Association of prevalent vertebral fractures, bone density, and alendronate treatment with incident vertebral fractures: effect of number and spinal location of fractures

Bone Vol. 25, No. 5 November 1999:613– 619

Association of Prevalent Vertebral Fractures, Bone Density, and Alendronate Treatment With Incident Vertebral Fractures: Effect of Number and Spinal Location of Fractures M. C. NEVITT,1 P. D. ROSS,2 L. PALERMO,1 T. MUSLINER,2 H. K. GENANT,3 and D. E. THOMPSON2 FOR THE FRACTURE INTERVENTION TRIAL RESEARCH GROUP 1

University of California Prevention Sciences Group, San Francisco, CA, USA Merck Research Laboratories, Rahway, NJ, USA 3 Department of Radiology, University of California, San Francisco, CA, USA 2

tures are a strong risk factor for future fractures, and consider treating such patients to reduce their risk of subsequent fractures. (Bone 25:613– 619; 1999) © 1999 by Elsevier Science Inc. All rights reserved.

Vertebral fractures are the most common osteoporotic fracture and are associated with significant pain and disability. Prior vertebral fracture and low bone mineral density (BMD) are strong predictors of new vertebral fracture. Using data from 6082 women, ages 55– 80 years, in the Fracture Intervention Trial (a randomized, placebo-controlled trial of the antiresorptive agent, alendronate), we explored the association of the number of prior vertebral fractures with the risk of new fractures and whether this association is influenced by the spinal location of fractures. The risk of future vertebral fractures increased with the number of prevalent fractures, independently of age and BMD; in the placebo group, more than half of the women with five or more fractures at baseline developed new vertebral fractures, compared to only 3.8% of women without prior vertebral fractures. The magnitude of association with an increased risk of future vertebral fractures was equal for prevalent fractures located in either the “lower” (T12–L4) (relative risk [RR] ⴝ 2.9; 95% CI ⴝ 1.9, 3.6) or “upper” (T4 –10) spine (RR ⴝ 2.6; 95% CI ⴝ 1.9, 3.6). We found no evidence that the effectiveness of alendronate in reducing the risk of future vertebral fracture was attenuated in women with up to five or more prevalent fractures, or that it varied by the location of prevalent fractures. However, prevalent vertebral fractures in any location were more strongly associated with risk of new fractures in the upper (RR ⴝ 5.2; 95% CI ⴝ 3.2, 8.3) than in the lower spine (2.3; 1.6, 3.3). In addition, each 1 SD decrease in spinal BMD was associated with a 2.1 (1.7, 2.6) times greater odds of new fracture in the upper spine, compared with 1.5 (1.3, 1.8) for the lower spine. These findings suggest that, in older women, osteoporosis may be a stronger risk factor for new fractures in the upper (vs. lower) thoracolumbar spine, although we found no evidence that the location of prior fractures should influence treatment decisions. Physicians should recognize that prior vertebral frac-

Key Words: Osteoporosis; Vertebral fracture; Bone density; Alendronate; Clinical trials. Introduction Fractures of the thoracolumbar spine are the most common osteoporotic fracture and are associated with significant pain and disability.33,34 Ten percent of 50 –54-year-old white women and nearly 50% of 80 – 84-year-old women have at least one vertebral fracture.31 Women with lower bone density have an increased risk of vertebral fracture,27 whereas women with preexisting vertebral fractures have an increased risk of further vertebral,3,12,35,37,41 hip,2,8,20,21 and other fractures,2,6,8,12–14 independently of low bone mass. At least two small studies suggest that the risk of new vertebral fractures rises continuously with an increasing number of prevalent fractures,35,37 but this issue deserves further study in a larger sample. In white women, the prevalence15,16,18,31,36,43 and incidence8,11,15,16,43 of fractures in the thoracolumbar spine are highest in the midthoracic (T7– 8) area in the upper spine and at the thoracolumbar junction (T12–L1) in the lower spine. The peaks in fracture occurrence at these two locations suggests that each is a biomechanically compromised area of the spine. However, different biomechanical patterns of spinal loading may contribute to fracture risk in these two locations.29 Bone mass in the spine may also vary by location. Volumetric density of cancellous bone assessed by quantitated computed tomography (qCT) is higher in the thoracic than lumbar spine in the same individuals, but the opposite has been observed for integral measures of bone density obtained by dual-energy X-ray absorptiometry (DXA) that are influenced by bone size— bone mineral density (BMD) is higher in the lumbar than in the thoracic spine.38 Whether the association of prior fracture with new vertebral fractures, or the relationship of BMD with fracture risk, varies according to the location of vertebral fractures in the spine is not known. In addition, whether antiresorptive agents that have been shown to reduce the risk of vertebral fractures1,26 have a

