It is suspected that estrogen depletion resulting from treatment may contribute to cognitive compromise in patients with breast cancer. However, the evidence for estrogen effects on cognition is inconclusive, and the consequences of hormonal changes for cognitive function in patients with cancer rarely have been investigated. In this study, the authors investigated the effects of treatment-induced menopause and antiestrogen therapy with tamoxifen and aromatase inhibitors (AIs) on cognitive function.
Cognitive performance was assessed in 101 patients with breast cancer before the start of cancer therapy (T1), toward the end of neoadjuvant chemotherapy (T2), and 1 year after baseline (T3) using 12 cognitive tests. Menopause occurred in a subgroup of patients, and an overlapping subgroup started antiestrogen therapy with tamoxifen or AIs. Linear mixed-effects models that made it possible to determine effects at group levels and individual levels simultaneously were used for statistical analysis.
At the group level, a significant favorable effect of induced menopause emerged in a test of executive function (P = .0035). Two additional group-level effects of induced menopause, both favorable, and 2 individual-level effects that were positive in some patients and negative in others were not significant when multiple testing was taken into account. No significant effects of tamoxifen or AIs on cognitive function were observed.
Starting in the middle 1990s, a series of studies reported cognitive compromise in subsets of breast cancer survivors who had been treated with cytostatic agents.1–10 Cognitive dysfunction in cancer patients consequently has been attributed almost unanimously to chemotherapy.11 However, many issues associated with the concept are unresolved,12 and some recent observations are inconsistent with this conclusion. The largest prospective trial conducted to date did not identify any detrimental effects of chemotherapy on the course of cognitive function.13 In addition, cognitive compromise has been demonstrated in cancer patients even before they receive cytostatic treatment.14–16 These findings, along with an increasing awareness of methodological problems11, 17, 18 that limit the validity of much of the evidence, have led to a broader view of the causation of cognitive dysfunction in cancer patients. Multiple causes may contribute to the condition colloquially dubbed ‘chemobrain.’12, 19 One of the factors suspected to affect breast cancer patients' cognitive functioning is treatment-induced hormonal changes.
To date, the role of estrogen in cognitive functioning still is not understood well. Most laboratory studies using animal models have indicated beneficial effects of estrogen on memory and learning,20, 21 and plausible biologic explanations for these effects have been suggested.22, 23 However, results of studies in humans are inconsistent. The population-based Rotterdam study tested the relation between physiologic estrogen levels and the risk of dementia in more than 1000 subjects aged ≥55 years. In women, higher levels of estradiol were associated with an elevated risk of dementia.24 Many studies have tested the effects of hormone-replacement therapy (HRT) on cognition. The Women's Health Initiative Memory Study (WHIMS) reported an increased incidence of dementia and more cognitive decline25–27 in women who were assigned randomly to receive HRT compared with women who were assigned to receive placebo. Before the WHIMS, beneficial effects of HRT on several cognitive domains and a reduction in the risk of dementia had been reported, as well as no effects of HRT on cognitive function.22, 28, 29
Very little is known about the effects of treatment-associated hormonal changes on cognitive function in patients with breast cancer. In the majority of premenopausal breast cancer patients, abrupt menopause is induced either by chemotherapy or by treatment with gonadotropin-releasing–hormone (GnRH) agonists. In most patients who are rendered postmenopausal in this way, and in the majority of naturally postmenopausal breast cancer patients, further estrogen depletion is achieved with antiestrogen agents, mainly tamoxifen or AIs.
To our knowledge, only 1 study has assessed the effects of treatment-induced menopause on cognitive function. In a prospective trial that investigated the course of cognitive function in breast cancer patients, Jenkins et al13 observed an increased risk of cognitive decline in patients who underwent treatment-induced menopause compared with patients who already had been postmenopausal at the start of therapy (P = .086).
