Article Text

Medications associated with fracture risk in patients with rheumatoid arthritis
  1. Gulsen Ozen1,
  2. Sofia Pedro2,
  3. Frederick Wolfe2,
  4. Kaleb Michaud1,2
  1. 1 University of Nebraska Medical Center, Omaha, Nebraska, USA
  2. 2 Forward, The National Databank for Rheumatic Diseases, Wichita, Kansas, USA
  1. Correspondence to Dr Kaleb Michaud, National Data Bank for Rheumatic Diseases, Wichita, KS 67214, USA; kmichaud{at}


Objective To examine the fracture risk with use of disease-modifying antirheumatic drugs (DMARDs), statins, proton pump inhibitors (PPIs), opioids, non-opioid analgesics and psychotropic medications in a US-wide observational rheumatoid arthritis (RA) cohort.

Methods Patients with RA without prior fracture from 2001 through 2017 in FORWARD, a longitudinal observational registry, were assessed for osteoporosis-related site fractures (vertebra, hip, forearm and humerus). DMARD exposure was assessed in four mutually exclusive groups: (1) methotrexate monotherapy-reference, (2) tumour necrosis factor-α inhibitors (TNFi), (3) non-TNFi biologics and (4) others. Non-DMARDs and glucocorticoids were classified as current/ever use and based on treatment duration. Fracture Risk Assessment Tool (FRAX) scores estimating for 10-year major osteoporotic fractures were calculated. Cox proportional hazard models stratified by FRAX were used to adjust for confounders.

Results During median (IQR) 3.0 (1.5–6.0) years of follow-up in 11 412 patients, 914 fractures were observed. The adjusted models showed a significant fracture risk increase with use of any dose glucocorticoids ≥3 months (HR (95% CI) for <7.5 mg/day 1.26 (1.07 to 1.48) and for ≥7.5 mg/day 1.57 (1.27 to 1.94)), opioids (for weak: 1.37 (1.18 to 1.59); strong: 1.53 (1.24 to 1.88)) and selective serotonin reuptake inhibitors (SSRIs) (1.37 (1.15 to 1.63)). Fracture risk with opioids increased within 1 month of use (1.66 (1.36 to 2.04)) and with SSRIs >3 months of use (1.25 (1.01 to 1.55)). Statins (0.77 (0.62 to 0.96)) and TNFi (0.72 (0.54 to 0.97)) were associated with reduction in vertebral fracture risk only. PPIs and other psychotropic medications were not associated with increased fracture risk.

Conclusion Use of opioids, SSRIs and glucocorticoids were associated with increased risk of any fracture in patients with RA, whereas statins and TNFi were associated with decreased vertebral fractures.

  • rheumatoid arthritis
  • fracture
  • opioids
  • antidepressants
  • statins
  • cohort study

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Key messages

What is already known about this subject?

  • Chronic inflammation predisposes patients with rheumatoid arthritis (RA) to fractures, cardiometabolic comorbidities and psychosocial problems which in turn lead to the use of several medications other than disease-modifying antirheumatic drugs such as opioids, non-steroidal anti-inflammatory drugs, antidepressants and statins.

  • In the general population opioids, antidepressants, psychotropic medications and proton pump inhibitors have been shown to increase fracture risk while statins have been found to decrease fracture risk. So far, only opioids have been investigated in RA and shown to increase fracture risk.

What does this study add?

  • This study demonstrates that fracture risk in patients with RA is also associated with the use of medications in RA such that opioids and selective serotonin reuptake inhibitors antidepressants increase all-site (vertebral and non-vertebral) fracture risk and statins and tumour necrosis factor-α inhibitors decrease vertebral fracture in patients with RA.

How might this impact on clinical practice or future developments?

  • Opioid use, which is still frequent in RA despite the presence of effective treatment strategies, should be minimised to avoid fracture risk and other devastating consequences of opioids. A regular review of medications in patients with RA may help deprescription of unnecessary and potentially high risk treatments.

  • Given the underuse of statins for cardiovascular disease in RA, the protective association against vertebral fractures may be encouraging for statin prescriptions in patients with RA.


