Osteoporosis International

, Volume 20, Issue 11, pp 1889–1894

Bone fragility in male glucocorticoid-induced osteoporosis is not defined by bone mineral density

Authors

  • K. Hayashi
    • Internal Medicine 1Shimane University Faculty of Medicine
    • Internal Medicine 1Shimane University Faculty of Medicine
  • Y. Murakawa
    • Internal Medicine 3Shimane University Faculty of Medicine
  • M. Yamauchi
    • Internal Medicine 1Shimane University Faculty of Medicine
  • H. Kaji
    • Division of Diabetes, Metabolism, and Endocrinology, Department of Internal MedicineKobe University Graduate School of Medicine
  • T. Yamaguchi
    • Internal Medicine 1Shimane University Faculty of Medicine
  • T. Sugimoto
    • Internal Medicine 1Shimane University Faculty of Medicine
Original Article

DOI: 10.1007/s00198-009-0901-6

Cite this article as:
Hayashi, K., Yamamoto, M., Murakawa, Y. et al. Osteoporos Int (2009) 20: 1889. doi:10.1007/s00198-009-0901-6

Abstract

Summary

Eighty-seven male Japanese subjects taking prednisolone ≥5 mg for more than 6 months and 132 age- and body mass index (BMI)-matched control subjects were examined. Multiple regression analysis adjusted for age and BMI showed that spinal bone mineral density (BMD) in the prednisolone group was not associated with prevalent vertebral fractures (VFs).

Introduction

Glucocorticoid (GC) treatment is known to increase the risk for bone fractures. However, the association between VFs and BMD in GC-treated male patients remains unclear.

Methods

Eighty-seven male subjects taking prednisolone ≥5 mg for more than 6 months and 132 age- and BMI-matched control subjects were examined using lateral thoracic and lumbar spine radiographs and spine dual energy X-ray absorptiometry.

Results

The presence of GC use was an independent risk factor for VFs adjusted for age and BMI (odds ratio 10.93, P < 0.001). By receiver operating characteristic analysis, the absolute BMD values for detecting VFs were higher and the sensitivity and specificity were lower in the GC group than in the control group (0.936 vs 0.825 g/cm2 and 53.5% vs 74.0%, respectively). Multiple regression analysis adjusted for age and BMI showed that spinal BMD in the GC group was not associated with prevalent VFs, even after adding current and past maximum GC doses as independent variables.

Conclusions

These results show that lumbar BMD values are not associated with prevalent VFs in GC-treated male patients, suggesting that bone fragility in male GC users is affected by bone quality rather than by BMD.

Keywords

Bone mineral densityGlucocorticoidMenOsteoporosisVertebral fractures

Introduction

Oral glucocorticoids (GC) have been used for the treatment of a range of autoimmune, respiratory, hematological, and renal diseases. A meta-analysis showed that patients treated with GC had a higher risk of fractures of the hip, vertebra, forearm, or any other site compared to control subjects [1]. Low bone mineral density (BMD) is a key predictor of fractures related to primary osteoporosis in both men and women [2]; however, vertebral fractures (VFs) occurred at much higher rates than expected on the basis of BMD [1], suggesting that the bone fragility of GC users is not defined by BMD. Indeed, the cutoff values of BMD at the lumbar spine (L), femoral neck (FN), and one third of radius (R) for VFs in female Japanese patients treated with GC were higher than those of controls in our previous study [3]. However, the relationship between BMD and fracture risk in GC-treated male patients is still unknown.

Several clinical aspects of male osteoporosis differ from those of female osteoporosis. Approximately half the male patients with osteoporosis presenting with symptomatic VFs have identifiable secondary causes [4]; GC therapy is a common cause of osteoporosis in older men [5]. The prevention of fractures in men seems to be much more important than in women because male mortality rates within 1 or 2 years after a fracture are higher than in females [6, 7]. In this study, to clarify the relationship between BMD and bone fragility in male GC users, we measured L-BMD in male GC users and age- and body mass index (BMI)-matched control subjects and compared the values of BMD Z scores between subjects with and without VFs. We also compared BMD cutoff levels and odds ratios (ORs) for VFs between GC users and controls.

