Figures
Abstract
Objective
We aimed to investigate whether and how corticosteroid use was associated with serious hip arthropathy.
Methods and Materials
This population-based cohort study analyzed the Taiwan National Health Insurance Research Database and screened the one-million random sample from the entire population for eligibility. The steroid cohort consisted of 21,995 individuals who had used systemic corticosteroid for a minimum of 6 months between January 1, 1997 and December 31, 2006. They were matched 1:1 in propensity score on the index calendar date with controls who never used steroid. All participants were followed up until occurrence of serious hip arthropathy that required arthroplasty, withdrawal from the national health insurance, or the end of 2011. Surgical indication was classified as fracture-related and -unrelated. The cumulative incidence of hip arthroplasty was estimated by the Kaplan Meier method. The association with steroid exposure was explored by the Cox proportional hazard model.
Results
Cumulative incidences of hip arthroplasty after 12 years of follow-up were 2.96% (95% confidence interval [CI], 2.73–3.2%) and 1.34% (95% CI, 1.2–1.51%) in the steroid users and non-users, respectively (P<0.0001). The difference was evident in fracture-related arthroplasty with 1.89% (95% CI, 1.71–2.09%) versus 1.10% (95% CI, 0.97–1.25%), but more pronounced in fracture-unrelated surgery, 1.09% (95% CI, 0.95–1.24%) versus 0.24% (95% CI, 0.19–0.32%). Multivariate-adjusted Cox regression analysis confirmed steroid use was independently associated with both fracture-related (adjusted hazard ratio [HR], 1.65; 95% CI, 1.43–1.91) and unrelated arthroplasty (adjusted HR, 4.21; 95% CI, 3.2–5.53). Moreover, the risk for fracture-unrelated arthropathy rose with steroid dosage, as the adjusted HR increased from 3.30 (95% CI, 2.44–4.46) in the low-dose subgroup, 4.54 (95% CI, 3.05–6.77) in intermediate-dose users, to 6.54 (95% CI, 4.74–9.02) in the high-dose counterpart (Ptrend<0.0001).
Citation: Kao F-C, Hsu Y-C, Lin C-FJ, Lo Y-Y, Tu Y-K (2017) Corticosteroid Is Associated with Both Hip Fracture and Fracture-Unrelated Arthropathy. PLoS ONE 12(1): e0169468. https://doi.org/10.1371/journal.pone.0169468
Editor: Ping-Hsun Wu, Kaohsiung Medical University Hospital, TAIWAN
Received: July 28, 2016; Accepted: December 17, 2016; Published: January 26, 2017
Copyright: © 2017 Kao et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Data for the analysis of this study was obtained from the National Health Research Institutes of Taiwan, which imposed legal restrictions on data release for the purpose of privacy protection. If any readers wish to access the data, please contact with Dr. Yao-Chun Hsu (E-mail: holdenhsu@gmail.com).
Funding: The work was supported by the Center for Database Research, E-Da Hospital (CFDR-B-103-1-1).
Competing interests: The authors have declared that no competing interests exist.
Introduction
Corticosteroid may induce bone disease that is associated with substantial morbidity and mortality[1]. For instance, osteoporosis is a well-recognized complication that can precipitate fracture of the weight-bearing bones [2, 3]. Previous studies have established that glucocorticoid use is a risk factor of vertebral and hip fracture[4]. However, the detrimental effect of corticosteroid on bone health extends beyond demineralization. It may also reduce osteocytes’ life span, impair the balance between osteoclast and osteoblast, and compromise skeletal angiogenesis[5–7]. As another complication of steroid exposure, osteonecrosis reportedly affects femoral heads in 8~10% of the users[8]. An epidemiological survey from Japan estimated that among 2,500~3,300 new cases of non-traumatic osteonecrosis every year, 34.7% were induced by corticosteroid[9]. How steroid duration, dosing frequency, and cumulated dosage might impact the risk of osteonecrosis, however, remains controversial[10–12].
