Abstract
Objective The World Health Organization fracture risk assessment tool (FRAX) algorithm for risk prediction of major osteoporotic and hip fractures accounts for several risk factors, including rheumatoid arthritis (RA), since individuals with RA have an excess burden of fractures. FRAX has not been validated in population-based RA cohorts in the US. We aimed to determine the accuracy of FRAX predictions for individuals with RA in the US.
Methods This retrospective population-based cohort study included residents of Olmsted County, Minnesota, who were followed until death, migration, or last medical record review. Each patient with RA (1987 American College of Rheumatology criteria met in 1980-2007, age 40-89 years) was matched 1:1 on age and sex to an individual without RA from the same underlying population. Ten-year predictions for major osteoporotic and hip fractures were estimated using the FRAX tool. Fractures were ascertained through follow-up, truncated at 10 years. Standardized incidence ratios (SIRs) and 95% CI were calculated to compare observed and predicted fractures.
Results The study included 662 patients with RA and 658 non-RA comparators (66.8% vs 66.9% female and a mean age of 60.6 vs 60.5 years, respectively). Among patients with RA, 76 major osteoporotic fractures and 21 hip fractures were observed during follow-up (median follow-up: 9.0 years) compared to 67.0 predicted major osteoporotic fractures (SIR 1.13, 95% CI 0.91-1.42) and 23.3 predicted hip fractures (SIR 0.90, 95% CI 0.59-1.38). The observed and predicted major osteoporotic and hip fracture risks were similar for patients with RA and non-RA comparators.
Conclusion The FRAX tool is an accurate method for estimating major osteoporotic and hip fracture risk in patients with RA.
Rheumatoid arthritis (RA) is a chronic autoimmune inflammatory disease with a prevalence of 0.5% to 1% that primarily affects synovial joints, leading to cartilage and bone destruction, although systemic extraarticular manifestations are also common.1,2 Sociodemographics (peak incidence of RA at 50 years of age, female predominance), chronic autoimmune inflammation, and frequent use of glucocorticoids (GCs) contribute to an increased risk of bone loss and a high prevalence (up to 30%) of osteoporosis in patients with RA.3,4
In the United States, osteoporosis affects an estimated 10 million people in their 50s and older.5 Several epidemiological studies concluded that patients with RA have a 2-fold increased risk of developing osteoporosis compared to the general population and an associated increased risk of fragility (ie, low-impact) fractures, especially of the spine or hip.6-9 Of note, only 2 of these studies accounted for GC use.6,7 The risk of fragility fracture is higher in patients with a longer RA duration and impaired functional status.10 Fragility fractures in patients with RA have been associated with a higher risk of mortality,11 cardiovascular disease,12 and functional disability.13
Vertebral fractures account for almost half of the fragility fractures in patients with RA,14 with about 25% of patients experiencing at least 1 vertebral fracture compared to 16% of controls in a recent case-control study.15 In the general population, hip fractures are predicted to occur in 1 in 3 women and 1 in 12 men throughout their lifetime and are associated with substantial socioeconomic impact and increased mortality.16 In patients with RA, the risk of hip fractures is 2-fold higher compared to the general population.17 A previous retrospective observational study from a Spanish population showed that the incidence of hip fractures in patients with RA has increased in both sexes by 3.1% per year between 1995 and 2015. This study did not account for the use of GCs.18 The increase was thought to be at least partly related to the increased life expectancy of patients with RA. The substantial and growing burden of osteoporosis and fragility fractures, with their significant detrimental effect on health outcomes and economic costs, dictates an urgent need for accurate risk prediction and effective prevention of fractures in patients with RA.
The World Health Organization fracture risk assessment tool (FRAX) was introduced in 2008 to help predict the individual risk of major osteoporotic and hip fractures, and has been validated and extensively used in the general population.19 The FRAX algorithm accounts for RA and GC use as binary (yes/no) variables, but it does not account for other important RA disease-related characteristics (eg, serostatus, disease activity/severity).20 Despite these uncertainties about its applicability for the RA population, the FRAX tool is frequently used in patients with RA.21,22 Several studies evaluated FRAX performance in patients with RA from different populations in Europe and Asia.23-25 However, studies validating the performance of FRAX in patients with RA in the US are lacking.
