Osteoporos Int (2007) 18:761770 DOI 10.

1007/s00198-006-0306-8

ORIGINAL ARTICLE

Enhanced prediction of fracture risk combining vertebral fracture status and BMD
E. S. Siris & H. K. Genant & A. J. Laster & P. Chen & D. A. Misurski & J. H. Krege

Received: 27 July 2006 / Accepted: 21 November 2006 / Published online: 24 January 2007 # International Osteoporosis Foundation and National Osteoporosis Foundation 2007

Abstract Summary Prevalent vertebral fractures are associated with increased fracture risk, but the magnitude of this effect across a range of BMD T-scores has not been quantified. In this analysis, for any given BMD T-score, incident fracture risk varied up to twelve fold when information regarding prevalent radiographic vertebral fracture status was considered.
This study was supported by Eli Lilly and Company. E. S. Siris (*) Metabolic Bone Disease Program, Toni Stabile Osteoporosis Center, Columbia University Medical Center, 180 Fort Washington Avenue, New York, NY 10032-3784, USA e-mail: es27@columbia.edu H. K. Genant Department of Radiology and Medicine, University of California San Francisco and Synarc, Inc, San Francisco, CA, USA e-mail: harry.genant@ucsf.edu A. J. Laster Arthritis and Osteoporosis Consultants of the Carolinas, Charlotte, NC, USA e-mail: ajlaster@aocc.md P. Chen : D. A. Misurski : J. H. Krege Eli Lilly and Company, Indianapolis, IN, USA P. Chen e-mail: chenpe@lilly.com D. A. Misurski e-mail: misurskida@lilly.com J. H. Krege e-mail: kregejh@lilly.com

Background Clinical fracture risk evaluation of older women usually includes assessment of bone mineral density (BMD) but often not vertebral fracture status. In this analysis, we quantified the impact of vertebral fracture burden on two year fracture risk across a range of BMD T-scores. Methods Data were from 2,651 postmenopausal women who were assigned to the placebo groups of the Fracture Prevention Trial (median observation 21 months) and the Multiple Outcomes of Raloxifene Evaluation Trial (MORE; observation 2 years). Using the Genant visual semiquantitative criteria, we defined prevalent vertebral fracture status as: a) presence or absence of fracture; b) fracture number; c) maximum semi-quantitative (SQ) score (normal=0, mild fracture=1, moderate fracture=2, severe fracture=3); and d) spinal deformity index (SDI) score (sum of SQ scores of T4 to L4 vertebrae). Incident fractures over two years were identified via lateral spine radiographs and outside the spine by questioning of patients and review of radiographs or radiographic reports. Results Femoral neck BMD T-score provided significant information regarding fracture risk. Across the range of T-scores, vertebral fracture status provided additional prognostic information. The risk increased with increasing number and severity of prevalent vertebral fractures and SDI, a summary measure of spine fracture burden. Across a range of BMD values, prevalent spine fracture burden as assessed by SDI increased the risk of incident vertebral fractures by up to 12-fold, nonvertebral fractures by about twofold, and any fractures by up to sevenfold. Conclusions These findings indicate that at any given BMD T-score, the risk of incident vertebral, non-vertebral, and any fracture depended heavily on prevalent radiographic vertebral fracture status. Assessment of vertebral fracture

762

Osteoporos Int (2007) 18:761770

status, in addition to BMD, provides practical and relevant clinical information to aid in predicting fracture risk in postmenopausal women. Keywords Non-vertebral fracture . Osteoporosis . Semiquantitative assessment . Spine deformity index . Spine radiograph . Vertebral fracture . Vertebral fracture assessment

Introduction The prevalence of vertebral fracture varies according to the criteria used to define a fracture and the population studied. In all populations, the prevalence of vertebral fracture increases with age and can exceed 50% in women age 85 years and older [1]. Many studies have demonstrated that prevalent vertebral fractures increase the risk of new vertebral fractures in postmenopausal women [26]. Future vertebral fracture risk is positively associated with the number of prevalent vertebral fractures [5, 7, 8], and the severity of the vertebral deformity [5, 8, 9]. Despite the value of having information about prevalent vertebral fractures, spine imaging is not typically performed in the clinical evaluation of postmenopausal women at risk for future fractures [10]. In this population, only one of three vertebral fractures is likely to present with clinical symptoms [11], and therefore come to clinical attention. Barriers to routine spine imaging by plain radiographs include cost, radiation exposure, accessibility and patient inconvenience. Another barrier to spine imaging in patients presenting with low bone mass or osteoporosis based upon bone mineral density (BMD) testing might be insufficient information regarding the degree to which spine fracture burden offers future fracture risk prediction beyond that provided by BMD assessment alone. Therefore, this analysis was directed at quantifying the fracture risk information conferred by vertebral fracture status beyond BMD. To do this, the impact of vertebral fracture status on vertebral, non-vertebral fragility, and any fracture risk was assessed across a range of BMD T-scores in postmenopausal women.

