Abstract

Introduction

Aromatase inhibitors (AIs) are increasingly used for the adjuvant treatment of hormone receptor-positive breast cancer instead of or in sequence with tamoxifen [1–6]. These medications are more effective than tamoxifen and have less risk of potentially life-threatening toxicities, including thromboembolic disease and endometrial cancer. However, AIs can cause short- and long-term side effects, including menopausal symptoms, bone loss, and musculoskeletal symptoms.

Arthralgias and myalgias have been noted in up to one-third of patients treated with an AI [1–5, 7–9]. The reported incidence of musculoskeletal side effects varies widely across studies due to differences in definitions of toxicity. Based on cross-study comparisons the three AIs appear similar in both the frequency of these symptoms and their characteristics when used in the adjuvant setting. In a direct comparison of letrozole and anastrozole in the metastatic setting, a similar incidence of bone pain was reported for the two drugs [10]. The reported symptoms can substantially impact quality of life, and occasionally lead to early discontinuation of AI therapy in patients who develop severe symptoms [11, 12].

The mechanisms underlying the development of this toxicity remain unknown, and few studies have carefully evaluated those who developed this type of toxicity using a pre-defined threshold for referral and a standardized rheumatologic evaluation. The COnsortium on BReast cancer phArmacogenomics (COBRA) is currently conducting a multicenter, prospective randomized trial primarily to evaluate the pharmacogenomics of two AIs, letrozole and exemestane, in postmenopausal women with early stage hormone receptor positive breast cancer. A secondary objective of this trial, and the focus of this manuscript, was a prospective evaluation of consecutive individuals who developed new or worsening musculoskeletal symptoms while on therapy. Since few data regarding these side effects were available at the time the clinical trial was developed, the purpose of this preliminary analysis was to characterize the musculoskeletal symptoms that develop in breast cancer patients treated with an AI, and to determine if modifications should be made to the assessment methods being utilized in this clinical trial. This report summarizes the musculoskeletal symptoms and rheumatologic findings in the first 100 women enrolled in this prospective clinical trial.

Subjects and methods

Results

Patient characteristics

The first 100 patients enrolled in the clinical trial were followed for at least 6 months from initiation of AI therapy, with a 12-month median time of follow-up (range 6.5–20.1 months). Baseline patient characteristics are listed in Table 1. Prior breast cancer therapies included surgical resection (100%), adjuvant chemotherapy (54%), adjuvant radiation therapy (76%), and adjuvant tamoxifen therapy (54%). Of the patients treated with tamoxifen, the median duration of therapy was 2 years (range 0.1–13 years).

Table 1

Baseline characteristics of patients according to referral for rheumatologic evaluation

Characteristic	All n = 100	Referred n = 44	Not referred n = 53	p valued
Demographics
Age, years; median (range)	59 (38–83)	58 (47–70)	60 (38–83)	0.5839e
Race
Caucasian, n (%)	89	38 (86.4)	48 (90.6)	0.8633e
African-American, n (%)	9	5 (11.4)	4 (7.6)
Asian, n (%)	2	1 (2.3)	1 (1.9)
Weight, kg; median (range)	74.0 (40.7–150.6)	75.3 (55.8–135.4)	73.3 (40.7–150.6)	0.3670e
BMI, kg/m2; median (range)	28.8 (17.7–53.4)	29.6 (22.2–53.4)	28.1 (17.7–48.5)	0.6506e
Prior breast cancer therapy
Axillary surgery, n (%)a
Sentinel LN biopsy	64 (76.2)	28 (75.7)	35 (79.5)	0.6765f
Axillary LN dissection	35 (41.7)	18 (48.6)	15 (34.1)	0.1841f
Chemotherapy, n (%)	54 (54.0)	22 (50.0)	29 (54.7)	0.9865f
Taxane, n (%)b	29 (31.8)	13 (34.2)	14 (28.0)	0.5315f
Radiation, n (%)	76 (76.0)	35 (79.6)	40 (75.5)	0.6334f
Tamoxifen, n (%)	54 (54.0)	25 (56.8)	27 (50.9)	0.5635f
Medical comorbidities
Diabetes, n (%)	8 (8.0)	5 (11.4)	3 (5.7)	0.4620g
Thyroid disorder, n (%)	12 (12.0)	5 (11.4)	6 (11.3)	1.0000g
Any arthritis, n (%)	36 (36.0)	18 (40.9)	16 (30.2)	0.2706f
Prior fracture, n (%)c	22 (22.2)	12 (27.9)	9 (17.0)	0.1978f

