Adherence with bisphosphonate therapy in US veterans with rheumatoid arthritis




Pharmacy Benefits Management program data for patients enrolled in the Veterans Affairs Rheumatoid Arthritis (VARA) registry were linked with clinical data to determine bisphosphonate adherence and persistence among US veterans with rheumatoid arthritis (RA) and to determine factors associated with adherence.


The primary outcome measures were the duration of bisphosphonate therapy and the medication possession ratio (MPR). Patients with an MPR <0.80 were classified as nonadherent. Potential covariates considered in the analysis included patient demographics, RA disease activity and severity parameters, and factors associated with osteoporosis risk. Associations of patient factors with duration of therapy and adherence were examined using multivariable regression modeling.


Bisphosphonates were prescribed to 573 (41.5%) of 1,382 VARA subjects. The mean ± SD duration of therapy for bisphosphonates was 39.2 ± 31.4 months. A longer duration of therapy correlated with older age, more years of education, and dual x-ray absorptiometry testing. The mean ± SD MPR of VARA subjects for bisphosphonate therapy was 0.69 ± 0.28; 302 (52.7%) were nonadherent. In multivariate analyses, nonadherence with bisphosphonate therapy was associated with a longer duration of RA disease (odds ratio [OR] 1.02, 95% confidence interval [95% CI] 1.00–1.04) and duration of bisphosphonate therapy >32 months (OR 1.63, 95% CI 1.04–2.57). Whites were less likely to have a low MPR compared with nonwhites (OR 0.52, 95% CI 0.30–0.88).


Nonadherence with bisphosphonates was common in this cohort of RA patients and was associated with nonwhite ethnicity, a longer duration of RA disease, and a greater duration of bisphosphonate therapy.


Osteoporosis is common in both women and men and predisposes to fragility fractures primarily of the hip, spine, or distal forearm. The lifetime fracture risk is reported to be nearly 40% for white women ages >50 years, but only 13% for white men of a similar age (1). Ethnic minorities have a smaller although still significant lifetime risk for osteoporotic fracture. A report from the California Hospital Discharge Data revealed a rate of hip fractures for non-Hispanic white women of 140.7/100,000 compared to rates of 49.7, 57.3, and 85.4 for African American, Hispanic, and Asian women, respectively (2). Men have a worse outcome following hip fractures, with mortality rates twice that of women in the first and second years following hip fractures (3).

Effective therapies for the prevention of both primary and secondary fragility fractures are available, and bisphosphonates are among the most common pharmacologic agents prescribed. Randomized controlled clinical trials have demonstrated the efficacy of alendronate, risedronate, ibandronate, and zoledronic acid in increasing bone mineral density (BMD) and reducing fractures (4–7).

Adherence with bisphosphonate therapy is associated with a measurable reduction in fractures compared with nonadherence (8). Most reports on bisphosphonate adherence have focused on women with postmenopausal osteoporosis, including African Americans and populations of European ancestry (9–12). In these prior studies, determinants of poor adherence with bisphosphonate treatment were found to include older age, daily versus weekly bisphosphonate dosing, number of comorbid conditions, male sex, the absence of BMD testing, side effects, and poor physical health status (10–13). It is unclear if these findings can be extrapolated to other groups, particularly men with rheumatoid arthritis (RA).

RA is associated with marginal joint erosions, periarticular osteopenia, and generalized bone loss or osteoporosis. Synovial cytokines, including macrophage colony-stimulating factor and RANKL, promote osteoclast differentiation and subsequent bone erosion (14). The mechanisms for generalized osteopenia distant from synovial inflammation remain controversial. Histomorphometric analysis of bone biopsies reveals decreased bone formation in these areas, but increased bone turnover is a factor (15). Other clinical factors that are contributory include the use of glucocorticoids and decreased mobility (16).

The purpose of this study was to examine adherence with oral bisphosphonate therapy in a cohort of RA patients utilizing data from the Veterans Affairs Rheumatoid Arthritis (VARA) registry and the Veterans Affairs (VA) Pharmacy Benefits Management (PBM), and to determine factors associated with nonadherence.