were randomized to receive daily self-administered, subcutaneous injections of placebo, teriparatide 20 mcg or teriparatide 40 mcg. Daily calcium (1000 mg) and vitamin D (400 1200 IU) supplementation was provided. In the placebo group of this trial, 448 of 544 (82.4%) women had adequate radiographic vertebral assessments at baseline and study endpoint. In the MORE Trial, ambulatory women who had been postmenopausal for at least two years were randomized to receive daily oral placebo, raloxifene 60 mg or raloxifene 120 mg. All women received daily calcium supplements of 500 mg and 400 to 600 IU of vitamin D. From the placebo group of this trial, 2,203 of 2,292 (96.1%) women had adequate radiographic vertebral assessments at baseline and at two years. Of these 2,203 women, 534 (24.2%) were osteopenic. The median observation period was 21 months in the Fracture Prevention Trial. Two-year MORE trial data were used for this analysis. The median observation period for the combined cohort was therefore 23 months. Fracture diagnosis In the Fracture Prevention Trial, lateral thoracic and lumbar spine radiographs were obtained at baseline and study endpoint and assessed at a central site (Osteoporosis and Arthritis Research Group, University of California San Francisco, San Francisco, CA, USA) by radiologists blinded to group assignment, but not to temporal sequence. Fracture severity was assessed using the visual semiTable 1 Baseline characteristics of women in the placebo groups of the MORE and fracture prevention trials (meanSD) who were included in this analysis1 MORE trial (2-year) (n=2203) Fracture Prevention Trial (n=448) 69.16.9 21.08.3 157.46.4 26.74.7 2.51.5 2.30.8 2.31.8 1.70.9 3.63.0 Combined (n=2651)

Age (years) Years postmenopausal Height (cm) BMI (kg/m2) Lumbar spine T-score Femoral neck T-score Number of prevalent vertebral fractures Maximum SQ score SDI score
1

66.57.0 18.78.3 159.06.5 25.23.9 2.61.1 2.30.5 0.731.4 0.610.9 1.10.9

67.07.0 19.08.4 158.8 6.5 25.54.1 2.61.2 2.30.6 0.981.5 0.801.0 1.62.7

Methods Study participants For this analysis, subjects were women from the placebo groups of the Fracture Prevention and Multiple Outcomes of Raloxifene Evaluation (MORE) Trials. Results of these trials have been previously published [2, 12]. In the Fracture Prevention Trial, ambulatory women with osteoporosis who had been postmenopausal for at least five years

All women had lumbar spine BMD measurements and spine radiographs at baseline and post-baseline. BMD, bone mineral density; BMI, body mass index; SQ, semi-quantitative; SDI, spine deformity index.

Osteoporos Int (2007) 18:761770 Fig. 1 The relationship between femoral neck T-score and the 23-month vertebral fracture risk without consideration of prevalent vertebral fracture status (a); considering presence versus absence of prevalent vertebral fracture(s) (b); considering maximum prevalent SQ (semiquantitative) vertebral deformity score (c); considering number of prevalent vertebral fractures (d); and considering prevalent SDI (spine deformity index) score (e)
Absolute Vertebral Fracture Risk (%) Absolute Vertebral Fracture Risk (%)
70 60 50 40 30 20 10 0 70 60 50 40 30 20 10 0

763

Women With and Without Prevalent Vertebral Fracture(s) (Entire Cohort)

b
Women With Prevalent Vertebral Fracture(s)

Women Without Prevalent Vertebral Fracture(s)

-1

-2

-3

-4

-5

-6

-1

-2

-3

-4

-5

-6

Femoral Neck T-Score Absolute Vertebral Fracture Risk (%) Absolute Vertebral Fracture Risk (%)
70 60 50 40 70 60 50 40 30 20 10 0

Femoral Neck T-Score SQ Score 3 2

Number of Prevalent Vertebral Fractures >3 2 1 0

30 20 10 0

1 0

-1

-2

-3

-4

-5

-6

-1

-2

-3

-4

-5

-6

Femoral Neck T-Score Absolute Vertebral Fracture Risk (%)

70 60 50 40 30 20 10 0

Femoral Neck T-Score SDI Score 7 6 5 4 3 2 1 0

-1

-2

-3

-4

-5

-6

Femoral Neck T-Score

quantitative vertebral deformity (SQ) scoring system of Genant [13, 14]. The SQ score represents an accurate and reproducible method of assessing fracture severity that has been tested and applied in many clinical trials and epidemiologic studies [1416]. SQ scores were assigned to each individual vertebra from T4 to L4 [12]. An SQ score of 0 was assigned to normal, non-fractured vertebra; 1 for a mild deformity (2025% reduction in anterior, middle or posterior vertebral height); 2 for a moderate deformity (2540% reduction); 3 for a severe deformity (>40% reduction). Discrepancies between the screening and final score for baseline radiographs were reconciled by consensus with a second radiologist. Vertebrae that were not radiographically evaluable due to kyphosis, fusion, or

other anomalies were not graded. A new vertebral fracture was recorded if a normal vertebra (grade 0) became deformed (grade 1). In the MORE Trial, radiographs were obtained at baseline, and after two, three, and four years. Because follow-up lateral spine radiographs in the Fracture Prevention Trial were obtained after a median of 21 months, the 2-year MORE Trial results were included in the current analysis to facilitate pooling of the data from these two trials. Radiographs were first assessed using the same visual SQ scoring system at the same central site described above. A score change from 0 to at least 1 was required for an incident fracture to be diagnosed. If neither a prevalent nor an incident fracture was identified, no

