Ovarian Function Recovery During Anastrozole in Breast Cancer Patients With Chemotherapy-Induced Ovarian Function Failure
Abstract
Background: Aromatase inhibitors (AIs) are given as adjuvant therapy for hormone receptor–positive breast cancer in post- menopausal women, also to those with chemotherapy-induced ovarian function failure. The current analysis reports on endocrine data of patients with chemotherapy-induced ovarian function failure who were included in the phase III DATA study assessing different durations of adjuvant anastrozole after tamoxifen.Methods: We identified all patients with chemotherapy-induced ovarian function failure. Women who underwent a bilateral ovariectomy or used luteinizing hormone-releasing hormone agonists before random assignment were excluded. Plasma es- tradiol and follicle-stimulating hormone levels were monitored until 30 months after random assignment at local laborato- ries. We aimed to determine the ovarian function recovery (OFR) rate during AI use by the cumulative incidence competing risk method and analyzed the trend of estradiol levels during AI use by a nested case-control approach in which a subset of control subjects were compared with the OFR patients excluding the value at OFR diagnosis.Results: The 329 eligible patients had a median age of 50.0 years (range ¼ 45–57 years) at random assignment. Thirty-nine patients developed OFR, corresponding with a 30-month recovery rate of 12.4%. Of these, 11 (28.2%) were age 50 years or older at AI initiation. The estradiol level decreased statistically significantly by 37.8% (95% CI ¼ 27.4% to 46.7%) over the initial 30 months of AI treatment in both groups. However, the estradiol levels in the women who experienced OFR remained statis- tically significantly higher (difference ¼ 20.6%, 95% CI ¼ 2.0% to 42.7%) prior to OFR diagnosis compared with those who did not experience OFR.
Conclusions: The risk of OFR during AI treatment in breast cancer patients with chemotherapy-induced ovarian function failure is relevant, even beyond 45 years. Furthermore, women experiencing OFR had statistically significant higher estradiol levels during AI treatment (before OFR) than those without, with potential consequences regarding efficacy.
In early-stage hormone receptor–positive breast cancer, aroma- tase inhibitors (AIs) are an established component of adjuvant endocrine therapy for postmenopausal women (1). By inhibiting the aromatase enzyme, the conversion from androgens to estra- diol (E2) is blocked, which leads to E2 deprivation, but only in women without functioning ovaries. In daily practice, AIs are also offered to patients with chemotherapy-induced ovarian function failure. However, the use of AIs in these women has not been studied extensively, and this could be very relevant be- cause it is difficult to determine to what extent their ovaries still function. If AIs are administered to women with functioning ovaries, the gonadotropin secretion is stimulated, which may result in a vigorous increase of E2 levels that abolishes the ex- pected anticancer effect of AIs and subsequently withholds pa- tients from effective breast cancer therapy (2,3).Some small studies showed that the recovery of ovarian function (OFR) occurred in 27% to 39% of women with chemotherapy-induced ovarian function failure receiving AI treatment, months to even years after prior adjuvant chemo- therapy (3–6). However, these studies were small and patients received various prior systemic treatments.The DATA trial is a large phase III randomized clinical trial assessing the efficacy of three vs six years of adjuvant anastro- zole in postmenopausal women with hormone receptor–posi- tive breast cancer who previously received two to three years of adjuvant tamoxifen (ClinicalTrials.gov No. NCT00301457) (7). Patients age 45 years or older with chemotherapy-induced ovar- ian function failure were also eligible for the DATA trial, pro- vided they had E2 levels within the postmenopausal range within the last three months before random assignment. After the publication of Smith and colleagues, which described OFR during AI use in women with chemotherapy-induced amenor- rhea, we recommended monitoring plasma E2 and follicle-stim- ulating hormone (FSH) levels serially at six monthly intervals during the first 30 months (4). The current analysis reports on the patients with chemotherapy-induced ovarian function fail- ure and aims to assess the frequency and timing of OFR while receiving AIs after prior tamoxifen. Moreover, we analyzed the trend of E2 levels during treatment with AIs in patients who experienced OFR vs those who did not.