Management for hip osteonecrosis depends on severity of the disease. A replacement arthroplasty is warranted for a symptomatic joint that is extensively degenerated or destructed[13]. The surgery of choice reflects the principle of “different operations for different stages” [14]. A hemi-arthroplasty is often sufficient for an intact acetabulum, but a total hip arthroplasty (THA) is usually required for a substantially degenerated joint[15]. It has been estimated that 5~12% of all THAs result from osteonecrosis[16], but how it relates to corticosteroid use has not been elucidated. Prior researches were predominantly hospital-based, without population-scale data. Besides, previous studies mostly focused on fracture-related outcomes and paid less attention to arthropathy unrelated to bone fracture.
In order to clarify the association between corticosteroid and risk of hip arthropathy, this population-based study analyzed a national healthcare research database which covers nearly the entire population of Taiwan. Individuals with and those without exposure to corticosteroid were followed up for the occurrence of hip arthroplasty over a 12-year study period.
Methods
Study design and data source
This is a retrospective open-cohort study based on analysis of the National Health Insurance Research Database (NHIRD). In Taiwan, the Ministry of Health and Welfare had launched the National Health Insurance (NHI) program since 1995 to cover healthcare of the 22.9 million residents in this country. Because the NHI is a single-payer system that implements compulsory and universal policy, it reimburses more than 99% of the national population. The Taiwan National Health Research Institute (NHRI) establishes and maintains NHIRD, which contains comprehensive registry and claim data for research purpose. In the NHIRD, diseases were coded according to the International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9-CM]. Our previous studies have demonstrated utility of the NHIRD in population-based research[17, 18]. In addition to the original NHIRD, the NHRI constructed and maintained the Longitudinal Health Insurance Database (LHID) as a representative data subset of all NHI enrollees. The expanded LHID 2000 was composed of 1,000,000 individuals randomly sampled from the years 1997 through 2011 registry of all beneficiaries. Its representativeness is supported by similar distribution in age, gender, and healthcare cost with the original NHIRD. Data acquisition and study conduction were approved by the institutional review board of E-Da Hospital (EMRP-103-011; EMRP-103-012) as well as Taiwan NHRI (NHIRD-103-116). The study was exempted from a full review by the institutional review board of the E-Da Hospital.
Definition of the study cohorts and outcomes
This cohort study analyzed the expanded LHID 2000 to identify study cohorts. Individuals exposed to oral or injected form of systemic corticosteroid for a minimum of 6 months between January 1, 1997 and December 31, 2006 were screened. The index day for baseline matching and observation inception was assigned at the date when corticosteroid had been used for 6 months. These steroid users were then matched 1:1 in the propensity score on the index day with controls who had not been exposed to any dose of corticosteroid. Individuals who had had hip hemiarthroplasty, total hip arthroplasty or hip joint disarticulation operated prior to or within 6 months after the index day were excluded from the study. Patients who were diagnosed with femoral neck fracture, femoral head fracture, acetabulum fracture, avascular necrosis (AVN) of femoral head, osteoarthritis (OA), rheumatoid arthritis (RA) or systemic lupus erythematous (SLE) before the index date were also excluded, given that these disorders might confound the association between steroid and hip arthroplasty. The corresponding disease and procedure codes were listed in the Appendix (S1 Table).
All participants were followed up until the day when hip arthroplasty was performed, withdrawal from the NHI, or the end of 2011. Therefore, without occurrence of the study outcome or termination from the NHI enlistment, the follow-up duration was at least five years. The indication for arthroplasty was categorized by whether or not it was related to fracture. Fracture-related arthroplasty was defined when the admission diagnosis included a relevant code of bone fracture (femoral neck fracture, femoral head fracture, acetabulum fracture, and traumatic injury from falling down). Otherwise the indication was defined as fracture-unrelated and further classified as OA or AVN.
Assessment of and adjustment for confounding factors
Comorbidity that could confound steroid use and hip arthropathy was considered. The analysis was adjusted for baseline conditions including cerebrovascular accident (CVA), osteoporosis[19], diabetes mellitus (DM) [20], cirrhosis[21], and chronic obstructive pulmonary disease (COPD). In order to delineate the dose-dependent relationship[12], the steroid users were grouped according to the dosage. All forms of systemic steroid reimbursed in the NHIRD were identified by the Anatomical Therapeutic Chemical codes (S2 Table, Appendix). The high-dose (upper quartile) subgroup indicated an average dose above 13.35mg/day, intermediate-dose (interquartile) between 1.09 and 13.35mg, and the low-dose (lower quartile) below 1.08mg per day.
