Osteoporosis International

, Volume 24, Issue 9, pp 2493–2498

Influence of recency and duration of glucocorticoid use on bone mineral density and risk of fractures: population-based cohort study

Authors

    • Department of MedicineUniversity of Alberta
    • 2 F1.24 Walter Mackenzie Health Sciences CentreUniversity of Alberta Hospital
  • S. N. Morin
    • Department of MedicineMcGill University
  • L. M. Lix
    • Department of Community Health SciencesUniversity of Manitoba
  • W. D. Leslie
    • Departments of Medicine and RadiologyUniversity of Manitoba
Original Article

DOI: 10.1007/s00198-013-2352-3

Cite this article as:
Majumdar, S.R., Morin, S.N., Lix, L.M. et al. Osteoporos Int (2013) 24: 2493. doi:10.1007/s00198-013-2352-3

Abstract

Summary

Although systemic glucocorticoids are commonly used, it is difficult to obtain accurate exposure history. In 50,000 patients, we confirmed that glucocorticoids were associated with reductions in bone mineral density (BMD) and increases in fracture and documented that recent and prolonged durations of exposure were particularly associated with adverse events—dose information did not improve risk prediction.

Introduction

Systemic glucocorticoid use, defined as ever having taken supra-physiologic doses for 90-days or more, is a risk factor for low BMD and fractures. This definition does not distinguish recent (vs remote) exposure.

Methods

Within a population-based clinical BMD registry in Manitoba, Canada, we identified all adults over age 40 years tested between 1998 and 2007 and then undertook a cohort study. We identified all oral glucocorticoid dispensations from 1995 to 2009 and stratified exposure by timing (“recent” if within 12 months vs “remote”) and duration (short [<90 days] vs prolonged [≥90 days]). Osteoporosis-related risk factors and treatments and major fractures were obtained using administrative health data.

Results

A total of 12,818 of 52,070 (25 %) subjects had used glucocorticoids prior to BMD testing; the most common exposure was remote short (n = 6453) vs recent prolonged (n = 2896) vs recent short (n = 2644) vs remote prolonged (n = 825). Compared to 39,252 never-users, only recent prolonged glucocorticoid use was significantly associated with femoral neck T-score (ANCOVA-adjusted difference −0.13, 95 % CI −0.16 to −0.10, p < 0.001). There were 2,842 major (566 hip) fractures over median 5-year follow-up. Compared with never-users, only recent prolonged glucocorticoid use was significantly associated with BMD-independent increases in risk of incident major fracture (5.4 vs 7.7 %, adjusted HR 1.25, 95 % CI 1.07–1.45, p = 0.004) and hip fracture (1.1 vs 1.8 %, adjusted HR 1.61, 95 % CI 1.18–2.20, p = 0.003).

Conclusions

Recent and prolonged glucocorticoid use (but neither remote nor recent short courses) was independently associated with reduced BMD and increased risk of fractures. These findings should permit clinicians to identify a high-risk group of patients that might benefit from osteoporosis prevention.

Keywords

BMDFracture riskGlucocorticoidsOsteoporosisPredictionSide effects

Introduction

Long-term systemic glucocorticoids are used frequently for chronic inflammatory conditions, and up to 1 % of adults are treated with glucocorticoids each year [1,2]. Glucocorticoid-induced osteoporosis is the most common cause of secondary (or iatrogenic) osteoporosis and has been associated with substantial morbidity related to major fractures as well as medico-legal liability [1,2]. Unlike most other etiologies of osteoporosis, glucocorticoid-induced bone loss occurs very rapidly and (independent of changes in bone mineral density [BMD]) the increased risk of fracture manifests within 3 months of starting therapy [25]. The increased risk of fracture with exposure to glucocorticoids, dependent on how exposure is defined, ranges anywhere from 80 to 500 % [48]. For this reason, glucocorticoid exposure is always considered a major risk factor for osteoporotic fracture, and it is captured within all risk prediction tools such as FRAX [9,10].

