Osteoporosis-related fractures persist as a global health issue despite the availability of effective treatments to reduce fracture risk. In the European Union, osteoporosis affects an estimated 22 million women and 5.5 million men, resulting in 3.5 million new fragility fractures annually and 37 billion euros per year in direct health care costs [1]. In Canada, there were 57,413 acute care admissions, 832,594 hospitalized days, and $1.2 billion in acute care costs attributable to osteoporosis fractures during 2007–2008 [2]. Patients who sustain fragility fractures experience significantly worse health-related quality of life (HRQoL) [35] and face an increased risk of mortality [6, 7].

There are several treatment options with demonstrated efficacy in reducing fracture risk in osteoporotic patients, including bisphosphonates. Bisphosphonates are the most commonly prescribed therapy. However, among patients who initiate treatment, adherence to and persistence with therapy is often poor in clinical practice. Among patients initiating bisphosphonate therapy, only approximately 40–60 % are adherent to therapy during the first year [810] and a similar proportion are non-persistent in their first year of treatment [1012]. The consequences of low adherence and persistence include greater fracture risk [10, 13] and the sequalae of higher health care utilization and costs [1416].

Gastrointestinal (GI) symptoms among patients on oral bisphosphonates may also contribute to worse clinical outcomes. Although controlled trials have found no differences in the rate of GI symptoms among patients (with and without pre-existing GI disorders) treated with oral bisphosphonates compared with placebo [17, 18], analyses of real-world data have noted upper GI complaints often manifesting as dyspepsia, gastroesophageal reflux, and nausea [19] and less frequently as esophagitis [20] and gastric ulcers [21]. Patients with a history of GI disorders are more likely to develop symptoms on oral bisphosphonate treatment [2224], and both pre-treatment gastrointestinal complaints and complaints while on treatment have been associated with discontinuation of oral bisphosphonates and other osteoporosis therapies [22, 2527]. Additionally, evidence from the POSSIBLE-US observational study of osteopenic and osteoporotic women suggests that patients who experience GI symptoms on treatment have lower treatment satisfaction than patients without GI complaints [26] and lower treatment satisfaction was associated with a higher risk of treatment discontinuation or switching [28]. GI complaints on treatment were also linked to lower HRQoL among patients newly initiating therapy [26]. However, the rate of GI symptoms and the relationship between GI symptoms and patient-reported outcomes has not been well explored in osteoporotic women in the EU and Canada.

This study focuses on postmenopausal women in the EU and Canada who were initiating or continuing oral pharmacological therapy for osteoporosis at the start of the study. The objectives were to identify treatment patterns, GI symptoms, and the association between GI symptoms and treatment adherence, treatment satisfaction, health-related quality of life, and health care resource utilization.


Study design and participants

The study design, recruitment, and patient baseline characteristics of the Medication Use Patterns, Treatment Satisfaction, and Inadequate Control of Osteoporosis Study (MUSIC-OS) have been previously reported [29]. In brief, MUSIC-OS investigators enrolled postmenopausal women aged 55 and older, with a physician diagnosis of osteoporosis, from 96 primary care and specialty clinics in Canada, France, Italy, the Netherlands, Sweden, and the UK. The study was carried out in accordance with the Declaration of Helsinki, the standards of good clinical practice, and either the local ethics boards of the participating institutions or a central institutional review board. All participants provided written informed consent prior to study enrolment. Potential enrollees were identified at each site at the time of a medical consultation. This analysis describes the results of data collected at the enrollment (baseline) visit for new users who were initiating oral pharmacological treatment for osteoporosis and experienced users who were continuing the same oral pharmacological treatment.

Qualifying oral pharmacologic treatment for osteoporosis in MUSIC-OS included bisphosphonates (e.g., alendronate, risedronate, ibandronate), calcitonin, strontium ranelate, and selective estrogen receptor modulators (SERMs [e.g., raloxifene, and bazedoxifene]). Calcium, vitamin D, and estrogen and other hormone replacement therapy were not considered pharmacologic treatments for osteoporosis in this study. New users were patients either initiating qualifying oral pharmacological therapy at enrollment or receiving qualifying oral pharmacological therapy for less than 3 months prior to enrollment with no previous history of any osteoporosis pharmacological therapy. Experienced users were receiving the same qualifying oral pharmacological therapy for at least 3 months continuously prior to enrollment and were continuing treatment at the time of enrollment. Each cohort was further categorized by the presence or absence of GI symptoms by asking patients at baseline if they had experienced any GI symptoms in the 6 months prior to enrollment.


