Randomized clinical trials have shown the superiority of aromatase inhibitors (AIs) over tamoxifen in improving disease-free and overall survival among women diagnosed with breast cancer.1–3 As a result of these studies, AIs have become the adjuvant hormone treatment of choice for postmenopausal women diagnosed with estrogen receptor–positive breast cancer and are recommended by the American Society of Clinical Oncology to be used in adjuvant therapy initially or after tamoxifen use among this group of breast cancer patients.4 However, AIs have also been shown to be associated with loss of bone mass, osteoporosis, and an increased risk of fractures, which result from the profound reduction in circulating estrogen concentrations that accompany AI therapy (reviewed in Santen5). Additionally, up to 80% of breast cancer patients taking AIs report the onset of or increase in musculoskeletal pain, with a localization in the wrists and palms,6 that can adversely affect the patient's functional abilities and quality of life.7 The adverse bone effects, including the potentially severe musculoskeletal symptoms, and their consequences, can lead to discontinuation of or nonadherence to treatment,7, 8 putting the breast cancer patient at risk for breast cancer recurrence and death.9
Measurement of bone turnover markers can be used to examine changes in bone turnover over a short period of time.5 In clinical trials, AI therapy has been shown to be associated with statistically significant increases in both bone formation markers (eg, osteocalcin) and bone resorption markers (eg, cross-linked N-telopeptides of bone type I collagen [NTXs]) over the first 3 to 24 months of treatment.10–13 In contrast, in these same clinical trials, either no change or statistically significant reductions in the concentrations of these markers were observed with tamoxifen treatment. Changes in bone formation and bone resorption markers occur simultaneously, and the results of studies examining AI-induced bone marker changes suggest that AIs are associated with an imbalance between resorption and formation, leading to a “net bone loss” and increased fracture risk.5 The bone turnover marker profile may be clinically useful in identifying patients at highest fracture risk who may require early intervention with antiresorptive agents or, potentially, patients requiring a change in treatment.
The detrimental effects of AI therapy on bone (as measured using bone mineral density, bone turnover markers, or fracture occurrence) have been shown in the large randomized clinical trials; however, most of these trials have examined AI therapy compared with tamoxifen, which is bone-sparing.10–13 Therefore, results of these clinical trials may exaggerate the actual differences in bone markers compared with an untreated group. There have been only a few studies examining bone changes of AI therapy versus placebo14 or investigating the effects of AIs on bone in the broader non–clinical trial patient population.15, 16 Whether accelerated bone loss as measured by bone markers is associated with AI-associated musculoskeletal symptoms has, to our knowledge, not been reported.
From 2006 to 2010, a 6-month prospective cohort study was conducted at Mercy Medical Center in Baltimore, MD, USA, to examine musculoskeletal symptoms among breast cancer patients initiating AI therapy and a group of women without a history of breast cancer of a similar age.6 In a previous publication from this study, it was reported that at the 3-month visit (3 months after initiating AI therapy for breast cancer patients and 3 months after the baseline evaluation in the comparison group), the breast cancer patients were more likely to report musculoskeletal pain compared with women in the comparison group.6 In addition, for the breast cancer patients, but not the women without a history of cancer, musculoskeletal pain severity increased significantly over the 6-month time period among women reporting pain at baseline.
To further investigate the musculoskeletal symptoms associated with the initiation of AI therapy, this study was conducted to: (1) examine 3- and 6-month changes in bone formation (osteocalcin) and bone resorption (NTXs) markers, as well as calcium, phosphorus, and intact parathyroid hormone (iPTH), among the breast cancer patients initiating AI therapy and the comparison group of women without a history of cancer; and (2) examine whether bone marker changes were associated with musculoskeletal pain among the breast cancer patients and/or the comparison group.
Patients and Methods
Data were analyzed from a 6-month prospective cohort study of breast cancer patients on AIs and a comparison group of postmenopausal women without a history of cancer. Detailed methods of this study are described elsewhere.6 Briefly, women with newly diagnosed nonmetastatic postmenopausal estrogen receptor–positive breast cancer whose adjuvant treatment plan included an AI and a comparison group of postmenopausal women with no history of cancer except possibly cervical cancer in situ or nonmelanoma skin cancer were recruited from the Women's Center for Health & Medicine and the Institute for Cancer Care, located in the same community-based outpatient setting. Women who reported a history of rheumatoid arthritis or fibromyalgia were not eligible for enrollment into either group. A total of 100 breast cancer patients and 200 women without a history of cancer were enrolled; however, one breast cancer patient did not initiate AI therapy as planned. The study was approved by the Institutional Review Board at Mercy Medical Center in Baltimore, MD, USA. Written informed consent was obtained from all participants.