Rheumatoid arthritis (RA) is associated with twofold increased risk of osteoporosis (OP) and fractures across all age groups compared with the general population.1 2 Osteoporotic fractures can lead to reduced quality of life, disability and increased mortality in patients with RA.3 The incident osteoporotic fractures and related costs are estimated to exceed 3 billion and US$25 billion per year, respectively, in the US general population by 2025.4 This growth in fracture burden is worrying in RA due to the already heightened fracture risk, increased life expectancy of patients with the advancements in RA management and suboptimal OP screening and treatment in RA.5

Chronic inflammation, physical inactivity and glucocorticoids mediating the increased OP and fracture risk also predispose patients with RA to multiple comorbidities including cardiometabolic, mood, sleep and gastrointestinal disorders, which in turn correlate with disability and mortality.6–12 To manage comorbidities and pain, many patients with RA require medications other than disease-modifying antirheumatic drugs (DMARDs) that might influence fracture development. In the general population, proton pump inhibitors (PPIs), selective serotonin reuptake inhibitors (SSRIs), benzodiazepines, antipsychotics, anticonvulsants, opioids and non-steroidal anti-inflammatory drugs (NSAIDs) have been reported to increase fracture risk13–17 while statins have been shown to decrease fracture risk.18 The concomitant use of these medications is quite common and more frequent in patients with RA19 20 where use of opioids reaches ~40%,19 statins 25%,20 antidepressants 30%,20 NSAIDs 45%–76%,21 and PPIs ~30%.22 Despite this, the data regarding these drugs’ influence on the fracture risk in RA are very limited and primarily focused on opioids. Studies using administrative data suggested an increased non-vertebral fracture risk with opioids in patients with RA.23–26 However, they did not include data regarding disease activity, disability, detailed DMARD treatment, body mass index (BMI), physical activity and smoking that can significantly affect fracture risk. There was also no information about patients’ baseline fracture risk estimated by WHO Fracture Risk Assessment Tool (FRAX).27 Furthermore, the fracture risk associated with statins, antidepressants, PPIs and new biological DMARDs (bDMARDs) remain unknown.

Given the important cost, disability and mortality consequences of fractures in RA, it is important to identify modifiable risk factors. Therefore, we performed a cohort study that aimed to assess the fracture risk in patients with RA with the use of statins, PPIs, opioids and non-opioid analgesics, psychotropic medications and DMARDs

Patients and methods

Patients were participants in FORWARD—The National Databank for Rheumatic Diseases, a longitudinal prospective observational study. Participants were primarily recruited from US rheumatologists and were followed with self-report semiannual questionnaires.28 Patients with RA ≥40 years old who did not have prevalent fracture and completed ≥2 semiannual questionnaires from January 2003 through December 2017 were included.

The primary outcome was incident fractures. By semiannual questionnaires, patients were asked about fractures of the finger, wrist, forearm, humerus, elbow, toe, foot, ankle, lower leg, knee, hip, pelvis, rib, skull and face, cervical, thoracic and lumbar vertebra that developed in the past 6 months. Among these, only the first of the OP-related site fractures, vertebral, hip, forearm and humerus, were included. The secondary outcomes were vertebral and non-vertebral fractures. Follow-up started at cohort entry and continued until the development of the first fracture or censoring at death, loss to follow-up or end of the study.

Fracture risk assessment

We calculated the 10-year probability of a major osteoporotic fracture (MOF) (hip, clinical spine, forearm or humerus) or hip fracture of patients with the US version of FRAX tool (V.4.0) ( without bone mineral density (BMD) results.27 29 This approach uses clinical risk factors (sex, race, age, BMI, previous fracture, parental history of hip fracture, current smoking, glucocorticoid ever use (≥5 mg/day prednisone >3 months), RA, other causes of secondary OP and alcohol consumption (≥3 units/day)) to estimate the 10-year fracture probability. The parental history of hip fracture was unavailable. Therefore, we replaced the missing data by simulations based on the conditional probability of the association of a risk factor with age, BMI and other dichotomous clinical variables by logistic regression as done previously.5 Using this method, we estimated 5.7% of the patients to have a parental history of hip fracture (online supplementary material). A FRAX score of ≤10% for a 10 year MOF was regarded as low-risk, 10%–20% medium-risk and ≥20% high-risk.