This study aims to evaluate FRAX performance in a population-based cohort of patients with RA overall and by subgroups (age, sex, GC use). We hypothesized that FRAX performs similarly well in RA and non-RA cohorts.
METHODS
Study population. Our retrospective population-based cohort study was conducted using the Rochester Epidemiology Project (REP), a unique medical records linkage system that provides virtually complete ascertainment of all clinically recognized individuals with RA in Olmsted County, Minnesota, and adjacent counties.26
The REP records were used to identify an inception cohort of patients with RA who fulfilled the 1987 American College of Rheumatology (ACR) classification criteria for RA between January 1, 1980, and December 31, 2007. A comparison cohort consisted of individuals without RA from the same underlying population. Each patient with RA was matched 1:1 with a non-RA comparator based on age, sex, and race using an ethnic-specific version. The index date for both cohorts was the date that RA criteria were met for patients with RA or the corresponding incidence RA date for matched comparators.
To match FRAX limits, 2 cohorts were created for analysis. Cohort 1 included individuals whose age at the RA index date was between 40 and 89 years and whose weight was < 125 kg on the index date. The baseline date for cohort 1 subjects was the index date. FRAX scores for cohort 1 were calculated without bone mineral density (BMD) using information at the time of index. Cohort 2 included individuals who had a dual-energy x-ray absorptiometry (DXA) scan done at the Mayo Clinic within 1 year before the index date or at any time during follow-up. The date of the DXA scan closest to the index date was chosen as the baseline date. Cohort 2 included individuals whose age at the DXA scan was between 40 and 89 years and whose weight was < 125 kg at the time of the DXA scan. FRAX scores for cohort 2 were calculated with BMD using information at the time of the DXA scan.
Ascertainment of fractures and FRAX score. Fragility fractures (ie, all nonpathologic fractures occurring because of no more than moderate trauma) from the index date until death or last follow-up were identified through medical record review by trained nurse abstractors, as previously described.12 Major osteoporotic fractures included fragility fractures of the proximal femur (hip), thoracic/lumbar vertebrae (spine), distal forearm (wrist), or proximal humerus (shoulder). Data on fragility fractures before index date were collected from the medical records if fractures were at any of the following sites: vertebral, proximal femur, distal forearm, and proximal humerus. Both symptomatic and incidental fractures before index date were included.
Ten-year predictions for major osteoporotic and hip fractures were estimated using the US version of FRAX. FRAX is based on individual patient models that consider age, BMI (calculated as weight in kilograms divided by height in meters squared), and clinical binary risk factors (yes/no), including prior fragility fracture at the 4 sites as listed above, parental history of hip fracture, long-term use of oral GCs (eg, ≥ 3 months), presence of RA, secondary osteoporosis, current cigarette smoking, and high alcohol consumption (≥ 3 units/day).25,27 Information on clinical risk factors, such as BMI, GC use, smoking, alcohol consumption, and prior fragility fracture, was gathered through manual medical record review or electronic data retrieval from electronic medical records.25,27 BMD at the femoral neck was used to calculate the FRAX score for cohort 2 only. Data on the parental history of hip fracture were not available in the medical records. Antiosteoporotic medications (ie, alendronate, pamidronate, risedronate, and ibandronate) with continuous use for ≥ 30 days at baseline were collected in patients with RA. Information on antiosteoporotic medications was not available for individuals without RA.
Statistical analyses. Patients with and without RA were compared using Wilcoxon rank-sum tests for continuous variables and chi-square tests for categorical variables. The observed follow-up was truncated at 10 years after the baseline date. For patients with < 10 years of follow-up, the obtained fracture risk prediction using the FRAX tool was adjusted proportionately. Standardized incidence ratios (SIRs), which are the ratios of the observed fracture events to the predicted fracture risk, were calculated assuming the observed fracture events follow a Poisson distribution and the predicted rates are fixed.28 Calibration plots were obtained by first subdividing patients according to deciles of FRAX and then reducing the number of groups until a minimum of 3 events were included in each group. This resulted in quintiles for cohort 1 and quartiles for cohort 2. For each subgroup, the Kaplan-Meier method was used to obtain the observed rate of major osteoporotic fractures at 10 years, which was plotted against the mean FRAX for each subgroup.