764

Osteoporos Int (2007) 18:761770

Table 2 Impact of prevalent vertebral fracture status on vertebral fracture risk at various femoral neck T-scores in postmenopausal women with osteoporosis Prevalent vertebral fracture Femoral neck T-Score 2 3 4 No Yes Prevalent vertebral fracture number 3 20.7 (8.6) 29.1 (7.9) 39.2 (6.9) Maximum prevalent SQ score Prevalent SDI score

2 13.3 (5.5) 19.4 (5.2) 27.4 (4.8)

2 13.2 (5.5) 18.8 (5.1) 26.3 (4.8)

3 23.0 (9.6) 31.5 (8.5) 41.3 (7.5)

3 8.7 (3.6) 12.4 (3.4) 17.4 (3.3)

5 20.7 (8.6) 27.9 (7.8) 36.4 (7.0)

7 29.3 (12.2) 38.0 (10.5) 47.9 (9.2)

2.3 12.3* (5.3) 3.8 18.9 (5.0) 6.2 28.0 (4.5)

2.4 7.2 (3.0) 3.7 10.9 (2.9) 5.7 16.1 (2.8)

2.4 7.4 (3.1) 3.7 10.9 (2.9) 5.5 15.8 (2.9)

2.4 6.9 (2.9) 3.6 9.8 (2.7) 5.2 13.9 (2.7)

*Fracture risk was derived from the logistic regression model. SQ, semi-quantitative; SDI, spine deformity index. Values shown are the absolute vertebral fracture risk in percent over a median observation period of 23 months. Relative risks for fracture in patients with prevalent vertebral fracture(s) versus those of patients with the same BMD but not having prevalent vertebral fracture are provided in parentheses.

further analyses were performed. For fractures observed at baseline or endpoint, a second radiologist determined whether a fracture was present and performed a quantitative morphometric (QM) analysis for each vertebra. For the QM analysis, an incident fracture was defined as a decrease in the anterior, mid, or posterior height of 20%, with a minimum change of 4 mm. At baseline, adjacent vertebrae were referenced for comparison. A fracture was reported if the SQ grade was greater than 0 in at least one vertebra and at least two of three techniques were in agreement. The SDI for a subject was defined as the sum of the individual vertebral deformity scores. By combining both the number and severity of vertebral fractures, the SDI is a summary measure of spine fracture burden and has proven to be predictive of future vertebral fracture risk [17, 18]. In both trials, patients were queried regarding the occurrence of non-vertebral fractures at each follow-up visit. Traumatic fractures and those involving the fingers, metacarpals, toes, face, and skull were excluded from the analysis. New non-vertebral fractures were confirmed by review of radiographs or radiologic reports [2, 12]. In this analysis, risk of any fracture refers to the risk of a patient experiencing a vertebral and/or a nonvertebral fragility fracture. BMD assessment In both trials, BMD was assessed by DXA using Hologic (Hologic Corp., Waltham, MA, USA), Norland (Norland Corp., Ft. Atkinson, WI, USA), and GE Lunar equipment (GE Lunar Corp., Madison, WI, USA). BMD analyses were assessed at a central site (Osteoporosis and Arthritis Research Group, University of California San Francisco, San Francisco, CA, USA). For this analysis, BMD T scores

at lumbar spine and femoral neck were considered. BMD values were converted to standardized units (expressed as milligrams per square centimeter) to eliminate differences attributable to the type of densitometer [19]. Statistical analysis A logistic regression model was used to relate the risk of a new fracture (vertebral, fragility non-vertebral, and any fracture) at two years (yes/no) to baseline prevalent vertebral fracture status, baseline femoral neck BMD, years postmenopausal, and age. We also examined the relationship between the femoral neck BMD T-score and fracture risk without considering the prevalent vertebral fracture status (i.e., risk assessment for the entire cohort based on femoral neck BMD T-score alone). To pool the data from two trials, the model included the interaction effects of baseline prevalent fracture status-by-trial, and baseline T-score-by-trial at the 10% level of significance. In the analyses of vertebral and any fractures prevalent vertebral fracture (no/yes), SQ score (0, 1, 2, 3), prevalent vertebral fracture number (0, 1, 2, and 3) and SDI score (0, 1, 2, 3, 4, 5, 6, 7) were all treated as categorical variables. For the non-vertebral fragility fracture outcome, there were fewer non-vertebral fragility fractures in each category, so prevalent vertebral fracture (yes/no), SQ score (0, 1, or 2), prevalent vertebral fracture number (0, 1, and 2) and SDI score (0, 13, 46, 7) were used in the model as categorical variables. The absolute risk of fracture was estimated using the logistic regression model. As there were no statistically significant interactions, the relationship between all measures of prevalent vertebral fracture status and incident fracture risk was statistically consistent between the two trials. All analyses are reported for

