Quantitative computed tomography of the lumbar spine, not dual x-ray absorptiometry, is an independent predictor of prevalent vertebral fractures in postmenopausal women with osteopenia receiving long-term glucocorticoid and hormone-replacement therapy




To determine which measurement of bone mineral density (BMD) predicts vertebral fractures in a cohort of postmenopausal women with glucocorticoid-induced osteoporosis.


We recruited 114 subjects into the study. All had osteopenia of the lumbar spine or hip, as demonstrated by dual x-ray absorptiometry (DXA), and were receiving long-term glucocorticoids and hormone replacement therapy (HRT). Measurements of BMD by DXA of the lumbar spine, hip (and subregions), and forearm (and subregions), quantitative computed tomography (QCT) of the spine and hip (n = 59), and radiographs of the thoracolumbar spine were performed on all subjects to assess prevalent vertebral fractures. Vertebral fracture prevalence, as determined by morphometry, required a ≥20% (or ≥4-mm) loss of vertebral body height. Demographic information was obtained by questionnaire. Multiple regression and classification and regression trees (CART) analyses were used to assess predictors of vertebral fracture.


Twenty-six percent of the study subjects had prevalent fractures. BMD of the lumbar spine, total hip and hip subregions, as measured by QCT, but only the lumbar spine and total hip, as measured by DXA, were significantly associated with prevalent vertebral fractures. However, only lumbar spine BMD as measured by QCT was a significant predictor of vertebral fractures. CART analysis showed that a BMD value <0.065 gm/cm3 was associated with a 7-fold higher risk of fracture than a BMD value ≥0.065 gm/cm3.


In postmenopausal women with osteoporosis induced by long-term glucocorticoid treatment who are also receiving HRT, BMD of the lumbar spine as measured by QCT, but not DXA, is an independent predictor of vertebral fractures.

Glucocorticoids are widely used for the treatment of various inflammatory and autoimmune diseases, with more than 1 million prescriptions dispensed annually in North America (1, 2). Unfortunately, despite the efficacy of glucocorticoids in the treatment of inflammatory diseases, long-term therapy is associated with a significant number of side effects. Bone loss, resulting in an increased risk of vertebral and nonvertebral fractures, is one of the most serious complications (1, 2). Glucocorticoids affect bone metabolism in many different ways. They decrease the activity and life span of osteoblasts and increase the activity of osteoclasts; they act both directly and indirectly through alterations in calcium absorption and reduced gonadal hormone levels (1–3). Initially, trabecular bone, with its high surface area and high turnover rate, is most affected by glucocorticoids, with reports of an 8% loss of trabecular bone as measured by quantitative computed tomography (QCT) in patients initiating prednisone therapy at a dosage of 10 mg/day for 6 months (4).

Cross-sectional studies have shown that measurements of bone mineral density (BMD) by dual x-ray absorptiometry (DXA) and QCT predict incident fracture and are strongly associated with prevalent fracture (5–7). However, few studies have evaluated the ability of DXA or QCT to predict vertebral fractures in glucocorticoid-treated patients (8). Very little research has been performed to assess which anatomic sites or bone measurement devices would best predict the risk of a vertebral fracture in glucocorticoid-treated patients.

Since glucocorticoid treatment can rapidly reduce trabecular bone mass (1–3, 9), we hypothesized that QCT of the lumbar spine that measures only trabecular bone may be a better predictor of vertebral fracture risk in glucocorticoid-treated patients than DXA of the spine, which is an integral measurement of BMD. Therefore, the present study was performed to characterize the associations of DXA and QCT of the lumbar spine and hip with vertebral fracture in glucocorticoid-treated subjects.


Study subjects

The study subjects were 114 women who had been recruited for a longitudinal study to determine if human parathyroid hormone 1-34 could reverse glucocorticoid-induced bone loss (10). Subjects were recruited from local newspaper advertisements and were referred from local rheumatologists practicing in the San Francisco area of California. Postmenopausal women between the ages of 50 and 90 years who had a chronic, noninfectious inflammatory disease were eligible for the study if they had osteopenia (defined as low bone mass [≥2.0 SD below the mean in healthy young women, as measured in the lumbar spine and femoral neck or in the entire hip] by DXA), were currently taking hormone replacement therapy (HRT) (0.625 mg/day of Premarin [Wyeth-Ayerst Laboratories, Philadelphia, PA] or an equivalent dosage of another estrogen or selective estrogen receptor modulator) for ≥1 year, and had been treated with prednisone at a dosage of 5–20 mg/day for the previous 12 months and were expected to continue glucocorticoid therapy for at least 2 years.