Subjects and methods

Subjects

We enrolled eighty-seven Japanese male patients (age range: 20–85 years, mean 52.1 years; BMI: 16.3–29.7 kg/m2, mean 22.4 ± 2.8 kg/m2) who were diagnosed with autoimmune disease, except for rheumatoid arthritis, who were treated with an oral GC [prednisolone (PSL) ≥ 5 mg/day] for more than 6 months (GC group), and who underwent L-BMD measurements at the outpatient clinic of Shimane University Hospital. We excluded patients with renal dysfunction who had serum creatinine levels higher than the normal range (0.56–1.23 mg/dl) and patients whose performance status was disturbed. We investigated patients’ history of GC treatment with regard to the duration of GC treatment and the current and past maximum dose. No patients were taking any drugs or hormones that are known to affect bone metabolism, including sex steroids, warfarin, or bisphosphonates. Basal diseases of the GC-treated patients were systemic lupus erythematosus (11 cases), polymyositis (11 cases), Behcet’s syndrome (seven cases), ulcerative colitis (seven cases), dermatomyositis (four cases), myasthenia gravis (three cases), and others (dermatological diseases, hematological diseases, and granulomatous and respiratory diseases). For the control group, 132 age- and BMI-matched men (age range: 22–86 years, mean 53.4 years; BMI: 14.0–32.9 kg/m2, mean 22.9 ± 3.7 kg/m2), who were suffering from lumbago or suspected osteoporosis without underlying conditions affecting the skeleton, were recruited. To verify that patients in the GC group have a higher risk of VFs than control subjects (expected fracture rate: GC group, 30%; control group, 10%, respectively), 122 control and 81 GC subjects were needed to reach the sufficient statistical power (α = 0.05 and 1-β = 0.8). No patients were taking drugs or hormones that affect bone metabolism, including sex steroids and calcitonin. Baseline characteristics of subjects are shown in Table 1. The mean current dose and past maximum dose of PSL were 11.1 ± 9.8 and 41.5 ± 21.0 mg/day, respectively, in the GC group. This study was cross-sectional, was approved by the ethical review board of our institution, and was in compliance with the Helsinki declaration. All subjects agreed to participate in the study and gave written informed consent.
Table 1

Background data in control and GC-treated groups

Control group

GC group

P

No. of subjects

132

87

 

No. of vertebral fractures

11 (8.3%)

30 (34.5%)

<0.001*

Age (years)

53.4 ± 18.4

52.1 ± 16.9

 

BMI (kg/m2)

22.9 ± 3.7

22.4 ± 2.8

 

L-BMD (g/cm2)

0.913 ± 0.158

0.941 ± 0.172

0.216

Z score

−0.41 ± 1.09

−0.31 ± 1.28

0.531

T score

−1.11 ± 1.32

−0.89 ± 1.45

0.257

Current dose of PSL (mg/day)

 

11.1 ± 9.8

 

Maximum dose of PSL (mg/day)

 

41.5 ± 21.0

 

Duration of GC treatment (months)

 

75 ± 83

 

GC glucocorticoids, BMI body mass index, L lumbar, PSL prednisolone

*P < 0.05, Chi square test

Densitometry

BMD values were measured by dual-energy X-ray absorptiometry using the QDR-4500 DEXA system (Hologic, Waltham, MA, USA) at L in all GC-treated and control subjects, and BMD measurements of the FN and R were performed in 73 and 58 GC subjects, respectively. Vertebrae with fractures or overt osteoarthritis were excluded from the analysis of L-BMD because these factors may increase BMD through artifacts. BMD was automatically calculated from the bone area (cm2) and bone mineral content (g). Values were also expressed relative to the standard deviation (SD) of age- and sex-matched normal Japanese mean values of BMD provided by the manufacturer (BMD Z score) [8]. The coefficients of variation of the L, FN, and R measurements were 1.0, 1.0, and less than 1%, respectively.

Assessment of fractures

In all subjects, conventional thoracic and spinal radiographs in lateral and antero-posterior projections were obtained. Following the classification by Genant et al. [9], VFs were diagnosed if a reduction of 20% or more was observed by two investigators who were blinded to each others’ readings. If judgment of VFs did not agree, the film was independently reassessed. If the reevaluated findings were again different, we regarded that case as a nonfracture.

Statistical analysis

All data are expressed as the mean ± SD for each index. The Mann–Whitney U test was used to compare parameters between subjects with and without VFs. P values < 0.05 were considered significant. To compare the strength of association between BMD values at each measurement point and the presence of VFs, we analyzed the areas under the receiver operating characteristic (ROC) curves [10]. For each of the BMD measurements at the L and for each of the VF groups, possible cutoff points for BMD were defined, and the proportion of subjects with fractures below these points (the sensitivity) and the proportion of subjects without fractures above these points (the specificity) were calculated. This calculation yields an ROC curve that illustrates the relationship between sensitivity and specificity for each BMD measurement as a discriminator between the normal and fracture groups. Statistical analyses were performed using the computer program StatView for Windows, version 5.0 (SAS Institute, Cary, NC, USA).