Women enrolled in the study completed a questionnaire at a baseline visit, which occurred before AI treatment for the breast cancer patients, and at 3 and 6 months after baseline. The questionnaire collected information on demographics, musculoskeletal symptoms, medical history, and health habits. In addition, participants were asked to donate a blood sample at all three time points. Of the 99 breast cancer patients initiating AI therapy who enrolled in the study, 63 donated a blood sample and completed a questionnaire at all three study visits. Among the 200 women in the comparison group, 141 donated a blood sample and completed a questionnaire at all three study visits.
Musculoskeletal pain was assessed at each time point by asking participants: “In the past 4 weeks, have you experienced any of the following types of pain: Joint, Muscle, Bone?” Women who responded “yes” for one or more of the types of pain were directed to a module with more detailed pain questions pertaining to that specific type of pain. Included in that module was a 10-cm visual analog scale (VAS) on which the participant was asked to mark her average pain severity for each type of pain reported. Participants who responded that they did not experience a specific type of pain or who scored less than 1 on the VAS for a specific type of pain were categorized as having “no pain” for that type of pain. Onset or worsening of pain (measured using participant ratings on the VAS) over the study period was examined as an outcome.
Data were collected on age, race, height, weight, bisphosphonate and supplement use, and prior cancer treatments using the self-administered questionnaire. Body mass index (BMI) was calculated using self-reported height and weight; for analyses, BMI was categorized as normal (<25.0 kg/m2), overweight (25.0–29.9 kg/m2), and obese (≥30.0 kg/m2).
Measurement of estradiol and biomarkers
Each blood sample was collected in a red top tube and serum was aliquotted and stored at −70°C until assayed. Serum samples from the breast cancer patients only were sent to the Core Endocrinology Laboratory at the Penn State Hershey Medical Center for measurement of estradiol levels. Estradiol was measured using a solid-phase radioimmunoassay (RIA) (Siemens Diagnostics, Los Angeles, CA, USA) based on estradiol-specific antibodies that are immobilized to the wall of polypropylene tubes and the use of a 125I-labeled estradiol tracer. The functional sensitivity of the estradiol assay is 5 pg/mL. The mean interassay coefficient of variation (CV) was 7.9%; the mean intraassay CV was 7.0%. Estradiol values were used to assess noncompliance along with the medical record, as described in the Treatment Compliance section below.
The bone biomarker and iPTH assays were conducted at the Clinical Research Unit Core Laboratory of the Institute for Clinical and Translational Research at Johns Hopkins Bayview Medical Center. Osteocalcin was measured using an immunoradiometric assay (Immutopics, San Clemente, CA, USA). The functional sensitivity of the osteocalcin assay is 0.06 ng/mL; the inter- and intraassay CVs were 6.30% and 2.97%, respectively. NTXs were assayed using an enzyme-linked immunosorbent assay (ELISA) (Inverness Medical, Princeton, NJ, USA); the recovery rate of this assay is 98%. The inter- and intraassay CVs for the NTXs measurements were 0.83% and 1.10%, respectively. Finally, iPTH was assayed using an ELISA kit (ALPCO, Windham, NH, USA). The functional sensitivity of the assay is 0.9 pg/mL and the inter- and intraassay CVs were 4.99% and 4.09%, respectively.
Compliance to AI therapy among the breast cancer patients was assessed by both medical chart review and the estradiol concentrations measured from the 3- and 6-month blood draws. A breast cancer patient was identified as being noncompliant if at least one of her follow-up estradiol measurements was above the limit of detection and her medical record indicated that she was not adherent to her AI therapy during the 6-month study period. Follow-up estradiol concentrations above the limit of detection alone were not sufficient to categorize a breast cancer patient as being noncompliant, as not all patients taking an AI experience total estrogen suppression.17, 18 Eight breast cancer patients (n = 8) were identified as being noncompliant.