Tamoxifen, a selective estrogen receptor modulator (SERM), has estrogen-antagonistic effects on some tissues and estrogen-agonistic effects on others. In studies that used neuroimaging, both estrogen-like positive effects on the brain30 and negative effects opposed to those of estrogen31 have been detected. Paganini-Hill and Clark mailed cognitive tests to a large group of breast cancer survivors. Current users of tamoxifen had significantly poorer scores than past users and never-users in 1 of 3 tests (P = 0.03), and there were no differences between never-users and past users.32 Several studies that assessed the effects of chemotherapy on cognitive function also determined the effects of tamoxifen, identifying either no effects4, 6 or adverse effects.3, 7, 33
The evidence regarding the effects of AIs on cognition is sparse.34 In a pilot study, patients with breast cancer who received either tamoxifen, or AIs (anastrozole), or both had poorer test results than healthy control individuals.35 In another study, patients who received AIs (anastrozole; n = 15) were compared with patients who received tamoxifen (n = 16). Significantly poorer performance in learning and memory measures was reported in those who received anastrozole.36
The current report is based on the multicenter Cognitive Impairment in Therapy of Breast Cancer study, which was designed to assess the effects of chemotherapy, induced menopause, and medication with erythropoietin on cognitive function in patients with breast cancer. Cognitive function was assessed before the start of neoadjuvant chemotherapy (T1), toward the end (T2), and about 7 months after the completion of neoadjuvant chemotherapy (T3). A previous report on the effects of chemotherapy and concomitant erythropoietin was based on data from the first 2 assessments.16 The current report focuses on the consequences of induced menopause and antiestrogen therapy and comprises the data from all 3 assessments.
MATERIALS AND METHODS
All patients were aged <66 years at T1; had nonmetastatic, invasive breast carcinoma that measured ≥2 cm or inflammatory lesions; and received 1 of 2 neoadjuvant chemotherapy regimens containing epirubicin, paclitaxel, and either cyclophosphamide alone or cyclophosphamide, methotrexate, and fluorouracil at 1 of 5 gynecologic or oncologic centers in Bavaria, Germany. Details of the participation criteria, patient accrual, and chemotherapy regimens have been reported previously.16 Chemotherapy was followed by surgery with or without radiotherapy. Patients with carcinomas that were positive for steroid hormone receptors received antiestrogen treatment with either tamoxifen (20 mg) or AIs (anastrozole 1 mg or letrozole 2.5 mg) and additionally GnRH agonists when appropriate. Only the data for patients who completed at least the first 2 assessments are considered here.
Written informed consent was obtained from all patients. The study was approved by the ethics committees of all involved universities.
Cognitive function was assessed before the start of neoadjuvant chemotherapy (T1), before the final chemotherapy session—ie, approximately 5 months later (T2)—and approximately 1 year after the baseline assessment (T3).
Twelve cognitive tests were used, including Logical Memory I and II (verbal memory) from the Wechsler Memory Scale Revised (WMS-R),37 Digit Span Forward (verbal short-term memory, attention), and Digit Span Backward (verbal working memory) from the WMS-R,37 Digit Symbol (psychomotor function, information processing speed) from the Wechsler Adult Intelligence Scale-Revised,38 the d2 Test39 (concentration), Trail Making Test Part A40 (TMT-A) (psychomotor function, selective attention), TMT-B40 (psychomotor function, divided attention, cognitive flexibility), and the Regensburg Word Fluency Test (RWT)41 (executive function) with 4 subtests (Category Fluency, Category Fluency With Switch of Category, Letter Fluency, and Letter Fluency With Switch of Letter). Anxiety and depression were measured using the Hospital Anxiety and Depression Scale (HADS).42 Premorbid intelligence was assessed once at T2 using the MWT-B,43 which is a validated, German language-related test. The tests and questionnaires were administered in a fixed order either in a hospital or in patients' homes by 1 of 2 university-qualified psychologists.