Results

Background data

Background data for the control and GC-treated subjects are presented in Table 1. There were 11 (8.3%) subjects with VFs in the control group and 30 (34.5%) in the GC group. The prevalence of VFs in the GC group was significantly higher than that of controls (P < 0.001, chi-square test); however, there was no difference in L-BMD between the control and GC groups. Multivariable logistic analysis adjusted for age, BMI, and L-BMD revealed that GC use was an independent risk factor for VFs [OR 10.93, 95% confidence interval (CI) 4.36–27.40, P < 0.001].

Comparison of various parameters between subjects with and without VFs in the GC group

We compared various parameters of GC patients with and without VFs. As shown in Table 2, patients with VFs were significantly older than those without VFs. There was no difference in other parameters, including L-BMD, between those with and without VFs.
Table 2

Comparison of various parameters between men with and without vertebral fractures in GC-treated group

GC group

Vertebral fractures

P

No

Yes

No. of subjects

57

30

 

Age (years)

48.7 ± 17.8

58.4 ± 13.5

0.029*

BMI (kg/m2)

22.4 ± 2.9

22.4 ± 2.6

0.978

L-BMD (g/cm2)

0.949 ± 0.148

0.927 ± 0.211

0.553

Z score

−0.315 ± 1.149

−0.287 ± 1.513

0.837

Current dose of PSL (mg/day)

9.9 ± 6.9

13.4 ± 13.6

0.420

Maximum dose of PSL (mg/day)

41.3 ± 21.2

41.0 ± 20.7

0.989

Duration of GC treatment (months)

61.8 ± 67.4

96.9 ± 104.8

0.177

GC glucocorticoids, BMI body mass index, L lumbar, PSL prednisolone

*P < 0.05, Mann–Whitney U test

Distribution of BMD in subjects with and without VFs

Figure 1 shows the distribution of L-BMD in all subjects as a function of age. In control subjects, those with VFs (black dots) are clearly grouped in the region with higher age and lower BMD. In contrast, subjects with VFs in the GC group were not associated with age or BMD.
https://static-content.springer.com/image/art%3A10.1007%2Fs00198-009-0901-6/MediaObjects/198_2009_901_Fig1_HTML.gif
Fig. 1

Distribution of BMD in men with and without vertebral fracture. White dots represent subjects without vertebral fractures, and black dots represent subjects with vertebral fractures. GC, glucocorticoids; VF, vertebral fracture

Cutoff values for VF

The cutoff value of L-BMD for VFs in the GC group was higher than that of controls (GC group vs control: 0.936 vs 0.825 g/cm2) (Table 3). The sensitivity and specificity of ROC analysis in the GC group were lower than those of the control group (53.5% vs 74.0%).
Table 3

Comparison of various parameters between men with and without vertebral fractures in GC-treated group

Independent variables

Cutoff value

BMD (g/cm2)

T score

(%)

Sensitivity (%)

Specificity (%)

Control group

L-BMD

0.825

−1.85

(79)

74.0

74.0

GC group

L-BMD

0.936

−1.03

(89)

53.5

53.5

Cutoff values as well as sensitivity and specificity were calculated by ROC analysis.

GC glucocorticoids, BMI body mass index, L lumbar, PSL prednisolone

Multivariable logistic regression analysis with the presence of VFs

Finally, we performed multivariable logistic regression analysis with the presence of VFs as the dependent variable and L-BMD adjusted for age and BMI as the independent variable. L-BMD in the control group was significantly associated with the presence of VFs (OR = 0.26, 95 % CI 0.09–0.74 per SD increase, P = 0.012). In contrast, there was no significant association in the GC group between VFs and L-BMD, even after adding current or past maximum PSL dose or duration of GC treatment as independent variables (Table 4).
Table 4

Association between the presence of vertebral fractures and L-BMD in male control and GC group

Independent variables

Presence of vertebral fractures

OR

(95% CI)

P

Control group

   

L-BMD

0.26

(0.09–0.74)

0.012

GC group

   

L-BMD

0.77

(0.47–1.26)

0.298

Multivariate logistic regression analysis adjusted for age and BMI. Unit of change, per SD increase.

GC glucocorticoids, BMI body mass index, L lumbar, PSL prednisolone

Discussion

This study revealed that the VF rate in the GC group was significantly higher than that of age-matched controls and that GC use was an independent risk factor for VFs when adjusted for age, BMI, and L-BMD. It has been established by meta-analysis that GC users have an increased risk of fractures compared to controls [1]. Kanis et al. reported that exposure to corticosteroids was associated with a significantly increased risk of fractures in both genders [11], which is compatible with our findings that male GC users had a higher risk of fractures than control subjects.