Significance & Innovations

  • A longer duration of bisphosphonate therapy is associated with greater years of education, older age, and receiving a dual x-ray absorptiometry test.

  • Poor adherence with bisphosphonate therapy is associated with nonwhite ethnicity.

  • Poor bisphosphonate adherence is associated with a longer duration of rheumatoid arthritis disease and a greater duration of bisphosphonate therapy.



The VARA registry is a longitudinal multicenter chronic disease cohort of US veterans who fulfill the 1987 American College of Rheumatology criteria for the classification of RA, and is described in detail elsewhere (17–19). In summary, the registry began in 2003 and currently enrolls patients from 11 VA medical centers. The registry database includes patient demographics, longitudinal markers of RA disease activity and severity, comorbid diseases, and BMD results for those undergoing dual x-ray absorptiometry (DXA) as part of their routine care. VARA subjects eligible for this analysis were required to have received at least 1 prescription for an oral bisphosphonate.

Bisphosphonate therapy.

The PBM database contains detailed information on every prescription filled within the VA Health Care System since 1999 and includes medication name, dose of tablet, date dispensed, number of tablets dispensed, expected duration of therapy, and prescribing directions (20, 21). At the time of the analysis, PBM data on VARA patients were available for 6 VA medical centers (Dallas, Denver, Jackson, Omaha, Salt Lake City, and Washington, DC). The PBM data were analyzed for the bisphosphonates alendronate, risedronate, etidronate, and ibandronate, which included all doses dispensed within the VA for these medications. Dosing regimens used for Paget's disease and parenteral preparations were not included for this study. For each dispensed prescription, an expected duration of therapy was calculated. The dispensing date and expected duration were used to calculate the end date for each prescription. Differences between the expected end of a prescription and the dispensing date of the subsequent medication refill were considered gaps in therapy. A patient was defined as receiving a course of bisphosphonate therapy if there was a <90-day gap between refills for a specific bisphosphonate. A course ended when a 90-day gap occurred as defined above or the medication was discontinued. The duration of the course was defined from the date of the first bisphosphonate prescription dispensed until the expected end date for the last prescription before a 90-day gap or the end of the observation period. The duration of therapy was measured from the date of the initial prescription until the last expected refill date of the final prescription, and therefore included the duration of all courses plus gaps for each patient.

Bisphosphonate adherence.

Medication adherence was assessed by calculating the medication possession ratio (MPR), defined as the proportion of treatment time that a patient had an available drug (22). The MPR was determined for the duration of therapy by calculating the time that the patient had medication available divided by the time from the date the medication was initially dispensed until the last expected refill date of the final prescription. The details of the extraction of medication courses from PBM were recently described for methotrexate; we used similar methodology for bisphosphonates (23). Current data indicate that fracture risk decreases as medication adherence increases, from 50% to 80%, but with no further reduction beyond 80% (24). Therefore, for this analysis, subjects were deemed adherent with bisphosphonate therapy if the MPR was ≥0.80 and nonadherent if the MPR was <0.80.

Primary outcome measures and patient characteristics.

Duration of bisphosphonate therapy and an MPR <0.80 were the primary outcome measures. To examine factors potentially contributing to the outcomes, we analyzed covariates, including sex, age (<70 versus ≥70 years), race/ethnicity (nonwhite versus white), education (<12 versus ≥12 years), current smoking status, comorbidity measured by the Charlson-Deyo Comorbidity Index (25), disease activity as measured by the mean Disease Activity Score in 28 joints (DAS28) (26), the mean Multidimensional Health Assessment Questionnaire (MDHAQ) score (27, 28), duration of RA, use of prednisone during VARA enrollment, duration of bisphosphonate therapy (for an MPR <0.80), and whether patients had DXA performed or a densitometric diagnosis of osteoporosis.

Statistical analysis.

