Defining the aromatase inhibitor musculoskeletal syndrome: A prospective study




To define the musculoskeletal syndrome associated with use of aromatase inhibitors (AIs), specifically, to describe its incidence, time to onset, risk factors, and clinical presentation.


Postmenopausal women with hormone-sensitive, nonmetastatic breast cancer starting AI therapy were enrolled in this prospective cohort study. They underwent complete rheumatologic evaluation and contrast-enhanced magnetic resonance imaging (MRI) of the hands and wrists prior to starting AI, at 3 and 6 months. The primary outcome was change in grip strength.


Twenty-eight (54%) of 52 women reported new or worsening musculoskeletal symptoms. Two discontinued AIs due to pain. Mean time to symptom onset was 6 weeks (range 2–18 weeks), and 75% of symptomatic patients developed symptoms by 8 weeks. Later-stage cancer and worse quality of life (QOL) pretreatment were significantly associated with symptom development. Sixty-eight percent of symptomatic subjects had involvement of the hands; however, there was no difference in the mean change in grip strength (−2.9 kg versus −1.3 kg; P = 0.6). Among symptomatic subjects, 46% had evidence of focal tenosynovitis of the hands and feet on examination. Although some symptomatic subjects had new MRI abnormalities, Rheumatoid Arthritis Magnetic Resonance Imaging Scoring did not significantly change.


The incidence of AI-associated musculoskeletal syndrome is more than 50%, with most women developing symptoms by 8 weeks. The key finding in symptomatic women was focal tenosynovitis of the hands and feet, without evidence of autoimmune disease or systemic inflammation. Later-stage cancer and poorer QOL were predictive of symptom development.


Third-generation aromatase inhibitors (AIs) are quickly replacing tamoxifen as the standard adjuvant endocrine therapy for hormone receptor–positive, postmenopausal breast cancer. Adherence to this crucial therapy has been limited by debilitating musculoskeletal side effects. Adverse event reporting from large randomized controlled trials (RCTs) has consistently shown higher rates of arthralgia in patients taking AIs (1–7). Of RCT participants, 5–10% discontinued therapy due to musculoskeletal symptoms (8). In several practice-based studies, the incidence rates of AI-associated arthralgia were higher than those reported in RCTs, ranging from 32–47%, with a concerning rate of discontinuation of 13–50% (9–11).

Despite its clinical significance, the AI-associated musculoskeletal syndrome remains poorly defined. Reports on time to onset of symptoms encompass a broad range, from 6 weeks to 6 months (7, 9). Case series have described arthralgia, myalgia, and stiffness in women receiving AIs, predominantly involving the hands and wrists (12–14). In 2 small studies, Morales et al described tenosynovitis presenting as trigger finger, carpal tunnel syndrome, and de Quervain's tenosynovitis (15, 16). Other studies, however, describe a more widespread anatomic involvement presenting as osteoarthritis and tendinitis (9). Risk factors for symptom development also have been discordant in the literature. While some studies have suggested an increased risk with lower body mass index (BMI), chemotherapy, and previous hormone replacement therapy (7, 10), others have not demonstrated this relationship (9).

We conducted a 6-month, prospective pilot study of breast cancer patients starting adjuvant AI therapy with the goal of describing the associated musculoskeletal syndrome. In this study, we aimed specifically to define the incidence, time to onset, and risk factors for symptoms development as well as to describe the clinical presentation.

To achieve our objective and provide a comprehensive description of the syndrome, we performed an extensive rheumatologic evaluation, including quality of life (QOL) questionnaires, physical examination, laboratory studies, and magnetic resonance imaging (MRI). Grip strength, a well-validated, objective measure of hand impairment, was chosen as the primary outcome measure based on the existing AI literature that suggests that the musculoskeletal effects of AIs are seen predominantly in the hands (15, 16). Choosing the same outcome measure allowed calculation of a rational sample size.

We anticipate that the description of this syndrome will guide physicians in counseling and treating patients with the hope of improving compliance with this important therapy.

Significance & Innovations

  • More than 50% of women starting aromatase inhibitors develop new or worsening musculoskeletal pain.

  • Focal tenosynovitis of the hand and foot is the most common presentation.

