever used in 24 months) was statistically significant and varied over time (P < 0.0001) with the tamoxifen group having the lowest FSH levels at 12 and 24 months. As expected, there was a difference in age with younger women having lower FSH levels. BMI was also statistically significant, and the magnitude of the difference varied over time (P < 0.0001). Obese women consistently had lower FSH levels than those with normal BMIs. Underweight women had higher FSH levels early on, but lower levels by 24 months. However, it should be noted that due to the low number of women in the underweight category, these estimates are inconclusive.

Fig. 3
figure 3

Mean Follicle-stimulating Hormone (FSH) Levels Respectively over Time in NSABP B-47. Levels are presented by: treatment group (a), chemotherapy regimen (b), HR status/endocrine therapy use (c), age group (d), and BMI (e). A higher FSH level indicates a late premenopausal/postmenopausal range. Adjusted least-square means for time points 3 months and beyond were obtained from a mixed model for repeated measures analysis of the FSH level. The model includes age, FSH baseline level, the variable of interest, time, and variable-by-time interaction.

Persistent amenorrhea

An estimate of persistent amenorrhea was also of interest. Supplementary Table 4 shows the distribution of amenorrhea status at 24 and 36 months among those amenorrheic at 12 months. Among those amenorrheic at 12 months, 79.0% remained amenorrheic at 24 months, and 70.0% remained amenorrheic at 36 months. Persistent amenorrhea in relation to postmenopausal levels of FSH and estradiol are also presented. Among those who were amenorrheic at 12 months with high FSH and low estradiol at 12 months, 84.1% and 74.8% remained amenorrheic at 24 and 36 months, respectively.

Amenorrhea and IDFS

Although underpowered in this trial subgroup, we investigated the prognostic effect of TRA using each patient’s amenorrhea status at 12 months. There were 1249 patients in the landmark population, and median follow-up time from 12 months post-randomization was 4.2 years, with 76 IDFS events. Five-year IDFS was 93.6% and 92.5% in the amenorrhea and no amenorrhea groups, respectively, HR 0.67 (95% CI = 0.36–1.28), P = 0.23 (Supplementary Fig. 2).

Discussion

In this large, prospective, observational study of pre- and perimenopausal women receiving contemporary chemotherapy regimens, we found no difference in TRA between the trastuzumab and non-trastuzumab groups. Blood specimens were also collected to measure menopausal reproductive hormones to evaluate their usefulness in tracking the onset of menopause, along with clinical documentation of menstrual status. Using a combination of estradiol levels in the postmenopausal range and elevated levels of FSH, at 6 months after random assignment, more than half of the women had postmenopausal hormone profiles; however, by 24 months reproductive hormone levels had improved such that 22% had postmenopausal levels. Among those with premenopausal hormone levels and amenorrhea, ET with tamoxifen was likely a contributing factor to persistent amenorrhea.

Adjuvant chemotherapy has known risks for the development of TRA in premenopausal women2,3. Although chemotherapy treatment regimen was selected by investigators from two established regimens, and was equally distributed across the two treatment arms in the entire B-47 trial1, among the younger women in the MH sub-study, the anthracycline-containing regimen was more often selected (69%). In addition, the anthracycline-containing regimen was more likely to be chosen for women with HR-negative tumors, positive nodes, and high-grade tumors (Table 1). Thus, in addition to randomized treatment group, intended chemotherapy regimen was also an important exploratory variable.

Only a few studies have examined the impact of adjuvant HER2-directed therapy on menstrual status4,5,6. Lambertini et al.4 recently reported on the impact of adjuvant lapatinib and/or trastuzumab (four treatment arms) in the BIG 2-06 trial, in which menopausal status data were collected at randomization and at the 37-week visit. TRA was described in 2682 premenopausal women, and the impact of amenorrhea on disease-free and overall survival were examined using landmark and time-dependent modeling. Prior to randomization, patients had received heterogeneous adjuvant chemotherapy regimens that were included in the stratification. In addition, diverse ET treatments were received in women with HR-positive tumors during the course of the HER2-directed therapy. To be included in the premenopausal sample, patients had to have had a menstrual period within 6 months of randomization or be <50 years. In the TRA cohort, those with amenorrhea at 37 weeks were significantly older, had higher nodal status, and were more likely to be HR-positive. These investigators found no significant difference across the four treatment arms for rates of TRA in univariate analysis, and in multivariable analysis, the factors associated with higher risk of TRA were older age at diagnosis, the addition of taxanes to anthracycline-based chemotherapy, administration of docetaxel, carboplatin and trastuzumab treatment, and adjuvant ET use. With a median follow-up of 6.9 years, in the landmark analysis, a statistically significant interaction between TRA and HR status was found for DFS and OS.

Unlike the Lambertini study4, in B-47 we were able to examine the independent effects of chemotherapy on risk for amenorrhea, with and without the use of trastuzumab, and did not find that trastuzumab contributed to TRA. Rather, age, intended chemotherapy regimen, HR status, and ET use were associated with persistent amenorrhea at 36 months post randomization. As noted in the NSABP B-30 trial2, women receiving ET were more likely to remain amenorrheic than those without ET. Our landmark analysis of survival outcomes in the B-47 population was underpowered to examine the effects of TRA on these outcomes, and thus we cannot confirm the findings noted by Lambertini.

To the best of our knowledge, this is one of the largest prospective studies of menstrual function in premenopausal patients with breast cancer receiving contemporary adjuvant chemotherapy, with or without trastuzumab, including assessment of reproductive hormones and amenorrhea. This report provides much new information about the trajectory of TRA in younger patients with breast cancer and provides insight into the biology of TRA in this setting. Return of menses does not guarantee residual fertility, and more detailed information about the ovarian reserve is necessary to make clinical decisions about fertility in this patient population. Additional evaluation of some B-47 MH study participants that has been completed explores the added value of measuring anti-Mullerian hormone in this setting, to assess its role in understanding TRA7.