Osteoporos Int (2007) 18:761770 Fig. 2 The relationship between femoral neck T-score and the 23-month non-vertebral fragility fracture risk without consideration of prevalent vertebral fracture status (a); considering presence versus absence of prevalent vertebral fracture(s) (b); considering maximum SQ (semi-quantitative) prevalent vertebral deformity score (c); considering number of prevalent vertebral fractures (d); and considering prevalent SDI (spine deformity index) score (e)
Absolute Nonvertebral Fracture Risk (%) Absolute Nonvertebral Fracture Risk (%) Women With and Without Prevalent Vertebral Fracture(s) (Entire Cohort) Women With Prevalent Vertebral Fracture(s)

765

10

10

a
8 6

b
8 6

Women Without Prevalent Vertebral Fracture(s)

0 0 -1 -2 -3 -4 -5 -6

0 0 -1 -2 -3 -4 -5 -6

Femoral Neck T-Score Absolute Nonvertebral Fracture Risk (%) Absolute Nonvertebral Fracture Risk (%)
10

SQ Score

10

Femoral Neck T-Score Number of Prevalent Vertebral Fractures

2 or 3

>2 1

1
6

0
4

0
4

0 0 -1 -2 -3 -4 -5 -6

0 0 -1 -2 -3 -4 -5 -6

Femoral Neck T-Score SDI Score Absolute Nonvertebral Fracture Risk (%)
10

Femoral Neck T-Score

>7

e
8 6

4-6 1-3

0
4

0 0 -1 -2 -3 -4 -5 -6

Femoral Neck T-Score

pooled data using SAS version 8.2 (SAS Institute, Cary, NC, USA).

Results Subject characteristics

prior vexrtebral fracture. New vertebral fractures occurred in 217 (8.2%) patients, including 64 (29.5%) from the Fracture Prevention Trial and 153 (70.5%) from the MORE trial. Incident non-vertebral fragility fractures occurred in 100 patients including 27 from the Fracture Prevention Trial and 73 from the MORE trial. Assessment of future vertebral fracture risk

Data were included from 2,651 postmenopausal women with low bone mass or osteoporosis including 448 (16.9%) from the Fracture Prevention Trial and 2,203 (83.1%) from the MORE trial. The clinical profile of the group is presented in Table 1. Of these, 44.5% (n=1181) had a

Figure 1a illustrates the inverse relationship between femoral neck T-score and 2-year vertebral fracture risk for the overall cohort. In Fig. 1b, the relationship between these variables is plotted for patients having versus not having a

766

Osteoporos Int (2007) 18:761770

Table 3 Impact of prevalent vertebral fracture status on nonvertebral fragility fracture risk at various femoral neck T-scores in postmenopausal women with osteoporosis Prevalent vertebral fracture Femoral neck T-Score 2 3 4 No 2.2 2.9 3.8 Yes 3.8* (1.7) 5.0 (1.7) 6.4 (1.7) Prevalent vertebral fracture number 0 2.2 2.9 3.7 1 3.6 (1.6) 4.7 (1.6) 6.0 (1.6) 2 4.0 (1.8) 5.2 (1.8) 6.7 (1.8) Maximum prevalent SQ score 0 2.3 2.9 3.6 1 3.3 (1.4) 4.1 (1.4) 5.2 (1.4) 2 or 3 4.4 (1.9) 5.5 (1.9) 6.8 (1.9) Prevalent SDI score 7

13

46

2.2 3.6 (1.6) 4.2 (1.9) 4.6 (2.1) 2.9 4.6 (1.6) 5.4 (1.9) 5.8 (2.1) 3.7 5.9 (1.6) 6.9 (1.9) 7.5 (2.1)

*Fracture risk was derived from the logistic regression model. SQ, semi-quantitative; SDI, spine deformity index. Values shown are the absolute nonvertebral fragility fracture risk in percent over a median observation period of 23 months. Relative risks for fracture in patients with prevalent vertebral fracture(s) versus those of patients with the same BMD but not having prevalent vertebral fracture are provided in parentheses.

