To determine the incidence rates and risk factors of cervical and trochanteric hip fractures (HFs) among patients with systemic lupus erythematosus (SLE) based on a nationwide population-based data set.
To determine the incidence rates and risk factors of cervical and trochanteric hip fractures (HFs) among patients with systemic lupus erythematosus (SLE) based on a nationwide population-based data set.
We conducted a cohort study using data from the Taiwan National Health Insurance database. Patients with SLE and their age- and sex-matched counterparts without SLE were identified. The primary end point was the first occurrence of HF. Cox proportional hazards model was used to evaluate the respective risk factors of cervical and trochanteric HFs in the lupus cohort.
Among 14,544 patients with SLE (90% women, mean age 38.1 years) with a mean followup of 6 years, 75 developed HF (incidence rate 8.60 per 10,000 person-years). Compared to controls, the incidence rate ratios (IRRs) for developing HF among lupus patients were 3.17 (95% confidence interval [95% CI] 1.92–5.39, P < 0.001) for cervical HF and 1.11 (95% CI 0.58–2.11, P = 0.571) for trochanteric HF. The IRRs for HF were 2.38 (95% CI 1.58–3.63, P < 0.001) for women and 1.06 (95% CI 0.21–4.93, P = 0.922) for men. Lupus patients with cervical HF were younger than controls with cervical HF (mean age 56.7 versus 67.8 years; P = 0.007). Multivariable Cox regression analyses showed that age, use of intravenous cyclophosphamide, higher dose of steroid, and stroke were associated with cervical HF, whereas age was the only associated factor for trochanteric HF.
SLE is associated with a higher risk for cervical but not trochanteric HF, and these 2 types of HFs have different risk factors.
Systemic lupus erythematosus (SLE) is a chronic autoimmune disorder that commonly affects the musculoskeletal system. Previous studies have reported a high prevalence of fragility fractures in different cohorts of patients with SLE ([1, 2]). However, none of these studies have focused on hip fractures, which represent the most devastating complication of osteoporosis.
Hip fractures can be classified into 2 distinct anatomic types: cervical fracture and trochanteric fracture. Several epidemiologic investigations suggest that these 2 types of fractures have different risk factors and different mechanisms ([3-6]). Studies have shown that women with trochanteric hip fractures have significantly lower bone mineral density (BMD) compared to those with cervical hip fractures ([4, 7]). Also, corticosteroid use is a stronger predictor for cervical hip fracture rather than trochanteric hip fracture (). Because patients with SLE tend to have reduced BMD and usually receive long-term corticosteroid therapy, the relationships between SLE and these 2 fracture types are complicated. Therefore, it is of great importance to identify the prevalence of each fracture type among lupus patients and to determine their associated risk factors, since treatment options and prognoses may differ.
Using a nationwide population-based data set from Taiwan, the aims of this study were to determine the incidence rates of cervical and trochanteric hip fractures among lupus patients, to calculate the incidence rate ratios (IRRs) of cervical and trochanteric hip fractures between lupus patients and their age- and sex-matched controls, and to perform a risk factor analysis of cervical and trochanteric hip fractures among lupus patients.
The Taiwan National Health Insurance (NHI) program, a mandatory single-payer health insurance system, provides coverage for approximately 98.3% of citizens (). The Taiwanese citizens are predominantly of Han Chinese descent, followed by aboriginal descent and a minority of foreign immigrants. The National Health Insurance Research Database (NHIRD), derived from the Taiwan NHI program and encrypted to protect patient confidentiality, has been provided to researchers since 1999 for research purposes. The International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes (2001 version) were used for disease coding in this system. Because the NHIRD consists of deidentified secondary data, the Institutional Review Board of Taipei Veterans General Hospital excused the study from review.
SLE was placed in the category of NHI-defined “catastrophic illness,” so those who were diagnosed with SLE were enrolled in the Registry of Catastrophic Illness. We conducted a retrospective study from January 1, 2000 to December 31, 2006 utilizing this registry data set. We included those who were diagnosed with lupus for at least 1 year, given the length of time needed for SLE and related treatments to be effective.