GC use was not assessed in the comparators, and they were assumed to have had no exposure. Since no data were available on parental history of fractures in our study population, the primary analyses assumed a parental fracture rate of 0%. To account for this likely underestimation, we conducted a sensitivity analysis, assuming a 20% parental fracture rate for individuals with RA and without RA, which exceeds the ~11% reported in another cohort in our population.29 A sensitivity analysis was performed to account for a possible dose-dependent detrimental effect of GC use on the risk of fracture for patients with an average daily dose ≥ 7.5 mg/day as recommended in the ACR guidelines for prevention and treatment of GC-induced osteoporosis.30 Multipliers of 1.15 for major osteoporotic fracture and 1.20 for hip fracture were applied to the FRAX scores if the closest GC dose to index date was ≥ 6 mg/day, as GC doses were captured in categories (eg, 0-5, 6-10, 11-15 mg/day) and information on the average cumulative dose over time was not available. Analyses were performed using SAS version 9.4 (SAS Institute) and R version 4.1.2 (R Foundation for Statistical Computing). Our study was approved by the institutional review boards of the Mayo Clinic (17-002593) and the Olmsted Medical Center (017-omc-17). The need for informed consent was waived. Patients who declined the use of their medical records for research purposes were not included in the study, in accordance with Minnesota law. This manuscript follows the Strengthening the Reporting of Observational studies in Epidemiology (STROBE) reporting guidelines for observational studies.31
RESULTS
Among 815 Olmsted County residents age ≥ 18 years who met criteria for incident RA in 1980-2007 and an equal number of comparators, 662 patients with RA and 658 patients without RA fulfilled inclusion criteria for this study, as outlined in the methods.
Among the 662 patients with incident RA, mean age was 60.6 years, 66.8% were female, and 95.6% were White. The comparison cohort of 658 patients without RA was similar in age, sex, BMI, and prevalence of prior fractures (Table 1). The prevalence of the following risk factors was higher in the RA cohort than in the non-RA cohort: current smoking, prior alcoholism, GC use, and secondary osteoporosis. GC use in RA varied by age group (Table 1).
Baselinea characteristics of individuals with and without RA.
The 10-year probability of major osteoporotic and hip fractures estimated by FRAX was significantly higher in patients with RA than in the non-RA comparators (Table 1). Table 2 summarizes the SIR for observed vs predicted major osteoporotic and hip fractures calculated without BMD data in the RA and non-RA cohorts. In the RA cohort, during the median follow-up of 9.0 (IQR 5.5-10.0) years, 76 major osteoporotic fractures and 21 hip fractures were observed, compared to the predicted 67.0 major osteoporotic fractures (SIR 1.13, 95% CI 0.91-1.42) and 23.3 hip fractures (SIR 0.90, 95% CI 0.59-1.38). The Figure (left) depicts the calibration plot comparing the observed and predicted major osteoporotic fractures by quintiles of FRAX.
Observed vs predicted fracturesa after RA incidence/index date using FRAX without BMD (cohort 1).
Observed vs predicted risk of major osteoporotic fractures for patients with RA in cohort 1 (without BMD) and cohort 2 (patients with BMD). Intervals that cross the identify line (dashed line) demonstrate good calibration. BMD: bone mineral density; MOF: major osteoporotic fracture; RA: rheumatoid arthritis.
In the non-RA cohort, during the median follow-up of 9.7 (IQR 5.7-10.0) years, 46 major osteoporotic fractures and 10 hip fractures were observed, compared to the predicted 40.5 major osteoporotic fractures (SIR 1.14, 95% CI 0.85-1.52) and 10.1 hip fractures (SIR 0.99, 95% CI 0.53-1.84). Overall, the observed and predicted major osteoporotic and hip fracture risks calculated without BMD were similar for patients with and without RA, with no statistically significant differences among the subgroups by sex, age group, GC use, and positivity for rheumatoid factor and/or anticyclic citrullinated peptide antibodies (Table 2).