Subjects were excluded from the study if they had osteoporosis secondary to any condition other than a rheumatic disease and glucocorticoid therapy, significant renal or hepatic dysfunction, or abnormalities on radiographs of the spine, including significant degenerative joint disease or multiple vertebral fractures, that would prevent accurate measurement of the lumbar spine at L1–L2 by QCT or at L1–L4 by DXA (10, 11). All subjects gave their informed consent. The study was approved by the Committee on Human Research of the University of California at San Francisco.

Data collection. Demographic information was collected by questionnaire at the first study visit. The following information was recorded: age, noninfectious inflammatory diseases requiring glucocorticoid treatment, duration of the chronic, noninfectious inflammatory diseases, number of years since menopause, family history of osteoporosis, current glucocorticoid dosage, duration of glucocorticoid therapy, average glucocorticoid dosage over the previous 12 months, cumulative glucocorticoid dose, other medical diseases, dosage and duration of estrogen therapy, and dosage and duration of calcium and vitamin D supplementation (10).

All 114 subjects underwent a physical examination, which included measurements of weight and height (by use of a Harbender stadiometer; Holtain, Dyfed, UK). All subjects underwent QCT (GE 9800 scanner; General Electric Medical Systems, Milwaukee, WI) measurement of the trabecular bone density of the lumbar spine at L1–L2. In addition, 59 of the subjects underwent a volumetric QCT scan of the hip, which allowed separate characterization of trabecular and cortical bone components (11). BMD of the lumbar spine (posteroanterior; L1–L4), the total hip (including subregions), and total forearm (including subregions) was measured by DXA, using a Hologic QDR 1000 (n = 51) or a Hologic QDR 4500 (n = 63) instrument (Hologic, Waltham, MA) (10–12).

Daily quality assurance data were collected from the DXA scanner to assess performance. The long-term in vivo precision error for both DXA instruments was 1% for the lumbar spine and total hip, 0.8–1.0% for the hip subregions, and 1% for the total forearm. The precision error for the QCT scanner was 1.5% for the lumbar spine (10, 11) and 0.6–0.8% for the integral and trabecular components of the hip (7).

Radiographs of the thoracolumbar spine were obtained by a standard technique and used to assess prevalent vertebral fractures. Vertebral fractures were defined as a ≥20% (or ≥4-mm) loss of vertebral body height and were confirmed by digitization of the involved vertebrae. Vertebral fractures identified by morphometry were classified as anterior wedge, midline or end-plate, and crush fractures (10–12).

Statistical analysis. Baseline demographic differences between the group of subjects with vertebral fractures and the group without vertebral fractures were tested for significance with Student's t-test for normally distributed variables. For categorical variables, either the chi-square test or, when necessary, Fisher's exact test was used. To ascertain which variables were independent predictors of vertebral fracture, a multivariate logistic regression analysis followed by a classification and regression trees (CART) analysis (using SyStat version 10 software; SPSS, Chicago, IL) was performed to determine the cutoff values of the independent predictors that best separated the subjects with a vertebral fracture from those without a vertebral fracture (13).


Characteristics of the study subjects at baseline. One hundred fourteen subjects who fulfilled the study inclusion criteria gave their informed consent to participate and were enrolled into the study. The demographic data for study subjects, dichotomized by the presence or absence of vertebral fractures, are shown in Table 1. The 2 groups were very similar in age, body mass index, dosage and duration of glucocorticoids, years postmenopausal, and years of estrogen therapy. The distributions of underlying chronic inflammatory diseases were similar in the 2 groups, with the majority of subjects in both groups having rheumatoid arthritis; however, systemic lupus erythematosus was present in 30% of the subjects with vertebral fractures and only 17% of those without vertebral fractures. Nearly 90% of the study subjects were being treated concurrently with glucocorticoid-sparing or disease-remittive agents for their glucocorticoid-requiring disease (data not shown).