prevalent vertebral fracture. At all BMD values, patients with prevalent vertebral fractures had greater risk than the overall group and those with no prevalent vertebral fractures had lower risk than the overall group. Figures 1c to e illustrate the relationship between femoral neck T-score and vertebral fracture risk for patients with varying vertebral fractures status. At all BMD values, increasing maximum severity (SQ score) (Fig. 1c) and number (Fig. 1d) of vertebral fractures were associated with greater future fracture risk. Because both number and severity of vertebral fractures are independent predictors of future fracture risk, the spinal deformity index, which integrates both of these variables, was also analyzed; at all BMD values, increasing SDI was associated with increasing vertebral fracture risk (Fig. 1e). To quantify the impact of considering prevalent vertebral fracture status, absolute vertebral fracture risk was analyzed for femoral neck T-scores of -2, -3, and -4 and various degrees of spine fracture burden (Table 2). At all BMD values, the risk in subjects with one vertebral fracture were about three times the risk in subjects without vertebral fractures. Two fractures resulted in about five times the fracture risk, while three or more prevalent vertebral fractures resulted in seven to nine times the risk. At all BMD values, the risk of an incident vertebral fracture was approximately tripled in subjects with an SQ score of 1 compared to those subjects with an SQ score of 0. An SQ score of 2 conveyed approximately five times the vertebral fracture risk, while an SQ score of 3 was associated with eight to ten times the risk. Similarly, higher prevalent SDI scores were associated with greater risks at all BMD values. Compared to SDI scores of 0, SDI scores of 7 or more resulted in nine to twelve times the risk for fracture. Assessment of future non-vertebral fragility fracture risk Figure 2a illustrates the inverse relationship between femoral neck T-score and 2-year non-vertebral fragility

fracture risk for the overall cohort, and in Fig. 2b, the relationship between these variables is plotted for patients having versus not having a prevalent vertebral fracture. Again, patients with prevalent vertebral fractures had greater risk than the overall group and those with no prevalent vertebral fractures had lower risk than the overall group. Figures 2c to e illustrate the relationship between femoral neck T-score and non-vertebral fragility fracture risk for patients with varying vertebral fractures status. At all BMD values, increasing maximum severity (Fig. 2c), number (Fig. 2d) of vertebral fractures and spinal deformity ndex (Fig. 2e) was associated with greater future fracture risk. Table 3 shows the impact of considering prevalent vertebral fracture status along with femoral neck BMD Tscores of -2, -3, or -4 on non-vertebral fragility fracture risk. At each BMD value, the relative risk in subjects with at least one prevalent vertebral fracture versus those without such a fracture was 1.7. Consideration of details of spinal fracture burden added some additional information, with greater number and severity of vertebral fractures, and greater spinal deformity index conferring slightly greater risk of incident non-vertebral fractures. Assessment of future any fracture risk For this analysis, the baseline vertebral fracture burden was used to assess the risk of an incident vertebral or nonvertebral fragility fracture across the range of baseline femoral neck T-scores. Figure 3a illustrates the expected inverse relationship between femoral neck T-score and 2year any fracture risk for the overall cohort, and in Fig. 3b, the relationship between these variables is shown for patients having versus not having a prevalent vertebral fracture. Again, patients with prevalent vertebral fractures had greater risk than the overall group and those with no prevalent vertebral fractures had lower risk than the overall group. Figure 3c to e illustrate the relationship between femoral neck T-score and any fracture risk for patients with

Osteoporos Int (2007) 18:761770 Fig. 3 The relationship between femoral neck T-score and the 23-month any fracture risk without consideration of prevalent vertebral fracture status (a); considering presence versus absence of prevalent vertebral fracture(s) (b); considering maximum SQ (semi-quantitative) prevalent vertebral deformity score (c); considering number of prevalent vertebral fractures (d); and considering prevalent SDI (spine deformity index) score
70 70

767

Absolute Fracture Risk (%)

50 40 30 20 10 0

Absolute Fracture Risk (%)

60

Women With and Without Prevalent Vertebral Fracture(s) (Entire Cohort)

60 50 40 30 20 10 0

Women With Prevalent Vertebral Fracture(s)

Women Without Prevalent Vertebral Fracture(s)

-1

-2

-3

-4

-5

-6

-1

-2

-3

-4

-5

-6

Femoral Neck T-Score

70

SQ Score 3 2 1 0 Absolute Fracture Risk (%)

70 60 50

Femoral Neck T-Score Number of Prevalent Vertebral Fractures

Absolute Fracture Risk (%)

60 50 40 30 20 10 0

>3 2

40 30 20 10 0

1 0

-1

-2

-3

-4

-5

-6

-1

-2

-3

-4

-5

-6

Femoral Neck T-Score

70

Femoral Neck T-Score SDI Score 7 6 5 4 3 2 1 0

Absolute Fracture Risk (%)

60 50 40 30 20 10 0

-1

-2

-3

-4

-5

-6

Femoral Neck T-Score

varying vertebral fractures status. At all BMD values, increasing maximum severity (Fig. 3c), number (Fig. 3d) of vertebral fractures and spinal deformity index (Fig. 3e) were associated with greater future fracture risk. Table 4 shows the impact of considering prevalent vertebral fracture status along with femoral neck BMD T-scores of -2, -3, and -4 on the risk of any fracture. At each BMD value, the relative risk in subjects with at least one prevalent vertebral fracture versus those without such a fracture was approximately 3 to 4 times greater. Again, consideration of number, severity, and spinal deformity index provided additional information regarding risk. Across the range of BMD values, spinal deformity index impacted future fracture risk by up to 5 to 7 fold.