Adherent and nonadherent patients were compared using chi-square tests for categorical variables and univariate one-way analysis of variance for continuous variables. Multivariate associations with nonadherence were examined using logistic regression analysis. Odds ratios (ORs) and their 95% confidence intervals (95% CIs) were calculated. Mean duration of therapy was measured and the SD was calculated. Multivariate associations of patient factors with duration of therapy were examined using linear regression analysis. Covariates with P values greater than 0.1 in the univariate analyses were not included in the multivariate analyses for both duration of bisphosphonate therapy and an MPR <0.80.

Human subjects review.

The VARA registry received Institutional Review Board approval at each site and all enrollees provided written informed consent. This study was approved by the Scientific and Ethics Advisory Committee for the VARA registry.


There were 1,382 subjects enrolled in the VARA at the 6 VA sites that had data available for analysis. Five hundred seventy-three subjects (41.5%) were dispensed at least 1 prescription for an oral bisphosphonate and served as the cohort for analysis. These patients were mostly men (89.9%), with a mean age of 68.7 years. The majority of subjects were white (81.7%), with African Americans and Hispanics accounting for 12.8% and 3.5% of the cohort, respectively. DXA testing was available in 258 subjects (45.0%), 66 (25.6%) of whom had a densitometric diagnosis of osteoporosis. Three hundred ninety-two individuals (68.4%) were prescribed prednisone. Apart from being more likely to have osteoporosis compared with non–bisphosphonate users, bisphosphonate users were more likely to be older, be white, have a DXA scan performed, receive prednisone, have higher Charlson-Deyo scores, have higher mean MDHAQ and mean DAS28 scores, and have a longer duration of RA (Table 1).

Table 1. Characteristics of bisphosphonate users and nonusers among Veterans Affairs Rheumatoid Arthritis enrollees*
 Total sample (n = 1,372)Subjects ever used bisphosphonates (n = 573)Subjects never used bisphosphonates (n = 799)
  • *

    DXA = dual x-ray absorptiometry; RA = rheumatoid arthritis; MDHAQ = Multidimensional Health Assessment Questionnaire; DAS28 = Disease Activity Score in 28 joints.

  • P < 0.001 (bisphosphonate users vs. nonusers).

  • P < 0.05 (bisphosphonate users vs. nonusers).

  • §

    P < 0.01 (bisphosphonate users vs. nonusers).

Age, mean ± SD years66.3 ± 11.568.7 ± 10.364.6 ± 12.0
Women, no. (%)127 (9.3)58 (10.1)69 (8.6)
Ethnicity, no. (%)   
 White1,053 (76.7)468 (81.7)585 (73.2)
 Nonwhite304 (22.2)100 (17.5)204 (25.5)
Education, mean ± SD years12.8 ± 2.612.8 ± 2.612.9 ± 2.7
Smoking (current), no. (%)367 (26.8)146 (25.5)221 (27.7)
Charlson-Deyo Index, mean ± SD1.8 ± 2.21.99 ± 2.081.72 ± 2.25
Prednisone use, no. (%)828 (60.4)392 (68.4)43 (5.4)
DXA testing, no. (%)495 (36.1)258 (45.0)237 (29.7)
Osteoporosis, no. (%)79 (16)66 (25.6)13 (5.5)
Duration of RA, mean ± SD years16.6 ± 12.319.1 ± 12.614.9 ± 11.8
MDHAQ score, mean ± SD§1.0 ± 0.51.05 ± 0.490.96 ± 0.51
DAS28, mean ± SD§3.63 ± 1.183.74 ± 1.183.54 ± 1.17

Four hundred ninety-seven patients (86.7%) received alendronate as their only bisphosphonate and 17 (3.0%) received risedronate; 59 patients (10.3%) received prescriptions for both of these bisphosphonates at some point during the period of observation. Seven of the 556 subjects prescribed alendronate received only a once daily preparation, and 1 of the 76 subjects prescribed risedronate received a once daily prescription. Three subjects received a prescription for etidronate and no subjects were prescribed ibandronate. These subjects given either etidronate or daily bisphosphonate therapy were excluded from further analysis, given the likely influence of this dosing schedule on adherence and the low number of subjects. Subjects received a mean ± SD of 2.3 ± 1.6 courses of therapy and 240 subjects (42%) were only dispensed a single course (range 1–11 courses).