Table 1. Baseline characteristics of the study subjects, by vertebral fracture group
CharacteristicAny vertebral fracture (n = 30)No vertebral fracture (n = 84)P
  • *

    Includes scleroderma, polymyalgia rheumatica, CREST syndrome (calcinosis, Raynaud's phenomenon, esophageal dysmotility, sclerodactyly, telangiectasias), and vasculitis.

Age, mean ± SD years62 ± 1162 ± 9.80.93
Body mass index, mean ± SD kg/m223.9 ± 4.025.6 ± 5.70.27
Prednisone (or equivalent) dosage, mean ±  SD mg/day7.5 ± 3.58.0 ± 4.40.60
Years since menopause, mean ± SD19.4 ± 10.716.3 ± 10.30.18
Years of estrogen therapy, mean ± SD11.8 ± 10.612.4 ± 12.70.81
Years of glucocorticoid therapy, mean ± SD13.6 ± 11.513.9 ± 10.70.91
Underlying disease, no. of subjects
 Rheumatoid arthritis11420.37
 Systemic lupus erythematosus9140.10
 Other rheumatic disease*8240.95

There were 128 prevalent vertebral fractures (from T4 to L5) in 30 of the 114 subjects; 23 of the 30 subjects had an average of 2 fractures, and 7 subjects had an average of ≥5 fractures. The distribution of prevalent vertebral fractures by morphometric classification is shown in Table 2. Fifty-six of the prevalent fractures were midline or end-plate fractures, 50 were anterior wedge fractures, and 22 were crush fractures.

Table 2. Distribution of prevalent vertebral fractures, by morphometric classification
Vertebral fracture typeNo. (%) of subjects (n = 114)
Midline or end-plate
 No fracture91 (80)
  1 fracture9 (39)
  2 fractures4 (17)
  3 fractures6 (26)
  ≥4 fractures4 (18)
Anterior wedge
 No fracture90 (79)
  1 fracture11 (10)
  2 fractures4 (4)
  3 fractures5 (4)
  ≥4 fractures4 (4)
 No fracture103 (90)
  1 fracture7 (6)
  2 fractures1 (1)
  3 fractures1 (1)
  ≥4 fractures2 (2)

BMD and vertebral fractures. BMD in the lumbar spine, total hip, femoral neck, and trochanteric region as measured by QCT was significantly lower in the fracture group compared with the nonfracture group (P = 0.0003, P = 0.007, P = 0.02, and P = 0.007, respectively). In contrast, there was no significant between-group difference in lumbar spine BMD as measured by DXA (P = 0.34) (Table 3). Total hip BMD measured by DXA was significantly different between the 2 groups. However, BMD in the subregions of the hip and in the total forearm and its subregion as measured by DXA were not significantly different between the fracture and nonfracture groups.

Table 3. BMD in postmenopausal women with glucocorticoid-induced osteoporosis, by vertebral fracture group*
BMD site, measurement techniqueAny vertebral fracture (n = 30)No vertebral fracture (n = 84)P
  • *

    Values are the mean ± SD. P values are 2-tailed. BMD = bone mineral density; DXA = dual x-ray absorptiometry; QCT = quantitative computed tomography.

  • A total of 59 subjects underwent QCT of the hip and subregions, 17 in the any vertebral fracture group and 42 in the no vertebral fracture group.

Lumbar spine
 DXA at L1–L4, gm/cm20.791 ± 0.160.806 ± 0.10.34
 QCT at L1–L2, gm/cm30.067 ± 0.020.096 ± 0.020.0003
Total hip
 DXA, gm/cm20.604 ± 0.280.716 ± 0.10.03
 QCT, gm/cm30.198 ± 0.020.223 ± 0.030.007
 QCT of trabecular bone, gm/cm30.078 ± 0.010.087 ± 0.010.01
Femoral neck
 DXA, gm/cm20.565 ± 0.080.587 ± 0.080.14
 QCT, gm/cm30.205 ± 0.030.228 ± 0.020.02
 DXA, gm/cm20.495 ± 0.090.509 ± 0.20.3
 QCT, gm/cm30.201 ± 0.020.225 ± 0.030.007
 DXA, gm/cm20.385 ± 0.1100.399 ± 0.090.52
 DXA one-third the distal radius, gm/cm20.575 ± 0.1050.591 ± 0.1070.48

This relationship persisted when the groups were analyzed by individual fracture type (anterior wedge, midline, and crush), with significant between-group differences in BMD of the spine and hip (including subregions) by QCT and no significant between-group difference in BMD of the spine by DXA (data not shown). However, since multiple fracture types occurred in the same subjects and since both anterior wedge and midline fractures were equally prevalent, we chose to present the data as any vertebral fracture versus no vertebral fracture.