For statistical analysis, t tests or binomial tests, as appropriate, were conducted for comparison with test norms representative of the German population,37–39, 41, 43 with 2 exceptions: For the TMT-A and TMT-B, test norms from a United States normative study were used.44
To determine the effects of hormonal changes on the test results, 2 factors were conceptualized: 1) ‘induced menopause,’ which included the transition from premenopausal status at T1 to perimenopausal or postmenopausal status at T2 or T3 and was effected either by chemotherapy or by GnRH agonists (patients with regular menses [last menstruation no more than 6 weeks previously when assessed] were considered premenopausal, and all other patients were labeled postmenopausal); and 2) ‘antiestrogen therapy,’ which comprised treatment with tamoxifen or AIs. Linear mixed-effects models were used to examine the effects of induced menopause and antiestrogen therapy on cognitive performance. Linear mixed-effects models are suitable for accommodating data from multiple longitudinal assessments, such as the 3 time points T1, T2, and T3 in this study, and for simultaneously determining the effects of multiple factors at a group level (‘fixed effects’) and at an individual level (‘random effects’). Two of 12 cognitive tests, Digit Span Forward and Backward, did not meet the statistical requirements for linear mixed models—namely, continuity of data. For these tests, rank-based linear models were used. With the RWTs, parallel versions were administered at the different assessments; therefore, the raw data could not be compared directly. Instead, percentage ranks corrected for age were used in the analysis.
In all models, 2 factor levels for induced menopause (present/not present) and 3 factor levels for antiestrogen therapy (tamoxifen/AIs/neither) were introduced. Additional factors that were entered into the models were premorbid intelligence, age, and time of assessment (T1/T2/T3). The inclusion of additional factors (HADS Anxiety, HADS Depression, previous HRT, GnRH agonists as a separate factor) and interactions (induced menopause with antiestrogen therapy) also were tested. These factors were kept in the model only if their inclusion produced significant results. In a second step, the significance of factor effects was corrected for multiple testing using the Bonferroni method.
Linear model analyses were carried out with the MIXED procedure (SAS, version 9.1; SAS Inc., Cary, NC). For all other analyses, the Statistical Package for the Social Sciences (SPSS, version 15; SPSS, Chicago, Ill) was used. All statistical tests were 2-sided, and significance was calculated with a 5% probability of error.
The T1 and T2 assessments were completed by 101 patients. At T3, 7 patients (7%) no longer met the participation criteria for the presence of metastases, 2 patients (2%) declined further participation, and 92 patients completed the third assessment. The mean age was 45.4 years for excluded patients and 48.7 years for nonexcluded patients (P = .71), and the mean premorbid intelligence quotient was 105 for excluded patients and 108 for nonexcluded patients (P = .58).
At T1, 48 patients (47.5%) were premenopausal (Fig. 1). At T3, menopause had occurred in 31 of these patients (Fig. 1) and was induced either by chemotherapy (22 patients) or by GnRH agonists (9 patients). These patients' last menstruation dated back a mean of 46 weeks at T3 (standard deviation [SD], 11.3 weeks). Of 11 patients who were premenopausal at T3, 8 had experienced an induced menopause at T2 and had resumed menses after the completion of chemotherapy. The patient groups that were formed according to menopausal status at T1 and T3 differed with regard to age (P < .001) and intelligence (P = .025) (Table 1).
Table 1. Characteristics of Patients Relative to Menopausal Status and Antiestrogen Therapy*
Premenopausal at T1 and T3
Induced Menopause at T3
Postmenopausal at T1 and T3
No Antiestrogen Therapy: Tam or AI
Antiestrogen Therapy: Tam, AI
T1 indicates assessment before the start of breast cancer therapy; T3, assessment 1 year after T1; Tam, tamoxifen; AI, aromatase inhibitor; SD, standard deviation; IQ, intelligence quotient; HRT, hormone replacement therapy.
The entire sample comprises all patients who underwent at least the T1 assessment and at the assessment toward end of chemotherapy (T2). Before T3, 9 patients were excluded. The menopausal status of 1 patient was indeterminable.