Assessment of future fracture risk considering lumbar spine T-score plus vertebral fracture burden The impact of prevalent vertebral fracture status on the risk of incident vertebral, non-vertebral, and any fracture was also evaluated using lumbar spine T-scores and the results were similar to those observed using femoral neck T-scores (data not shown).

Discussion The present analysis quantifies the impact of vertebral fracture status on BMD-based assessments of 2-year fracture risk. For any given BMD T-score, the risk of an

768

Osteoporos Int (2007) 18:761770

Table 4 Impact of prevalent vertebral fracture status on any fracture risk at various femoral neck T-scores in postmenopausal women with osteoporosis Prevalent vertebral fracture Prevalent vertebral fracture number 0 1 2 3 Maximum prevalent SQ score 0 1 2 3 Prevalent SDI score 7

Femoral Neck T-Score 2 3 4

No

Yes

4.5 7.2 11.2

16.0* (3.6) 23.8 (3.3) 33.9 (3.0)

4.6 7.0 10.6

10.8 (2.3) 16.0 (2.3) 23.0 (2.2)

17.6 (3.8) 25.1 (3.6) 34.5 (3.3)

24.0 (8.6) 33.2 (7.9) 43.8 (6.9)

4.7 6.9 10.2

10.4 (2.2) 15.0 (2.2) 21.2 (2.1)

17.5 (3.7) 24.5 (3.6) 33.2 (3.3)

27.2 (5.8) 36.3 (5.3) 46.6 (4.6)

4.6 6.8 9.9

10.4 (2.3) 14.9 (2.2) 20.7 (2.1)

11.7 (2.5) 16.7 (2.5) 23.1 (2.3)

23.3 (5.1) 31.3 (4.6) 40.6 (4.1)

33.4 (7.3) 43.0 (6.3) 53.1 (5.4)

*Fracture risk was derived from the logistic regression model. SQ, semi-quantitative; SDI, spine deformity index. Values shown are the absolute any fracture risk in percent over a median observation period of 23 months. Relative risks for fracture in patients with prevalent vertebral fracture(s) versus those of patients with the same BMD but not having prevalent vertebral fracture are provided in parentheses.

incident vertebral, non-vertebral fragility, and any fracture differs by up to twelve times, 2 times, and 7 times, respectively, when information regarding spine fracture burden is considered. In the absence of knowledge about the prevalent vertebral fracture status, assessments based solely on BMD may under- or over-estimate the true risk of a patient experiencing an incident fracture. This is the first analysis to examine the impact of fracture burden on fracture risk across a range of T-scores treated as a continuous variable. Furthermore, we assessed fracture burden in four different ways, allowing for comparisons of the impact of dichotomization of presence or absence of vertebral fractures, number or severity of vertebral fractures, and spinal deformity index as predictors of future fractures across a range of T-scores. SDI provided the greatest range of risk at each T-score. Data provided in this report can be used to assist clinicians in estimating the probability of future fractures. By coordinating the baseline BMD T-score with the line representing the appropriate prevalent vertebral fracture status, the absolute risk of a fracture over approximately two years can be estimated. To illustrate, a femoral neck T-score of -2 and an SDI score of 0 results in about a 2% absolute vertebral fracture risk, while at the same BMD, an SDI of 2 corresponds to about an 8% absolute vertebral fracture risk, and an SDI score of 6 results in approximately a 19% vertebral fracture risk (Fig. 1e). Some general features of the analysis are remarkable. At all T-scores, patients without vertebral fractures had relatively lower risk for fractures during the ensuing two years. Conversely, it is evident that the highest risk was observed in patients with both low BMD and greater spine fracture burdens. However, patients with a high burden of vertebral fractures have a great deal of risk even if their BMD is relatively preserved. For example, a patient with a T-score of -2 with a spinal deformity index of

at least 7 had approximately a 33% risk of any fracture over 2 years. Bone strength has been described as a reflection of the integration of bone density and bone quality [20]. Since vertebral fracture status impacts the risk of future fractures at all BMD values, spine imaging may represent a test of bone quality, and vertebral fractures provide evidence of impaired bone quality. In support of this, a recent analysis showed that postmenopausal women with increasing severity of vertebral fractures had progressively worse microstructural integrity as assessed by histomorphometry or microcomputed tomography indices of iliac crest biopsy specimens [21]. Across the range of BMD T-scores, spine fracture burden had up to a 12-fold impact on the predicted vertebral fracture risk but only approximately twofold impact on predicted non-vertebral fragility fracture risk. This finding is consistent with a review of more than 40 studies showing that while prior fractures at any site are predictive of fractures at other sites, they are most predictive of future fractures at that particular site [22]. Illustratively, an analysis from the Fracture Prevention Trial showed that in placebo patients with zero, one, and two or more prior nonvertebral fragility fractures, 3.6%, 8.2%, and 18.0%, respectively, experienced new non-vertebral fragility fractures during the trial (P<0.001), representing a fivefold range of risk [8]. Hence, in assessing future fracture risk, the assessment of a patients future non-vertebral fracture risk should probably include a consideration of the number of prior non-vertebral fractures in addition to the spine fracture burden. In this study, prevalent vertebral fracture status was assessed semi-quantitatively via lateral spine radiography, a widely used gold standard for identifying vertebral fractures. However, there are several practical considerations