The exclusion criteria consisted of 1) patients with hip fractures (ICD-9-CM codes 820.xx) before enrollment, 2) a followup period of <1 year, 3) patients with pathologic fractures (ICD-9-CM codes 733.14 and 733.15), and 4) age <18 years.
For the comparison (i.e., control) cohort, we used the claims data of 1 million NHI beneficiaries randomly selected from 21,400,826 total enrollees throughout Taiwan in 2000. This control cohort was age and sex matched to the lupus cohort at a 1:1 ratio. The same exclusion criteria described above were applied to the matched cohort. The cohort entry started on January 1, 2000.
The primary end point was the first occurrence of a diagnosed hip fracture (i.e., newly diagnosed), as retrieved from the inpatient data set (ICD-9-CM codes 820.xx, but not subtrochanteric fracture [code 820.22 or 820.32]). Followup continued until the occurrence of a hip fracture, permanent disenrollment from the NHI, or the end of the study (December 31, 2006), whichever came first. Death within 1 year following the index hip fracture event was also recorded.
Fracture site–specific types of hip fractures, including cervical and trochanteric fractures, were stratified to further analyze their differences and clinical relevance. Menopause status was unknown in the NHI database. Therefore, a subgroup analysis was performed to test whether age ≥50 years increased the fracture risk, given that the mean age at menopause is ∼50 years in Taiwan (). Moreover, alternative age strata (<30, 30–39, 40–49, and ≥50 years) were employed in the female subgroup to determine the fracture risk in each age group and to identify whether the fracture risk increases significantly at a younger age, given that premature menopause is not uncommon among women with SLE ([11, 12]).
Age, sex, preexisting comorbidities, and medication use were recorded at the time of cohort entry (i.e., January 1, 2000). Comorbidities, including type 2 diabetes mellitus, hypertension, chronic heart failure, end-stage renal disease, chronic obstructive pulmonary disease (COPD), stroke, and/or thyroid disease, were ascertained based on the ICD-9-CM codes of the inpatient and outpatient data sets. Oral medications that were prescribed before enrollment and taken for at least 28 days were extracted for analysis and included hydroxychloroquine, azathioprine, cyclophosphamide, mycophenolate mofetil, cyclosporine, and warfarin. The use of intravenous cyclophosphamide prior to study entry was also recorded. Medications for osteoporosis, such as calcitonin, bisphosphonates, selective estrogen receptor modulators, and teriparatide, are covered by the NHI only for those who have osteoporosis (T score −2.5 or less) as diagnosed using dual x-ray absorptiometry (DXA) or those who have twice sustained osteoporotic fractures. Therefore, these medications were excluded from analyses because susceptibility bias rendered the interpretation of their effects difficult. To analyze the effect of steroids, we identified the cumulative doses and average daily doses of oral corticosteroids (prednisolone or equivalent), but the former was not included in the analysis in light of possible immortal time bias. Use of intravenous methylprednisolone pulse therapy within 180 days prior to the index hip fracture, or at the end of the time of followup, was also recorded to evaluate its impact on hip fractures.
Descriptive statistics of the patient characteristics are shown as the number (percentage) of cases and the mean ± SD. The significance of differences was evaluated using the independent t-test, the chi-square test, or Fisher's exact test, where appropriate. The incidence rate of fracture was calculated as the number of fractures per 10,000 person-years and was computed for age-, sex-, and fracture site–specific subgroups. The IRR and its 95% confidence interval (95% CI) were also calculated to compare the differences between the lupus and control cohorts. Hip fracture–free survival was plotted using the Kaplan-Meier method. Risk factors of cervical and trochanteric hip fractures in lupus patients were evaluated using univariate and multivariable Cox proportional hazards models, respectively, and summarized as the hazard ratio (HR) and 95% CI. P values less than 0.05 were considered to be statistically significant. The extraction and computation of data were performed using the Perl programming language, version 5.12.2. The SQL Server 2005 (Microsoft) was used for data linkage, processing, and sampling. All of the statistical calculations were performed using SPSS software for Windows, version 18.0.