  • Later-stage breast cancer and poorer quality of life are predictive of symptom development.



Postmenopausal women with stage 1–3 hormone receptor–positive breast cancer starting adjuvant AI therapy were enrolled. Exclusion criteria included previous use of AIs, corticosteroid use in the 4 weeks prior to enrollment, other active cancer not in remission (excluding nonmelanomatous skin cancer), or previous diagnosis of any systemic rheumatic diseases or crystal-induced arthritides.

Study design.

In this prospective, single-center, observational cohort study, patients were evaluated by a rheumatologist before initiating AI therapy, at 3 months and 6 months. Symptomatic patients were defined as those who either responded in the affirmative at their 3- or 6-month scheduled study visit to the question, “Have you developed new or worsening muscle and/or joint symptoms since starting AIs?” or contacted an investigator to report new or worsening symptoms at any time. If symptoms were reported between scheduled study visits, the patient was brought in for an additional interim visit within 7 days of symptom report.

All patients without contraindication had contrast-enhanced MRI of the bilateral hands and wrists at baseline and 6 months. Symptomatic participants had an additional interim MRI of the hands and wrists only if they also had an abnormal hand or wrist rheumatologic examination. The interim MRI was performed within 7 days of symptom report.

Time to symptom onset was calculated from the start date of the AIs to the first day the patient noticed symptoms, regardless of when they reported the symptoms. Patients were censored once they became symptomatic.

Clinical outcome measures.

The primary outcome measure was grip strength, measured in kilograms using a Jamar Hydraulic Hand Dynamometer (Sammons Preston Rolyan). Three readings were taken for each hand and were added together for a total score. This outcome measure was chosen based on findings that patients started on AIs were >2 times more likely to have a decrease in grip strength than those started on tamoxifen. Grip strength was a clinically and statistically significant correlate of tenosynovial changes on MRI (16). Furthermore, this well-validated tool is an objective marker of change in function in the hands, the site most commonly involved according to the existing data.

Secondary outcome measures included a standardized 66/68 swollen and tender joint count and a nonstandardized symptom-directed rheumatologic examination (in symptomatic patients) performed by board-certified rheumatologists, a visual analog scale (VAS) for global pain, the Health Assessment Questionnaire (HAQ) disability index (DI), Short Form 36 (SF-36), Australian Canadian Osteoarthritis Hand Index (AUSCAN), Functional Assessment of Cancer Therapy breast subset with endocrine subscale (FACT-B-ES) quality of life instrument, and serum markers for creatine phosphokinase, erythrocyte sedimentation rate, C-reactive protein level, antinuclear antibodies, anti-SSA antibodies, anti-SSB antibodies, rheumatoid factor, and anti–citrullinated protein antibodies.


MRI of both hands and wrists was performed at 1.5T (GE, Excite HDx) simultaneously using a body phased-array coil. Images were acquired using axial and coronal fast spin-echo T1, coronal and sagittal fast spin-echo inversion recovery, and axial and coronal LAVA (Liver Acquisition with Volume Acceleration) before and after contrast administration with fat saturation. Gadopentetate dimeglumine contrast was used at 0.5 moles/liter weight based at 0.2 ml/kg.

MRIs were interpreted using a modified Outcome Measures in Rheumatology Rheumatoid Arthritis MRI Scoring system with tenosynovitis subscale (RAMRIS-TS). The RAMRIS-TS has 4 scores: erosion, edema, synovitis, and tenosynovitis. At each joint in the hand and wrist, erosion is graded from 0–10, and edema, synovitis, and tenosynovitis are graded from 0–3. Combined (right and left) scores range from 0–460, 0–130, 0–42, and 0–60, respectively. Based on the distribution of scores in our population, we used a modified RAMRIS scoring system; combined erosion, edema, and synovitis scores >6 and combined tenosynovitis scores >1 were considered positive.

Statistical analysis.

Descriptive baseline analyses included all enrolled patients. Prospective analyses were performed on all patients who were started on an AI and were seen at least once in followup. Baseline MRI analyses were performed on all patients with a baseline MRI. Prospective imaging analyses were performed on all patients with either a paired interim or a 6-month MRI. For prospective analyses, the last value carried forward was used to impute missing data.