We also showed that age-matched BMD (BMD Z score) did not differ between GC-treated patients with and without VFs. The cutoff value of L-BMD for VFs in the GC group was higher compared to that of the control group. In addition, the sensitivity and specificity of cutoff values for VFs in the GC group were lower than those of the control group. These results are in agreement with our previous report on female GC-treated patients [3]. However, the cutoff value of L-BMD T score for VFs in male GC users was higher than that in female users (−1.03 vs −1.88), and the sensitivity and specificity of the cutoff L-BMD value in men were lower than those of the cutoff L-BMD value in women (53.5% vs 61.5%), suggesting that assessing bone fragility by measuring BMD is more difficult for male patients taking GCs than for female GC users.

Several reports have also noted that L-BMD is not sensitive enough to assess the risk of VFs. Orwoll et al. suggested that osteophytic calcification exerted an important influence on the measurement of L-BMD in men because the presence of osteophytes obscured the relationship of BMD to age [12]. The prevalence of osteoporosis became lower when estimated by L-BMD instead of by FN-BMD in both genders [13], suggesting that L-BMD was less sensitive for the diagnosis of osteoporosis. Thus, FN- or R-BMD might be more suitable measures for evaluating the risk of VFs in GC users than L-BMD. We compared the FN-BMD values of 132 control subjects and 73 GC users, as well as the R-BMD of 132 controls and 58 GC subjects. Compared to normal subjects, cutoff values of FN- and R-BMD for VFs in the GC group were higher than those of the control group (FN: 0.724 vs 0.644 g/cm2; R: 0.726 vs 0.621 g/cm2), and the sensitivity and specificity in the GC group were both low (FN: 52.0 vs 75.2%; R: 56.0 vs 84.3%). Thus, FN- and R-BMD seem to give results similar to those found by L-BMD. Taken together, these results indicate that it is difficult to assess bone fragility in male GC-treated patients by measurement of BMD at any site, suggesting that bone strength of GC users was associated with bone quality rather than BMD.

The pathophysiologic mechanisms of impaired bone quality in GC users are still unknown. Bone quality is determined by architecture, turnover, microdamage accumulation, mineralization, and bone matrix proteins, such as collagen [14]. In vivo studies have shown that GC administration caused low bone turnover by the suppression of osteoblast function, the induction of apoptosis in osteoblasts [15], and a major loss of trabecular connectivity [16]. GC also affected bone geometry by reducing bone formation rate on the periosteal surface [17]. These results demonstrated that GC administration led to deterioration in bone structural properties. A limited report about the effect of GC treatment on bone material properties indicated that steroid administration reduced bone mineralization and elastic modulus of bone surrounding osteocyte lacunae [18]. These skeletal factors might have led to the observed bone fragility of GC users in this study.

In this study, the current and past maximum dose were not significantly different between subjects with and without VFs in the GC group. Van Staa et al. revealed that the increased risk of fractures in GC users was related to the daily dose of GC [19]. This discrepancy might partly be explained by insufficient statistical power due to the smaller number of patients in this study.

This study has some limitations. First, the study was not population-based and the sample size was not large enough to draw definite conclusions. Second, we only analyzed the control subjects who attended Shimane University Hospital, a tertiary care center, for evaluation or treatment of back pain or osteoporosis. Therefore, the control subjects enrolled in this study might have had a relatively severe case of the disorder and might not be representative of normal Japanese men. Third, we did not confirm that all of GC-treated men did not have VFs when GC was administrated. Several underlying autoimmune diseases would negatively affect bone health and compound the effect of GCs. Osteopenia and osteoporosis in systemic lupus erythematosus are associated more closely with increased disease duration than cumulative GC use [20]. Indeed, approximately 30% of patients enrolled into the Risedronate GIOP prevention study had VFs at baseline [21]. These reports suggested that inflammatory diseases themselves are risk factors for osteoporosis and that all VFs in this study were not caused by GC use alone. Finally, hypogonadism is the most common cause of osteoporosis, and low testosterone levels are associated with VFs in men [14, 22]. However, we did not assess dysfunction of the hypothalamus–pituitary–gonadal axis in this study.

In conclusion, this report shows that the prevalent fracture rate and the cutoff value of BMD for VFs in male GC-treated subjects are higher than those in control subjects and that BMD values could not distinguish GC-treated subjects with fractures from those without fractures, suggesting that bone strength in male GC users would be defined by bone quality rather than bone density. Further studies are needed to clarify how GC administration influences bone quality.

Funding

No funding was received.

Conflicts of interest

None.

Copyright information

© International Osteoporosis Foundation and National Osteoporosis Foundation 2009