Osteoporos Int (2007) 18:761770

769

including cost, radiation exposure and patient inconvenience that might preclude obtaining spine radiographs in all patients with low bone mass or osteoporosis. While our analysis did not use data generated by this approach, lateral spine imaging performed by DXA - vertebral fracture analysis or VFA - at the time of BMD testing in women found to have low bone mass or osteoporosis may provide a practical solution to routine imaging of the spine in clinical practice. VFA involves substantially less radiation exposure and cost, and is less subject to issues of parallax distortion, although it historically has generated images of lower resolution compared with lateral spine radiography [23, 24]. Ongoing refinements in this technology include improvements in resolution. VFA and routine lateral spine radiographs have shown good agreement for identifying moderate and severe vertebral fractures by semi-quantitative assessment. There has been less agreement for mild, grade 1, vertebral fractures [2529]. The results presented here were statistically consistent across the two clinical trials and are also consistent with many other studies. They are likely to be relevant to women with low bone mass or osteoporosis who share characteristics similar to those of the participants in these clinical trials. However, the important prognostic information provided by spine imaging might not be relevant to some populations of women. For example, younger women with high BMD and older women with very low BMD may not require spine imaging. Nonetheless, in this analysis, it is noteworthy that consideration of age had no significant impact on fracture risk after BMD and vertebral fracture status were considered. The analyses were conducted based on 2-year data and therefore reflect 2-year fracture risk. It is probable that the fracture risk associated with any given prevalent vertebral fracture status/BMD T-score combination would increase with time. In conclusion, across a range of BMD, women without vertebral fractures had a relatively low 2-year risk for incident fracture while those with vertebral fractures had an increased risk. The risk for incident fracture increased with greater number and severity of prevalent vertebral fractures and with increasing SDI. Across a range of BMD, prevalent vertebral fracture status impacted incident fracture risk by up to 12 times. Since bone strength is the integration of BMD plus bone quality, these findings suggest that spine imaging provides a means to assess an element of bone quality and that vertebral fractures provide evidence of poor bone quality. Risk assessments based solely on BMD may overestimate the true risk of future fractures in patients without vertebral fractures and underestimate the true risk of future fractures in patients with vertebral fractures. Given the numerous pharmacologic and nonpharmacologic options available to the clinician to treat

patients with broadly varying levels of risk for future fractures, clinical decision making may be facilitated by considering both BMD and prevalent vertebral fracture burden in assessing short term fracture risk in postmenopausal women.
Acknowledgements The authors thank the investigators and patients from both of the trials. Data were analyzed at Lilly Research Laboratories, Eli Lilly and Company. This study was supported by Eli Lilly and Company. All authors had access to all of the relevant data for this analysis. Conflicts of interest statement Dr. Siris serves as a consultant and receives honoraria from Eli Lilly and Company, Merck, Novartis, Procter & Gamble, Amgen, Pfizer and Glaxo-Smith Kline. Dr. Genant serves as a consultant and receives an honorarium from Eli Lilly and Company, Novartis, Servier, Roche, GSK, BMS and Wyeth. Dr. Laster serves as a consultant and receives honoraria or financial support from Eli Lilly and Company, Procter & Gamble, Merck, Roche/GSK, Abbott, and Genentech. Drs. Chen, Misurski and Krege are employees of Eli Lilly and Company.

References
1. Cummings SR, Melton LJI (2002) Epidemiology and outcomes of osteoporotic fractures. Lancet 359(9319):17611767 2. Ettinger B, Black DM, Mitlak BH et al (1999) Reduction of vertebral fracture risk in postmenopausal women with osteoporosis treated with raloxifene: results from a 3-year randomized clinical trial. Multiple Outcomes of Raloxifene Evaluation (MORE) Investigators. JAMA 282(7):637645 3. Black DM, Cummings SR, Karpf DB et al (1996) Randomised trial of effect of alendronate on risk of fracture in women with existing vertebral fractures. Fracture Intervention Trial Research Group. Lancet 348(9041):15351541 4. Cummings SR, Black DM, Thompson DE et al (1998) Effect of alendronate on risk of fracture in women with low bone density but without vertebral fractures: results from the Fracture Intervention Trial. JAMA 280(24):20772082 5. Black DM, Arden NK, Palermo L, Pearson J, Cummings SR (1999) Prevalent vertebral deformities predict hip fractures and new vertebral deformities but not wrist fractures. Study of Osteoporotic Fractures Research Group. J Bone Miner Res 14 (5):821828 6. Lindsay R, Silverman SL, Cooper C et al (2001) Risk of new vertebral fracture in the year following a fracture. JAMA 285 (3):320323 7. Ross PD, Davis JW, Epstein RS, Wasnich RD (1999) Pre-existing fractures and bone mass predict vertebral fracture incidence in women. Ann Intern Med 114(11):919923 8. Gallagher JC, Genant HK, Crans GG, Vargas SJ, Krege JH (2005) Teriparatide reduces the fracture risk associated with increasing number and severity of osteoporotic fractures. J Clin Endocrinol Metab 90:15831587 9. Delmas PD, Genant HK, Crans GG et al (2003) Severity of prevalent vertebral fractures and the risk of subsequent vertebral and nonvertebral fractures: results from the MORE trial. Bone 33 (4):522532 10. Genant HK, Li J, Wu CY, Shepherd JA (2000) Vertebral fractures in osteoporosis: a new method for clinical assessment. J Clin Densitom 3(3):281290