Descriptive statistics were calculated for variables of interest. Univariate analyses included Wilcoxon's rank sum test for continuous and ordinal variables and Fisher's exact test for categorical variables. All analyses were performed using SAS software, version 9.1.


Patient disposition.

Patients were enrolled between January 2009 and March 2010. Ninety-eight patients were referred to the study; 55 were enrolled. Three patients were not included in the prospective analysis: 2 did not start AIs and 1 was lost to followup before the 3-month visit. Of 43 referred but not enrolled, 26 patients declined enrollment and 17 were ineligible (12 with previous AI use, 2 with Lyme disease, 1 with psoriatic arthritis, 1 with polymyalgia rheumatica, and 1 with steroid use). Fifty-two patients were analyzed; 51 completed 6 months of followup and 1 patient was lost to followup after the 3-month visit.

Baseline demographics.

The mean age of enrollees was 60.3 years, and 87% of the participants were white (Table 1). Fifty-four percent had switched from tamoxifen to AIs. Ninety-two percent were prescribed anastrozole and 8% were prescribed letrozole. Sixty-two percent had stage 1 disease. Patient demographics were similar between the groups.

Table 1. Demographic and baseline cancer characteristics*
 Total (n = 55)Symptomatic (n = 28)Asymptomatic (n = 24)P
  • *

    Values are the number (percentage) unless otherwise indicated. BMI = body mass index.

  • Fifty-five patients were enrolled; 2 did not start aromatase inhibitors and 1 was lost to followup. Fifty-two patients were analyzed at 6 months (28 symptomatic and 24 asymptomatic).

  • Comparing symptomatic to asymptomatic.

  • §

    Lymph node dissection was defined as ≥5 lymph nodes removed.

Age, mean (range) years60.3 (44–76)59.8 (44–76)61.5 (45–76)0.52
 White48 (87.3)22 (78.6)23 (95.8)0.06
 Asian4 (7.3)4 (14.3)0 
 African American1 (1.8)01 (4.2) 
 Other2 (3.6)2 (7.1)0 
Hispanic4 (7.3)3 (10.7)1 (4.2)0.61
BMI, kg/m2   0.40
 ≤2528 (50.9)12 (42.9)14 (58.3) 
 ≥2627 (49.1)16 (57.1)10 (41.7) 
 Less than high school graduate2 (3.6)1 (3.4)1 (4.2) 
 High school graduate12 (21.8)5 (17.8)7 (29.2)0.72
 College graduate20 (36.6)11 (39.4)8 (33.3) 
 Graduate degree21 (38.2)11 (39.4)8 (33.3) 
Drug   1.00
 Anastrozole49 (89.1)26 (92.9)22 (91.7) 
 Letrozole4 (7.3)2 (7.1)2 (8.3) 
Previous tamoxifen20 (36.4)10 (35.7)9 (37.5)1.00
Cancer stage   0.04
 134 (61.8)13 (46.3)19 (79.1) 
 217 (30.9)12 (42.9)4 (16.7) 
 34 (7.3)3 (10.7)1 (4.2) 
Her2/Neu positive7 (12.7)4 (14.3)3 (12.5)1.00
Surgery   0.39
 Lumpectomy35 (63.5)16 (57.1)17 (70.9) 
 Mastectomy20 (36.4)12 (42.9)7 (29.1) 
Lymph node dissection§20 (36.6)13 (46.4)6 (25)0.15
 Any24 (43.6)15 (53.6)8 (33.3)0.17
 Paclitaxel based15 (27.3)10 (35.7)4 (16.7) 
Radiation39 (70.9)19 (67.9)18 (75)0.76

Twenty-eight patients (54%) developed new or worsening musculoskeletal symptoms and 2 (3.8%) stopped AIs due to musculoskeletal symptoms. Five participants discontinued the AI for nonmusculoskeletal symptoms: 2 for uterine bleeding and 3 for hot flashes. Sixty-eight percent of symptomatic participants had involvement of their hands and/or wrists and 32% had involvement of their feet and/or ankles. Relatively few subjects had hip, knee, or back pain. The mean time to symptom onset was 6 weeks (median 4 weeks, range 2–18 weeks). Seventy-five percent of symptomatic patients presented with their symptoms by 8 weeks.