In the DATA trial, postmenopausal women with stage I to III hormone receptor–positive invasive early breast cancer were randomly assigned, after two to three years of adjuvant tamoxi- fen, to three or six years of anastrozole 1 mg once daily (ClinicalTrials.gov No. NCT00301457). Eligibility criteria for the DATA trial were the following: no recurrence at random assign- ment and postmenopausal status at the time of random assign- ment according to the following criteria: bilateral ovariectomy irrespective of age, age 55 years or older and natural amenorrhea for one or more years, or between age 45 and 54 years with amenorrhea and E2 levels within the postmeno- pausal range, confirmed by local postmenopausal laboratory values, assessed within the last three months before random assignment, also applicable for those with chemotherapy- induced ovarian function failure. FSH levels were not consid- ered at this point because all women were using tamoxifen until random assignment, which is known to suppress FSH levels (5,8–10). Local postmenopausal reference values for E2 and FSH
were gathered for each participating hospital (n ¼ 76) because the hospitals used various assays to measure the E2 levels. Confirmation in a central laboratory was not available. Trained registration clerks reported the laboratory values and the occur- rence of menstrual bleeding on electronic case report forms.The DATA trial protocol was initially dated per July 3, 2006, and it was amended on January 23, 2007, recommending moni- toring plasma E2 and FSH levels serially in patients under the age of 55 at six monthly intervals up to the visit at around 30 months (þ/— one month) after random assignment. During evaluation, we noticed that there were some patients older than age 55 years whose E2 and FSH levels had been checked regularly.
Therefore, for the current subanalysis, we selected patients from the DATA trial who were age 57 years or younger and were initially pre-/perimenopausal and who developed chemotherapy-induced ovarian function failure after receiving adjuvant chemotherapy that was still persisting at random as- signment with E2 measurements within postmenopausal ranges (11,12). Patients who underwent an ovariectomy or used a luteinizing hormone-releasing hormone (LHRH) agonist prior to random assignment were not eligible for the current analysis. Lastly, we excluded patients who did not have E2/FSH monitor- ing during follow-up because without data here on we could not determine whether these patients experienced OFR or not.Patients provided written informed consent before enrollment. The study was conducted in accordance with the Declaration of Helsinki and the principles of Good Clinical Practice. The study was approved by the Ethics Committee of the Radboud University Medical Center, Nijmegen, the Netherlands, in agreement with the Dutch law code for medical research on humans.The primary end point for the current analysis was OFR within 30 months after random assignment. OFR was considered when any of the following events occurred: 1) return of menstrual bleeding with premenopausal E2/FSH levels or 2) premeno- pausal E2/FSH levels only (according to the local reference val- ues in the participating hospitals). The secondary aim was to analyze the evolution of E2 levels during anastrozole use over time in the women whose ovarian function did and did not re- cover in the period preceding OFR. Furthermore, we aimed to identify baseline risk factors of OFR considering age, E2 values, body mass index (BMI), type of adjuvant chemotherapy, and du- ration of tamoxifen. The cumulative risk for OFR was estimated with the cumulative incidence competing risk (CICR) method (13). The occurrence of a disease-free survival (DFS) event including (non-)invasive breast cancer recurrences (local, regional, distant), second primary (non-)invasive (breast) cancers other than basal cell or squamous cell carcinoma of the skin and carcinoma in situ of the cervix, and death of any cause hindered the observation of an OFR and were therefore considered a competing risk. The time to recovery was censored at the last E2/FSH monitoring date or at an earlier date if recovery could not be assessed because of prior ovariectomy.Risk factors of OFR were assessed by Cox regression analysis and included age at random assignment, body mass index (BMI), type of previous (neo-)adjuvant chemotherapy, and dura- tion of previous tamoxifen therapy (5,6,14). They were presented as hazard ratios (HRs) with 95% confidence intervals (CIs).
For the explorative analyses on the trend in E2 levels, we used a nested case-control approach in which a subset of con- trols were compared with the OFR patients. Because different hospitals used different tests with a wide range of reference val- ues, we could not use the absolute E2 values. The E2 levels were logarithmic-transformed and analyzed in a linear mixed model with random effects for hospital and patient to take into ac- count the dependency of measurements within the same pa- tient and within the same hospital. Hereby, we compared the E2 levels of the women with and without OFR per hospital to over- come the differences in local reference values between the hos- pitals. A model with linear trend in time and the interaction of the group with time was analyzed. The interaction term reflects the difference in trend in E2. Because we were interested in the development of the E2 levels preceding OFR, we did not consider the high E2 levels at the time of OFR diagnosis and thereafter. The reported P values are calculated with Wald tests. All sta- tistical tests were two-sided, and a P value of less than .05 was considered statistically significant.