The mean ± SD duration of oral bisphosphonate therapy was 39.2 ± 31.4 months. Thirty-five patients receiving once weekly oral bisphosphonates were switched to intravenous zoledronic acid and 6 were switched to teriparatide. The univariate and multivariate associations of treatment duration are shown in Table 2. In the multivariate analysis, a longer duration of therapy was associated with age ≥70 years (OR 1.79, 95% CI 1.04–3.08), education past twelfth grade (OR 2.12, 95% CI 1.09–4.14), and receiving DXA (OR 1.74, 95% CI 1.05–2.89).

Table 2. Predictors of prescribed duration of bisphosphonate therapy of >32 months for Veterans Affairs Rheumatoid Arthritis subjects*
 Univariate OR (95% CI)Adjusted OR (95% CI)
  • *

    OR = odds ratio; 95% CI = 95% confidence interval; DXA = dual x-ray absorptiometry; RA = rheumatoid arthritis; MDHAQ = Multidimensional Health Assessment Questionnaire; DAS28 = Disease Activity Score in 28 joints.

  • Adjusted for age ≥70 years, female sex, enrollment site, number of bisphosphonate courses, education ≥12 years, DAS28 (mean), Charlson-Deyo Index, prednisone use, and DXA.

  • P < 0.01.

  • §

    P < 0.05.

  • Not included in the multivariate model because P > 0.10.

Age ≥70 years1.70 (1.22–2.37)1.79 (1.04–3.08)§
Female sex0.73 (0.42–1.26)0.74 (0.33–1.66)
Nonwhite1.32 (0.85–2.03)1.32 (0.70–2.48)
Education ≥12th grade1.29 (0.78–2.13)2.12 (1.09–4.14)§
Smoking (current)0.57 (0.39–0.83)0.64 (0.36–1.15)
Charlson-Deyo Index0.98 (0.95–1.01)1.03 (0.99–1.03)
Prednisone use0.86 (0.60–1.23)0.73 (0.41–1.29)
DXA1.16 (0.83–1.61)1.74 (1.05–2.89)§
Osteoporosis1.04 (0.59–1.82) 
Duration of RA0.98 (0.96–0.99)1.01 (0.99–1.03)
MDHAQ score, mean1.04 (0.74–1.46)1.26 (0.72–2.19)
DAS28, mean1.03 (0.89–1.19)0.87 (0.68–1.11)

The mean ± SD MPR of VARA subjects receiving bisphosphonate therapy was 0.69 ± 0.28. Approximately half (302 [52.7%]) were nonadherent (MPR <0.80). If gaps >90 days were removed from the duration of therapy and the MPR calculated for each course the patient received, giving more weight to longer courses, the mean ± SD MPR would be 0.90 ± 0.13; this MPR was not used for further analyses. In univariate analyses, nonadherence with bisphosphonate therapy was associated with <12 years of education (OR 1.67, 95% CI 1.00–2.79), a longer duration of RA (OR 1.02, 95% CI 1.01–1.04), a higher mean DAS28 score (OR 1.17, 95% CI 1.01–1.35), and a duration of bisphosphonate therapy ≥32 months (OR 1.98, 95% CI 1.42–2.77). In the multivariate analysis, whites were less likely to be nonadherent (OR 0.52, 95% CI 0.30–0.88). The MPR <0.80 for bisphosphonates was associated with the duration of RA (OR 1.02, 95% CI 1.00–1.04) and bisphosphonate therapy prescribed for ≥32 months (OR 1.63, 95% CI 1.04–2.57). The univariate and multivariate analyses for MPR are shown in Table 3.

Table 3. Univariate and multivariate associations between demographic and clinical variables and MPR*
 MPR <0.80MPR ≥0.80MPR <0.80, OR (95% CI)
  • *

    Values are the number (percentage) unless otherwise indicated. MPR = medication possession ratio; OR = odds ratio; 95% CI = 95% confidence interval; DXA = dual x-ray absorptiometry; RA = rheumatoid arthritis; MDHAQ = Multidimensional Health Assessment Questionnaire; DAS28 = Disease Activity Score in 28 joints.