Patients were interviewed, and their medical chart history was reviewed during the enrolment visit (i.e., baseline visit) to obtain information on qualifying oral pharmacological osteoporosis therapy (defined above), concomitant medications, and GI symptoms. Concomitant medications included other medications/supplements taken for osteoporosis (e.g., calcium and/or vitamin D, estrogen and hormone replacement therapy, parathyroid hormone), gastrointestinal conditions (e.g., antacids), and drugs linked to gastrointestinal conditions, e.g., acetaminophen, non-steroidal anti-inflammatory drugs). The presence and type of GI symptoms were assessed by asking the patient if she had experienced any of a defined list of upper (heartburn/acid reflux, upset stomach/indigestion, nausea/vomiting, pain behind breastbone, pain or difficulty swallowing, stomach pain above navel) or lower (diarrhea or constipation, stomach pain below navel, bloating) GI problems in the last 6 months. Health care utilization in the 3 months prior to enrolment was assessed by asking patients to report medical services utilized specific to osteoporosis-related concerns and those specific to gastrointestinal-related concerns. Service categories captured were visits to family physician/general practitioner and specialists.

Patients completed four instruments to assess treatment adherence, treatment satisfaction, and HRQoL. Adherence was assessed among experienced users with the 12-item adherence evaluation of osteoporosis (ADEOS) treatment questionnaire, which has been validated in women with osteoporosis [30]. Scores range from 0 to 22 with scores ≥20 predicative of a high probability of medication persistence. Treatment satisfaction was measured with the Osteoporosis Treatment Satisfaction Questionnaire for Medications (OPSAT-Q), which uses Likert scales to assess four domains (treatment convenience, confidence in daily functioning, overall satisfaction with treatment, and side effects) with the total score reported on a 0–100 scale (higher scores indicating greater satisfaction) [31]. The OPSAT-Q has been validated in women taking bisphosphonates [31]. HRQoL was evaluated with the EuroQol-5 dimension (EQ-5D) utility which assesses health state [32], and the disease-specific Osteoporosis Assessment Questionnaire-Short Version (OPAQ-SV) which assesses health status in three domains: physical function, emotional status, and back pain [33].

Statistical analyses

The recruitment goal was 3300 treated patients to achieve a sample size of 2640, assuming 20 % loss to follow-up. This sample size is estimated assuming a normal distribution to provide 95 % confidence intervals at a 0.05 significance level. We computed descriptive statistics for GI symptoms. The difference in the proportion of new and experienced users reporting GI symptoms was examined by chi-squared test. The association between the presence/absence of GI symptoms and ADEOS, OPSAT-Q, EQ-5D, and OPAQ-SV scores was assessed with general linear models. Odd ratios for GI- and OP-related health care utilization were computed by logistic regression models. Separate models were constructed for each cohort (experienced vs. new user) and each measure (ADEOS, OPSAT-Q, EQ-5D, OPAQ-SV) and each component (general practitioner or specialist visit) of GI- and OP-related health care utilization. Modeling was conducted in two phases: an initial full model followed by a final reduced model. The initial full models were adjusted for the following patient characteristics: age group in years (50–59 [reference]; 60–69, 70–79; ≥80), race (white [reference], non-white), BMI category (underweight <18.5; normal 18.50–24.99 [reference]; overweight 25.00–25.99; obese ≥30), duration of osteoporosis at study entry in years (<1 [reference]; 1–<5, 5–10; >10), duration of pharmacological treatment for osteoporosis at study entry in years (<1 [reference]; 1–<5, 5–10; >10; this variable was excluded from the model for new users), history of falls in the past 12 months (yes, no [reference]), history of osteoporotic fracture at any time in the past (yes, no [reference]), presence of any selected comorbid conditions (yes, no [reference]; conditions included hypertension, diabetes, chronic kidney disease, rheumatoid arthritis, hypothyroidism, anorexia nervosa, celiac disease, inflammatory bowel disease, lactose intolerance, lupus, asthma, stroke, dementia, chronic obstructive pulmonary disease), combination treatment (receiving bisphosphonates and non-bisphosphonates, receiving bisphosphonates [reference]), treatment class (receiving non-bisphosphonates, receiving bisphosphonates [reference]), and concomitant medication use (yes, no [reference]). Because not all participants had values for each covariate (e.g., race information was missing in France), the final model was a reduced model selected by using the subset of covariates in the full model that was significant at P ≤ 0.15 by backward elimination. The ADEOS and OPSAT-Q scores were not modeled for new users because these patients were either initiating osteoporosis therapy at study entry or had <3 months exposure to therapy. Covariates with unstable confidence intervals were excluded from models.