Breast cancer patients and women in the comparison group who reported being treated with a bisphosphonate at any time during their participation in the study were excluded from the analyses (14 breast cancer patients and 24 women in the comparison group). Baseline characteristics of the breast cancer patients and the women without a history of breast cancer were compared using χ2 tests and Fisher's exact tests. Biomarker values did not conform to a normal distribution and mathematical transformations failed to normalize the sample due to outliers. As a result, the biomarker comparisons were examined using nonparametric tests including: the Mann-Whitney U test for differences by exposure group at the baseline, 3-, and 6-month follow-up visits and the Kruskal-Wallis test for differences across follow-up time from baseline. In order to examine biomarker changes across time, a change variable was computed between baseline and the time of follow-up for each biomarker. Test of change between women in the breast cancer group who were identified as being compliant with their AI therapy (n = 41) and women in the comparison group (n = 117) were analyzed using independent sample t tests, and differences across time were analyzed with paired t tests. Finally, the associations between the change in biomarkers and the onset or increase in musculoskeletal pain over time among the breast cancer patients (n = 47; 2 patients were missing data on pain at both follow-up time points) were analyzed with independent sample t tests and paired t tests. An analysis for potential confounders of the associations was conducted by analyzing the associations between baseline biomarkers concentrations and the baseline characteristics (age, race, BMI, supplement use) among the comparison group; however there were no statistically significant differences found and, therefore, no confounders identified. A p < 0.05 was considered statistically significant.
Baseline characteristics of the breast cancer patients and the women in the comparison group are shown in Table 1. The mean ages of the breast cancer patients and women in the comparison group in the analytic study sample were 60.1 (SD 7.8) years and 59.8 (SD 7.7) years, respectively (p = 0.81). The breast cancer patients did not differ significantly from the women in the comparison group with regard to age, race, BMI, calcium or vitamin D supplement use, or baseline musculoskeletal pain. Approximately 45% of the breast cancer patients reported being previously treated with chemotherapy; 24.5% reported being treated with radiation. Breast cancer patients had significantly higher median concentrations of calcium (9.9 mg/dL versus 9.5 mg/dL; p = 0.007) and phosphorus (3.8 mg/dL versus 3.6 mg/dL; p = 0.007) at baseline than women in the comparison group. There were no statistically significant differences between the two groups in terms of baseline iPTH, NTXs, or osteocalcin concentrations. The median estradiol concentration at baseline among the breast cancer patients was <5.0 (range, <5.0 to 23.4) pg/mL. All of the breast cancer patients in the analytic sample initiated treatment with a nonsteroidal AI (anastrozole: 55.1% or letrozole: 44.9%).
Table 1. Baseline Characteristics of Breast Cancer Patients and Women in the Comparison Group
Percentages may not equal 100% for certain characteristics due to missing values.
iPTH = intact parathyroid hormone; NTXs = cross-linked N-telopeptides of bone type I collagen; BCE = bone collagen equivalents.
Value of p derived using χ2 tests and Fisher's exact tests for categorical variables and Student's independent sample t tests and Mann-Whitney U tests for continuous variables.
Age, years, mean (SD)
Race, n (%)
Body mass index, kg/m2, n (%)
Type of aromatase inhibitor treatment, n (%)
Prior chemotherapy, n (%)
Prior radiation, n (%)
Calcium supplement use only, n (%)
Vitamin D supplement use only, n (%)
Calcium and Vitamin D supplement use, n (%)
Estradiol, pg/L, median (range)
Any musculoskeletal pain, n (%)
Joint pain, n (%)
Muscle pain, n (%)
Bone pain, n (%)
Calcium, mg/dL, median (range)
Phosphorus, mg/dL, median (range)
iPTH, pg/mL, median (range)
NTXs, nmol BCE, median (range)
Osteocalcin, ng/mL, median (range)
Changes in biomarker concentrations from baseline to 3 and 6 months among the breast cancer patients and women in the comparison group are shown in Fig. 1. Among the breast cancer patients, calcium concentrations decreased significantly (−7.8% change; p = 0.013) and concentrations of NTXs increased significantly from baseline to 6 months (9.6% change; p = 0.012). There were no statistically significant changes in biomarker concentrations over the follow-up period for women in the comparison group. Statistically significant differences in percent change between the breast cancer patients and the women in the comparison group were observed for calcium at 6 months (−7.8% versus 0.0%; p = 0.025), phosphorus at 6 months (−5.1% versus 16.7%; p = 0.003), NTXs at 6 months (9.6% versus −0.7%; p = 0.017), and osteocalcin at 6 months (11.5% versus −3.6%; p = 0.016). iPTH concentrations did not change significantly over time for either group nor did they differ between groups at either 3 or 6 months.