In univariate analysis, having stage 2 or 3 cancer was significantly associated with developing musculoskeletal symptoms (P = 0.04) (Table 1). There was no association between the development of AI-associated pain and age, education level, BMI, previous tamoxifen use, Her2/Neu positivity, or exposure to any chemotherapy, including paclitaxel.

All subscales of the FACT and the SF-36 were lower at baseline in the symptomatic group (lower scores indicate poorer QOL) (Table 2). The physical component summary (PCS) score of the SF-36 and both the general FACT and the FACT-B were statistically significantly associated with developing musculoskeletal symptoms (P = 0.01 and 0.03, respectively).

Table 2. Baseline quality of life questionnaire scores, joint counts, and serologic evaluation*
 Total (n = 55)Symptomatic (n = 28)Asymptomatic (n = 24)P
  • *

    Values are the mean unless otherwise indicated. VAS = visual analog scale; HAQ = Health Assessment Questionnaire; DI = disability index; SF-36 = Short Form 36; PCS = physical component summary; MCS = mental component summary; FACT = Functional Assessment of Cancer Therapy; FACT-G = FACT general; FACT-B-TOI = FACT breast cancer trial outcome index; FACT-ES = FACT endocrine symptoms; AUSCAN = Australian Canadian Osteoarthritis Hand Index; ESR = erythrocyte sedimentation rate; CRP = C-reactive protein; CPK = creatine phosphokinase; ANA = antinuclear antibody; RF = rheumatoid factor; ACPA = anti–citrullinated protein antibody.

  • Patients with abnormal joint counts (>1 swollen and/or tender joint).

  • Participants with laboratory values outside the limits of normal.

  • §

    Abnormal ANA titer defined as ≥1:80.

VAS pain (range 0–100)9.6710.3390.45
Grip strength, kg44.844.644.90.96
HAQ DI (range 0–3)
 PCS score (range 0–100)56.855.961.90.01
 MCS score (range 0–100)53.854.154.40.69
 FACT-G total (range 0–108)92.691.495.50.03
 FACT-B-TOI (range 0–92)75.675.7877.90.12
 FACT-B total (range 0–144)119.6117.8123.90.03
 FACT-ES total (range 0–176)152.6152.4154.80.56
 Pain (range 0–20)
 Stiffness (range 0–4)0.350.290.330.75
 Physical function (range 0–36)2.42.571.830.84
 Total (range 0–60)
66/68 joint count, no. (%)    
 Swollen8 (14.6)3 (10.7)5 (20.8)0.45
 Tender12 (21.8)6 (21.4)5 (20.8)1.00
Serologic analysis    
 ESR, mean mm/hour (no. [%])15.87 (4 [7.3])18.25 (3 [10.7])13.7 (1 [4.2])0.61
 CRP level, mean mg/dl (no. [%])0.38 (0 [0])0.5 (0 [0])0.29 (0 [0]) 
 CPK, mean units/liter (no. [%])129.6 (3 [5.5])114 (2 [7.1])151 (1 [4.2])1.00
 ANA, no. (%)§5 (9.1)2 (7.1)3 (12.5)0.65
 RF, no. (%)4 (7.3)2 (7.1)2 (8.3)1.00
 ACPA, no. (%)0 (0)0 (0)0 (0) 
 SSA, no. (%)0 (0)0 (0)0 (0) 
 SSB, no. (%)0 (0)0 (0)0 (0) 

Clinical outcomes.

Grip strength.

There was a trend toward a greater decrease in combined grip strength in the symptomatic group, but these results did not meet statistical significance (−2.89 kg versus −1.39 kg; P = 0.6) (Table 3).

Table 3. Change in outcome variables at 6 months or time of symptom onset*
 Symptomatic (n = 28)Asymptomatic (n = 24)P
  • *

    Values are the mean difference unless otherwise indicated. VAS = visual analog scale; HAQ = Health Assessment Questionnaire; DI = disability index; SF-36 = Short Form 36; PCS = physical component summary; MCS = mental component summary; FACT = Functional Assessment of Cancer Therapy; FACT-G = FACT general; FACT-B-TOI = FACT breast cancer trial outcome index; FACT-ES = FACT endocrine symptoms; AUSCAN = Australian Canadian Osteoarthritis Hand Index; ESR = erythrocyte sedimentation rate; CRP = C-reactive protein; CPK = creatine phosphokinase; ANA = antinuclear antibody; RF = rheumatoid factor; ACPA = anti–citrullinated protein antibody.