In the clinical context, it is often difficult to capture a patient’s glucocorticoid history, particularly as it relates to dose (either current daily dose or past cumulative dose). This is even tacitly acknowledged within the FRAX schema as the risk factor is worded as: “Patient is currently exposed to oral glucocorticoids or has been exposed to oral glucocorticoids for more than 3-months at a dose of prednisolone of 5 mg daily or more” [9,10]. Indeed, in the most recent iteration of FRAX, information regarding dose has been omitted altogether [9]. Furthermore, in clinical practice, this information is usually captured as “Have you ever used oral steroids for longer than 3-months?” or perhaps even more commonly as “Have you ever used steroids?” We sought to examine dose-independent information about the timing of glucocorticoid exposure, reasoning that this information might be more easily and accurately captured from patients as well as from administrative or other health-related databases.

Therefore, we set out to determine the influence of recency and duration of glucocorticoid use on BMD and risk of major osteoporotic fractures. We defined “recent” use as within the past year (vs remote use, more than 1 year ago) and “prolonged” course as 90 days or longer (vs short course, less than 90 days). Compared with those who had never used systemic glucocorticoids, we hypothesized that recent and prolonged use of glucocorticoids would be associated with the greatest reductions in femoral neck BMD and increases in the risk of fracture.

Methods

Study design

We undertook a population-based cohort study of all adults 40 years of age and older in Manitoba, Canada, who underwent a baseline (index) BMD test between 1998 and 2007. Manitoba has a population of 1.3 million residents with universal healthcare coverage that includes comprehensive services regardless of age or income. The Health Research Ethics Board of the University of Manitoba approved the study and the Manitoba Health Information Privacy Committee approved data access.

Data sources

Population health research data repository

This is a comprehensive collection of continuously updated and longitudinal population-based health services datasets provided by the provincial government to the Manitoba Centre for Health Policy in an anonymized format that permits linkage using unique identifiers [1,1114]. Data include sociodemographic characteristics, vital statistics, physician claims and hospital discharge diagnoses, and prescription drugs (including drug name and date dispensed from all retail pharmacies since 1995). These datasets are well validated and have been used extensively in previous research [1114].

Manitoba bone density program

This is a population-based clinical registry of all BMD tests performed with central DXA in the province and has been linked to the aforementioned databases; it is well validated and has been previously described [11,12]. Briefly, all clinical DXA tests are performed within a single provincial program that maintains uniform data collection with respect to osteoporosis risk factors (including many components of the World Health Organization FRAX® risk calculator [9,10]) as well as testing indications and reporting standards. All densitometers (GE Lunar DPX prior to 2000, GE Lunar Prodigy after 2000, Madison, WI, USA) undergo daily evaluation of stability using anthropometric spine phantoms, and all equipment and personnel are subject to province-wide quality assurance programs. All BMD test results have been captured since the first instrument was installed and the registry database itself is better than 99 % complete [11,12]. Femoral neck T-scores (number of SDs above or below young adult mean BMD) were calculated based on NHANES-III reference data for US white females as recommended by the WHO Collaborating Centre for Metabolic Bone Diseases[12].

Glucocorticoid exposure

Using linkages to the province-wide retail pharmacy network, we identified each subject’s total oral glucocorticoid dispensations from 1995 through to the time of their baseline BMD test performed between 1998 through 2009. We defined recent exposure as use within 1 year of the BMD test (vs remote use) and we defined prolonged course as 90 days or more of continuous treatment (vs short course). Based on these patterns of glucocorticoid exposure before the baseline BMD test, we derived the following independent variables: never-use [reference group for all analyses], recent short course, recent prolonged course, remote short course, and remote prolonged course. For purposes of this study, we did not capture doses or types of oral glucocorticoids, although prior work in this population has reported that prednisone is the most commonly prescribed agent (93 %) and that two thirds of all dispensations are for more than 5 mg daily [1].