A total of 17,989 patients with SLE were identified from the catastrophic illness database of the Taiwan NHIRD from 2000–2006. After excluding patients who had hip fractures before enrollment (n = 39), a followup of <1 year (n = 2,154), pathologic fractures (n = 164), and age <18 years (n = 1,088), a total of 14,544 adult patients with SLE were included for analyses. The characteristics of the lupus patients and the age- and sex-matched control cohort are shown in Table 1. The mean age of the study cohorts was 38.1 years; nearly one-fifth of the patients (18.9%) were age ≥50 years. Women comprised 90% of the study cohorts. The mean followup period for the lupus patients was 6.0 years, which was statistically different from the followup period for the controls (7.7 years; P < 0.001). Despite a shorter followup period, there were more hip fractures in the lupus cohort compared to the controls (75 versus 43; P = 0.003). Two-thirds (66.7%) of the hip fractures occurred at age ≥50 years in the lupus cohort and in the majority (88.4%) of the controls. The difference in mean age at hip fracture was significant between the lupus and control cohorts (59.6 versus 69.4 years; P = 0.002). Such significant differences in frequency and age were observed in cervical hip fracture, but not in trochanteric hip fracture.
|Variables||SLE patients (n = 14,544)||Control cohort (n = 14,544)||P|
|Age, mean ± SD years||38.1 ± 13.5||38.1 ± 13.5||1|
|Women, no. (%)||13,087 (90.0)||13,087 (90.0)||1|
|Followup, mean ± SD years||6.00 ± 2.24||7.67 ± 1.24||< 0.001|
|Hip fracture, no. (%)/F:M|
|Overall||75 (0.5)/71:4||43 (0.3)/38:5||0.003|
|Cervical fracture||57 (0.38)/53:4||22 (0.15)/19:3||< 0.001|
|Trochanteric fracture||18 (0.12)/18:0||21 (0.15)/19:2||0.647|
|Age at hip fracture, mean ± SD years|
|Overall||59.6 ± 16.0||69.4 ± 15.2||0.002|
|Cervical fracture||56.7 ± 15.3||67.8 ± 17.6||0.007|
|Trochanteric fracture||68.8 ± 15.2||71.0 ± 12.6||0.629|
|1-year mortality following index hip fracture, no. (%)||12 (16.0)||5 (11.6)||0.515|
The 1-year mortality rate following the index hip fracture event among the lupus patients was 16.0% (overall: 12 of 75, cervical: 9 of 57, trochanteric: 3 of 18), whereas for controls, the 1-year mortality rate following the index hip fracture event was 11.6% (overall: 5 of 43, cervical: 3 of 22, trochanteric: 2 of 21). The difference in mortality rates between the 2 cohorts, however, was not statistically significant (P = 0.515).
Kaplan-Meier curves for cervical and trochanteric hip fracture–free survival in the lupus cohort versus the matched cohort are shown in Figures 1A and B, respectively. Compared to the matched cohort, the incidence of hip fractures was significantly higher in the lupus cohort, with an IRR of 2.23 (95% CI 1.51–3.32, P < 0.001) (Table 2). As for the 2 main hip fracture types, the lupus cohort had a significantly higher incidence rate of cervical fractures, with an IRR of 3.17 (95% CI 1.92–5.39, P < 0.001). There were no significant differences in the incidences of trochanteric fracture between the 2 study cohorts. The incidence increased dramatically in the age ≥50 years group (<50 years: 3.47, ≥50 years: 33.04), whereas a higher IRR was observed in those ages <50 years (<50 years: 6.29, ≥50 years: 1.81). Similar differences were seen among women; specifically, the incidence of hip fracture and the incidence of cervical hip fracture were significantly higher in women with lupus. However, the small case number of male lupus patients with hip fractures rendered the subgroup analyses unreliable; therefore, the data are not shown.