The findings from B-47 demonstrate the unreliability of amenorrhea in assessing menopausal status in younger breast cancer patients after adjuvant chemotherapy. This has implications for choice of ET, as well as considerations related to contraception. Laboratory assessment of estradiol and FSH should be performed to document menopausal status before considering initiation of aromatase inhibitors in this setting, because persistent amenorrhea to 36 months after treatment does not ensure postmenopausal status in premenopausal women.

Methods

MH study population

In addition to the main trial eligibility/exclusion criteria1, sub-study participant eligibility included: women with an intact uterus and at least one ovary were eligible if they had at least one menstrual period in the past 12 months; no use of current oral contraceptive hormones or other hormone replacement therapy. After written informed consent, eligible women were enrolled in the MH sub-study upon random assignment to the parent trial. Approximately 1500 eligible women were expected based on the planned total sample size.

Treatments

Patients were randomly assigned to receive chemotherapy with or without trastuzumab, and were stratified by HER2 IHC score (1+ vs 2+), pathological nodal status (0–3, 4–9, ≥10), HR status (ER+ or PgR+ vs both negative), and intended chemotherapy regimen. Investigator choice between two chemotherapy regimens was declared upon study entry. An anthracycline (doxorubicin, cyclophosphamide followed by WP) or non-anthracycline regimen (docetaxel plus carboplatin) were the two options1. Patients received adjuvant radiotherapy and ET, as clinically indicated.

Data collection

Study coordinators assessed MH status at baseline and at 3, 6, 12, 18, 24, 30, and 36 months of follow‐up using a standard questionnaire designed to assess menstrual bleeding history, any reported changes in menstrual cycle length, and history of hysterectomy and/or bilateral oophorectomy2. In the MH sub-study, consent was required for a blood specimen to measure reproductive hormones at baseline through 3, 6, 12, 18, and 24 months. To replicate assessments similar to routine clinical practice, blood samples were not timed to the menstrual cycle. Investigators were instructed to collect the sample within a one-week window surrounding the corresponding MH assessment with preference given to the same day. Baseline and follow‐up blood specimens were sent to a central clinical laboratory (Covance Laboratories, Indianapolis, IN) where standard laboratory assessment methodologies were used for estradiol and FSH. MH assessments and specimen collection were not continued for patients with an invasive breast cancer recurrence or second primary cancer.

ET use was assessed annually. If a patient reported any tamoxifen use during the time between randomization and the time point in question, then she was counted as a tamoxifen user. If any other ET medication was reported, she was included in an “other category.” All those with unknown ET status were excluded. For this analysis, a combination variable of HR status and ET use was constructed. The variable was defined with four categories: HR‐negative, HR‐positive with no ET use, HR‐positive with tamoxifen use, and HR‐positive with other use. Note that almost all the patients in the “other” category reported the use of aromatase inhibitors.

Statistical considerations

Amenorrhea status was evaluated separately at the 12, 24, and 36‐month time points. Women whose last reported period was ≥6 months prior to the specific time point were categorized as amenorrheic. Patients who received a hysterectomy or bilateral oophorectomy prior to the specific time point were removed from the corresponding analyses. Amenorrhea status at each time point was compared between treatment groups using logistic regression. Logistic regression (univariable/multivariable) was used to examine other potential explanatory amenorrhea predictors including age, race, ethnicity, BMI, chemotherapy regimen, IHC score, number of positive nodes, HR status, histologic grade, and HR status/ET use. All comparisons were adjusted for age, as a known variable associated with amenorrhea. To investigate the prognostic effect of amenorrhea, we conducted a 12-month landmark Kaplan-Meier analysis for IDFS, stratified by treatment group, IHC score, number of positive nodes, HR status, and chemotherapy regimen.

Estradiol and FSH levels were dichotomized based on predetermined thresholds for identifying levels in postmenopausal ranges. An estradiol level of <20 pg/mL and an FSH level of >50 mIU/mL were used as cutoffs. The latter was chosen as a conservative measure of late menopausal transition as defined by the Stages of Reproductive Aging Workshop criteria8, for which amenorrhea >60 days and FSH > 25 define this stage. In addition, we examined the simultaneous occurrence of low estradiol and high FSH as a stronger indicator of loss of ovarian function. The dichotomized groups were then compared between treatment arms and other categories of explanatory variables as described above for amenorrhea. We also examined the relationships between hormone levels and TRA at both 12 and 24 months. A Cohen’s simple kappa statistic for the rate of agreement was calculated to determine whether amenorrheic patients were more likely to have hormone suppression, without regard to treatment assignment or other predictors.

The laboratory reported the lowest levels of estradiol as either <19 or <15 pg/mL, preventing exploration of mean values. However, FSH was reported as a continuous value. Thus, we explored mean FSH levels over time by treatment group, intended chemotherapy regimen, HR status/ET use, age, and BMI. Mixed model analyses with repeated measures were used to obtain least-square means which were plotted over time points.

Amenorrhea analyses included patients who completed the baseline MH questionnaire and at least one follow‐up questionnaire. Hormone analyses included patients with baseline samples and at least one follow‐up time point. All analyses followed an intention‐to‐treat principle for both trastuzumab and chemotherapy regimen. All assessments were based on a two-sided test with α of 0.05.

This multicenter trial was approved by the National Cancer Institute’s Central Institutional Review Board and/or local human investigations committees or institutional review boards at institutions participating in this multicenter trial in accordance with National Cancer Institute policies and procedures and assurances filed with and approved by the Department of Health and Human Services.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.