Statistical analysis
Continuous variables were summarized in mean and standard deviation, and categorical variables in number and proportion. The propensity score for steroid use was generated by the logistic regression built on baseline characteristics that could either affect steroid prescription or arthropathy risk, and included age, gender, osteoporosis, CVA, DM, liver cirrhosis, and COPD. The cumulative incidences of hip arthroplasty were estimated by the Kaplan-Meier method, and the between-cohort difference was tested by the 2-tailed log-rank test. The association between steroid exposure and fracture-unrelated hip arthroplasty was explored by the Cox proportional hazard model that took into account age, gender, and baseline comorbidity. Univariate and multivariate-adjusted hazard ratios (HRs) were reported. Dose-response relationship was evaluated by the Cochran–Armitage trend test. Data management and calculation of HRs were performed using the Statistical Package for the Social Sciences (version 17.0; SPSS Inc, Chicago, IL). Calculated results were expressed with the point estimates along with their 95% confidence intervals (CIs). All statistical tests were defined as significant with a P value less than 0.05.
Results
Baseline characteristics of the study population
From 1,000,000 individuals randomly sampled from the entire Taiwan population, we identified 21,995 patients free of any exclusion criteria (steroid cohort) and matched them in propensity score with 21,995 individuals who had not been exposed to systemic corticosteroid (non-steroid cohort). Baseline characteristics were comparable between the two study cohorts (Table 1).
Incidences of hip arthroplasty between study cohorts
The steroid and non-steroid cohorts were followed up for a mean duration of 13.11 (range, 0.03–16.00) and 13.69 (range, 0.16–16.00) years, respectively. During the observation period, 711 (3.23%) corticosteroid users and 372 (1.40%) non-users underwent hip arthroplasty (Table 1). Cumulatively at 12 years, 2.96% (95% CI, 2.73–3.20%) in the steroid users received hip arthroplasty, significantly higher (P<0.0001) than 1.34% (95% CI, 1.20–1.51%) in the non-users (Fig 1). The difference between the steroid and non-steroid cohorts was evident in fracture-related arthroplasty (Fig 2A), 1.89% (95% CI, 1.71–2.09%) versus 1.10% (95% CI, 0.97–1.25%), but even more pronounced in fracture-unrelated surgery (Fig 2B), 1.09% (95% CI, 0.95–1.24%) versus 0.24% (95% CI, 0.19–0.32%).
The steroid cohort had a higher incidence of hip arthroplasty not only for bony fracture (2A, upper panel) but also for fracture-unrelated arthropathy (2B, lower panel).
Multivariate-adjusted association between corticosteroid use and hip arthroplasty
In the unadjusted analysis, the risk for receiving hip arthroplasty was associated with steroid exposure, older age, osteoporosis, CVA, DM, liver cirrhosis, and COPD (Table 2). The multivariate-adjusted Cox proportional hazard model confirmed that steroid use (adjusted HR, 2.09; 95% CI, 1.84–2.37), age (adjusted HR, 1.05 per year; 95% CI, 1.05–1.06), osteoporosis (adjusted HR, 2.18; 95% CI, 1.37–3.45), and DM (HR, 1.52; 95% CI, 1.08–2.13) were independent risk factors for hip arthropathy that necessitated surgical treatment (Table 2).
Association of steroid with fracture-related and -unrelated hip arthroplasties
After adjustment for potential confounding factors (Table 3), steroid exposure was associated with excessive risks of both fracture-related (adjusted HR, 1.65; 95% CI, 1.43–1.91) and fracture-unrelated hip arthropathy (adjusted HR, 4.21; 95% CI, 3.2–5.53).
The risk of undergoing hip arthroplasty, whether it was related to fracture or not, was significantly higher in all steroid dosage subgroups as compared with the non-steroid cohort (Table 4). However, the risk of fracture-unrelated arthropathy rose significantly with the dose, with adjusted HR of 3.30 (95% CI, 2.44–4.46) in the low-dose, 4.54 (95% CI, 3.05–6.77) in the intermediate-dose, and 6.54 (95% CI, 4.74–9.02) in the high-dose subgroup (Ptrend<0.0001 by the Cochran–Armitage test). On the other hand, steroid was similarly associated with the surgery performed for hip fracture across all three dosage subgroups., without a dose-response relationship.