Address for correspondence and reprints: Michael C. Nevitt, Ph.D., University of California Prevention Sciences Group, 74 New Montgomery Street, Suite 600, San Francisco, CA 94105. E-mail: mnevitt@psg. ucsf.edu © 1999 by Elsevier Science Inc. All rights reserved.

613

8756-3282/99/$20.00 PII S8756-3282(99)00202-1

614

M. C. Nevitt et al. Vertebral fracture risk by spinal region

similar effect on fractures in different locations of the spine has not been studied. Differences in these relationships by spinal location could indicate that fractures in different areas of the spine have different etiologies or associations with osteoporosis. We used data from the Fracture Intervention Trial (FIT), a large randomized, placebo-controlled trial of alendronate, to address these questions. We examined the relationship of the number of prevalent fractures in the thoracolumbar spine, and the location of prevalent fractures, with the risk of new vertebral fractures in the placebo group. We also determined whether prevalent vertebral fractures and bone density predict new fractures in the upper or lower spine equally well. Finally, we determined whether treatment with alendronate had a similar effect on the risk of new fractures in the two spinal regions. Methods

Bone Vol. 25, No. 5 November 1999:613– 619

phometry or visual inspection were excluded from the calculation of baseline lumbar spine BMD. Lateral spine radiographs were obtained at baseline, and were repeated as described in what follows, using standard protocols.1,10 Vertebrae with level-specific vertebral height ratios more than 3 SD below the population mean were classified as prevalent fractures at baseline,4 based on morphometric measurements.10,15,19 New vertebral fractures were defined as a decrease in any vertebral height of at least 20% and at least 4 mm, relative to baseline.1 New fractures included further height reductions in existing prevalent fractures of at least 20% and at least 4 mm (6.3% of new fractures). Prevalent fractures by morphometry were compared with a parallel, binary semiquantitative grading16 and discrepancies reviewed for point placement and nonfracture anomalies by the technicians and study radiologist (H. K. G.). All new fractures by morphometry were reviewed by the study radiologist to exclude nonfracture anomalies.

Participants in FIT Participants were recruited from 11 metropolitan areas of the USA using population-based listings. Details of the recruitment process have been published previously,1,5 and are summarized briefly here. Women were eligible if they were between the ages of 55 and 81 years, had femoral neck BMD of 0.68 g/cm or less (measured by Hologic Model QDR-2000, Waltham, MA), and had been postmenopausal for at least 2 years. The femoral neck BMD cutoff used corresponds to approximately 1.6 SD below the mean for young white women.24 Women were excluded if they had active serious peptic ulcer disease during the past year, a recent history of abnormal renal function, uncontrolled hypertension, severe malabsorption, myocardial infarction during the previous 6 months, unstable angina, or medical problems that would interfere with participation for the 3– 4 year study duration. Women who had used fluoride or bisphosphonates at any time in the past, or had used estrogen or calcitonin during the previous 6 months, were also excluded. The study was approved by the institutional review board at each center, and all participants gave written informed consent. Participants with a daily calcium intake of less than 1000 mg at baseline, as measured by a food-frequency questionnaire,9 were given a supplement containing 500 mg calcium (as the carbonate) and 250 IU of vitamin D; this represented 83% of the placebo group and 81% of the alendronate group.1 The alendronate dose was 5 mg/day for the first 24 months, but was increased to 10 mg/day thereafter, while maintaining the double-blind procedure, in view of other studies showing significantly greater increases in BMD with similar tolerability for the higher dose.23 Compliance was high; for example, among those with preexisting vertebral fractures at baseline (Vertebral Fracture Arm), 89% of surviving participants were still taking the study medication at the time of the closeout visit.1 Similar compliance was observed among women without vertebral fractures at baseline (Clinical Fracture Arm); 83% were still taking study medication at the closeout visit. Measurements BMD of the lumbar spine, femoral neck, and total hip was measured at baseline, and annually thereafter. For quality assurance, all densitometry operators received central training and were certified, and a random sample of BMD scans was reviewed centrally. To further insure stability and comparability of densitometer measurements, local BMD phantoms were scanned daily, and study-wide phantoms were circulated between centers. Lumbar spine BMD was calculated as the average of BMD measured at L2– 4. Vertebrae with preexisting fractures by mor-