|SLE IRa||Control IRa||IRR (95% CI)||SLE IRa||Control IRa||IRR (95% CI)|
|Overall||8.60||3.86||2.23 (1.51–3.32)b||9.00||3.78||2.38 (1.58–3.63)b|
|Cervical||6.53||2.06||3.17 (1.92–5.39)b||6.72||1.89||3.55 (2.07–6.35)b|
|Trochanteric||2.29||2.06||1.11 (0.58–2.11)||2.28||1.89||1.21 (0.60–2.43)|
|<50||3.47||0.55||6.29 (2.36–21.03)b||3.65||0.48||7.53 (2.59–29.86)b|
|≥50||33.04||18.23||1.81 (1.16–2.84)c||35.66||18.90||1.89 (1.19–3.03)c|
Risk factors for cervical and trochanteric hip fractures among patients with SLE are shown in Tables 3 and 4, respectively. Type 2 diabetes mellitus, hypertension, COPD, and stroke were all associated with both fracture types. The average oral dose of steroids, intravenous pulse steroid therapy within the past 180 days, ever use of intravenous cyclophosphamide at baseline, and end-stage renal disease were negatively associated with cervical hip fractures, while chronic heart failure was negatively associated with trochanteric hip fractures. In multivariable analyses, independent predictors of cervical hip fractures were older age, a higher average oral dose of steroids, a history of stroke, and a higher incidence of the use of intravenous cyclophosphamide. In contrast, age was the only significant risk factor for trochanteric hip fractures after adjustment.
|Variable||Univariate analysis||Multivariable analysis|
|HR (95% CI)||P||HR (95% CI)||P|
|Age||1.07 (1.05–1.09)||< 0.001||1.09 (1.07–1.10)||< 0.001|
|Diabetes mellitus||3.90 (2.02–7.53)||< 0.001||–||–|
|Hypertension||2.73 (1.60–4.65)||< 0.001||–||–|
|Chronic heart failure||2.62 (0.95–7.24)||0.064|
|End-stage renal disease||2.13 (1.01–4.49)||0.048||–||–|
|Stroke||5.49 (2.49–12.12)||< 0.001||2.29 (1.02–5.15)||0.045|
|Thyroid disease||1.16 (0.50–2.72)||0.725|
|Mycophenolate mofetil||0.05 (0–2.48 × 109)||0.811|
|Intravenous||2.63 (1.24–5.55)||0.011||3.35 (1.57–7.14)||0.002|
|Recent pulse therapya||1.88 (1.03–3.44)||0.040||–||–|
|Average oral doseb||1.06 (1.02–1.09)||0.002||1.08 (1.05–1.11)||< 0.001|
|Variables||Univariate analysis||Multivariable analysis|
|HR (95% CI)||P||HR (95% CI)||P|
|Age||1.14 (1.10–1.17)||< 0.001||1.15 (1.11–1.19)||< 0.001|
|Diabetes mellitus||4.76 (1.56–14.50)||0.006||–||–|
|Chronic heart failure||7.15 (2.06–24.78)||0.002||–||–|
|End-stage renal disease||0.77 (0.10–5.76)||0.796|
|COPD||7.42 (2.77–19.84)||< 0.001||–||–|
|Thyroid disease||2.89 (0.95–8.80)||0.062|
|Mycophenolate mofetil||0.05 (0–2.04 × 1017)||0.891|
|Cyclosporine||0.05 (0–3.90 × 107)||0.773|
|Recent pulse therapya||0.31 (0.04–2.31)||0.252|
|Average oral doseb||0.99 (0.93–1.06)||0.836|
Of note, in further analyses among different age strata, the risk of cervical hip fracture in women with lupus was elevated, starting at age 40 years (HR 3.25, 95% CI 1.02–10.37; P = 0.046) (Table 5). In contrast, the risk of cervical hip fracture in the female controls did not become apparent until age 50 years. Also, the risk of trochanteric hip fracture in both cohorts was also increased at age ≥50 years (data not shown).
|SLE, HR (95% CI)||Controls, HR (95% CI)|
|Age strata, years|
|40–49||3.25 (1.02–10.37)a||1.30 (0.08–20.83)|
|≥50||14.64 (5.18–41.40)b||28.98 (3.86–217.74)a|
To our knowledge, our study is the first to describe the incidence rates of hip fractures among lupus patients. We found a 2-fold increase in hip fracture occurrence in our lupus cohort of 14,544 patients compared to the age- and sex-matched control cohort during a followup period of at least 6 years. Remarkably, the incidence rate of cervical, but not trochanteric, hip fracture was significantly higher in the lupus cohort than in controls. Moreover, our findings suggest that although the incidence of hip fractures increased with age, the lupus patients sustained cervical hip fractures at a younger age than controls. Only women with SLE carried a significantly higher hip fracture risk than controls.