Medication exposure

Treatment exposure was measured at study enrolment and every 6 months by study questionnaires. Glucocorticoid treatment was examined in time-varying dose (prednisone equivalents) and duration-combined categories: not-using (reference), <7.5 mg/day for <3 months, <7.5 mg/day for ≥3 months, ≥7.5 mg/day for <3 months and ≥7.5 mg/day for ≥3 months. We assessed DMARD exposure in four mutually exclusive hierarchical groups: (1) methotrexate monotherapy (reference), (2) tumour necrosis factor-α inhibitors (TNFi) (infliximab, adalimumab, etanercept, golimumab and certolizumab), (3) non-TNFi bDMARD (abatacept, rituximab, tocilizumab, anakinra) and (4) other synthetic DMARDs (sDMARDs). Patients could contribute to different DMARD groups over time when they switched therapy.

Statins, antidepressants, PPIs, opioids, NSAIDs, anticonvulsants, antipsychotics and benzodiazepines were assessed as follows: (1) current use versus non-use, (2) ever-use versus never-use and (3) based on treatment duration with non-use-referent; ≤1 month; 1–3 months and >3 months of use. We categorised antidepressants as SSRIs and non-SSRI antidepressants, opioids as weak and strong opioids and NSAIDs as cyclooxygenase-2 (COX-2) inhibitors and non-selective NSAIDs. Weak opioids included hydrocodone, tramadol, codeine, pentazocine and propoxyphene. Strong opioids included hydromorphone, dihydromorphinone, oxymorphone, butorphanol, methadone, morphine, oxycodone, meperidine and fentanyl.


Variables potentially confounding the association between RA and non-RA related medications and fracture risk were adjusted for in multivariable models. Covariables were obtained from self-reported questionnaires. These included age (40–49 years-reference, 50–64 years,≥65 years), sex, location of residence (rural vs urban), education level (years), health insurance (any vs none), BMI in WHO categories (normal weight-reference), smoking history (ever vs never), physical activity, rheumatic disease comorbidity index,30 prior OP diagnosis, RA duration, patient-reported RA severity and activity measures (individual components of patient activity scale (PAS), Health Assessment Questionnaire (HAQ), pain and patient global scores (0–10)), use of OP medications (bisphosphonates, hormone replacement therapy and teriparatide), number of previous DMARDs, calendar year, all-cause hospitalisations in the past 6 months and annual influenza vaccination as proxy measures of healthcare utilisation. Also, a dichotomous variable representing depression and anxiety assessed by mental component summary (MCS) score of short form-36 (SF-36) (MCS ≤38—probable major depressive/generalised anxiety disorder) was included.31

Statistical analysis

Baseline characteristics of patients with RA by incident fractures and medication use were compared. Fracture crude incidence rates were calculated by dividing the number of events per 1000 patient-years of follow-up with 95% CI. For the assessment of fracture risk, we constructed multivariable Cox proportional hazards models to adjust for above-mentioned confounding factors. We stratified each model according to baseline FRAX 10-year MOF risk category. By using similar multivariable models, we also examined the influence of opioid strength, antidepressant and NSAID types, treatment durations and ever-use of the medications on the fracture risk. In sensitivity analyses, we restricted the analysis to patients with RA who never had an OP diagnosis or OP treatment.

All assessed covariables and medication exposures were time-dependent. Missing covariables were replaced by using multiple imputations by chained equations to create multiple imputed datasets for analyses.32 All tests were two-sided and considered statistically significant when p<0.05. All statistical analyses were performed using Stata V.14.0 (Stata, College Station, Texas, USA).


The study included 11 412 patients with RA with a baseline mean (SD) age of 61.4 (10.8) years and disease duration of 15.6 (12.7) years. Patients who developed fractures were older, more likely to be female, had longer disease duration, higher HAQ, disease activity by PAS, comorbidity and FRAX scores and more frequent glucocorticoid (higher doses and longer durations), antidepressant, PPI and opioid use at baseline than patients who did not (table 1). Baseline characteristics of patients with RA by medication use are shown in the online supplementary table S1.