  • Adjusted for age <70 years, female sex, white ethnicity, enrollment site, DXA, education ≥12 years, smoking (current), Charlson-Deyo Index, prednisone use, DXA, duration of RA, MDHAQ score (mean), DAS28 (mean), and duration of bisphosphonate therapy.

  • P < 0.01.

  • §

    P < 0.05.

  • Osteoporosis was not included in the multivariate analysis due to the small numbers.

  • #

    P < 0.0001.

Age, years    
 ≥70137 (45.4)130 (48.0)ReferentReferent
 <70165 (54.6)141 (52.0)1.11 (0.80–1.54)1.36 (0.86–2.23)
 Male274 (90.7)241 (88.9)ReferentReferent
 Female28 (9.3)30 (11.1)0.82 (0.48–1.41)0.74 (0.37–1.47)
 Nonwhite66 (22.0)34 (12.7)ReferentReferent
 White234 (78.0)234 (87.3)0.51 (0.33–0.81)0.52 (0.30–0.88)§
Education, years    
 ≥12172 (78.2)168 (85.7)ReferentReferent
 <1248 (21.8)28 (14.3)1.67 (1.00–2.79)§1.64 (0.92–2.93)
 No217 (73.6)196 (74.2)ReferentReferent
 Yes78 (26.4)68 (25.8)1.04 (0.71–1.51)1.20 (0.72–1.99)
Charlson-Deyo Index, mean ± SD2.18 ± 2.221.78 ± 1.891.01 (0.98–1.04)1.03 (0.99–1.08)
Prednisone use    
 No82 (28.2)83 (31.2)ReferentReferent
 Yes209 (71.8)183 (68.8)1.16 (0.80–1.66)1.25 (0.77–2.04)
 Never166 (55.1)146 (54.3)ReferentReferent
 Ever135 (44.9)123 (45.7)0.97 (0.69–1.34)1.03 (0.66–1.60)
 No98 (72.6)94 (76.4)Referent 
 Yes37 (27.4)29 (23.6)1.22 (0.70–2.15) 
Duration of RA, mean ± SD years20.6 ± 13.017.4 ± 12.01.02 (1.01–1.04)1.02 (1.00–1.04)§
MDHAQ score, mean ± SD1.07 ± 0.531.04 ± 0.441.12 (0.80–1.58)1.01 (0.63–1.63)
DAS28, mean ± SD3.84 ± 1.213.63 ± 1.131.17 (1.01–1.35)§1.15 (0.93–1.42)
Duration of bisphosphonate, months    
 <32126 (41.7)159 (58.7)ReferentReferent
 ≥32176 (58.3)112 (41.3)1.98 (1.42–2.77)#1.63 (1.04–2.57)§


Clinical studies show that use of oral bisphosphonates reduces the risk of osteoporotic fractures. However, adherence to therapy is required for improved outcomes and appears to represent a major treatment obstacle in US veteran patients with RA. In our VARA cohort, we demonstrated a lower adherence in nonwhite patients and patients with a longer duration of RA or in whom bisphosphonates were prescribed for a longer period of time.

Much of the literature pertaining to adherence with bisphosphonate therapy is from randomized clinical trials enrolling primarily women, which may not reflect clinical practice, or from studies utilizing prescription benefit claims, which lack clinical data. A major strength of the current study is the merging of electronic databases, i.e., clinical data from the VARA registry with the detailed dispensing data from the PBM. The mean MPR of 0.69 was comparable to that of other studies that included predominantly women and reported 43–70.5% of subjects to be adherent with oral bisphosphonates (8, 29, 30). We included all gaps in the duration of therapy when calculating the MPR; if we removed gaps of >90 days and calculated the MPR for each course, giving more weight to longer courses, the mean ± SD MPR would be 0.90 ± 0.13. We believe the method used for calculating the MPR in our analysis was more appropriate, since bisphosphonates are used for long-term prophylaxis against fractures; therefore, long-term adherence is more important.