The following characteristics strengthen the validity of this study. First, the Taiwan NHI database utilized in this study offered a large sample size, which reduced the selection bias substantially and allowed us to calculate the incidence of hip fracture (including the 2 main subtypes) among lupus patients. Second, only patients with a catastrophic illness certificate for SLE were included. The NHI covers the treatment costs incurred by this disease to alleviate these patients' financial burdens. A strict verification process is mandated for the catastrophic illness certificate application, including a clinical assessment by a review committee. These features make the SLE diagnosis in our study complete and reliable.
In line with studies of fragility fractures among patients with SLE ([1, 2, 13]), the incidence of hip fractures increased with age among our lupus cohort, with the majority occurring at age ≥50 years. Given that 90% of our lupus cohort was female, menopause may be a likely contributor to hip fracture risk, in addition to the normal effects of aging. This hypothesis was based on the observation that postmenopausal women had a higher incidence of hip fracture ([14, 15]) and that, in a population-based study, the risk of hip fractures increased incrementally in postmenopausal women after cessation of hormone therapy (). Intriguingly, we found that the risk of cervical hip fracture became apparent at age ≥40 years in a subgroup analysis of women with SLE. This observation was multifactorial, but may be partially attributed to a higher tendency of premature menopause in lupus patients ([11, 12]). In contrast, Yee et al reported that age was a predictor of fragility fracture among patients with SLE and that menopause was associated with osteoporosis, but not fracture (). Both types of hip fractures increased with age, but the cervical fracture group was younger than the trochanteric fracture group in our study. This finding was consistent with previous studies performed on the general population that concluded that aging might presumably predispose a person to trochanteric fracture ([6, 7, 15]).
In terms of the sex differences in our study, compared to controls, men with SLE had a similar incidence of hip fracture, while women with SLE carried a significantly higher risk. Such differences might arise from a higher prevalence of osteoporosis, a smaller peak bone mass, and greater cortical bone loss during aging in women than in men ([16-19]). Previous epidemiologic studies also showed sex differences in the incidence rate of hip fractures ([3, 20]). Nonetheless, the association between sex and hip fracture was not evident in our Cox regression analyses, which might be due to the few cases of male lupus patients with hip fractures in our study.
The incidence rate of cervical hip fractures among our lupus cohort was significantly higher than that of controls, but the incidence rates of trochanteric hip fractures were similar. This finding confirms previous observations that cervical and trochanteric hip fractures have a different pathogenesis ([3-6]). Alterations in bone geometry and focal bone loss of the femoral neck play a role in the pathogenesis of cervical hip fractures ([4, 5]), whereas age and generalized bone loss may be stronger predictors of trochanteric hip fractures (). Alele et al reported significant differences in bone geometry at all hip subregions between women with SLE and controls in addition to lower hip BMD in women with SLE (). Collectively, these findings suggest that SLE can reduce BMD as well as bring about changes in bone geometry that may lead to increased fragility, thus rendering these patients more susceptible to cervical hip fractures. Further research is warranted to prove this hypothesis.
The average oral dose of corticosteroids was unfavorably associated with cervical hip fracture among patients with SLE, but was not associated with trochanteric hip fracture. Comparably, previous studies have shown that steroid use was associated with cervical hip fracture, instead of trochanteric hip fracture, in the general population ([8, 22]). Longitudinal studies showed a dose-dependent relationship between corticosteroid use and bone loss at the lumbar spine, but not at the hip region, among SLE patients ([23, 24]). Therefore, BMD loss may not adequately explain the risk of hip fracture in SLE patients receiving prolonged corticosteroids. On the other hand, corticosteroid use may result in altered bone geometry, particularly cortical compartment, which might augment a decrease in bone strength ([25, 26]). Crabtree et al reported an association between altered cortical bone geometry and cervical hip fracture using peripheral quantitative computed tomography (). In contrast, a higher daily steroid dose might be a marker of inflammation, which in turn predisposes the occurrence of cervical hip fractures.
