Introduction
Rheumatoid arthritis (RA) is the most common form of inflammatory arthritis in adults and is characterized by chronic, progressive, systemic inflammation leading to substantial pain, disability, and other morbidities [
1‐
3]. It is well accepted that patients with RA are at an increased risk of osteoporosis and osteoporotic fracture [
4‐
11], even though a previous US population-based study based on the data from the Third National Health and Nutrition Examination Survey (1988 to 1994) did not find a difference in femoral neck bone mineral density (BMD) between RA and non-RA patients [
12].
Osteoporosis, particularly in patients with RA, is a multifactorial condition. Some studies have suggested the association between osteoporosis and proinflammatory cytokines such as TNF-α, IL-1 and IL-6, as these cytokines play an important role in bone resorption [
13‐
19]. Positive correlations between osteoporosis and C-reactive protein (CRP), a marker of active inflammation, have been observed, although not always confirmed, in a number of epidemiologic studies [
16,
17,
20‐
22]. Other known risk factors for osteoporosis include older age, female sex, menopause, lower body mass index, glucocorticoids use, high RA disease activity, long RA disease duration, and decreased physical activity [
8,
9,
19,
22‐
27]. Osteoporotic fracture, particularly at the hips, is associated with the risk of falling [
28,
29]. Fall-related risk factors such as impaired heel-toe walking and inability to do stand-ups without arm use [
30] were more common in patients with RA than non-RA patients [
24], probably related to impaired balance and poor lower limb muscle strength. In addition, RA patients have chronic polyarticular pain, which increases a risk of falls [
31]. A high prevalence of osteoporosis is observed as 50% of 925 female RA patients in a large Italian multicenter cross-sectional study had osteoporosis defined as BMD T-score lower than -2.5 in at least one region of measurement, although it might have been overestimated due to referral bias [
23].
Although an increased risk of osteoporosis in RA patients is well-reported, little information is available with regard to the population-based frequency of incident osteoporotic fractures in RA patients and their risks relative to different age groups, sex, anatomic site, and glucocorticoid use. We studied a very large cohort from a health care utilization database: to estimate the incidence rate (IR) of typical osteoporotic fractures among RA patients relative to age, sex and anatomic site; to assess the risk of typical osteoporotic fractures among RA compared with non-RA patients; and to evaluate the effects of rheumatoid factor (RF) and acute phase reactants on fracture risk among patients with RA.
Materials and methods
Data source
We conducted a cohort study using the administrative claims data in the HealthCore Integrated Research Database (HIRD) for the period 1 January, 2001 to 30 June, 2008. This database contained longitudinal claims information including medical diagnoses, procedures, hospitalizations, physician visits, and pharmacy dispensings on more than 28 million fully-insured subscribers, with medical and pharmacy coverage, to 14 Blue Cross/Blue Shield health plans across the USA. Results for outpatient laboratory tests, including CRP, erythrocyte sedimentation rate (ESR), and RF were available on a subset of beneficiaries. No information was available on the results of BMD tests or on other radiologic procedures.
Personal identifiers were removed from the dataset before the analysis to protect subject confidentiality. Patient informed consent was, therefore, not required. This study was approved by Brigham and Women's Hospital's Institutional Review Board and Data Use Agreements were in place with HealthCore, Inc.
Study cohort
Adults aged 18 years or older with at least two visits for RA identified with the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD 9-CM) code, 714.xx, were eligible for this study. Subjects who did not have a diagnosis of RA at any time during the entire study period were eligible to be part of the non-RA cohort. From this eligible non-RA cohort, five patients, matched on age and sex were selected for every subject with RA. The follow-up period began at the index date, defined as the date of second RA diagnosis for RA patients and the date of the first medical claim for the non-RA patients. We further required all subjects to have at least 12 months of continuous health plan eligibility before the start of follow-up. Subjects were then followed until occurrence of outcomes, loss of eligibility, end of study database, or death.
Outcome definitions
The definitions of fracture were based on diagnoses and procedure codes contained within the study database [see Additional file
1]. We included hip, wrist, humerus, and pelvis fractures, because these are considered to be typical sites of osteoporotic fracture and can be accurately defined in administrative claims databases [
32]. A composite of fractures at these four sites ('any osteoporotic fracture') was also considered. Patients were censored at their first fracture in the 'any osteoporotic fracture' analysis.