Open in a separate window

LN lymph node

^aAxillary surgery data missing for seven referred patients and nine non-referred patients. Some patients underwent more than one procedure

^bTaxane data missing for six referred patients and three non-referred patients

^cPrior bone fracture data missing for one referred patient

^dp value for referred versus non-referred patients

^eWilcoxon rank-sum test

^fChi-square goodness-of-fit test

^gFisher’s exact test

Ninety-seven patients were eligible for referral to rheumatology; two were ineligible because of elevated baseline HAQ scores and one discontinued participation in the study prior to initiating protocol-directed therapy because of poor venous access. Forty-four patients (45.4%) qualified for referral to rheumatology, and 38 (39.2%) were evaluated. Four patients initially declined referral but later consented to see the rheumatologist because of persistent or worsening symptoms. Six patients declined referral and never underwent rheumatologic evaluation; one of these subjects subsequently discontinued participation in the study because of musculoskeletal toxicity.

Of the first 100 patients enrolled, 23 discontinued therapy with an AI. Rheumatologic toxicity was the identified cause for 13 of the discontinuations. Patients who stopped therapy or switched to a different AI because of musculoskeletal toxicity did so after a median 6.1 months (range 2.2–13 months). The remaining patients discontinued therapy for other reasons, including recurrent menses (3), non-compliance with therapy (3), dizziness, restlessness, depression, and inability to obtain blood samples.

There were no obvious baseline characteristics that distinguished those patients who met criteria for referral versus those who did not (Table 1). The patients included in the referred subset included all patients who met objective criteria for referral, regardless of whether they consented to rheumatologic evaluation. There was no statistically significant difference in baseline weight, body mass index, or concomitant medical illness. Similarly, there was no statistically significant difference in prior therapy for breast cancer, including type of axillary surgery, radiation therapy, prior tamoxifen, or prior chemotherapy, including taxanes.

Symptom characterization

Median time from initiation of AI to onset of symptoms was 1.6 months (range 0.4–10 months). Symptoms in four patients pre-dated study enrollment. Median time to rheumatology referral was 3.7 months (range 0.9–12.6 months), and 59% of patients were referred to rheumatology after taking an AI for 3–6 months (Fig. 1). Four patients were referred specifically because of symptoms that developed or worsened between scheduled clinic visits. Increased HAQ score led to referral in 17 patients. New onset pain or pain reported as much worse or very much worse led to referral of 20 patients. Six patients were referred because of a combination of pain and decreased function.

Open in a separate window

Fig. 1

Distribution of rheumatology referrals over time. Light gray bars represent patients eligible for referral who did not qualify for referral at a given timepoint. Black bars represent patients eligible for referral who did qualify for referral at a given timepoint. Percentage at top of each bar represents percent of patients referred to rheumatology at each timepoint

All patients completed HAQ and VAS assessments at baseline (Table 2). Referred patients had statistically significantly higher baseline HAQ scores (p = 0.0440), although the median baseline HAQ score was 0 for both the referred and not referred cohorts. There was a trend toward higher baseline VAS scores for referred patients (p = 0.0664). At the time of referral, the median HAQ score for referred patients was 0.375, with a median absolute increase in HAQ score of 0.3125. Median VAS score at time of referral was 51, with a median absolute increase in VAS score of 32.5.

Table 2

Rheumatologic parameters of patients according to referral for rheumatologic evaluation

Characteristic	Referred (n = 44)	Not referred (n = 53)	p value
Health Assessment Questionnaire
Baseline	0 (0–1.0)	0 (0–1.0)	0.0440b
At referral	0.375 (0–1.625)	–
Absolute change	0.3125 (–0.375 to 1.375)	–
Visual Analog Scale
Baseline	10 (0–59)	2.5 (0–80)	0.0664b
At referral	51 (0–79)	–
Absolute change	32.5 (–12 to 78)	–
Months to symptomsa	1.6 (0.4–10.0)
Months to referral	3.7 (0.9–12.6)
Months to evaluation	5.4 (1.3–13.0)
Months to study discontinuation	6.1 (2.2–13.0)

Open in a separate window

Median values given, with ranges in parentheses

^aExcluding patients with symptoms prior to AI initiation (n = 4)

^bWilcoxon rank-sum test

At time of rheumatology evaluation, the primary symptoms were joint pain and achy and stiff joints (Fig. 2). Other reported symptoms included muscle pain, morning stiffness, tingling, numbness, and joint swelling. In descending order, the joints primarily involved per patient report were hands and wrists (14 patients), shoulders (12), knees (11), feet and ankles (9), spine (7), and hips (5). Four patients reported specific muscle pain.