Follow-up Participants with prevalent vertebral fractures were assigned to the Vertebral Fracture Arm of the FIT to have 36 months of follow-up, with radiographs repeated at 24 and 36 months after the baseline visit.1,5 Follow-up of this group of participants was ended early when a planned interim analysis revealed a substantial reduction in the incidence of new vertebral fractures. This early termination for ethical reasons resulted in an average of 2.9 years of follow-up for participants in the Vertebral Fracture Arm. Participants without vertebral fractures at baseline were assigned to the Clinical Fracture Arm of the FIT with a planned follow-up of 48 and 54 months (actual average follow-up was 50 months), with radiographs at baseline and again at 48 months.5 Randomization was successful in that there were no meaningful differences in the distributions of BMD or other variables between the alendronate and placebo groups at baseline. Analyses of alendronate effects were performed using the intention-to-treat approach; regardless of compliance, surviving participants who did not withdraw were followed to the end of the study whenever possible, and their data are included in the analyses.1 Analysis The analyses presented here were limited to 6082 women who were randomized and had lateral spine radiographs at baseline and at least one additional visit. This represents 94.2% of the women initially randomized. The randomized women who were not included in the present analysis (because of missing follow-up radiographs) did not differ from the other randomized women in age and lumbar spine, femoral neck, or total hip BMD at entry; however, mean weight was 1.2 kg lower and mean height was 0.13 cm higher at entry in these women. Analyses were performed separately in the alendronate and placebo groups, or were restricted to the 3042 women in the placebo group, as indicated in the text and tables. Statistical tests for trend by increasing number of prevalent fractures were performed by including an ordinal variable for number of prevalent fractures, truncated at five or more, in regression models. Interaction between number of prevalent fractures and treatment group was evaluated by including an interaction term in the logistic regression model, calculated as the product of the two variables (alendronate and an indicator for the number of prevalent fractures). For analyses of fracture location, the “upper” spine was defined as vertebrae T4 –10 (seven vertebrae) and the “lower” spine as T11–L4 (six vertebrae); two subregions (midthoracic, or T7– 8, and T-L junction, or T12–L1) were also used in some analyses. These choices were based on

Bone Vol. 25, No. 5 November 1999:613– 619

M. C. Nevitt et al. Vertebral fracture risk by spinal region

615

Table 1. Characteristics of participants by number of prevalent fractures

0

1

2

3–4

5⫹

p value for trend

4133 (67.9%) 67.6 (6.1) 160.5 (6.0) 64.4 (10.7) 0.841 (0.131) 0.593 (0.060) 0.707 (0.081) 4.3 (0.3)

1358 (22.3%) 70.3 (5.6) 160.1 (6.1) 65.9 (11.5) 0.809 (0.139) 0.577 (0.065) 0.677 (0.088) 2.9 (0.2)

329 (5.4%) 71.3 (5.3) 159.0 (6.2) 64.8 (12.1) 0.782 (0.130) 0.557 (0.066) 0.653 (0.083) 2.9 (0.2)

177 (2.9%) 72.8 (4.9) 157.8 (6.1) 61.9 (11.1) 0.726 (0.120) 0.529 (0.069) 0.609 (0.089) 2.9 (0.2)