Patients with available BMD data. A total of 356 patients with RA and 264 non-RA comparators had available BMD data from DXA scans, which were assessed, on average, 6.7 years after the RA diagnosis and 8.0 years after the corresponding index date. The patients with RA were younger (P = 0.003) and more were male (P < 0.001) than the comparators, though the femoral neck t scores were similar between the 2 groups (Table 1).
Table 3 summarizes the SIR for observed vs predicted major osteoporotic fractures and hip fractures calculated with BMD data in RA and non-RA cohorts. In patients with RA, during the median follow-up of 6.8 (IQR 3.8-8.9) years, 43 major osteoporotic fractures and 13 hip fractures were observed, compared to the predicted 41.5 major osteoporotic fractures (SIR 1.04, 95% CI 0.77-1.40) and 12.8 hip fractures (SIR 1.01, 95% CI 0.59-1.74). The observed and predicted major osteoporotic and hip fracture risks calculated with BMD were similar for patients with RA in most subgroups except for the major osteoporotic fracture risk in the 40- to 59-year-old age group (SIR 1.65, 95% CI 1.03-2.66). The majority (73.8%) of patients in this 40–59-year-old age group were GC users. The predictions were accurate for patients with different RA duration. The Figure (right) depicts a calibration plot comparing the observed and predicted major osteoporotic fractures by quartiles of FRAX.
Observed vs predicted fracturesa using FRAX with BMD (Cohort 2) and sensitivity analysis accounting for history of parental hip fracturesb.
In non-RA comparators, during the median follow-up of 6.8 (IQR 4.4-9.7) years, 32 major osteoporotic fractures and 8 hip fractures were observed, compared to the predicted 18.4 major osteoporotic fractures (SIR 1.74, 95% CI 1.23-2.46) and 4.4 hip fractures (SIR 1.84, 95% CI 0.92-3.67). Findings were consistent across age and sex subgroups in the non-RA cohort, with the rate of observed major osteoporotic fractures being higher than predicted (Table 3).
Sensitivity analyses with predefined parental hip fracture risk. Results of the sensitivity analyses of major osteoporotic and hip fracture risk assuming a prevalence of parental hip fracture of 20% for patients with RA and non-RA comparators are shown in Table 3. The observed and predicted major osteoporotic and hip fracture risks were similar for patients with RA overall and by subgroup, except for the major osteoporotic fracture risk in GC nonusers (SIR 0.37, 95% CI 0.15-0.89). In non-RA comparators, the observed risk of major osteoporotic fractures was higher than predicted in both sexes and in patients aged ≥ 80 years. In this older age group, the observed risk of hip fracture also exceeded the predicted estimates (Table 3).
Sensitivity analyses accounting for GC dosing. The ACR-recommended multipliers were applied to 210 patients with RA in cohort 1 and 116 patients in cohort 2. Results were similar to the primary analysis overall and in each subgroup. For example, the SIR for cohort 1 overall was as follows: major osteoporotic fracture risk was 1.05 (95% CI 0.84-1.32) and hip fracture risk was 0.81 (95% CI 0.53-1.25). For GC users, risk for major osteoporotic fracture and for hip fracture was 0.90 (95% CI 0.65-1.24) and 0.60 (95% CI 0.32-1.12), respectively. For nonusers, the overall risk of osteoporotic fractures and hip fractures was 1.24 (95% CI 0.91-1.69) and 1.19 (95% CI 0.66-2.15), respectively. For cohort 2 overall, major osteoporotic fracture risk was 0.97 (95% CI 0.72-1.30) and hip fracture risk was 0.93 (95% CI 0.54-1.60). For GC users, major osteoporotic fracture risk was 1.16 (95% CI 0.84-1.59) and hip fracture risk was 1.01 (95% CI 0.56-1.82). For nonusers, risk of major osteoporotic fracture and hip fracture was 0.43 (95% CI 0.18-1.03) and 0.65 (95% CI 0.16-2.58), respectively.