  • Patients with abnormal joint counts (>1 swollen and/or tender joint).

  • Participants with laboratory values outside the limits of normal.

  • §

    Abnormal ANA titer defined as ≥1:80.

VAS pain (range 0–100)14.6−2.60.002
Grip strength, kg−2.89−1.390.6025
HAQ DI (range 0–3)
 PCS score (range 0–100)−2.09−1.80.87
 MCS score (range 0–100)−0.420.640.45
 FACT-G total (range 0–108)−0.90.390.63
 FACT-B-TOI (range 0–92)−1.481.560.37
 FACT-B total (range 0–144)−
 FACT-ES total (range 0–176)−1.76−0.580.78
 Pain (range 0–20)2.25−0.17< 0.001
 Stiffness (range 0–4)0.68−0.308< 0.001
 Physical function (range 0–36)1.8600.03
 Total (range 0–60)4.79−0.25< 0.001
66/68 joint count, no. (%)   
 Swollen4 (14.3)2 (8.3)0.67
 Tender9 (32.1)2 (8.3)0.04
Serologic analysis, no. (%)   
 ESR, mm/hour4 (14.3)1 (4.2)0.36
 CRP level, mg/dl0 (0)0 (0) 
 CPK, units/liter5 (17.9)3 (12.5)0.71
 ANA§2 (7.1)3 (12.5)0.65
 RF2 (7.1)2 (8.3)1.00
 ACPA0 (0)0 (0) 
 SSA0 (0)0 (0) 
 SSB0 (0)0 (0) 


Symptomatic patients had a 14.6-point worsening from baseline on the VAS global pain scale (range 0–100), compared to a 2.6-point improvement in the asymptomatic group (P = 0.002) (Table 3). The symptomatic group had greater worsening in all components of the AUSCAN (Figure 1). There was no significant difference in the changes in the HAQ DI, SF-36, and FACT scores between the symptomatic and asymptomatic groups (Table 3). There was no statistically significant association between musculoskeletal symptoms and endocrine symptoms as measured by the FACT-ES. Furthermore, of the 5 participants who stopped AIs due to endocrine symptoms, none had musculoskeletal symptoms.

Figure 1.

Change in Australian Canadian Osteoarthritis Hand Index (AUSCAN) scores at 6 months.

Physical examination and laboratory findings.

A greater proportion of symptomatic patients had an increase in their tender joint count (32.1% versus 8.3%; P = 0.04). There was no difference in swollen and total joint counts. Symptom-directed physical examination revealed tenosynovitis in 13 patients, defined as signs of tendon sheath inflammation and/or enthesitis. There was a positive Finkelstein test for de Quervain's tenosynovitis in 3 patients, tenderness over the digital flexor tendons in 4 patients with locking of the digit in 1 patient, tenderness over the plantar fascia in 4 patients, and tenderness over the Achilles tendon in 1 patient. There was no association between clinical symptoms and any change in serologic markers (Table 3).

MRI findings.

Forty-three of the 52 enrolled patients had baseline MRIs performed (9 patients had a contraindication to MRI). Five patients with baseline MRIs refused the followup MRI, leaving 38 paired MRIs available for analysis. Twenty-one MRIs were performed on symptomatic patients. Of those, 15 were performed at 6 months and 6 were performed as an interim MRI. Mean time from symptom onset to MRI for all patients was 13.5 weeks (range 1–21 weeks), and was 12.5 weeks (range 1–21 weeks) for patients with hand symptoms.

There was no difference in either the mean or the proportion of abnormal baseline MRI RAMRIS-TS scores between the symptomatic and asymptomatic groups (data not shown). Furthermore, there was no difference in the change in modified RAMRIS-TS scores between the groups (Table 4). In a subanalysis comparing patients with symptoms localized to the hands with all other patients (symptomatic without hand involvement and asymptomatic), there was likewise no difference in MRI scores.