86 (1.4%) 73.1 (5.3) 153.7 (6.3) 61.5 (11.0) 0.713 (0.122) 0.517 (0.074) 0.586 (0.093) 2.9 (0.1)

— 0.0001 0.0001 0.0995 0.0001 0.0001 0.0001 —

Number of prevalent vertebral fracturesa Characteristic N Age (yr) Height (cm) Weight (kg) Spine bone mineral density (BMD) (g/cm2) Femoral neck BMD (g/cm2) Total hip BMD (g/cm2) Mean follow-up (yr) a

Women without prevalent fractures were assigned to the Clinical Fracture Arm of the FIT, with radiographs at baseline and again at 48 or 54 months; women with prevalent fractures were assigned to the Vertebral Fracture Arm of the FIT, with radiographs at baseline and again at 36 months. Values in parentheses are standard deviations, except the first row, which is percent of all women.

this and previous studies that reported a peak in frequency of fractures centered around T12–L1 and T7– 8,7,16,18,31,36,39,43 with a nadir in frequency around T-10. We recognize that other groupings by vertebral location are possible. To assess the sensitivity of our findings by location to the way levels were grouped, we also analyzed two subregions grouped by T6 –9 and T12–L3. Our findings with these alternative groupings were essentially the same as those for upper and lower spine, as defined earlier. We tested whether the odds ratios for predicting fractures in different locations differed significantly, by location, using generalized estimating equations (GEE) to test for interactions between predictor variables and the location of new fractures and to adjust the p values for within-subject correlation between the compared outcomes.22 Associations with BMD were expressed relative to a 1.0 SD decrease. Published SD values were used because the low BMD entry requirement reduced the variability of BMD among the FIT women; SD values were 0.112 g/cm2 for femoral neck and 0.134 g/cm2 for total hip.25 For lumbar spine, the SD of 0.110 g/cm2 was obtained from the manufacturer’s reference database.

one, and 4.0% had two or more fractures in the upper spine (T4 –10), whereas 14.5% had one, and 4.5% two or more fractures in the lower spine (T11–L4); 5.8% of women had one or more fractures in both regions. The number and location of prevalent fractures did not differ between treatment groups (data not shown). New vertebral fractures (n ⫽ 462) occurred in 344 women during an average of 3.8 years follow-up. Among women on placebo (329 fractures in 223 women), the pattern of new fracture incidence by spinal location was similar to that for

Results Participant Characteristics On average, women with existing (prevalent) vertebral fractures at baseline were approximately 3 years older and had 7% lower total hip BMD at baseline, compared to women without prevalent fractures (data not shown). These differences between women with and without prevalent vertebral fractures may be partly related to differences in recruitment strategies. The differences in age and BMD increased significantly with the number of prevalent fractures such that women with five or more prevalent fractures were almost 3 years older, and had 13% lower total hip BMD, compared to women with just one fracture at baseline (Table 1). Women with multiple prevalent fractures also weighed less, but this trend was not significant. Number and Location of Prevalent and New Fractures The FIT was designed so that approximately one third of participants would have prevalent vertebral fractures at baseline (Table 1).5 The prevalence of vertebral fractures at baseline, by spinal location, is shown in Figure 1a. There were two maxima in the distribution: a peak of approximately 7% at T12–L1, and another of about 5% at T7– 8. Prevalences at the locations with the highest frequencies were three to four times greater than at the levels with the lowest frequencies. Also, 12.2% of women had

Figure 1. (a) Prevalence of vertebral fracture at baseline, by spinal location. (b) Incidence of vertebral fracture during follow-up, by spinal location, in the placebo (solid) and alendronate-treated (open) groups.