Discussion
This population-based study revealed a significant association of corticosteroid with subsequent hip arthropathy that entailed surgical intervention. Cumulatively at 12 years of follow-up, 2.96% (95% CI, 2.73–3.20%) of steroid users would require hip arthroplasty, significantly higher than 1.34% (95% CI, 1.2–1.51%) of their propensity score-matched controls (P < .0001). Our study further revealed that the excessive risk conferred by steroid use was more pronounced for fracture-related arthropathy, although it was also significant for hip fracture. A four-fold risk (adjusted HR, 4.21; 95% CI, 3.2–5.53) of receiving fracture-unrelated hip surgery (mainly driven by AVN) was shown in the multivariate-adjusted analysis that took age and comorbidity in to account. Moreover, the association was significant in a dose-dependent manner. Collectively, these findings contribute to bridge the knowledge gap pertaining how this extensively prescribed medication may raise the risk of serious hip diseases at a population scale. Our data not only caution against prolonged exposure to steroid, but also indicate the unmet need of effective strategies capable of preventing adverse orthopedic events in individuals who cannot spare steroid.
Osteonecrosis appears to underpin the association between steroid use and fracture-unrelated hip arthropathy[13]. This biologically plausible explanation is corroborated by the fact that AVN was the major diagnosis of hospitalization for fracture-unrelated hip surgery, although we could not radiologically or pathologically ascertain the diagnosis. Corticosteroid has been shown to inhibit angiogenesis and revascularization[13], compatible with its angiostatic property[22]. Besides, the femoral head is particularly vulnerable because of the weight-bearing stress and the reverse blood supply. Other pathogenic mechanisms linking osteonecrosis and steroid include hypertrophy of fat cells[15, 23–26], fat embolization to susceptible subchondral vessels[26, 27], and induction of cytotoxic factors[28–30]. Hypertrophy of fat cells can lead to venous obstruction, vascular occlusion, compression of small veins, and elevation of intraosseous pressure. Blood stasis, ischemia, and bone death may then ensue. Fat embolization could block the subchondral vasculature, resulting in bone ischemia and elevated intraosseous pressure.
Earlier clinical studies have been inconsistent in the association between steroid use and femoral head osteonecrosis. Similar to our findings, Vreden and colleagues found that 4 (2.4%) out of 167 patients on corticosteroid replacement therapy for adrenal insufficiency developed clinically apparent femoral head osteonecrosis[31]. By using magnetic resonance image to detect early signs of osteonecrosis, Sakamoto and colleagues reported abnormal signals in as many as 31 hips (32%) of 17 patients from a total of 48 study participants[32]. However, there were negative reports as well. For instance, Colwell and colleagues prospectively observed 71 patients who received corticosteroid for asthma or arthritis and concluded no direct relationship between corticosteroid and development of femoral head osteonecrosis after 10 years of follow-up[33]. All of these prior researches were either small in sample size or short in observation duration. More importantly, their enrollees were selected patients with a specific disease. In contrast, our study observes more than 20,000 patients for a mean duration longer than 10 years in the setting of general population.
The dose-dependent effect of steroid on the risk of fracture-unrelated hip arthropathy suggests a causal relationship. To the best of our knowledge, this is the first population-based cohort study focusing on this issue. Our study agrees with a hospital-based research conducted by Lausten and colleagues who reported that a reduction in the cumulative dose of steroid paralleled a decrease in the rate of osteonecrosis among 498 patients with renal transplantation[34]. Our results, nonetheless, should not be misinterpreted as low-dose steroid could be safe, because excessive risk was already evident in individuals who had been exposed to an average dose below 1.08 mg/day over 6 months. Clearly, it is advisable to prescribe steroid at a lowest possible dose for a shortest possible duration, if the drug is unavoidable at all.