Covariates
Variables potentially related to a future fracture were assessed using the data from the 12 months prior to the index date. These variables included demographic factors (age and sex), osteoporosis-related factors (osteoporosis diagnosis, osteoporosis medications, prior fracture, BMD test, Parkinson's disease, Alzheimer's disease, prior falls, and other comorbidities), use of other medications likely to be associated with bone metabolism or fall risks (oral glucocorticoids, anticonvulsants, benzodiazepines, selective serotonin reuptake inhibitors (SSRIs), beta blockers, proton pump inhibitors, and opioids), and health care utilization factors (number of physician visits, acute care hospitalizations, and number of different medications). To quantify patients' comorbidities, we calculated the Deyo-adapted Charlson comorbidity index based on
ICD-9-CM [
33,
34]. The comorbidity index is a summative score, based on 19 major medical conditions including myocardial infarction, pulmonary disease, renal disease, hepatic disease, diabetes, cancer, HIV infection, etc. A score of 0 represents absence of comorbidity and a higher score indicates a greater number of comorbid conditions.
To explore a potential association between the disease status and fracture risk in patients with RA, we calculated the Claims-based Index for RA Severity (CIRAS) scores [
35], based on age, sex, number of tests for inflammatory markers, number of chemistry panels and platelet counts ordered, RF, the number of rehabilitation and rheumatology visits, and Felty's syndrome diagnosis, and examined outpatient laboratory data such as acute phase reactants (i.e., ESR or CRP) and RF levels in a subgroup of the RA cohort.
Statistical analyses
We compared the baseline characteristics between the RA and non-RA cohorts. Fracture IRs with 95% confidence intervals (CI) were calculated for all patients, and then stratified by age and sex. Rate ratios (RRs) were estimated by dividing the IR among RA patients by the IR among non-RA [
36]. Similar analyses were carried out for specific anatomic site fractures, and then stratified by baseline oral glucocorticoid use. Finally, to adjust for potential confounders, separate Cox proportional hazard models were used to compare the risks for any fracture and fracture at each site among RA patients with those in non-RA patients. Additional Cox proportional hazard models focused on the risks relative to age and sex. Finally, we conducted subgroup analyses to examine whether positive RF and elevated acute phase reactants, either ESR or CRP, increased a risk of fracture in RA patients. All analyses were performed using SAS 9.1 Statistical Software (SAS Institute Inc., Cary, NC, USA).
Discussion
Osteoporotic fractures are more common in patients with RA than the general population. We studied the IRs of osteoporotic fracture of the humerus, wrist, hip, and pelvis in a cohort of 47,034 RA patients using claims data. The IRs of fracture were 1.5 times higher among patients with RA compared with non-RA patients regardless of age, sex, or anatomic sites. This association between RA and fracture weakened after multivariable adjustment for known risk factors of osteoporosis including oral glucocorticoid use.
Chronic inflammation has been recently recognized as a potential risk factor for osteoporosis and fracture [
14]. The Health Aging and Body Composition Study showed that elevated inflammatory markers, such as IL-2, IL-6, CRP, and TNF-α, were associated with osteoporotic fracture [
18]. In a study of 74 post-menopausal women with RA, high disease activity, measured by high CRP levels and ESR, and an elevated IL-6 were associated with an increase in periarticular as well as systemic bone resorption [
19]. We performed subgroup analyses adjusted for either RF or acute phase reactants, in which the activity of RA or inflammation can be better, but not completely, taken into consideration. Both RF positivity and elevated acute phase reactant levels increased the risk of fracture but the association was not statistically significant. The association between oral glucocorticoid use and fracture risk remained increased among the subgroup of RA patients in whom these laboratory data were available. Nonetheless, our subgroup analyses might not have sufficient power to assess the independent association between these potential risk factors (i.e., RF, acute phase reactants, oral glucocorticoid use, and other medications) and fracture risks.
The recently developed World Health Organization fracture risk assessment tool, FRAX
®, computes the 10-year probability of hip fracture or a major osteoporotic fracture, based on individual patient models that integrate the risks associated with BMD at the femoral neck as well as clinical risk factors such as age, sex, smoking status, use of glucocorticoids, history of osteoporosis and prior fall, and RA [
37]. Fracture risk related to RA independent of BMD (RR 1.73, 95% CI 0.94 to 3.20) incorporated in this tool is higher than our results (HR 1.26, 95% CI 1.15 to 1.38) [
38]. This difference is probably associated with baseline characteristics of study population and degree of confounding adjustment.