Outcomes

We had two dependent variables of interest. First, we examined associations with BMD, specifically the femoral neck T-score. Subjects who did not have a femoral neck BMD measurement were excluded from all analyses. Second, we examined associations with major osteoporotic fractures as defined under the FRAX formulation (any non-traumatic fragility fracture of the forearm, humerus, clinical vertebrae, or hip) occurring after the baseline BMD test. We used validated case definitions of osteoporosis-related fractures based on diagnostic codes captured in physician claims and/or hospital discharges [1114]. In addition, we required that hip and forearm fractures be accompanied by site-specific fracture reduction, fixation, or casting codes to enhance diagnostic and temporal specificity for acute fracture.

Potential confounders

We determined established risk factors for osteoporosis and fracture including age, sex, body mass index (BMI in kg/m2), prior major non-traumatic osteoporotic fracture, rheumatoid arthritis, chronic obstructive pulmonary disease (COPD, a proxy measure for smoking), and a diagnosis of alcohol (or other substance) abuse. Lastly, we identified dispensations of prescription osteoporosis treatments (i.e., bisphosphonates, calcitonin, systemic estrogen products, raloxifene, and teriparatide) in the 1 year prior to the BMD test.

Analyses

We conducted descriptive analyses and present all available baseline characteristics according to the five pre-specified categories of glucocorticoid exposure. We then undertook two sets of multivariable regression analyses, one with femoral neck T-score as the outcome and the other with major fracture as outcome. Each model was adjusted for age, sex, BMI, prior major fracture, rheumatoid arthritis, COPD, alcohol abuse, and prescription osteoporosis treatment prior to the baseline BMD test. Time varying covariates were not introduced for glucocorticoid exposure or other characteristics as we were interested in examining the predictive performance of information available to clinicians at the time of BMD assessment or glucocorticoid initiation when attempting to risk-stratify their patients.

To examine the independent association between prior glucocorticoid exposure (five exposure categories represented by four dummy variables and with never-use of glucocorticoids as the reference group) and femoral neck T-scores, we used unadjusted and multivariable analysis of covariance (ANCOVA). The output from these models can be directly interpreted as the adjusted difference in femoral neck T-score when compared with never-use of glucocorticoids. In a secondary analysis, we used lumbar neck T-scores as the dependent variable of interest.

To examine the independent association between glucocorticoid exposure prior to BMD test and incidence of major fractures, we used unadjusted and adjusted Cox proportional hazards models. Data were censored at the time of the first fracture, death, disenrollment from the provincial health plan, or study end date (median 5-year follow-up). In addition to the covariates described earlier, we included femoral neck T-score because we were primarily interested in the BMD-independent fracture risk associated with the amount of glucocorticoid exposure. Proportional hazards assumptions were evaluated using log-minus-log survival plots and analysis of rescaled Schoenfeld residuals; there were no violations. We then repeated these analyses using incidence of hip fracture as the dependent variable. All analyses were conducted using Statistica (Version 10, StatSoft Inc, Tulsa, OK, USA).

Results

There were 52,070 subjects in the study cohort, among whom 12,818 (25 %) had used glucocorticoids prior to their baseline (index) BMD test. The most common pattern of glucocorticoid exposure was a remote short course (n = 6,453 [50 %]) vs recent prolonged (n = 2,896 [23 %]) vs recent short (n = 2,644 [21 %]) vs remote prolonged (n = 825 [6 %]). Compared to the 39,252 never-users, those exposed to glucocorticoids tended to be younger (63 vs 64 years), more likely to be male (14 vs 6 %), had greater BMI, and had more comorbidities (Table 1).
Table 1

Characteristics of 52,070 patients undergoing index bone mineral density testing according to different glucocorticoid exposures

Characteristics at the time of index, BMD test

No gluco-corticoid exposure, (N = 39,252)

Recent short course, (N = 2,644)

Recent prolonged course, (N = 2,896)

Remote short course, (N = 6,453)