DISCUSSION
Although the FRAX tool is frequently used in patients with RA, to our knowledge, no studies to date have assessed how well it predicts the likelihood of major osteoporotic and hip fractures in longitudinal population-based US cohorts. The findings of our study show the FRAX tool, with or without BMD, is an acceptable method for estimating the risk of major osteoporotic fractures and hip fractures in patients with RA. Among patients with RA, when applying FRAX without BMD, the predicted and observed fracture risks were similar across all the subgroups (eg, age, sex, GC use). When using FRAX with BMD in patients with RA, some underestimation of major osteoporotic fracture risk was observed only in patients aged 40-59 years, underscoring the disproportional excess fracture risk in this subgroup. Since GCs are known to decrease trabecular bone density more than cortical bone,32 this will not be detected by the isolated measurement of femur BMD used in the FRAX score. Although GCs increase the risk of osteoporosis and two-thirds of patients aged 40-59 years were GC users, the lack of change in risk estimates after adjusting for GC dosing makes it unlikely that higher than expected major osteoporotic fracture risk in this group was a result of GC use. On the other hand, this excess risk mirrors the larger gap in prevalence of multimorbidity in younger RA patient subgroups (< 50 years) vs older RA patient subgroups (≥ 50 years) that has been shown in our previous studies.33 The disproportional excess fracture risk in young and middle-aged patients may also reflect the impact of RA disease severity34 and impaired functional status,35,36 which are not accounted for in the FRAX algorithm and distinguish individuals with RA from their non-RA counterparts. This is assuming a parental fracture risk of 20% resolved this discrepancy in major osteoporotic fracture risk prediction. Concordantly, prior studies have shown that the risk of osteoporotic fractures is significantly influenced by a family history of osteoporotic fractures.37 Taken together, these findings suggest that familial genetic risk of osteoporosis and fragility fractures is of the utmost importance for accurate fracture risk prediction. This highlights the need for raising clinicians’ awareness of the possible underestimation of major osteoporotic fracture risk in patients with RA who are 40-59 years old, especially if parental hip fracture history is unknown. Interestingly, FRAX underestimation of major osteoporotic fractures was also found in systemic lupus erythematosus (SLE). This was reported in a recent retrospective cohort study that included 229 Chinese patients with SLE, mostly middle-aged women, and used DXA testing with BMD as a baseline date. This underestimation was despite the proper adjustment for the daily dose of GCs.
Whereas underestimation of fracture risk using the FRAX tool was previously reported in patients with longer RA disease duration (ie, ≥ 10 years),35 FRAX predictions with BMD data were accurate in our study for patients with various RA disease durations, including those with RA duration ≥ 10 years. In patients with RA, the risk of major osteoporotic fracture when using FRAX with BMD tended to be somewhat overestimated in GC nonusers, possibly due to more optimally controlled RA in patients taking disease-modifying antirheumatic drugs, which reduce inflammation and slow the rate of bone loss in patients with RA.38 Several previous publications examined FRAX performance in European and Asian populations of patients with RA and produced heterogenous results, showing overestimation of the fracture risk in RA in some studies and superior performance of FRAX with BMD data over FRAX without BMD data.23-25 These findings appear discrepant with our results showing that FRAX, and particularly FRAX without BMD, performed generally well in patients with RA. The reasons for these differences are unclear and may be at least in part explained by differences in population characteristics, such as racial and/or ethnic differences, as well as variations in lifestyle risk factors, RA disease severity, and RA treatments that affect BMD. Assessment of the effect of disease activity metrics and biologics use is out of scope for this study but will be an important question to address in future studies. Taken together, these findings underscore the importance of the validation of FRAX in different populations of patients with RA.
The risk of major osteoporotic fracture using the FRAX tool with BMD in participants without RA in our study was underestimated in both sexes, which appears to be driven by inaccurate prediction for the ≥ 80 years-old age category. This discrepancy may be because of selection bias, as individuals without RA who have an indication for a DXA scan are likely high-risk due to various underlying causes for osteoporosis. When the prevalence of parental hip fracture was assumed to be 20% instead of 0%, the differences in observed and predicted risks of major osteoporotic fracture attenuated, except for in the ≥ 80 group.