Among the 47 breast cancer patients with both musculoskeletal pain data and biomarker data at all three time points, 40 (85.1%) reported the onset or increase in the severity of pain over the study period. Comparisons of biomarker values within these groups across time and between the groups at each time point are shown in Table 2. Breast cancer patients who reported new onset or increase in the severity of pain during the study period had borderline statistically significant higher baseline NTXs concentrations than those who reported not experiencing pain or no increase in pain over the 6-month time period (mean 15.4 nmol bone collagen equivalents [BCE] versus 11.8 nmol BCE; p = 0.08). There were no other statistically significant differences in mean baseline, 3-month, and 6-month concentrations of other biomarkers between those reporting new onset or increase in the severity of pain versus those reporting no pain or no increase in pain.
Table 2. A Comparison of Mean Bone Biomarker Concentrations and Percent Change From Baseline Among the Breast Cancer Patients Reporting the Onset or Increase in Severity of Musculoskeletal Pain Over the Study Period (n = 40) and the Breast Cancer Patients Reporting No Onset or Increase in Pain Over the Study Period (n = 7)
However, among those with new onset or increase in severity of pain, there were statistically significant decreases in both calcium and phosphorus concentrations from baseline to 6 months (calcium: −9.7%, p = 0.004; phosphorus: −7.7%, p = 0.049); in contrast, little or no change in calcium (0.0%) and phosphorus (2.5%) was observed over this time period among breast cancer patients with no pain or no increase in pain. In addition, among those reporting no pain or no increase in pain over the study period, there was a statistically significant increase in NTXs from baseline to 6 months (42.4%; p = 0.004). Only a small, non–statistically significant increase in NTXs was found among the patients with new onset or increase in severity of pain from baseline to 6 months (2.6%; p = 0.40). There were no statistically significant changes in iPTH over the study period among either pain group nor were there statistically significant differences in iPTH at 3 and 6 months between the groups. Further, neither decreasing estradiol concentrations over the study period or a change in estradiol status from above the limit of detection of the assay to at the limit of detection of the assay over the study period was associated with the onset or increase in severity of musculoskeletal symptoms (data not shown) among the breast cancer patients.
The results of this prospective cohort study showed that AI treatment was associated with increases in bone turnover, as measured using serum osteocalcin concentrations and concentrations of NTXs, among breast cancer patients followed for 6 months after initiating therapy. The percent changes from baseline were greatest for the two bone turnover biomarkers at the 6-month visit, and the percent change values for both biomarkers were significantly different at this time point compared with a group of women without a history of cancer who were followed over the same time period. Coinciding with the changes in the bone resorption and bone formation marker concentrations among the breast cancer patients was a statistically significant decrease in mean calcium concentration from baseline to 6 months; no change in iPTH was observed. The changes (or lack thereof in the case of iPTH) in the biomarkers observed over the first 6 months of AI treatment among breast cancer patients likely reflect the effects of the profound decrease in estrogen concentrations occurring with AI treatment on bone health. Similar effects, albeit much more accelerated in the case of AI therapy, are observed during natural menopause, when cessation of ovarian function and, thus, decreased estrogen secretion, leads to increased bone resorption and an efflux of calcium from bone, but with compensatory increased renal calcium excretion, decreased intestinal calcium absorption, and partially suppressed parathyroid hormone secretion.19–22
No associations were observed between changes in bone turnover markers and onset of musculoskeletal pain among patients initiating AI therapy, with the exception of baseline NTXs being higher among breast cancer patients with new onset or increase in severity of pain compared with breast cancer patients with no pain or no increase in pain over the study period. To our knowledge, prior studies have not examined the association between symptoms and bone biomarkers. In this study, musculoskeletal symptoms overall were also not correlated with serum vitamin D concentrations, as reported in a previous publication,6 although limited data in the published literature suggests that low vitamin D levels may be associated with new onset of pain associated with initiation of AI therapy.6, 23–25 However, the underlying explanation for AI-associated musculoskeletal symptoms, which have a distinct distribution (hands and feet),6 remains uncertain. Changes in estradiol concentrations were not associated with the onset or increase in severity of musculoskeletal symptoms among the breast cancer patients in this study.