Results
Of the 1860 patients included in the DATA trial, 841 patients were age 45 to 57 years at random assignment (Figure 1). Of them, 51 had not received chemotherapy, 167 had their last menstrual bleeding more than one year before diagnosis and were consid- ered definitely postmenopausal, 104 underwent a bilateral ovari- ectomy or used LHRH agonists before inclusion, 166 had no baseline or follow-up E2 measurements available, and 24 patients had premenopausal E2 levels at random assignment. Hence, a to- tal of 329 patients were included in the current analysis.The baseline characteristics are presented in Table 1. The me- dian age at diagnosis was 48.0 years (range ¼ 41–54 years), and at random assignment the median was 50.0 years (range ¼ 45–57 years). Eighty-three percent had received (neo-)adjuvant chemo- therapy consisting of anthracyclines without taxanes prior to study entry (median number of five cycles, range ¼ 3–6 cycles).Among the 329 included patients, the recovery rate was 8.1% (n¼ 26) at 12 months after AI initiation; 0.6% had a DFS event. At 30 months, 12.4% (n ¼ 38) had experienced OFR and 4.1% had a DFS event (Figure 2A). The figure implies that this number could have increased even further with longer follow-up. Of the total of 39 patients experiencing OFR, 11 (28.2%) were age 50 years or older at AI initiation. Considering the patients being age 50 years or older (n ¼ 209) at random assignment, the 30-month re-covery rate was 5.1%, vs 25.2% for patients younger than age 50years (Figure 2B). The DFS event rates were 4.7% and 3.0%, re- spectively, for these age groups.
Nineteen out of 39 patients (48.7%) reported vaginal bleeding. Adjuvant endocrine treat- ment was adjusted in 27 (69.2%) patients: 12 patients received an LHRH agonist combined with AIs, seven underwent an oo- phorectomy, six switched to tamoxifen (in two patients com- bined with an LHRH agonist), and two received LHRH agonist monotherapy. In 12 patients, no treatment adjustments were made. In the univariate analysis, age younger than 50 years atAI initiation was the only factor that was related to OFR (HR ¼4.85, 95% CI ¼ 2.41 to 9.75, P < .001) (Table 2).The pattern of E2 levels over time of 38 patients with OFR and 102 patients without OFR from 21 hospitals was analyzed. The baseline characteristics of these patients were similar to the overall study population. As previously mentioned, the E2 level by which OFR was diagnosed was not included in the analysis. The relative decrease of E2 over time in patients who developed OFR during follow-up did not statistically significantly differfrom the relative decrease in patients without OFR (P ¼ .82)(data not shown). Over the 30 months after AI initiation, the E2 levels decreased statistically significantly by 37.8% (95% CI ¼ 27.4% to 46.7%, P < .001) (Figure 3). However, the E2 level inthe OFR group, prior to OFR diagnosis, was statistically signifi-cantly higher (difference ¼ 20.6%, 95% CI ¼ 2.0% to 42.7%,P ¼ .03) in comparison with the non-OFR group (Figure 3). Discussion 45 and 57 years at random assignment with chemotherapy- induced ovarian function failure experienced OFR while on ad- juvant anastrozole after adjuvant tamoxifen. Women under the age of 50 years at AI initiation were at a statistically significantly higher risk of OFR. Yet, of the 39 patients experiencing OFR, 11 (28.2%) were age 50 years or older at AI initiation.To our knowledge, the current analysis reports on the largest population to date on the occurrence of OFR in a relatively old pa- tient group. Earlier but smaller studies have also reported on OFR during AI use in breast cancer patients with chemotherapy- induced ovarian function failure and reported OFR rates of 27% to 39% (3–6). These studies are summarized in Table 3. The outcomes of our study are in line with earlier findings; however, in our study all patients were older than age 45 years at AI initiation, and all re- ceived two to three years of prior adjuvant tamoxifen. Moreover, our data underscore that amenorrhea can still be present despite biochemical OFR because only half of the women experiencing OFR in our patient cohort reported menstrual bleeding. Of note, in 12 OFR patients (30.7%), no treatment adjustments were made by the responsible physician. This stresses the need for a guideline on monitoring these patients during AI treatment, a clear defini- tion of OFR, and how treatment should be adjusted in case of OFR. Monitoring E2 and FSH levels can be challenging because the performances of most E2 assays are poor and diverse tests are used in different laboratories (15). Additionally, measuring E2 levels in patients receiving steroidal AIs (eg, exemestane) can be problematic because there is cross-reaction of various metabo- lites of this drug, even in specialized immunoassays (16). Guerrero and colleagues showed, however, that different assays when applied to the same patient resulted in a similar incidence of OFR and time to OFR, indicating that one can trust the refer- ence values of the local test (3). Based on our findings, we sug- gest monitoring E2 and FSH levels at three-monthly intervals for the first year of AI therapy because OFR occurred in 67% of the patients within the first year, and at six-month intervals during the second and third years of therapy. Furthermore, we suggest as a definition of OFR: “menstrual bleeding confirmed by premenopausal E2/FSH levels or biochemical OFR by preme- nopausal E2/FSH levels.” In a combined analysis of data from the Tamoxifen and