Table 4. Change in MRI scores at 6 months or time of symptom onset*
ScoreHand involvementNo hand involvement
Symptomatic (n = 21)Asymptomatic (n = 17)Hand symptoms (n = 16)No hand symptoms (n = 22)
  • *

    Values are the number (percentage) of participants with worsening in their magnetic resonance imaging (MRI) scores at 6 months or at time of symptom onset. No difference in scores met P < 0.05.

Erosion02 (11.8)02 (9.1)
Synovitis1 (4.8)4 (23.5)1 (6.25)4 (18.2)

Clinical course.

Of the 2 patients who stopped AIs due to musculoskeletal symptoms, 1 developed de Quervain's tenosynovitis at week 10. Although the RAMRIS-TS scores were not different between the groups, in this particular patient, an interim MRI showed fluid within the abductor pollicis longus and extensor pollicis brevis tendons (Figure 2). Symptoms continued for 1 month off of AIs despite conservative therapy, and resolved after corticosteroid injection. The second patient developed generalized body aches at 8 weeks. She stopped AIs at 5 months. She remained symptomatic at her final visit 1 month after AI discontinuation.

Figure 2.

Magnetic resonance imaging of the hands at baseline (A and C) and at 6 months (B and D) in a patient with new-onset de Quervain's tenosynovitis. The figure shows abnormal increased T2 signal on coronal stir (arrow; B) and enhancement on axial postcontrast images (arrow; D) in the first extensor compartment, including the right abductor pollicis longus and extensor pollicis brevis tendons.


In this prospective study, we described the AI-associated musculoskeletal syndrome in a large cohort of women. More than half of the enrolled women developed new or worsening musculoskeletal symptoms. Seventy-five percent of the symptomatic women presented within 8 weeks of initiation of AI. Our incident rate was consistent with other practice-based studies but higher than those from RCTs (1–10). In a cross-sectional study, 47% of 200 women questioned in an oncology waiting room reported AI-associated joint symptoms (9). This is comparable to findings from a nested cohort study of the Consortium on Breast Cancer Pharmacogenetics RCT, where 45% of participants developed musculoskeletal symptoms (8). The increased publicity surrounding the musculoskeletal side effects of AIs and the heightened awareness of those willing to participate in a study of these symptoms may account for the higher incident rates than those seen in RCTs.

Nonadherence with AIs due to musculoskeletal symptoms was lower in our study (3.8%) than previously reported (5–13%) (8, 11). This may in part be due to women's unwillingness to stop hormonal therapy given the increasing awareness of the benefit of AIs on breast cancer outcomes.

We were able to identify several possible predictors of symptom development. Later-stage cancer correlated with symptom development. However, contrary to findings from previous studies, development of musculoskeletal symptoms did not seem to be related to chemotherapy or, in particular, paclitaxel exposure (7, 10).

Women with poorer baseline QOL as measured by the SF-36 and FACT-B were also more likely to develop musculoskeletal symptoms. Interestingly, the mean difference in the baseline SF-36 PCS score of 6 between symptomatic and asymptomatic women is larger than the minimum clinically important difference (17). If women who develop pain have worse physical function at baseline, this is a potentially modifiable risk factor for the AI syndrome.

In our study, greater than 50% of the symptomatic patients had involvement of the hand and wrist. Despite this, there was no difference in change in grip strength, our primary outcome chosen based on the existing literature (10). However, the profound effects on the hands were evident by the impressive change in the AUSCAN.

A key finding in our study is that symptoms appeared to reflect tenosynovial pathology that was localized to the hands and feet. This presentation seems to be the defining clinical feature of this syndrome. Thirteen of 28 patients with symptoms had evidence of tenosynovitis. The absence of autoantibody development or a correlation between pain and inflammatory markers suggests a localized enthesitis/tenosynovitis rather than a systemic inflammatory process. Furthermore, the low frequency of lower back and knee symptoms, which are typical sites of postmenopausal pain, suggest pathology other than osteoarthritis of menopause.