Several strengths of this study are worth noting. We examined a very large cohort of RA and non-RA patients in a population that is representative of the US commercially-insured population. Our multivariable Cox models were simultaneously adjusted for more than 20 risk factors of osteoporotic fracture. Various subgroup analyses enabled us to provide specific risks relative to age, sex, anatomic site, and laboratory results. Our results are consistent with a population-based study of 30,262 RA patients in the UK [
9]. The authors noted that patients with RA had an increased risk of fracture compared with the non-RA patients (RR 1.5, 95% CI 1.4 to 1.6). Similar to our findings, the RR of fracture was highest for hip fracture and lowest for wrist.
There are, however, limitations to our study. First, this cohort study is likely to be subject to residual confounding by race, body mass index, calcium and vitamin D intake, frailty, and other unmeasured risk factors. Although we assessed variables potentially related to a future fracture using the data from the 12 months prior to the index date, this time period might not be long enough to capture all the information on potential confounders. We used both the comorbidity index and CIRAS scores to minimize the effect of such confounders. The comorbidity index has been widely used to measure comorbidity in various medical fields since its development [
33,
39‐
41]. Previous research showed moderate correlations between the CIRAS and a previously validated medical records-based index of severity [
35]. The substantial change in point estimates after multivariable adjustment indicates that further improved adjustment may explain our findings. We also conducted additional analyses on a subgroup of RA patients, in whom laboratory data were available, to assess whether the severity of RA affects the risk of fracture, and observed an increased risk associated with positive RF and elevated acute phase reactants, although it was not statistically significant. The analyses of laboratory test results need to be interpreted with caution as ordering laboratory tests in clinical practice is not a random process but often related to the disease status. Second, there could be misclassification with the diagnoses of RA and osteoporotic fractures as we mainly relied on diagnosis and procedure codes to identify them. Both the
ICD codes for RA and the
ICD codes and/or procedure codes for fractures have been used in a number of studies [
32,
42‐
45]. Third, we relied on prescription dispensing records in the database to determine patients' drug exposures including oral glucocorticoids. It may not be the most accurate way to verify individuals' daily drug exposures, but it is still considered as one of the best ways to ascertain drug exposure status in non-experimental settings [
46].
Finally, as true in most epidemiologic studies, patients were not randomly exposed to drugs in our study. Therefore, we cannot exclude the possibility of confounding by indication with regard to the effect of glucocorticoids on fracture risk in patients with RA. The detrimental effect of glucocorticoids on fracture might have been confounded if the drug was selectively given to patients with a higher degree of systemic inflammation and RA severity, and if the degree of systemic inflammation and the severity of RA correlated with the risk of fracture. Several studies recently reported a potentially beneficial effect of low-dose, short-term systemic glucocorticoids on BMD in RA due to its anti-inflammatory effects [
25,
47‐
49]. Although this study was not designed to address a potential role of disease-modifying anti-rheumatic drugs including TNF-α inhibitors in the risk of osteoporotic fracture among RA patients, there is some evidence suggesting a beneficial effect of such drugs on bone loss [
15,
48,
50]. The exact effects of glucocorticoids or disease-modifying anti-rheumatic drugs on the risk of fractures in RA should be further studied.
Acknowledgements
This study was supported by the National Institutes of Health K24 (AR055989) grant. Dr Kim is supported by the National Institutes of Health (T32 AR 055885) grant. Dr Schneeweiss is principal investigator of the Brigham and Women's Hospital DEcIDE Center on Comparative Effectiveness Research, funded by the Agency for Healthcare Research and Quality, and of the Harvard-Brigham Drug Safety and Risk Management Research Contract, funded by the US Food and Drug Administration. Dr Schneeweiss is a paid member of scientific advisory boards for HealthCore and ii4sm and has received consulting fees from WHISCON, RTI Health Solutions, the Lewin Group, and HealthCore. S Schneeweiss received research grants from Pfizer, Inc. Dr Solomon is supported by the National Institutes of Health (K24 AR055989, P60 AR047782, R21 DE018750, and R01 AR056215) grants. Dr Solomon has received research support from Abbott Immunology, Amgen, and BMS.
Competing interests
SY Kim, J Liu, GW Daniel, C-L Chang, and K Garneau declare that they have no competing interests. S Schneeweiss received research grants from Pfizer, Inc. DH Solomon received research grants from Abbot, BMS, Merck & Co., Inc. Novartis, and Amgen, Inc.
Authors' contributions
All authors participated in the study conception. SYK, SS, and DHS participated in the study design. SYK, SS, GWD, C-LC, KG, and DHS participated in data acquisition. SYS, SS, JL, and DHS participated in data analysis and interpretation. All authors participated in manuscript preparation and revision. All authors read and approved the final manuscript.