Exemestane Trial (TEXT) and the Suppression of Ovarian Function Trial (SOFT), ovarian function suppression in addition to exemestane resulted in an improved DFS, but not in an over- all survival difference (17,18). However, the results of the SOFT- EST trial (19), showing that 34.2% of patients had E2 levels above the threshold at least once during 12 months of follow-up, un- derline the importance of monitoring the extent of ovarian function suppression by gonadotropin-releasing hormone (GnRH) agonists in premenopausal women treated with an AI. In the ABCSG-12 trial, among premenopausal women who re- ceived GnRH agonists and either tamoxifen or anastrozole, those who received anastrozole actually had a statistically sig- nificantly worse overall survival (20). The ABCSG investigators hypothesized that the effect might be the result of incomplete E2 level suppression by AIs in overweight women. Therefore, until further results on the survival of these women are known, we would advise performing either a bilateral ovariectomy while continuing an AI or switching to tamoxifen monotherapy. Moreover, Guerrero and colleagues showed that E2 levels un- der AI treatment are statistically significantly higher in younger women (mean ¼ 48.1 years, range ¼ 41–55 years) than in women older than age 60 years (3). We also observed high E2 levels dur-ing AI treatment in a substantial number of patients in our study, which may compromise the efficacy of endocrine treat- ment. However, to what extent the suboptimal E2 suppression influenced breast cancer survival is not known yet. This leads to the more fundamental question of whether a more powerful de- crease in E2 levels leads to an improvement in terms of survival. To our understanding, there are few data about the clinical con- sequences of the extent of E2 reduction during AI treatment. Letrozole has been shown to decrease plasma E2 levels to a greater extent in comparison with anastrozole, although its effi- cacy in terms of survival has not showed superiority over anas- trozole or exemestane (21–23).In the current study, all patients were treated with tamoxi- fen for two to three years before they received adjuvant AI treat- ment. The ovarian response to tamoxifen is not completely understood, and its effect on the incidence of chemotherapy- induced ovarian function failure is not well defined (24). Walshe and colleagues reviewed studies focusing on amenorrhea in early breast cancer patients treated with various regimens of adjuvant chemotherapy (25). Some studies reported no impact of tamoxifen on chemotherapy-induced amenorrhea, whereas larger studies reported higher rates of chemotherapy-induced amenorrhea. A later prospective trial, including 595 patients age 20 to 45 years, showed that adjuvant tamoxifen was less likely associated with monthly menstrual bleeding during the firstyear following chemotherapy (odds ratio [OR] ¼ 0.50, 95% CI ¼0.37 to 0.67) (26). Tamoxifen seemed responsible for the persis- tence of amenorrhea after adjuvant chemotherapy in 15% of women. A meta-analysis showed that the use of adjuvant ta- moxifen statistically significant increased the incidence of chemotherapy-induced amenorrhea (OR ¼ 1.48, 95% CI ¼ 1.28 to1.70) (27). The mechanism for this phenomenon remainsunclear though. It is thought that tamoxifen has estrogen-like actions on gonadotropin, which causes a disruption of the hy- pothalamic ovarian feedback loop, resulting in a decrease in FSH and E2 synthesis (8–10).Previous studies ascribed the increased probability of OFR in patients with chemotherapy-induced ovarian function failure during AI treatment to the increased secretion of gonadotropin caused by the reduced feedback of estrogen to the hypothala- mus and pituitary (2–6,28). This seems a logical explanation be- cause AIs additionally proved to be effective in inducing ovulation in women with polycystic ovarian syndrome (28,29). Nevertheless, OFR could theoretically also be triggered by the discontinuation of tamoxifen itself, rather than solely by AI use. It is thought that by discontinuing tamoxifen, the negative feed- back on the hypothalamic ovarian feedback loop diminishes and consequently OFR is more likely to occur. An important limitation of the current analysis was that the E2 measurements were not standardized between the different hospitals participating in this study. Different assays were used, with different reference values, which made it difficult to com- pare the measurements. Therefore, the E2 measurements in OFR patients were compared with those in no-OFR patients from the same hospital in a linear mixed model; however, it was not possible to give a recommendation about absolute E2 cutoff values for OFR. Nevertheless, others have showed that different assays when applied to the same patient resulted in a similar incidence of OFR and time to OFR, indicating that one can trust the reference values of the local test (3).The results of this study show that we need to be aware of the risk of OFR during AI treatment in early breast cancer pa- tients with chemotherapy-induced ovarian function failure, even in women older than age 45 years. Furthermore, women with (vs without) OFR had statistically significantly higher E2 levels during AI treatment, which has potential consequences regarding the optimal efficacy. Overall, AIs should be used with caution in women with chemotherapy-induced ovarian func- tion failure. When considering AIs in these women, the hormone levels should be monitored Tamoxifen closely.