We enrolled 3335 patients from 96 sites. The number of patients (% of total study enrollment) and number of sites by country, respectively were UK, 873 patients (26.2 %), 22 sites; Canada, 760 patients (22.8 %), 15 sites; France, 661 patients (19.8 %), 27 sites; Italy, 593 patients (17.8 %), 22 sites; Netherlands, 225 patients (6.7 %), 7 sites; Sweden, 223 (6.7 %), 3 sites. Of the 3335 enrolled patients, 2980 were protocol eligible, treated patients. After excluding 21 patients with missing data for GI symptoms, there were 2275 patients in the experienced user cohort and 684 patients in the new user cohort.

Treatment patterns

Bisphosphonates were the most common treatment overall (79.9 %) and among each user cohort and GI symptom group (Table 1). Among new users, 44.8 % were not receiving pharmacological osteoporosis therapy at the baseline visit. Although inclusion in this analysis required that all patients be treated with qualifying osteoporosis medications, the percentage of new users on pharmacological therapy at enrollment is less than 100 because some new users were prescribed their treatment on the day of the enrollment visit. Overall, 10.2 % of patients were taking only non-pharmacological oral medications (calcium and/or vitamin D treatment, estrogen, and/or hormone replacement therapy), and the majority of these patients were new users. Among all patients, 71.8 % reported taking a calcium and/or vitamin D supplement (data not shown). Experienced users had been taking oral pharmacologic therapy for approximately 4 years.

Table 1 Osteoporosis treatment patterns at enrolment

Frequency of gastrointestinal symptoms

There were 2015 (68.1 %) patients who reported GI symptoms in the 6 months prior to enrollment (Table 2). The proportion of experienced users reporting GI symptoms was higher than new users (69.1 vs. 64.6 %; P < 0.03). The most common upper GI symptoms reported by all patients were heartburn/acid reflux (35.0 %) followed by upset stomach/indigestion (28.4 %). Diarrhea or constipation (38.2 %) and bloating (37.3 %) were the most frequently reported lower GI symptoms.

Table 2 Patient-reported gastrointestinal symptoms during 6 months prior to enrollment

Adherence and treatment satisfaction

Among experienced users (n = 2228), 49.2 % had an ADEOS score ≥20, indicative of a high probability of medication persistence (data not shown). ADEOS scores ≥20 were less frequent among experienced users with GI symptoms compared with users without GI symptoms (45.5 vs. 57.6 %; data not shown). The association between GI symptoms and ADEOS and OPSAT-Q scores, adjusted for patient demographic and clinical characteristics, is shown in Table 3. The adjusted mean difference between the GI and no GI symptom groups was −0.43 (P < 0.001) with lower mean ADEOS scores in the GI symptom group. Treatment satisfaction measured by OPSAT-Q score was also significantly lower among experienced users with GI symptoms: The mean adjusted difference between the GI and no GI symptom groups was −5.68 (P < 0.0001).