Variables potentially associated with vertebral fractures were initially evaluated by multivariate logistic regression. The dependent variable was vertebral fractures, and the independent variables were as follows: age, weight, height, body mass index, family history of osteoporosis, duration of glucocorticoid therapy (years), average daily dose of glucocorticoids in the previous 12 months, total dose of glucocorticoids, years postmenopausal, duration of HRT (years), baseline BMD of the lumbar spine, hip, and forearm measured by DXA, and baseline BMD of the lumbar spine and hip measured by QCT. The only significant predictor of a vertebral fracture was BMD of the lumbar spine as measured by QCT.

Based on the lumbar spine BMD distribution as measured by QCT, we performed a CART analysis to identify a cutoff value for risk of any type of vertebral fracture. Results of the CART analysis showed that the risk of any type of vertebral fracture in these glucocorticoid-treated subjects with a BMD value by QCT that was <0.065 gm/cm3 was nearly 7 times higher than that in subjects with a BMD value by QCT that was ≥0.065 gm/cm3 (P < 0.0005). In addition, utilizing CART analysis, 0.065 gm/cm3 was the cutoff point for lumbar spine BMD values for all 3 morphometric vertebral fracture types (Table 4). The 0.065 gm/cm3 cutoff value identified by CART analysis had high specificity but low sensitivity. The positive predictive value was 65%, and the negative predictive value was 86%. This lumbar spine BMD value measured by QCT is therefore a better predictor of who will not have a fracture compared with who will have a fracture.

Table 4. Classification and regression trees analysis for values of lumbar spine BMD by QCT that best predict a vertebral fracture, by fracture type*
 Type of vertebral fracture
Anterior wedgeMidlineCrushAny type
  • *

    BMD = bone mineral density; QCT = quantitative computed tomography; OR = odds ratio; 95% CI = 95% confidence interval.

Lumbar spine BMD by QCT
 <0.065, gm/cm3116107710116
 ≥0.065, gm/cm3138213824911976
95% CI3.6–37.52.8–28.53.9–65.92.4–22.9
Sensitivity, %46436437
Specificity, %93929093


Vertebral fractures are common among postmenopausal women with glucocorticoid-induced bone loss. Midline (end-plate), anterior wedge, and crush fractures are the most common morphometric types of vertebral fractures. We found that BMD of the lumbar spine as measured by QCT was a significant predictor of prevalent vertebral fractures, whereas the posteroanterior DXA was unable to discriminate between the fracture and nonfracture groups. While some studies have shown that the underlying rheumatic disease itself can significantly increase the risk of vertebral fractures independently of glucocorticoid use (14), the distribution of these diseases was similar in both the fracture and nonfracture groups.

Lumbar spine posteroanterior DXA is commonly used to measure bone mass, diagnose osteoporosis, and predict the risk of vertebral fractures (5, 6). While some studies have demonstrated a significant correlation between lumbar spine DXA and prevalent vertebral compression fractures in postmenopausal women with osteoporosis (5, 6), several other studies have shown that there are limitations of posteroanterior DXA in discriminating and predicting fractures (15–17). Still other investigators have found trabecular BMD assessment by QCT to be a better and important predictor of vertebral fractures in postmenopausal women ages ≥60 years compared with lumbar spine DXA (7, 8). Our findings in postmenopausal women with glucocorticoid-induced bone loss are in agreement with these reports.

Lumbar spine DXA provides an areal estimate of integral BMD consisting of both cortical and trabecular bone. These areal estimates are dependent on bone size. However, QCT determines the true volumetric bone density independently of bone size, and it separates trabecular bone from the bone of the end plates and posterior elements, which are subject to hypertrophy and other degenerative distortions (8, 14–16). In addition, QCT can now provide separate measurements of trabecular and cortical bone at the spine and hip (11, 18).