Table 1

Baseline characteristics of patients with rheumatoid arthritis by incident fractures*

During 55 482 patient-years (median (IQR) 3.0 (1.5–6.0) years) of follow-up, 914 first-time fractures were observed, yielding a crude incidence rate (95% CI) of 16.5 (15.4–17.6) per 1000 patient-years. Fracture incidence rates were higher in patients receiving glucocorticoids, antidepressants, opioids, antipsychotics and benzodiazepines than patients receiving other medications (table 2).

Table 2

Crude incidence rates of fractures by medications in patients with rheumatoid arthritis

The fully adjusted models showed a significant fracture risk increase with weak (HR 1.37 (1.18 to 1.59)) and strong opioids (HR 1.53 (1.24 to 1.88)), SSRIs (HR 1.37 (1.15 to 1.63)) and glucocorticoid use of any dose for ≥3 months (HR for <7.5 mg/day 1.26 (1.07 to 1.48); for ≥7.5 mg/day 1.57 (1.27 to 1.94)). In comparison with methotrexate monotherapy, other sDMARDs were associated with decreased fracture risk (HR (95% CI) 0.82 (0.69 to 0.99)). Other medications and bDMARDs did not significantly alter the fracture risk (table 3). Similar to current use, ever-use of opioids (HR 1.45 (1.25 to 1.68)) and SSRIs (HR 1.18 (1.00 to 1.39)) were associated with increased fracture risk compared with never-use (table 4). The fracture risk increase with opioids started even after 1–30 days of use (HR 1.66 (1.36 to 2.04)), which was more prominent than longer-term use (HR for 1–3 months 1.46 (1.20 to 1.78); >3 months 1.35 (1.12 to 1.64)). As for the SSRIs, fracture risk increase started after 3 months of use (HR 1.25 (1.01 to 1.55)) (table 4). Ever-use of other medications and different durations with further assessment of more extended use (≥12 months) were not associated with fracture risk in patients with RA (data not shown).

Table 3

Association of medications with incident fractures in patients with rheumatoid arthritis

Table 4

Risk of incident fractures in patients with rheumatoid arthritis by opioid and SSRI treatment duration

In the analysis of patients who never had OP or OP treatment (240 incident fractures in 6899 patients), opioids (HR 1.26 (1.01 to 1.72)) and SSRIs (HR 1.86 (1.32 to 2.62)) were again associated with fracture risk increase. However, only strong opioids were associated with fracture risk (HR 1.95 (1.28 to 2.96)). Regarding treatment duration, fracture risk increase with opioids was observed only with short-term (≤1 month) use (HR 1.48 (1.00 to 2.20)), whereas the risk increase with SSRIs started after 1 month of use (HR 1.96 (1.29 to 2.98)).

When we analysed vertebral and non-vertebral fractures separately, we identified 438 vertebral (incidence rate 7.7 (7.0 to 8.4) per 1000 patient-years) and 569 non-vertebral (incidence rate 10.1 (9.3 to 10.9) per 1000 patient-years) first-time fractures. Both vertebral and non-vertebral fracture risks were associated with current use of opioids (table 5). However, opioid-associated risk increase was higher for vertebral fractures than non-vertebral fractures. Again, SSRIs were associated with vertebral fracture risk but not non-vertebral (table 5). Of note, statins (HR 0.77 (0.62 to 0.96)) and TNFi (HR 0.72 (0.54 to 0.97)) were associated with decreased vertebral fracture risk.

Table 5

Association of medications with vertebral and non-vertebral fractures in patients with rheumatoid arthritis*


Osteoporotic fractures are the third greatest cause of mortality in patients with RA, after pulmonary and cardiac disease, and the second cause of disability after depression.30 Thus, identifying the modifiable risk factors for fractures, particularly medications in RA, is important. Opioids and SSRIs were associated with ~1.5-fold fracture risk increase in patients with RA after stratification according to baseline fracture risk by FRAX. Strong opioids and early treatment (<30 days of use) posed a higher risk than weak opioids and long-term use. SSRI-associated risk increase started after 3 months of use. While glucocorticoid use of any dose for ≥3 months increased fracture risk, TNFi and statins were associated with vertebral fracture risk reduction.