Table 3 Multivariate analyses of the association between GI symptoms and treatment adherence (ADEOS) and treatment satisfaction (OPSAT-Q) scores for experienced users

Health-related quality of life

The association between GI symptoms and measures of HRQoL, adjusted for patient characteristics, is shown in Table 4. Compared with patients without GI symptoms, patients with GI symptoms had significantly lower mean adjusted EQ-5D utility scores in both the new user (mean difference −0.04, P < 0.0099) and experienced user (mean difference −0.06, P < 0.0001) cohorts. For the disease-specific OPAQ-SV, GI symptoms in new users were associated with significantly lower adjusted mean scores in the domains of emotional status (mean difference −4.28, P < 0.01) and back pain (mean difference −5.82, P < 0.01). Experienced users with GI symptoms had lower scores in all three domains with adjusted mean differences of −4.17 for physical function, −5.68 for emotional status, and −11.33 for back pain (all P < 0.0001).

Table 4 Association between GI symptoms and measures of health-related quality of life for new and experienced users

Health care resource utilization

The likelihood of osteoporosis-related and GI-related health care utilization among patients with GI symptoms (reference: no GI symptoms), adjusted for patient characteristics, is shown in Fig. 1. Among new users with GI symptoms, there was a non-significant trend for higher odds of visiting general practitioners and specialists for osteoporosis-related services compared with new users without GI symptoms. However, experienced users with GI symptoms had a significantly higher risk of osteoporosis-related service utilization for general practitioner visits (odds ratio [OR]: 1.35, 95 % CI = 1.05–1.73, P < 0.03) but not specialist visits. The most pronounced difference between experienced users with and without GI symptoms was in GI-related service utilization. Experienced users with GI symptoms had approximately ninefold higher risk (95 % CI = 4.58–17.86, P < 0.0001) for GI-related general practitioner visits and 4.2 higher odds (95 % CI = 1.79–9.93, P = 0.001) of GI-related specialist visits compared with experienced users with no GI symptoms.

Fig. 1
figure 1

a Likelihood of osteoporosis-related health care resource use for new and experienced users with and without GI symptoms. Values are adjusted odds ratios from the logistic regression model with backward elimination. Adjustment variables retained in the model for GP visits were as follows: among new users: treatment class and concomitant medication use; among experienced users: duration of osteoporosis, history of fractures, comorbidities, and treatment class. Adjustment variables retained in the model for specialist visits were as follows: among new users: treatment class, concomitant medication use, and comorbidities; among experienced users: age, race, duration of osteoporosis, history of fractures, and treatment class: GI gastrointestinal, GP general practitioner. b Likelihood of gastrointestinal-related health care resource use for experienced users with and without GI symptoms. Values are adjusted odds ratios from the logistic regression model with backward elimination. Adjustment variables retained in the model for GP visits were as follows: age group, duration of osteoporosis, history of fractures, history of falls, treatment class, and concomitant medication use. Adjustment variables retained in the model for specialist visits were body mass index, duration of osteoporosis treatment, history of fractures, and concomitant medication use. GI gastrointestinal, GP general practitioner


The results of this baseline analysis of treated patients in MUSIC-OS reveal the high rate of GI symptoms in this cohort and the negative association of GI symptoms with patient-reported outcomes. The presence of GI symptoms was associated with small but consistently negative decrements in medication adherence, treatment satisfaction, and HRQoL.

At baseline, 68.1 % of patients reported GI symptoms in the last 6 months which is higher than observed in either POSSIBLE-EU [34] or POSSIBLE-US [26]. This is likely a result of methodological differences in capturing GI disorders. In the POSSIBLE-EU study, 22 % of patients had current or prior upper GI disorders and 9 % had current or prior lower GI disorders [34]. However, POSSIBLE-EU employed a narrower list of specific upper (GERD reflux, dyspepsia) and lower GI (irritable bowel syndrome, Crohn’s disease) disorders than MUSIC-OS, and site investigators (not patients) reported GI disorders which may have limited reporting to those GI disorders that resulted in medical consultation or treatment. In the POSSIBLE-US study, 21 % of patients reported a constellation of upper and lower GI symptom similar to those defined in the MUSIC-OS protocol [26]. However, our results are based on a 6-month recall period, and POSSIBLE-US symptoms were captured “at study entry.” The recall period is a likely contributor to the difference in rate of GI symptoms between the two studies. For example, 31 % of POSSIBLE-US participants had a history of gastroesophageal reflux disease, a subset of upper GI symptoms, which is higher than the rate (21 %) reported for all upper and lower GI symptoms at study entry. The proportion of patients using GI medications was similar among the two studies: 20.7 % of MUSIC-OS patients used GI medications in the last 12 months compared with approximately 25 % of POSSIBLE-US participants at study entry, suggesting a similar rate of GI conditions treated with medications although the difference in recall period (last 12 months vs. study entry) precludes a direct comparison. Irrespective of methodological differences, the results of POSSIBLE-US and MUSIC-OS suggest that GI complaints are quite common among women initiating or continuing osteoporosis therapy.