At T3, 62 patients (67.4%) had been receiving antiestrogen therapy with tamoxifen (49 patients; 53.3%) or with AIs (anastrozole, 9 patients; letrozole, 4 patients; total, 14.1%) for a mean of 19 weeks (SD, 6.3 weeks). More recipients than nonrecipients of antiestrogen therapy had previously used HRT (P = .047) (Table 1).
The mean interval between the T1 and T2 assessments was 21 weeks (SD, 1.3 weeks). The T3 assessment took place a mean of 32 weeks (SD, 3.3 weeks) after T2, a mean of 53 weeks (SD, 3.3 weeks) after T1, and a mean of 31 weeks (SD, 4 weeks) after the completion of chemotherapy.
The test results for the T1 and T2 assessments have been reported previously.16 At T3, the performance of the entire sample and of all subgroups, which were composed according to menopausal status and antiestrogen therapy, ranged within or above the test norms on all cognitive tests (see Table 2 for test results).
Table 2. Test Results at Assessment 1 Year After Baseline Relative to Menopausal Status and Antiestrogen Therapy*
Premenopausal at T1 and T3
Induced Menopause at T3
Postmenopausal at T1 and T3
No Antiestrogen Therapy
T1 indicates assessment before the start of cancer therapy; T3, assessment 1 year after baseline; SD, standard deviation; TMT-A, Trail Making Test Part A; TMT-B, Trail Making Test Part B; RWT, Regensburg Word Fluency Test.
The means and SD are given for normally distributed test results; medians and quartiles are given for test results without a normal distribution. The menopausal status of 1 patient was indeterminable.
Results are in seconds, with smaller numbers signifying better performance.
Significant negative effects of age were evident in all cognitive tests except for the RWTs; with these tests, the data introduced into the model already were corrected for age (Table 3). Significance persisted in all but 1 test after correction for multiple testing. Positive effects of intelligence and positive effects of the assessment time retained significance in 9 tests and 5 tests, respectively, after correction for multiple testing.
Table 3. Factor Effects on Results of Cognitive Tests*
Antiestrogen Therapy: Tam, AI
Time of Assessment: T1, T2, T3
Tam indicates tamoxifen; AI, aromatase inhibitor; T1, assessment before the start of breast cancer therapy; T2, assessment toward the end of chemotherapy; T3, assessment 1 year after baseline; df, degrees of freedom; TMT-A, Trail Making Test Part A; TMT-B, Trail Making Test Part B; RWT, Regensburg Word Fluency Test; NA, not applicable.
Factors that were included in single linear models only are not shown (anxiety, previous hormone replacement therapy, interaction of induced menopause with antiestrogen therapy). Random effects are not shown.
Significance was maintained after correction for multiple testing.
Data that already were corrected for age were entered into the model.
Fixed effects of induced menopause were significant in 3 tests. These effects were positive. After correction for multiple testing, significance still persisted in 1 test (RWT Letter Fluency With Switch of Letter) (Fig. 2). There also were 2 random effects of induced menopause (Logical Memory I; P = .0199; RWT Letter Fluency With Switch of Letter, P = .0286)—ie, in these tests, induced menopause had individual effects that varied in a normally distributed way (positive and negative) around the mean (fixed) effect of induced menopause. Significance did not persist for these effects after correction for multiple testing. Entering therapy with GnRH agonists as a separate factor into the models did not reveal any significant effects.
No significant effects of antiestrogen therapy emerged. The nonsignificant effects of antiestrogen therapy and of tamoxifen and AIs separately were partly positive and partly negative (data not shown); thus, no trend was identified.
Further effects were identified that lost significance after correction for multiple testing: a negative fixed effect of previous HRT (d2 Test; P = .0201), a negative fixed effect of anxiety (Digit Symbol; P = .0298), and a positive fixed effect of induced menopause with antiestrogen therapy (RWT Category Fluency With Change of Category; P = .0466).