Open in a separate window

Fig. 2

Frequency of symptoms reported by patients referred to rheumatology (n = 38)

For management of symptoms, the vast majority of patients had used non-steroidal anti-inflammatory medications or acetaminophen. In addition, four patients had been prescribed tramadol or opioids. A majority of patients reported an increase in analgesic usage after initiating therapy with an AI.

Diagnoses

At the time of evaluation, the majority of patients were judged by the evaluating rheumatologists as having moderate-intensity, non-inflammatory regional musculoskeletal disorders. Seven patients were considered to have mild pain not interfering with function, whereas 31 patients were considered to have moderate to severe pain that interfered with function or activities of daily living. The change in HAQ score between AI initiation and rheumatology referral was statistically significantly higher in subjects diagnosed with moderate/severe pain compared to those subjects with mild pain (p = 0.0285). There was no statistically significant correlation between degree of pain and functional status impairment on clinical assessment and HAQ or VAS scores at baseline or at time of referral, or absolute change in VAS between AI initiation and time of referral.

The most frequent diagnoses based on clinical evaluation were tendonitis or tenosynovitis, osteoarthritis, and carpal tunnel syndrome (CTS) (Table 4). Multiple rheumatologic conditions were diagnosed in 37% of patients. A few cases represented exacerbations of previously existing abnormalities, but most patients reported new symptoms since initiation of therapy. Three of the nine patients with clinical symptoms consistent with CTS were tested with electroneuromyography, and all three had changes consistent with CTS. In 73% of subjects, musculoskeletal symptoms were judged as being definitely (18%) or possibly (55%) attributable to AI therapy.

Table 4

Clinical rheumatologic diagnoses for patients referred to rheumatology (n = 38)

Diagnosis	Number of patients (%)
Bursitis	8 (21.1%)
Trochanteric	6 (15.8%)
Carpal tunnel syndrome	8 (21.1%)
Osteoarthritis	11 (28.9%)
Knee	3 (7.9%)
Hand	2 (5.3%)
Tendonitis	14 (36.8%)
Rotator cuff/shoulder	8 (21.2%)
Wrist	3 (7.9%)
Elbow	2 (5.3%)
Patellofemoral syndrome	7 (18.4%)

Open in a separate window

Some subjects had more than one diagnosis

Discussion

Aromatase inhibitors are increasingly used for adjuvant treatment of postmenopausal women with breast cancer. In this prospectively followed and evaluated, highly annotated cohort, 45.4% of subjects met pre-determined criteria for rheumatologic evaluation, 39.2% underwent a complete rheumatologic evaluation, and 13% discontinued therapy because of intolerable musculoskeletal toxicity. We were unable to identify any baseline patient characteristics, including pre-existing co-morbid conditions or specific breast cancer therapies, that were statistically significantly associated with the development of symptoms. It remains unclear whether these symptoms are due to a class effect or a specific medication. Therefore, we have recently amended our protocol to allow patients who develop intolerable symptoms to switch from one drug to the other, provided their initial symptoms completely resolve.

Overall, the clinical presentation of rheumatologic side effects was most consistent with AIs inducing a non-inflammatory musculoskeletal syndrome or the development of inflammation localized to the tenosynovial structures, including the carpal tunnel structures. Since there was no clear correlation between presence of symptoms and laboratory test abnormalities, we do not recommend routine clinical evaluation of the laboratory studies that were assessed in this trial. Because the distribution of reasons for referral were evenly divided between increased pain and worsened function, and there was a correlation between severity of pain and functional limitation on clinical assessment and increase in HAQ score with therapy, both measures for assessment of patients will continue to be utilized for the remaining patients accrued to this and other COBRA-conducted clinical studies.

The incidence of arthralgias and musculoskeletal symptoms in this study is similar to that reported in prior clinical trials of AIs in the adjuvant setting [1–7, 16, 17], although the drug discontinuation rate is higher in our study than previously reported. An increased incidence of CTS has been reported for both steroidal and nonsteroidal AIs compared to tamoxifen (IES trial: exemestane 2.8%, tamoxifen 0.3%; ATAC trial: anastrozole 3%, tamoxifen 1%) [7, 16]. A recent study using ultrasound and magnetic resonance imaging to characterize the wrists of women with severe wrist symptoms during AI therapy revealed flexor tenosynovitis. However, this study was quite small and all evaluated patients were symptomatic at the time of assessment [18]. These results suggest that AI therapy may lead to localized inflammation around tendons and nerves that may result in development of symptoms. Whether these changes are due to AI therapy, whether they only occur at the wrist, and how best to manage them remains to be determined.