Among all experienced users (with and without GI symptoms), only 49.2 % had an ADEOS score ≥20, which is indicative of a high probability of medication persistence. Although ADEOS scores were not well correlated with adherence as measured by medication possession ratio (MPR) in the validation study [30], our results are within the range of adherence to osteoporosis therapy observed in administrative claims studies that calculated adherence as MPR. An MPR threshold of 0.80 is commonly used to distinguish between adherence and non-adherence; in the first year of therapy, typically 34–55 % of patients are adherent to pharmacological osteoporosis treatments [810]. GI symptoms were associated with a small but significant decrease in ADEOS scores. The association between GI symptoms and lower ADEOS scores is consistent with previous studies of osteoporosis medication persistence during follow-up. In POSSIBLE-US, women with GI symptoms at study entry were 38 % more likely to discontinue therapy within 6 months [26]. A higher risk of osteoporosis therapy discontinuation has been observed among patients taking GI medication [35, 36]. MUSIC-OS patients are being followed for 12 months, and future analysis will explore the relationship between recurrent GI symptoms, therapy discontinuation, and adherence as measured by ADEOS scores.

GI symptoms were also associated with consistently lower treatment satisfaction and HRQoL scores. In US studies, decrements in treatment satisfaction have been linked with increasing severity of side effects [28] and with the presence of GI side effects in particular [26]. Women treated with bisphosphonates were more likely to report GI side effects than women treated with other osteoporosis therapies [26], and upper GI symptoms during bisphosphonate therapy are a prominent risk factor for early treatment discontinuation [27]. The lower HRQoL scores we observed in patients with GI symptoms are also consistent with POSSIBLE-US results: Among new users, GI side effects at 6 months were associated with lower mean OPAQ-SV emotional status at 6 months [26] and lower OPAQ-SV physical function at study entry predicted non-persistence with initial therapy [37]. Our results coupled with previous studies suggest that the selection of pharmacologic treatment for osteoporotic women should account for pre-existing GI symptoms and the potential for posttreatment GI symptoms to minimize the risk of therapy discontinuation.

MUSIC-OS is an observational study, and there are inherent limitations. In order to ensure robust implementation of the protocol, we screened sites based on their experience and ability to conduct observational research which may have biased the sample to physicians who are more engaged in the clinical management of osteoporosis. Self-selection bias may have occurred for both physicians who chose to participate and patients who elected to enroll. GI symptoms in the 6 months prior to baseline were patient-reported and may be subject to recall bias. Further, the presence of GI symptoms did not require evidence of medical consultation or treatment; thus, the severity of symptoms cannot be ascertained. Although GI symptoms were associated with lower scores on patient-reported outcomes, we cannot infer causation. Further, although scores in patients with GI symptoms were consistently and significantly lower than patients without GI symptoms, the differences were small and may not be clinically significant. We adjusted scores for multiple patient demographic and clinical characteristics; however, there may have been other unmeasured factors that could have influenced results.


Our results suggest that GI symptoms may deter patients from complying with therapy and have an adverse effect on treatment satisfaction, quality of life, and use of health care resources. MUSIC-OS will follow treated patients for 1 year and provide a comprehensive assessment of contemporary treatment patterns and patient-reported outcomes in Europe and Canada. Future analyses will explore overall medication discontinuation/switching patterns, patient rationale for discontinuation, adherence, treatment satisfaction, HRQoL, fracture incidence, and health care utilization. The association between treatment-emergent GI symptoms and these outcomes will also be assessed. These results will provide additional insight to improve clinical management of osteoporotic women.