Opioids have been linked with increased fracture risk in the general population.33 34 Only a few studies examined this in RA through administrative data, and they showed an increased non-vertebral fracture risk with current use of opioids (RR 1.8–4.9).23–26 Although the risk increase we observed with opioids was similar to previously reported general population results,34 it was lower than the risk found for patients with RA.23–25 However, the other studies differed in study population (inclusion of mostly elderly patients with osteoarthritis and RA (~10% RA)), comparators (NSAIDs), outcome (non-vertebral fractures) and lack of baseline fracture risk (FRAX) and clinical data that can influence fracture risk (disease activity, disability, DMARD treatment, BMI, physical activity and smoking).23–26 Our study also indicated that opioids increase the risk of vertebral fractures more than that of non-vertebral fractures in patients with RA. A similar trend has been demonstrated in the general population,34 but the only study showing vertebral and non-vertebral fracture risk increase with opioids in RA did not report the quantitative extent.23

The most commonly suggested mechanisms for increased fracture risk with opioids are fall risk increase caused by psychomotor effects, influence on bone metabolism through opioid-induced hypogonadism and direct interference with bone formation.35 Consistent with non-RA studies,33 34 the greatest fracture risk increase during the first month of opioid treatment suggests that the risk was mediated by increased fall risk rather than bone loss associated with long-term use. This is supported by the fracture risk only increasing with short-term use of opioids in patients who never had OP diagnosis.

Our study also showed that patients with RA receiving SSRIs were more likely to have fractures. The increased fracture risk with SSRIs has been previously reported in several studies in the general population.13 36 This risk increase with SSRIs has been attributed to increased fall risk and also decreased BMD even in patients without depression.37 38 Moreover, depression itself has been implicated as a risk factor for falls and fractures.38 To differentiate the disease and drug effect on fracture risk, we adjusted our models for mood changes measured by SF-36 MCS score. Although we found an increased fracture risk with depression (HR 1.25 (1.06 to 1.46), data not shown), SSRI use was still associated with fracture risk increase. The more prominent risk increase with long-term (>3 months) SSRI use could be caused by decreased BMD rather than an increased fall risk.

To our knowledge, the fracture risk with statins has not been examined in patients with RA. We found that statins were associated with decreased vertebral fracture risk with no effect on non-vertebral fractures. Many observational studies or meta-analyses evaluated the association of statins with OP and fracture risk in the general population.18 39 40 Most but not all concluded that statin use was associated with an improved BMD and reduced risk of fracture.18 39 40 The evidence suggests that statins have anabolic and antiosteoclastic effects by inducing proliferation, differentiation and protection of osteoblasts and reducing osteoclastogenesis.41 Moreover, in patients with RA, statins exert anti-inflammatory effects42 which might indirectly affect bone health. Thus, it is biologically plausible that the pleiotropic effects of statins may result in reduced fracture risk. It is also possible that the observed associations emerged due to confounding by indication, unmeasured or residual confounding such as BMI, other comorbidities or differences in health behaviours like healthy-adherer effect. Although our study included several covariables associated with fractures and healthcare utilisation, further research is warranted to better understand the effects of statins on OP and fractures in patients with RA.

Our study confirms the glucocorticoid-fracture association and also adds to the existing literature about fracture risk with DMARDs.10 23 26 43–45 We assessed TNFi and non-TNFi and vertebral and non-vertebral fractures separately, which had not previously been done comprehensively. We showed that TNFi was associated with a significant reduction in vertebral fracture risk compared with methotrexate monotherapy. As chronic inflammation is one of the main reasons for the increased risk of OP and fractures in RA, it is thought that effective anti-inflammatory medication might decrease bone loss and consequently fracture risk. However, consistent with our findings, previous studies mostly using administrative data could not demonstrate a reduction in non-vertebral fracture risk either with TNFi or all bDMARDs assessed together compared with non-bDMARDs.23 26 43–45 One study examining vertebral fracture risk with TNFi reported no statistical association even though a risk reduction was found in all site fractures with TNFi monotherapy.43 The discrepancy in the results can be explained by differences in included covariables related to OP and fracture risk, comparison group (non-users) and the TNFi exposure assessment (current or ever).43

Given that BMD improvement with TNFi has been shown in both vertebral and non-vertebral sites including hand and femur,46 it was surprising to find only vertebral fracture risk reduction with TNFi. The reason for this can be potential inclusion of non-vertebral traumatic fractures as we did not have enough data regarding the nature of the activity causing the fracture. Additionally, different pathophysiological and biomechanical mechanisms might be involved in vertebral and non-vertebral fracture aetiology.