Glucocorticoid-induced bone loss occurs initially and predominantly in the metabolically active trabecular bone, since trabecular bone has a higher turnover rate than cortical bone (3, 9). The vertebral bodies and the proximal femur are rich in trabecular bone (3, 9). Alterations in the trabecular bone microarchitecture caused by glucocorticoid treatment initially include a thinning of the trabecular plates, followed over time with perforation of the trabecular plates and a reduction in the number of trabeculae (3, 9).

Loss of trabecular bone structure is associated with an overall reduction in BMD and results in collapse or compression of the vertebrae. Measurement of lumbar spine BMD by QCT determines only the trabecular mineral content of the vertebral bodies, excluding the posterior vertebral elements as well as any extravertebral calcifications. It is therefore a more sensitive measure of the trabecular bone that is predominantly lost with glucocorticoid therapy.

Using CART analysis, we found a vertebral fracture threshold value of 0.065 gm/cm3 for lumbar spine BMD by QCT. This is very similar to the vertebral fracture threshold value of 0.072 gm/cm3 for spine BMD by QCT described by Lafferty et al (7). However, those investigators used a slightly more stringent criterion of ≥30% loss of vertebral height to define vertebral fractures. The fracture threshold value we identified is also consistent with the results reported by Cann et al (18). In their cross-sectional study of women ages 50–80 years with and without fractures, as determined by lumbar spine QCT, they found that 41% of subjects with fractures and 97% of subjects without fractures had vertebral mineral values >0.070 gm/cm3. They concluded that individuals with a lumbar spine QCT threshold value <0.060–0.07 gm/cm3 had a high risk of vertebral compression fractures, while those with a value of >0.110 gm/cm3 are unlikely to sustain a fracture.

Univariate analysis showed that BMD of the hip as measured by DXA and QCT was significantly associated with vertebral fractures. However, when the multivariate analysis was performed, DXA of the hip was no longer a significant predictor. Sambrook et al (19) reported that hip BMD as well as spine BMD as measured by dual-photon absorptiometry was reduced in glucocorticoid-treated subjects (19). Cummings et al (5) also reported that in postmenopausal Caucasian women, hip BMD measurement by DXA is also a strong predictor of vertebral fractures. Since only half of our subjects had hip BMD measurements by QCT, additional studies of hip BMD by QCT will need to be performed to determine what additional information these measurements provide in the assessment of vertebral fracture risk.

This study has several limitations. First, this was a cross-sectional study of postmenopausal women who were recruited because they were taking long-term estrogen and glucocorticoid therapy, and they had osteopenia identified by DXA of the lumbar spine and hip. Since the osteopenia may be the result of both the estrogen deficiency and the glucocorticoid use, these findings may not be generalizable to other glucocorticoid-treated populations. Because this was a cross-sectional study evaluating predictors of vertebral fractures at baseline, information about incident fractures could not be ascertained. In addition, our results cannot be compared with those of other trials that did not use the same definition of prevalent fractures (≥20% [or ≥4-mm] loss of vertebral body height, a definition developed by Black et al [ 12]). We divided vertebral fractures into morphometric types (anterior wedge, midline, and crush) and evaluated predictors of these fracture types both individually and together. However, since multiple types of fractures occurred in the same subjects, the results are potentially confounded and highly correlated with each other. We did not perform adjusted analyses, because the numbers of subjects were small. Last, the number of subjects in the entire study population was small (n = 114). However, we had statistical power of >80% to assess an association between both hip and lumbar spine BMD by DXA and prevalent vertebral fractures.

In summary, we found that glucocorticoid-induced osteoporosis in postmenopausal women was associated with a high prevalence of vertebral fractures. BMD of the lumbar spine as measured by QCT, but not by DXA, was found to be an independent predictor of all types of vertebral fractures. Therefore, physicians may want to consider obtaining a BMD measurement of the lumbar spine by QCT to assess vertebral fracture risk in postmenopausal women being treated with glucocorticoids. Further studies are needed to determine the correlation between DXA and QCT measurements of lumbar spine BMD to allow physicians to better predict which of their patients is at risk of fractures and would benefit from medical intervention.