A meta-analysis including 66 studies on BMD and 23 studies on clinical fracture concluded that oral steroid of 5mg (prednisolone equivalent) per day would lead to loss of BMD and rapidly (within 3 to 6 months) increase the risk of bone fracture[35]. Our result corroborates the association of steroid with hip fracture although the risk did not increase with dosage. Whereas the well-established osteoporotic effect of steroid would expose users to a higher long-term risk of receiving hip arthroplasty, our analyses further point out the adverse outcomes of steroid exposure can occur without an accompanying traumatic event.
The following limitations are acknowledged. First, we did not separately analysed oral or injected forms of steroid, because sequential or intermittent use of both administrative routes is common in daily practice. Second, laboratory, radiographic, and pathological data were regrettably unavailable in the NHIRD. Therefore, we resorted to admission diagnosis to define the presence of osteonecrosis. Similarly, this study could not detect subtle change of the joint or bone, such as minor fracture, but had to rely on unequivocal events to define outcomes. Nevertheless, using stringent definition and the consequence of under diagnosis would have biased the result toward null association instead of creating a spurious one, because under diagnosis was more likely to occur in those who did not seek medical attention. For instance, a diagnosis of OA would be less easily missed in patients who received steroid for relevant symptoms than in those who never used steroid. This potential misclassification should have biased the analysis toward a higher incidence of hip arthroplasty (because of undiagnosed OA) in the steroid non-users, and thus could have attenuated the association between steroid exposure and hip arthroplasty rather than have exaggerated it. Besides, this study was specifically designed to elucidate the influence of steroid, but not to comprehensively cover risk factors for hip arthropathy. Comorbid conditions were explored as covariates for the purpose of multivariate adjustment, but unmeasured confounding remained possible. The analysis could not examine potential influence by lifestyle, body weight, exercise activity, cigarette smoking, alcohol drinking, and dietary habit, since such information were unavailable in the NHIRD. As a means to overcome this limitation, we used relevant clinical diseases as surrogates. For instance, COPD is relevant for smoking and liver cirrhosis for drinking. Potential confounding by these factors were then adjusted in both the generation of propensity score and the multivariate Cox analysis.
In conclusion, corticosteroid used for 6 months or longer is associated with an increased risk of receiving hip arthroplasty in the long run. The risk results not only from hip fracture but also, more pronouncedly in fact, from arthropathy unrelated to fracture. The multivariate-adjusted association between steroid and long-term risk of fracture-unrelated hip arthropathy was significant in a dose-dependent manner. Findings from this population-based research remind healthcare professionals to minimize use of corticosteroid whenever possible. Novel strategy beyond merely strengthening the bone is needed for hip protection in patients who can’t spare steroid.
Supporting Information
S1 Table. Coding for diseases and procedures in the study.
https://doi.org/10.1371/journal.pone.0169468.s001
(DOCX)
S2 Table. Details regarding the dosage and formula of steroid used in this study.
https://doi.org/10.1371/journal.pone.0169468.s002
(DOCX)
Acknowledgments
This study is based on data(NHIRD-103-116) from the National Health Insurance Research Database provided by the Bureau of National Health Insurance, Department of Health and managed by National Health Research Institutes. The interpretation and conclusions contained herein do not represent those of Bureau of National Health Insurance, Department of Health or National Health Research Institutes. The authors are grateful to Ms. Chen Tzu-Shan for her efficient assistance.
Author Contributions
- Conceptualization: FCK.
- Formal analysis: YYL.
- Funding acquisition: YCH.
- Investigation: FCK YCH.
- Methodology: YCH.
- Project administration: YCH.
- Resources: YCH YKT.
- Software: YYL.
- Supervision: YKT YCH.
- Validation: CFJL.
- Visualization: YCH.
- Writing – original draft: FCK.
- Writing – review & editing: FCK YCH YKT.