Remote prolonged course, (N = 825)

Age, years (SD)

64 (11)

62 (12)

63 (12)

64 (11)

65 (13)

Female, %

94

88

69

93

85

Prior fracture, %

13

11

10

14

15

Rheumatoid arthritis, %

2

9

23

4

16

COPD, %

17

34

29

26

26

Alcohol abuse, %

3

4

4

3

3

Prior osteoporosis treatment, %

18

17

22

17

23

Body mass index, kg/m2 (SD)

27 (5)

28 (6)

28 (6)

28 (6)

27 (6)

Femoral neck BMD, T-score (SD)

−1.4 (1.0)

−1.3 (1.0)

−1.4 (1.1)

−1.3 (1.0)

−1.4 (1.0)

Lumbar spine BMD, T-score (SD)

−1.3 (1.5)

−1.0 (1.6)

−1.1 (1.6)

−1.1 (1.5)

−1.1 (1.6)

Bone mineral density and glucocorticoid exposure

The average femoral neck T-score was −1.4 (SD 1.0) and, in unadjusted analyses, was similar across glucocorticoid exposure groups (Table 1). However, in multivariable ANCOVA regression analyses, only recent prolonged courses of glucocorticoids were associated with a statistically significantly lower femoral neck T-score when compared with never-users (−0.13 adjusted difference in T-score, 95 % CI −0.16 to −0.10, p < 0.001; Table 2). A similar pattern was evident at the lumbar spine, where again only recent prolonged courses of glucocorticoids were associated with statistically significantly lower BMD (−0.18 adjusted difference in T-score, 95 % CI −0.23 to −0.13, p < 0.001; Table 2).
Table 2

Adjusted differences in bone mineral density according to different glucocorticoid exposures when compared with never-users

Glucocorticoid exposure subgroups

Adjusteda difference in T-score (95 % CI), femoral neck

Adjusteda difference in T-score (95 % CI), lumbar spine

Never use (N = 39,562)

0 [reference]

0 [reference]

Recent short course (N = 2,644)

0.02 (−0.01 to 0.05)

0.05 (0.0 to 0.10)

Recent prolonged course (N = 2,896)

−0.13 (−0.16 to −0.10)

−0.18 (−0.23 to −0.13)

Remote short course (N = 6,453)

0.09 (0.09 to 0.11)

0.07 (0.04 to 0.11)

Remote prolonged course (N = 825)

−0.03 (−0.08 to 0.02)

0.04 (−0.04 to 0.12)

aMultivariable analysis of covariance (ANCOVA) models adjusted for age, sex, BMI, prior major fracture, rheumatoid arthritis, COPD, alcohol abuse, and prescription osteoporosis treatment prior to the index BMD test

Major fractures and glucocorticoid exposure

Over 5 years of follow-up, there were 2,842 (5 %) individuals with one or more major osteoporotic fractures, including 566 (1 %) fractures of the hip. The median time to major fracture was 3.7 years (interquartile range [IQR] 1.6–6.3), and the median time to fracture was longer for never-users compared with any glucocorticoid exposure (3.9 years [IQR 1.7–6.3] vs 3.1 years [IQR 1.3–6.1], p = 0.005). The highest rate of major fracture and hip fracture occurred in the 2,896 subjects with recent prolonged glucocorticoid exposures, while the 9,097 subjects exposed to remote short courses or recent short courses had fracture rates similar to that of never-users (Fig. 1). Compared with never-use of glucocorticoids, in multivariable Cox proportional hazards analysis, only recent prolonged courses were independently associated with an increased risk of major fracture: 5.4 % for never-use vs 7.7 % for recent prolonged course, adjusted hazard ratio (aHR) 1.25, 95 % CI 1.07–1.45, p = 0.004 (Table 3). The same relationship was observed for hip fracture: 1.1 % for never-use vs 1.8 % for recent prolonged course, aHR 1.61, 95 % CI 1.18–2.20, p = 0.003 (Fig. 2; Table 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs00198-013-2352-3/MediaObjects/198_2013_2352_Fig1_HTML.gif
Fig. 1