In the current study, the effects of treatment-induced hormonal changes on the course of cognitive function were assessed in a group of patients with breast cancer, some of whom experienced induced menopause and/or started antiestrogen treatment with tamoxifen or AIs. No negative effects on cognition emerged.
Age and intelligence were strong predictors of cognitive performance, as expected. In addition, clear effects of the assessment time in many tests indicated that improvement occurred over time, debatably suggesting preservation or even enhancement of cognitive function despite chemotherapy. Still, the study did not include a control group of patients who were not receiving chemotherapy, and data regarding practice effects during the third administrations of the cognitive tests used are lacking. Therefore, it is not possible to determine whether patients who were not treated with chemotherapy would have improved to the same extent or more; consequently, this observation cannot be interpreted.
A favorable influence of induced menopause observed in 25% of the tests still was significant in 1 test after conservative correction for multiple testing. This test is the most demanding of the 4 word fluency tests that were used. Word fluency tests are an excellent way to determine whether and how well individuals organize their thinking,45 because they require building a strategy to guide the search for words. Such tests are indicators of quite complex executive functioning.
In addition, individual effects of induced menopause that were favorable in some patients and unfavorable in others were observed in 2 tests. Although these effects were not significant when multiple testing was taken into account, the possibility cannot be excluded that menopause has individual effects, including some decline in a subgroup of patients.
In contrast to the findings of the current study, the only previous report to our knowledge on the effects of treatment-induced menopause described an increased rate of cognitive decline in affected patients compared with initially postmenopausal patients.13 However, the difference was not significant at the 95% level, and it had decreased further at a follow-up assessment 1 year later.
No influence of treatment with tamoxifen or AI on any cognitive domain was identified. This may be because of a lack of statistical power. However, not even a trend was evident—either for effects of antiestrogen therapy as a whole or for tamoxifen and AIs separately.
Some (but not all) previous studies reported detrimental effects of tamoxifen3, 7, 32, 33 or AIs,36 or both35, 46 on cognitive function. With the exceptions of 1 small longitudinal study7 and 1 follow-up study33 that reported minimal differences, these studies were limited by their cross-sectional design.
The current study also had some limitations. First, it did not assess menopausal status by measuring hormone levels but relied solely on patients' reports of their menses. Therefore, the possibility cannot be fully excluded that the classification of menopausal status was incorrect in individual women. Second, treatment-induced hormonal changes dated back only a short interval at the assessments. Thus, this study does not provide any evidence concerning long-term effects of hormonal changes caused by breast cancer therapy.
In contrast to what is believed widely, the current findings suggest that women may fare even better with postmenopausal estrogen levels in terms of cognitive function. Unfavorable effects of estrogen on cognition have been reported previously. High natural estrogen levels have been associated with an increased risk of dementia,24 and negative effects of HRT on cognitive function were evident in the WHIMS.25–27 However, the women who were included in those studies were well beyond the menopausal transition, and the conclusions probably cannot be generalized to younger women. Many studies have reported beneficial effects of HRT on cognitive function in women during or shortly after the menopausal transition, but much of this evidence is limited because of methodological problems. In nonrandomized studies, better health status and healthier lifestyle among HRT recipients even before the start of HRT were not taken into account,47 resulting in a ‘healthy user bias.’ In case-control studies that retrospectively assessed the effect of HRT on the risk of dementia, participants with dementia systematically underestimated their previous HRT use,48 and proxy respondents may not have been aware of it.29 Most randomized controlled trials on this subject have been small and produced mixed findings. Thus, the evidence for effects of estrogen on cognition is far from conclusive.
The current study identified favorable effects of induced menopause on specific cognitive capacities. There also was some indication of individual effects of induced menopause, including deterioration in a subgroup of patients. In contrast, antiestrogen treatment with tamoxifen or AIs did not affect cognition.
The authors are grateful to Franziska Neufeld for conducting many of the assessments and to the patients, physicians, and research nurses in Erlangen, Munich, Ulm, and Rosenheim, Germany, who participated in the Cognitive Impairment in Therapy of Breast Cancer study.