References
- 1. Weinstein R. S. (2011). "Clinical practice. Glucocorticoid-induced bone disease." N Engl J Med 365(1): 62–70. pmid:21732837
- 2. Canalis E., Mazziotti G., Giustina A. and Bilezikian J. P. (2007). "Glucocorticoid-induced osteoporosis: pathophysiology and therapy." Osteoporos Int 18(10): 1319–1328. pmid:17566815
- 3. Van Staa T. P., Laan R. F., Barton I. P., Cohen S., Reid D. M. and Cooper C. (2003). "Bone density threshold and other predictors of vertebral fracture in patients receiving oral glucocorticoid therapy." Arthritis Rheum 48(11): 3224–3229. pmid:14613287
- 4. Steinbuch M., Youket T. E. and Cohen S. (2004). "Oral glucocorticoid use is associated with an increased risk of fracture." Osteoporos Int 15(4): 323–328. pmid:14762652
- 5. Weinstein R. S., Jilka R. L., Parfitt A. M. and Manolagas S. C. (1998). "Inhibition of osteoblastogenesis and promotion of apoptosis of osteoblasts and osteocytes by glucocorticoids. Potential mechanisms of their deleterious effects on bone." J Clin Invest 102(2): 274–282. pmid:9664068
- 6. O'Brien C. A., Jia D., Plotkin L. I., Bellido T., Powers C. C., Stewart S. A., Manolagas S. C. and Weinstein R. S. (2004). "Glucocorticoids act directly on osteoblasts and osteocytes to induce their apoptosis and reduce bone formation and strength." Endocrinology 145(4): 1835–1841. pmid:14691012
- 7. Weinstein R. S., Wan C., Liu Q., Wang Y., Almeida M., O'Brien C. A., Thostenson J., Roberson P. K., Boskey A. L., Clemens T. L. and Manolagas S. C. (2010). "Endogenous glucocorticoids decrease skeletal angiogenesis, vascularity, hydration, and strength in aged mice." Aging Cell 9(2): 147–161. pmid:20047574
- 8. Asano T., Takahashi K. A., Fujioka M., Inoue S., Satomi Y., Nishino H., Tanaka T., Hirota Y., Takaoka K., Nakajima S. and Kubo T. (2003). "Genetic analysis of steroid-induced osteonecrosis of the femoral head." J Orthop Sci 8(3): 329–333. pmid:12768474
- 9. Mirzai R., Chang C., Greenspan A. and Gershwin M. E. (1999). "The pathogenesis of osteonecrosis and the relationships to corticosteroids." J Asthma 36(1): 77–95. pmid:10077138
- 10. Harrington K. D., Murray W. R., Kountz S. L. and Belzer F. O. (1971). "Avascular necrosis of bone after renal transplantation." J Bone Joint Surg Am 53(2): 203–215. pmid:4927167
- 11. Mont M. A. and Hungerford D. S. (1995). "Non-traumatic avascular necrosis of the femoral head." J Bone Joint Surg Am 77(3): 459–474. pmid:7890797
- 12. Kubo T., Fujioka M., Yamazoe S., Yoshimura N., Oka T., Ushijima Y., Hasegawa Y. and Hirasawa Y. (1998). "Relationship between steroid dosage and osteonecrosis of the femoral head after renal transplantation as measured by magnetic resonance imaging." Transplant Proc 30(7): 3039–3040. pmid:9838339
- 13. Gebhard K. L. and Maibach H. I. (2001). "Relationship between systemic corticosteroids and osteonecrosis." Am J Clin Dermatol 2(6): 377–388. pmid:11770392
- 14. Cruess R. L. (1981). "Steroid-induced osteonecrosis: a review." Can J Surg 24(6): 567–571. pmid:7326617
- 15. Solomon L. (1981). "Idiopathic necrosis of the femoral head: pathogenesis and treatment." Can J Surg 24(6): 573–578. pmid:7326618
- 16. Hungerford D. S. (2002). "Osteonecrosis: avoiding total hip arthroplasty." J Arthroplasty 17(4 Suppl 1): 121–124.