Association between recency and duration of different glucocorticoid exposures and major osteoporotic fractures

Table 3

Unadjusted and adjusted incidence of fracture by glucocorticoid exposure

 

No gluco-corticoid exposure (N = 39,252)

Recent short course (N = 2,644)

Recent prolonged course (N = 2,896)

Remote short course (N = 6,453)

Remote prolonged course (N = 825)

Follow-up time, years (SD)

5.2 (3.0)

4.6 (3.0)

5.4 (3.1)

4.1 (2.8)

4.6 (2.9)

Major fractures, N (%)

2120 (5.4)

140 (5.3)

223 (7.7)

297 (4.6)

55 (6.7)

Major fractures, adjusted HR (95 % CI)a

1.0 (reference)

1.16 (0.97–1.38)

1.25 (1.07–1.45)

1.14 (1.00–1.29)

1.29 (0.99–1.69)

Hip fractures, N (%)

432 (1.1)

32 (1.2)

52 (1.8)

52 (0.8)

9 (1.1)

Hip fractures, adjusted HR (95 % CI)a

1.0 (reference)

1.26 (0.87–1.83)

1.61 (1.18–2.20)

0.98 (0.72–1.32)

1.05 (0.54–2.03)

aMultivariable Cox proportional hazards model that includes covariates of age, sex, BMI, prior major fracture, rheumatoid arthritis, COPD, alcohol abuse, prescription osteoporosis treatment, and femoral neck T-score

https://static-content.springer.com/image/art%3A10.1007%2Fs00198-013-2352-3/MediaObjects/198_2013_2352_Fig2_HTML.gif
Fig. 2

Association between recency and duration of different glucocorticoid exposures and hip fractures

Discussion

In a population-based registry of more than 50,000 subjects undergoing a baseline BMD test, we found that oral glucocorticoid exposure was very common—fully one quarter of those tested had received glucocorticoids in the past. Over 5 years of follow-up, 5 % of subjects suffered a major fracture and 1 % suffered a hip fracture. In adjusted analyses, we found that only recent (within 1 year) and prolonged (greater than 90 days) courses of systemic glucocorticoids were associated with significantly lower BMD (femoral neck and lumbar spine) and with significantly increased risk of major fractures (25 % relative increase) and hip fractures (61 % relative increase). Reassuringly, whether remote or recent, short durations of glucocorticoid use (71 % of all exposures) were not associated with lower BMD nor increased risk of fracture.

Many observational studies and some randomized trials have examined the risks to bone health associated with systemic glucocorticoids [29]. Although the studies are not entirely consistent with each other, to very broadly summarize, in terms of glucocorticoid exposure and BMD: that there is rapid (within 3 months) loss of BMD that continues with ongoing exposure, that “ever” glucocorticoid exposure is not necessarily associated with reductions in BMD, that the rate of BMD loss tends to slow down after 1 year of exposure, that soon after glucocorticoid cessation there is partial and rapid recovery of BMD, that there does not seem to be a daily dose threshold below which BMD is unaffected and even high-dose inhaled glucocorticoids can be associated with reduced BMD, that cumulative dose may be more strongly correlated with BMD than daily dose, and that much of the risk of fracture associated with glucocorticoids cannot be explained by changes in BMD [49].

In terms of glucocorticoid exposure and major fractures, to broadly summarize: there is a rapid (within 3–6 months) increase in the risk of fracture that continues with ongoing exposure, the increased risk of fracture remains as long as there is ongoing exposure, that “ever” glucocorticoid exposure is not consistently found to be a risk factor for fracture, that soon after glucocorticoid cessation there is fairly rapid reduction in fracture risk with a return to near-baseline risk within 1 year, that there does not seem to be a lower daily dose threshold below which fracture risk is unaffected, that daily dose may be more strongly correlated with fracture than cumulative dose, and that much of the risk of fracture associated with glucocorticoids is independent of BMD [49].