Previous studies have also described tenosynovitis of the hands (10, 15, 16). In a cross-sectional study of 12 symptomatic women receiving AIs, the most common symptoms were those of carpal tunnel syndrome and trigger finger (15). MRIs showed tendon sheath enhancement. In a followup prospective study by the same authors, 12 women receiving AIs were more likely to have MRI changes at 6 months consistent with tendon pathology of the hands than the 5 control patients receiving tamoxifen (16). A subsequent review of the anastrozole alone or in combination with tamoxifen versus tamoxifen alone (ATAC) trial showed that clinical carpal tunnel syndrome was reported as an adverse effect more frequently in the anastrozole group compared with the tamoxifen group (2.6% versus 0.7%; P = 0.0001) (18). In our study, tenosynovial pathology was not isolated to the hand. To our knowledge, we are the first to describe symptoms and physical findings of the foot and ankle consistent with plantar fasciitis and Achilles tendinitis in this patient population.

Despite our clinical findings, as well as MRI changes in individual symptomatic patients, tenosynovial pathology was not consistently demonstrated on MRI using the RAMRIS-TS, even among patients with tenosynovial symptoms. There are several possible explanations. There were delays between the onset of musculoskeletal symptoms and both the reporting of these symptoms as well as obtaining MRIs. Therefore, MRIs may not have been performed at the time of peak symptoms. Additionally, our protocol was limited to bilateral hand images using a 1.5T MRI. Individually, 3T MRI imaging of each hand and wrist may have improved capture image quality and increased the radiologist's ability to identify subtle tenosynovial changes. Furthermore, we chose to use a validated scoring system for musculoskeletal MRI of the hands; the RAMRIS, however, may not have been the ideal scoring system. The RAMRIS was designed and validated for use in rheumatoid arthritis, a systemic disease. It includes evaluation of multiple anatomic sites, resulting in a wide range of total scores. Although we used our data to derive meaningful cutoffs to dichotomize positive and negative scores, inflammation in our population was focal, and the RAMRIS-TS may not provide enough discrimination for such a relatively localized tenosynovial process.

The etiology of AI-associated arthralgia remains elusive. Estrogen deprivation has been proposed as the major contributor to joint symptoms. Low estrogen states have been associated with arthralgias as early as 1925, when Cecil and Archer first described “arthralgia of menopause” (19). In a retrospective analysis of the ATAC trial, participants who reported musculoskeletal and vasomotor symptoms at their 3-month study visit had lower cancer recurrence rates. The authors suggest that this indicates that lower estrogen levels achieved in this population account for their musculoskeletal symptoms (20). Contrary to this hypothesis, we found no correlation between endocrine symptoms on the FACT-B-ES and musculoskeletal symptoms. In fact, none of the women who stopped AIs due to endocrine side effects developed musculoskeletal pain. Given these findings, it is important to entertain the idea that other mechanisms aside from estrogen depletion contribute to symptoms. More research needs to be done to pursue alternate possible etiologies.

There are several limitations to our study. Recruitment was from an oncology unit at a single institution. The relatively homogenous patient group reflects the local patient mix, which limits generalizability. Due to prescribing patterns at that center, the majority of participants were taking anastrozole, and no conclusions can be made about the other third-generation AIs. Another limitation is the lack of a control group. Although women being started on tamoxifen as well as those with hormone receptor–negative breast cancer were entertained as possible controls, at our center, the former are premenopausal and younger and the latter undergo more aggressive treatment regimens. Both were not considered an appropriate control group.

In conclusion, the results shown from this large prospective study describe the incidence, time to onset, risk factors, and clinical characteristics of the AI-associated musculoskeletal syndrome. More than half of women prescribed this important hormonal therapy develop musculoskeletal symptoms, most by 8 weeks after starting AIs. Our data suggest that poorer QOL may be a risk factor for symptom development, and it is tempting to speculate there may be a benefit of improving physical function before starting AIs. Oncologists prescribing AIs can use this information to counsel women prior to initiation of these drugs.

Most notably, the defining clinical presentation of the syndrome is that of a focal tenosynovitis, most often involving the hands and feet, not associated with evidence of systemic inflammation or autoimmune disease. Knowledge of the localized nature of this syndrome will aid treating rheumatologists in streamlining their evaluations and focusing management of these patients.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Singer had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Singer, Cigler, Moore, Mandl.

Acquisition of data. Singer, Moore, Levine, Hentel, Belfi, Mandl.

Analysis and interpretation of data. Singer, Do, Mandl.