The findings from our study related to changes in the bone turnover markers are consistent with most of the clinical trials and observational studies that have reported on the associations between AI therapy and bone turnover, although the magnitude of changes in the AI group in the present study were, in general, less than observed in previous studies, and, for osteocalcin, not statistically significant at either the 3- or 6-month time point. For example, the Intergroup Exemestane Study (IES), a randomized controlled study comparing the AI exemestane with tamoxifen among women with breast cancer, showed an approximate 20% increase in osteocalcin concentrations at 6 months compared with baseline in the exemestane group; the percent increase in this bone formation marker was highest at 12 months and then decreased.12 No change was observed over time in the tamoxifen group.12 In the Tamoxifen Exemestane Adjuvant Multicentre (TEAM) trial, which compared 78 breast cancer patients initially treated with exemestane to 83 treated with tamoxifen, exemestane was associated with a 43.5% mean increase in osteocalcin levels at 3 months and an approximate 40% mean increase at 6 months.13 In comparison, tamoxifen was associated with a mean change of −16.9% at 3 months and an approximately −25% change at 6 months.
The bone turnover changes observed in our study as well as others indicate that bone loss occurs early on in the initiation of AI treatment and, in the clinical trials, bone loss has been shown to translate into increased fracture rates among women treated with AIs compared with those treated with tamoxifen.3, 26 These findings underscore the issue of bone loss prevention among breast cancer patients treated with AIs. Before initiating AI therapy, breast cancer patients should be assessed for fracture risk using known clinical risk factors, include evaluation of bone density, and, based on findings, appropriate treatment should be administered and follow-up conducted. Although calcium and vitamin D supplementation is recommended for breast cancer patients on AI therapy,23 results from our study showed no statistically significant correlations between the changes in circulating calcium and vitamin D concentrations and the changes in the bone turnover markers over the 6-month time period (data not shown).
Several limitations should be noted when interpreting the results of the present study. First, the study period was limited to 6 months, as the overall purpose of the study was to examine the early onset of musculoskeletal symptoms among breast cancer patients on AIs. Thus, we were not able to determine whether there were greater changes in the bone biomarkers among the breast cancer patients or greater differences between the two groups after the 6 month time point. Second, bone density data were not collected as part of the study and, therefore, we were not able to analyze the associations between AIs, bone biomarker values, and bone density. Finally, the overall numbers of breast cancer patients and women in the comparison group in this study were small, especially for those who completed all three clinic visits and who were not treated with bisphosphonates during the study period. This limited the power of the study to detect small differences between the groups or within each group across time.
Findings from this study indicate that AIs cause changes in bone turnover during the first 6 months of treatment; however, these changes do not appear to be associated with AI-associated musculoskeletal pain. Because increased bone turnover has been shown to be associated with a higher risk of fracture, breast cancer patients initiating AI therapy should have a fracture risk assessment, using known clinical risk factors, and the results of this assessment should guide treatment decisions. Further, future studies should assess the long-term effects of AI treatment on bone health, and whether interventions can be implemented or treatments administered to patients to avoid any adverse outcomes.
All authors state that they have no conflicts of interest.
Support for this study was provided, in part, by grants from AstraZeneca (IRUSANAS0073), Susan G. Komen for the Cure Foundation (POP0601174), and the National Center for Research Resources (NCRR), a component of the National Institutes of Health and NIH Roadmap for Medical Research (UL1 RR 025005).
Authors' roles: Study design: ER and KJH. Study conduct: LG, RM, BW, and KJH. Data collection: LG, RM, NSF, BW, and KJH. Data analysis: RM. Data interpretation: LG, RM, ER, NSF, and KJH. Drafting manuscript: LG. Revising manuscript content: RM and KJH. Approving final version of the manuscript: LG, RM, BW, NSF, ER, and KJH. RM takes responsibility for the integrity of the data analysis.