For patients aged ≥ 80 years, a higher prevalence of comorbid conditions and possibly unfavorable sociodemographics that are associated with increased absolute fracture risk and are not accounted for by FRAX might be contributing to the discrepancies in predicted fracture risk in the oldest age group.
Indeed, FRAX has previously shown a relatively low accuracy in predicting hip fracture in the oldest age group (age ≥ 80 years) in White and Asian populations39,40 and major osteoporotic fracture in Asian patients with RA.24 Further, FRAX has been shown to overestimate the risk of major osteoporotic fracture in almost all postmenopausal women between the ages of 50 to 79 years with different racial backgrounds, including White women. The greatest overestimation was among Asian women, followed by Black women. FRAX without BMD also overestimated the 10-year hip fracture risk in all racial groups except American Indian.41 In our study, apart from the underestimation of the risk of major osteoporotic fractures in the ≥ 80 years-old age group, FRAX was accurate when predicting major osteoporotic and hip fracture risk in a population of predominantly White patients.
In our study, the observed and predicted risks were not significantly different for hip fracture, although there were only 8 hip fractures among patients without RA in cohort 2 overall and only 2 in the ≥ 80 years-old age group. The number of major osteoporotic fractures in individuals without RA who were ≥ 80 years was also small (n = 8); thus, these results should be interpreted with extreme caution.
Our study has several strengths, including a well-characterized, population-based cohort of patients with incident RA and a cohort of matched comparators from the same underlying population; all individuals were ascertained using the REP resources. The study takes advantage of the long and complete follow-up for both cohorts and the availability of complete (inpatient and outpatient) medical record data from all medical care providers in the community, allowing for comprehensive data collection and manual verification of the records on fragility fractures.
Our study has several potential limitations. First, as with any retrospective study, only information available in the medical records was included. However, the complete and comprehensive resources of REP likely minimize this shortcoming. Second, BMD data were not available for all patients, which was addressed in a separate analysis of patients with available BMD (cohort 2). Third, data on the family history of parental hip fractures was not available. However, the sensitivity analysis assuming a 20% prevalence of parental fracture in RA and non-RA subsets produced results consistent with the main analyses. Fourth, we were not able to use the cutoff of > 7.5 mg for GC dosing for the use of multipliers to the FRAX score as recommended by the ACR guideline.30 We used a cutoff of ≥ 6 mg instead, and the results were similar to the primary analysis overall and by subgroup.30 Individuals without RA in this study were presumed to have had no exposure to GCs, although an estimated 1.2% of Americans aged > 20 years use GCs.42 As this percentage is small, it is unlikely to be statistically relevant in our controls and unlikely to affect the results. The management of RA has changed significantly over the recent decades. Examining the effect of antirheumatic and antiosteoporotic medication changes on FRAX performance was out of the scope of our study. Finally, the Olmsted County population is predominately White; therefore, our results may not be generalizable to more ethnically diverse populations, as the FRAX tool may disadvantage racial and ethnic minorities by yielding inaccurate risk predictions in these groups.43,44
In conclusion, the FRAX tool is an acceptable method for estimating the risk of major osteoporotic and hip fractures in patients with RA. Increased vigilance is needed when using FRAX with BMD in younger patients with RA (40-59 years old) with unknown parental hip fracture history, as the risk may be underestimated. More studies on the use of FRAX are needed in ethnically diverse populations.
Footnotes
This work was supported by grants from the National Institutes of Health, National Institute on Aging (R01 AG068192, R01 AG034676, K24 AG078179) and National Institute of Arthritis and Musculoskeletal and Skin Diseases (R01 AR46849). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The funders had no role in the study design, collection, analysis, or interpretation of data, or writing or submitting of the manuscript.
JMD has received research grants from Pfizer and royalties from Ghirihlet, and has a patent pending (provisional US patent application no. 63/243,933) entitled, “Methods and Materials for Assessing and Treating Arthritis.” The remaining authors declare no conflicts of interest relevant to this article.
- Accepted for publication June 12, 2023.
- Copyright © 2023 by the Journal of Rheumatology