770 11. Cooper C, Atkinson EJ, OFallon WM, Melton LJ III (1992) Incidence of clinically diagnosed vertebral fractures: a populationbased study in Rochester, Minnesota, 19851989. J Bone Miner Res 7(2):221227 12. Neer RM, Arnaud CD, Zanchetta JR et al (2001) Effect of parathyroid hormone (134) on fractures and bone mineral density in postmenopausal women with osteoporosis. N Engl J Med 344 (19):14341441 13. Genant HK, Wu CY, van Kuijk C, Nevitt MC (1993) Vertebral fracture assessment using a semiquantitative technique. J Bone Miner Res 8(9):11371148 14. Genant HK, Jergas M, Palermo L et al (1996) Comparison of semiquantitative visual and quantitative morphometric assessment of prevalent and incident vertebral fractures in osteoporosis The Study of Osteoporotic Fractures Research Group. J Bone Miner Res 11(7):984996 15. Grigoryan M, Guermazi A, Roemer FW, Delmas PD, Genant HK (2003) Recognizing and reporting osteoporotic vertebral fractures. Eur Spine J 12(suppl 2):S104S112, Sep 11 16. Harris ST, Watts NB, Genant HK et al (1999) Effects of risedronate treatment on vertebral and nonvertebral fractures in women with postmenopausal osteoporosis: a randomized controlled trial. Vertebral Efficacy With Risedronate Therapy (VERT) Study Group. JAMA 282(14):13441352 17. Crans GG, Genant HK, Krege JH (2005) Prognostic utility of a semiquantitative spinal deformity index. Bone 37(2):175 179 18. Genant HK, Siris E, Crans GG, Desaiah D, Krege JH (2005) Reduction in vertebral fracture risk in teriparatide-treated postmenopausal women as assessed by spinal deformity index. Bone 37(2):170174 19. Lu Y, Ye K, Mathur AK, Hui S, Fuerst TP, Genant HK (1997) Comparative calibration without a gold standard. Stat Med 16 (16):18891905

Osteoporos Int (2007) 18:761770 20. NIH Consensus Development Panel (2001) Osteoporosis prevention, diagnosis, and therapy. JAMA 285:785795 21. Genant HK, Delmas PD, Chen P, Yiang Y, Eriksen EF, Dalsky GP, Marcus R, San Martin J. Severity of vertebral fracture reflects deterioration of bone microarchitecture. Osteoporosis International (in press) 22. Klotzbuecher CM, Ross PD, Landsman PB, Abbott III TA, Berger M (2000) Patients with prior fractures have an increased risk of future fractures: a summary of the literature and statistical synthesis. J Bone Miner Res 15:721739 23. Guermazi A, Mohr A, Grigorian M, Taouli B, Genant HK (2002) Identification of vertebral fractures in osteoporosis. Semin Musculoskelet Radiol 6(3):241252 24. Faulkner KG (2001) Clinical use of bone densitometry. In: Marcus R, Feldman D, Kelsey J (eds) Osteoporosis. Academic Press, San Diego, CA, USA, pp 433458 25. Rea JA, Li J, Blake GM, Steiger P, Genant HK, Fogelman I (2000) Visual assessment of vertebral deformity by X-ray absorptiometry: a highly predictive method to exclude vertebral deformity. Osteoporos Int 11(8):660668 26. Binkley N, Krueger D, Gangnon R, Genant HK, Drezner MK (2005) Lateral vertebral assessment: a valuable technique to detect clinically significant vertebral fractures. Osteoporos Int 16 (12):15131518 27. Vokes TJ, Dixon LB, Favus MJ (2003) Clinical utility of dualenergy vertebral assessment (DVA). Osteoporos Int 14(11):871878 28. Ferrar L, Jiang G, Barrington NA, Eastell R (2000) Identification of vertebral deformities in women: comparison of radiological assessment and quantitative morphometry using morphometric radiography and morphometric X-ray absorptiometry. J Bone Miner Res 15(3):575585 29. Wu C, Genant H, von Ingersleben G et al (2004) Validation of lateral DXA imaging for assessment of vertebral fractures. J Bone Miner Res 19(suppl 1):S295