Thus, the available literature is fairly consistent with respect to both losses of BMD and increases in fracture risk accorded to glucocorticoids—except for the issue of the relative importance and inter-relationships between daily dose and cumulative dose. For the clinician, the various instructions given to patients related to tapering doses over time, alternate-day use, and dispensed but not taken “on-demand use” for patient-diagnosed acute exacerbations of disease, calculating a daily or cumulative dose is not at all straightforward, particularly for more remote exposure. On the other hand, it seems far easier to query patients about recent use (within the year) and duration of exposure (longer than 3-months). Given that there is no obvious safe-to-skeletal-health lower dose of glucocorticoids [29], our findings suggest that most of the fracture risk associated with glucocorticoids can be determined on the basis of recency of use and duration of exposure. That said, in our study, the subgroup of remote use and prolonged duration was very small and constituted only 6 % of all glucocorticoid users. This small sample size, coupled with a lower rate of glucocorticoid-associated major fractures than previously reported in the literature [4,5], may have led to a lack of statistical significance when in fact an association might be present.

Our study has several important limitations. First, although population-based, our registry cohort was drawn from only those subjects who were referred for a BMD test. Second, ascertainment of symptomatic fractures was based on validated administrative data algorithms. We did not have radiographic confirmation of fractures and we have no data with respect to clinically silent vertebral fractures that might have been captured with surveillance radiographs. Third, while we adjusted for prescription osteoporosis treatment and many osteoporosis-related risk factors, we did not adjust analyses for predicted probability of fracture according to FRAX. Nevertheless, FRAX already captures some information about glucocorticoid use, and so using FRAX would have led to “double-counting” and overadjustment. Lastly, we do not have doses or types of glucocorticoids dispensed, although given our stated objectives we do not consider this as a major limitation.

What, if any, are the possible clinical implications of our findings? The subgroup of patients with recent prolonged use of glucocorticoids, representing about one quarter of all glucocorticoid users in our study, is an easy-to-identify high-risk group that should be targeted for aggressive measures related to glucocorticoid-induced osteoporosis prevention. Obviously, clinicians do not necessarily always know which of their patients may require longer than 3 months of treatment at the time of initial prescription, but this suggests that either osteoporosis preventive measures be started at the time of initial prescription (predictable prolonged use) or at least if patients continue to take glucocorticoids at any dose beyond 3 months (unpredictable prolonged use), it needs to be addressed at that point.

In conclusion, we confirmed that oral glucocorticoid use is associated with reductions in BMD and increases in the risk of major osteoporotic and hip fractures. However, our most clinically important finding was that the highest risk subgroup of patients in everyday practice was the 23 % of glucocorticoid users who were exposed to recent and prolonged courses of treatment. This information should be helpful to clinicians and could inform both risk stratification related to new initiations or continued use of glucocorticoids as well as for a more general fracture risk assessment in terms of how one addresses the presence or absence of this risk factor.

Acknowledgments

The authors are indebted to Manitoba Health for the provision of data (HIPC File No. 2007/2008-49). The results and conclusions are those of the authors, and no official endorsement by Manitoba Health is intended or should be inferred. This article has been reviewed and approved by the members of the Manitoba Bone Density Program Committee. WD Leslie had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the analysis.

Conflicts of Interest

The authors declare no potential conflicts of interest with respect to this work. No external funding source was required for this research. SRM receives salary support from the Alberta Heritage Foundation for Medical Research—AIHS (Health Scholar) and holds the Endowed Chair in Patient Health Management (Faculties of Medicine and Dentistry and Pharmacy and Pharmaceutical Sciences, University of Alberta). SNM is chercheur-clinicien boursier des Fonds de la Recherche en Sante du Quebec. LML is supported by a Manitoba Research Chair.

Copyright information

© International Osteoporosis Foundation and National Osteoporosis Foundation 2013