Women were not more adherent than men in our cohort, but the relatively small number of women (10.1% of the cohort) limited the power of our calculations in examining sex-related differences. Nonwhites were less adherent in our study, an observation that was not found by others (9, 12); however, the predominance of whites in these studies could have obscured any racial differences. Economic factors are unlikely to have accounted for the racial differences in our study, given the small copayment (no more than $9 per month at the time of the study) for all veterans eligible for care in the VA Health Care System.

Poor bisphosphonate adherence is reported to be associated with older age, more comorbid conditions, more nonosteoporosis medications, and institutionalization in a nursing home (9). Our data set did not include nonosteoporosis or non-RA medications, but the burden of comorbidity measured by the Charlson-Deyo Index was not associated with adherence in either univariate or multivariate analyses.

A retrospective review of male veterans in Wisconsin found nonadherence with bisphosphonates to be associated with tobacco use and side effects related to the bisphosphonates, whereas men undergoing bone density testing were less likely to be nonadherent (13). This led us to postulate that patients with osteoporosis or DXA testing would be more adherent with bisphosphonate therapy. However, our data did not duplicate this finding. DXA was not systematically obtained prior to the initiation of bisphosphonate therapy, and the results of repeat DXA were not routinely entered in the VARA registry. This precluded analysis of the effect repeat DXA had on either duration of therapy or adherence.

Osteoporosis is typically an asymptomatic condition, and patients, particularly those with fewer years of education, as was seen in our cohort, may find it difficult to understand the risk of fracture and benefits of therapy. Other barriers to adherence with oral bisphosphonates include inconvenience and side effects, neither of which was included in our VARA data or analysis (12).

The mean duration of bisphosphonate therapy in VARA patients was 39.2 months, similar to the median duration of 3.27 years reported from a study of the Australian Department of VA data set (31). In our cohort, a duration of bisphosphonate therapy of more than 32 months was associated with older age, more years of education, and having a DXA scan performed. In contrast, a large study of 152,777 Danish patients with osteoporotic fractures reported that older subjects were more likely to discontinue or change treatments (10).

There are limitations to this retrospective cohort study. First, the subjects in the VARA cohort may not be representative of all patients with RA, since our patients are mostly men and older. Second, adherence may be overestimated in the current study, since we calculated the adherence based on medication dispensed to the patients and did not perform pill counts, although it may be logical to assume that patients are unlikely to refill medications they are not taking. Third, some patients may participate in dual care and may be receiving prescriptions from non-VA pharmacies, data that would not be captured in PBM. Another limitation of this study is the absence of fracture data, which could influence adherence; an analysis of 101,038 new bisphosphonate users found that the mean bisphosphonate adherence was 7% lower in the 6 months following a hip fracture compared to the 6 months preceding the hip fracture (29). Finally, our study did not measure side effects from bisphosphonates, which other studies have reported to be a major reason for poor adherence and persistence of bisphosphonate therapy (12).

Factors contributing to bisphosphonate nonadherence in patients with postmenopausal osteoporosis include lack of DXA testing, poor physical function, and older age, and may not apply to RA patients with secondary osteoporosis. Similar findings were demonstrated in our cohort, where nonwhite ethnicity, duration of RA, and duration of bisphosphonate therapy were the major risk factors for nonadherence; poor physical function and higher disease activity of RA did not explain bisphosphonate nonadherence.

In summary, in a cohort of veterans with RA, oral bisphosphonate adherence is subpar, and our findings provide baseline data to permit a targeted approach for improvement in fracture prevention interventions. Our findings provide useful baseline information for the development of initiatives for an improved risk management approach for the prevention of osteoporotic fractures in VA patients with RA.


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. Mr. Richards 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. Richards, Cannon, Kerr.

Acquisition of data. Richards, Cannon, Hayden, Amdur, Lazaro, Mikuls, Reimold, Caplan, Johnson, Schwab, Cherascu, Kerr.

Analysis and interpretation of data. Richards, Cannon, Amdur, Kerr.


The authors wish to thank Jeffrey Huang for his assistance in collating the data for analysis.