A higher incidence of the use of intravenous cyclophosphamide at baseline was associated with a higher incidence of cervical fractures among lupus patients. This association may be attributed to the adverse effect of premature ovarian failure due to cyclophosphamide (), which in turn leads to osteoporosis and, subsequently, osteoporotic fractures. On the other hand, patients receiving intravenous cyclophosphamide have more active disease (such as lupus nephritis and other systemic manifestations); therefore, such an association may in fact reflect the negative impact of inflammation on bone loss. However, we do not know why this association was seen only in cervical fractures, and not in trochanteric fractures.
Type 2 diabetes mellitus, hypertension, COPD, and stroke were associated with both types of hip fractures in our lupus patients. These findings are consistent with previous studies on these specific comorbidities with hip fractures ([29-34]). Of these, only stroke conferred an excessive risk for cervical hip fractures among lupus patients after adjustment for age and other risk factors. Several investigations have shown an increased risk of hip fractures following stroke ([33, 34]). A greater propensity for falls and decreased mobility might account for this increased risk. The concurrence of diabetes mellitus, hypertension, heart failure, and COPD among patients with SLE may have resulted from aging, the adverse effects of steroids, or the disease itself, thereby diminishing their significance in multivariable analyses.
It has been well documented that the 1-year mortality rate increases considerably following a hip fracture among the elderly ([20, 35]), but the mortality rate following a hip fracture among the nonelderly is not known. The 1-year mortality rate following a hip fracture was 16% in our lupus cohort, a figure that is apparently higher in comparison to certain malignancies, such as breast cancer (1-year mortality rate of 2.7–7.1% in Taiwan, as of 2008) (). The differences in the 1-year mortality rate between the 2 types of hip fractures were not comparable because of the small case numbers of death following hip fractures in our study. Previously, one epidemiologic study reported that the mortality rate was higher in trochanteric fractures (), but a later study showed no significant differences between the 2 types of fractures (). Overall, age was the most significant predictor of mortality following hip fractures ().
Given that women with SLE are prone to cervical hip fracture and that the 1-year mortality rate following a hip fracture is high, we advocate that women with SLE should undergo a risk assessment for cervical hip fractures after diagnosis, particularly those who are taking higher doses of steroids, have ever received intravenous cyclophosphamide, or have a history of stroke. Specifically, the incorporation of hip structure analysis to DXA scans may be warranted to evaluate the patient's bone geometric parameters ([23, 37]). For those with unfavorable bone geometry features, it is reasonable to start preventive treatment against hip fracture, regardless of BMD measurements. There is evidence that bisphosphonates may improve not only BMD, but also geometry in the proximal femur (), a finding that might justify a clinical trial to study its efficacy in preventing cervical hip fractures in high-risk lupus patients. Also, appropriate calcium and vitamin D supplementation is recommended to improve the bone health, to optimize the benefits of bisphosphonates, and to avoid drug-induced hypocalcemia, particularly in those who take higher steroid doses ([39-42]).
Some limitations to our study are worth noting. Data regarding disease activity, disease severity, BMD, body mass index, frequency of falls, and the menopause status of individual patients were not available in the claim-based study. Lifestyle risk factors, such as calcium/vitamin D supplements, alcohol consumption, or cigarette smoking, were also unknown. Therefore, we cannot analyze the effects of these factors on hip fractures among lupus patients. Also, the ethnic background of the patients in our study is predominantly Asian, which precludes the generalizability of the results to other ethnic populations.
In conclusion, our findings show that women with SLE are at a greater risk of hip fractures, starting at a younger age than the general population. In addition, our study showed that cervical, but not trochanteric, hip fracture is associated with SLE, implying an etiology involving more than mere BMD reduction among lupus patients. Further research is needed to better understand the pathophysiologic mechanisms between these 2 types of hip fractures in patients with lupus.
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. Dr. Shuu-Jiun Wang 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. Shu-Hung Wang, Chang, Liu, Lai.
Acquisition of data. Shu-Hung Wang, Chang, Tzeng-Ji Chen.
Analysis and interpretation of data. Shu-Hung Wang, Chang, Wei-Sheng Chen, Shuu-Jiun Wang.
The authors would like to thank Ms Chiu-Mei Yeh for her statistical assistance.