The effects of other breast cancer therapies, such as surgery, radiation therapy, chemotherapy, and tamoxifen, might also play a role in the development of AI-related musculoskeletal symptoms. Other authors have suggested that treatment with chemotherapy alone might lead to myalgias and arthralgias [19–23]. Another study demonstrated an inverse association between prior tamoxifen therapy and development of AI-related joint stiffness [8]. Development of musculoskeletal symptoms in patients treated with chemotherapy and/or tamoxifen prior to an AI could therefore be influenced by either treatment regimen [8, 24]. In our study, no correlation was noted between surgery, radiation therapy, chemotherapy, or tamoxifen and the development of symptoms. This lack of demonstration of a statistically significant association may be due to insufficient power, since only a subset of women received each treatment modality.

The mechanism underlying the development of AI-related musculoskeletal symptoms is unknown, although estrogen deprivation has been proposed as one etiology. Arthralgias increase in incidence as women gradually progress through the menopausal transition, and are more common in women who undergo surgical menopause [25, 26]. Arthralgias and myalgias have also been noted in 25% of premenopausal women treated with gonadotropin-releasing hormone agonists[27]. However, a prior study showed no correlation between estrone level and development of joint pain in menopausal women [26]. The association between estrogen deprivation and development of musculoskeletal symptoms in this patient population therefore remains unclear.

Autoimmunity has also been proposed as a possible underlying mechanism. Researchers found that aromatase-knockout mice develop a Sjögren’s syndrome-like autoimmune lymphoproliferative disease[28]. However, in a small study, women who developed AI-induced musculoskeletal symptoms had no evidence of anti-doublestranded DNA antibodies to suggest an autoimmune etiology [29]. Similarly, in our study population, there was no evidence of autoimmunity based on clinical criteria from history and physical examination, and the majority of patients did not demonstrate auto-antibody production of RF or ANA.

In summary, the musculoskeletal symptoms that occur with AI therapy are clinically important as they lead to discontinuation of therapy in more than 10% of patients. In this study, no baseline characteristics predicted this toxicity, and these symptoms do not appear to be consistent with any defined clinical or serologic rheumatologic syndrome. These important negative findings support the need for more research into the mechanisms underlying this significant toxicity and its management.

Acknowledgments

Footnotes

Contributor Information

N. Lynn Henry, Breast Oncology Program, University of Michigan, Comprehensive Cancer Center, 1500 East Medical Center Drive, 6312 CCGC, Ann Arbor, MI 48109-0942, USA.

Jon T. Giles, Division of Rheumatology, Johns Hopkins University, Baltimore, MD, USA.

Dennis Ang, Department of Medicine, Indiana University School of Medicine, Indianapolis, IN, USA.

Monika Mohan, Division of Rheumatology, Department of Internal Medicine, University of Michigan Medical School, Ann Arbor, MI, USA.

Dina Dadabhoy, Division of Rheumatology, Department of Internal Medicine, University of Michigan Medical School, Ann Arbor, MI, USA.

Jason Robarge, Department of Medicine, Indiana University School of Medicine, Indianapolis, IN, USA.

Jill Hayden, Breast Oncology Program, University of Michigan, Comprehensive Cancer Center, 1500 East Medical Center Drive, 6312 CCGC, Ann Arbor, MI 48109-0942, USA.

Suzanne Lemler, Department of Medicine, Indiana University School of Medicine, Indianapolis, IN, USA.

Karineh Shahverdi, Breast Cancer Program, Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University, Baltimore, MD, USA.

Penny Powers, Breast Cancer Program, Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University, Baltimore, MD, USA.

Lang Li, Department of Medicine, Indiana University School of Medicine, Indianapolis, IN, USA.

David Flockhart, Department of Medicine, Indiana University School of Medicine, Indianapolis, IN, USA.

Vered Stearns, Breast Cancer Program, Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University, Baltimore, MD, USA.

Daniel F. Hayes, Breast Oncology Program, University of Michigan, Comprehensive Cancer Center, 1500 East Medical Center Drive, 6312 CCGC, Ann Arbor, MI 48109-0942, USA.

Anna Maria Storniolo, Department of Medicine, Indiana University School of Medicine, Indianapolis, IN, USA.

Daniel J. Clauw, Division of Rheumatology, Department of Internal Medicine, University of Michigan Medical School, Ann Arbor, MI, USA.

References