Contrary to the general population studies, we did not find any association between fracture risk and PPI or other psychotropic use.15 16 This might be due to the low number of patients on other psychotropic medications. Although PPI use was common in our cohort, the use of bisphosphonates in this group was also higher than the overall cohort (35% vs 16%) which may explain the lack of association.

Our study has some limitations. As an observational cohort, the patients were not randomly assigned to the assessed medications. Despite the inclusion of several fracture-related covariables and a patient-reported disease activity measure in the analyses, there might be residual confounding bias due to the unmeasured factors. One important measure we lacked was fall risk, which might explain the association observed for opioids and SSRIs. However, our analyses included factors related to fall risk in RA such as patient-reported disease activity, disability and pain.47 Also, due to the lack of data about trauma level, some of the traumatic fractures might be misclassified as osteoporotic fractures. Despite that, our fracture incidence rates were comparable to the reported rates in a recent meta-analysis of 25 cohort studies in patients with RA (pooled incidence rates of fragility fractures 15.31 (95% CI 10.43 to 22.47) per 1000 person-years).48 Additionally, the majority of the vertebral fractures are not clinically recognisable at early stages, and the diagnosis of these fractures requires ≥2 radiographic assessments at appropriate intervals. As we did not have radiographic data, the incidence of vertebral fractures would be higher than our findings.

In conclusion, our study showed that opioids, SSRIs and glucocorticoids were associated with increased vertebral and non-vertebral fracture risk in patients with RA whereas statins and TNFi were associated with decreased vertebral fracture risk. The risk observed with opioids and SSRIs might be related to increased fall risk. In our view, OP, fracture and fall risks in RA and their severe consequences are underestimated. Given the frequent occurrence of chronic pain, mood disorders, polypharmacy and already heightened fracture and fall risks in patients with RA, we suggest a regular review of the necessity of the medications being used. In the opioid epidemic era, to avoid fractures and other devastating consequences, opioid use should be minimised for pain management in RA.


We thank Drs John Kanis and Helena Johansson for their help in FRAX risk calculations.


Supplementary materials

  • Lay summary

    Disclaimer : This is a summary of a scientific article written by a medical professional (“the Original Article”). The Summary is written to assist non medically trained readers to understand general points of the Original Article. It is supplied “as is” without any warranty. You should note that the Original Article (and Summary) may not be fully relevant nor accurate as medical science is constantly changing and errors can occur. It is therefore very important that readers not rely on the content in the Summary and consult their medical professionals for all aspects of their health care and only rely on the Summary if directed to do so by their medical professional. Please view our full Website Terms and Conditions.
    Copyright © 2019 BMJ Publishing Group Ltd & European League Against Rheumatism. Medical professionals may print copies for their and their patients and students non commercial use. Other individuals may print a single copy for their personal, non commercial use. For other uses please contact our Rights and Licensing Team.


  • Handling editor Josef S Smolen

  • Contributors GO, SP and KM had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: all authors. Analysis of data: GO and SP. Interpretation of data and drafting of the manuscript: all authors.

  • Funding This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient and public involvement statement This research was done without patient involvement. Patients were not invited to comment on the study design and were not consulted to develop patient relevant outcomes or interpret the results. Patients were not invited to contribute to the writing or editing of this document for readability or accuracy.

  • Patient consent for publication Not required.

  • Ethics approval This study was approved by Via Christi Hospitals Wichita, Inc. Institutional Review Board (IRB#: IRB00001674 FWA#: FWA00001005).

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement No data are available.

  • Author note The lead author (the manuscript’s guarantor) affirms that the manuscript is an honest, accurate and transparent account of the study being reported; that no important aspects of the study have been omitted and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.