- 17. Hsu Y. C., Lin J. T., Ho H. J., Kao Y. H., Huang Y. T., Hsiao N. W., Wu M. S., Liu Y. Y. and Wu C. Y. (2014). "Antiviral treatment for hepatitis C virus infection is associated with improved renal and cardiovascular outcomes in diabetic patients." Hepatology 59(4): 1293–1302. pmid:24122848
- 18. Hsu Y. C., Lin J. T., Chen T. T., Wu M. S. and Wu C. Y. (2012). "Long-term risk of recurrent peptic ulcer bleeding in patients with liver cirrhosis: a 10-year nationwide cohort study." Hepatology 56(2): 698–705. pmid:22378148
- 19. Russell L. A. (2013). "Osteoporosis and orthopedic surgery: effect of bone health on total joint arthroplasty outcome." Curr Rheumatol Rep 15(11): 371. pmid:24085661
- 20. King K. B., Findley T. W., Williams A. E. and Bucknell A. L. (2013). "Veterans with diabetes receive arthroplasty more frequently and at a younger age." Clin Orthop Relat Res 471(9): 3049–3054. pmid:23649224
- 21. Solaymani-Dodaran M., Card T. R., Aithal G. P. and West J. (2006). "Fracture risk in people with primary biliary cirrhosis: a population-based cohort study." Gastroenterology 131(6): 1752–1757. pmid:17087953
- 22. McNatt L. G., Lane D. and Clark A. F. (1992). "Angiostatic activity and metabolism of cortisol in the chorioallantoic membrane (CAM) of the chick embryo." J Steroid Biochem Mol Biol 42(7): 687–693. pmid:1380295
- 23. Sambrook P. N., Hassall J. E. and York J. R. (1984). "Osteonecrosis after high dosage, short term corticosteroid therapy." J Rheumatol 11(4): 514–516. pmid:6481723
- 24. Powell E. T., Lanzer W. L. and Mankey M. G. (1997). "Core decompression for early osteonecrosis of the hip in high risk patients." Clin Orthop Relat Res(335): 181–189. pmid:9020217
- 25. Wang G. J., Moga D. B., Richemer W. G., Sweet D. E., Reger S. I. and Thompson R. C. (1978). "Cortisone induced bone changes and its response to lipid clearing agents." Clin Orthop Relat Res(130): 81–85. pmid:639410
- 26. Wang G. J., Sweet D. E., Reger S. I. and Thompson R. C. (1977). "Fat-cell changes as a mechanism of avascular necrosis of the femoral head in cortisone-treated rabbits." J Bone Joint Surg Am 59(6): 729–735. pmid:908695
- 27. Fisher D. E. and Bickel W. H. (1971). "Corticosteroid-induced avascular necrosis. A clinical study of seventy-seven patients." J Bone Joint Surg Am 53(5):
- 28. Glimcher M. J. and Kenzora J. E. (1979). "Nicolas Andry award. The biology of osteonecrosis of the human femoral head and its clinical implications: 1. Tissue biology." Clin Orthop Relat Res(138): 284–309. pmid:445912
- 29. Glimcher M. J. and Kenzora J. E. (1979). "The biology of osteonecrosis of the human femoral head and its clinical implications: II. The pathological changes in the femoral head as an organ and in the hip joint." Clin Orthop Relat Res(139): 283–312. pmid:455846
- 30. Glimcher M. J. and Kenzora J. E. (1979). "The biology of osteonecrosis of the human femoral head and its clinical implications. III. Discussion of the etiology and genesis of the pathological sequelae; commments on treatment." Clin Orthop Relat Res(140): 273–312. pmid:477083
- 31. Vreden S. G., Hermus A. R., van Liessum P. A., Pieters G. F., Smals A. G. and Kloppenborg P. W. (1991). "Aseptic bone necrosis in patients on glucocorticoid replacement therapy." Neth J Med 39(3–4): 153–157. pmid:1791877
- 32. Sakamoto M., Shimizu K., Iida S., Akita T., Moriya H. and Nawata Y. (1997). "Osteonecrosis of the femoral head: a prospective study with MRI." J Bone Joint Surg Br 79(2): 213–219. pmid:9119845
- 33. Colwell C. W. Jr., Robinson C. A., Stevenson D. D., Vint V. C. and Morris B. A. (1996). "Osteonecrosis of the femoral head in patients with inflammatory arthritis or asthma receiving corticosteroid therapy." Orthopedics 19(11): 941–946. pmid:8936529
- 34. Lausten G. S., Jensen J. S. and Olgaard K. (1988). "Necrosis of the femoral head after renal transplantation." Acta Orthop Scand 59(6): 650–654. pmid:3063055
- 35. van Staa T. P., Leufkens H. G. and Cooper C. (2002). "The epidemiology of corticosteroid-induced osteoporosis: a meta-analysis." Osteoporos Int 13(10): 777–787. pmid:12378366