616

M. C. Nevitt et al. Vertebral fracture risk by spinal region

Bone Vol. 25, No. 5 November 1999:613– 619

Table 2. Vertebral fracture incidence at any location (T4 –L4), by number of prevalent fractures, and odds ratios adjusted for age, weight, and total hip bone mineral density (BMD)

Model #a 1

2

Placebo (n ⫽ 3042)

Alendronate (n ⫽ 3040)

Number of prevalent vertebral fractures

Incidence (%)

Odds ratio (95% CI)b

Incidence (%)

Odds ratio (95% CI)b

0 1 2 3–4 5⫹ ⱖ1

3.8 8.9 19.4 30.8 54.2 15.0

1.0 (reference) 1.9 (1.3–2.8) 4.2 (2.6–6.7) 5.6 (3.3–9.6) 16.4 (8.6–31.2) 3.0 (2.2–4.1)

2.1 5.2 12.8 16.3 18.4 7.9

1.0 (reference) 2.0 (1.3–3.2) 4.7 (1.7–8.4) 6.0 (3.0–11.9) 5.8 (2.3–14.7) 2.9 (1.9–4.3)

a

There were two models for each treatment group; the first model included four categories representing number of prevalent fractures (1, 2, 3– 4, 5⫹), and the second model used a single variable to represent one or more prevalent fractures. The reference category in both models was women without prevalent fractures (first row). b Adjusted for age, weight, and total hip BMD; p ⬍ 0.0001 for trend with number of prevalent fractures.

fracture prevalence at baseline, with one peak at T12–L1 and a less pronounced peak at T7–9 (Figure 1b). In the placebo group, 3.6% of the women had one or more new fractures in the “upper” spine (T4 –10), and 4.8% had one or more new fractures in the “lower” spine (T11–L4), whereas 1.2% of women had fractures in both locations. Association Between Number of Prevalent Fractures and Incident Fractures Women with at least one prevalent vertebral fracture had an approximately threefold higher odds of experiencing a new vertebral fracture, compared to women without a prevalent fracture, after adjusting for age, total hip BMD, and weight at baseline (Table 2). The risk of new vertebral fractures increased progressively with the number of prevalent fractures at baseline, up to five or more. Excluding worsening fractures in previously fractured vertebrae had no effect on the association between the number of prevalent and new fractures (data not shown). These analyses probably underestimate the association of new fractures with prevalent fractures, because the women with prevalent fractures had an average of 1.4 fewer years of follow-up (Table 1). The association between prevalent and new fractures was almost identical for the placebo and alendronate groups, except for women with five or more prevalent fractures (p ⫽ 0.016 for interaction between treatment group and five or more fractures). Over half of the women in the placebo group with five or more fractures at baseline experienced new vertebral fractures, more than ten times the incidence observed among women without prevalent fractures. The incidence of new fractures in the alendronate group was approximately half that in the placebo group for most categories of number of prevalent fractures (Table 2). Association of Prevalent Fractures With Incident Fractures, by Location of Fractures, in the Placebo Group To explore the possibility that fractures in certain locations of the spine might be stronger indicators of osteoporosis than other locations, we examined whether the association of prevalent fractures with new fractures, and baseline BMD with new fractures, differed by location of fractures in the upper spine (T4 –10) compared with the lower spine (T11–L4). These analyses were limited to women in the placebo group. The increase in a woman’s risk of a new fracture occurring anywhere in the spine was nearly identical for prevalent fractures in the upper and prevalent fractures in the lower spine, after

adjusting for age, total hip BMD, and weight (Table 3). The risk of a new fracture showed a similar increase with an increasing number of prevalent fractures in the two regions (p ⬎ 0.05 for all comparisons by location). Prevalent midthoracic (T7– 8) and T-L junction (T12–L1) fractures also did not differ significantly in their association with new fractures (data not shown). A different picture emerged for the association of prevalent fractures with the risk of new fractures occurring in specific locations (Table 4). Although preexisting fractures anywhere in the spine increased a woman’s risk of a new fracture at locations in both the upper and lower spine, this increase in risk was significantly greater for new fractures in the upper compared to new fractures in the lower spine. This difference was most pronounced among women with two or more prevalent fractures at baseline. Preexisting fractures were also significantly more strongly associated with new midthoracic compared with new T-L junction fractures. We examined whether this apparent preferential association of preexisting fractures with new upper spine fractures was consistent regardless of the location of the preexisting fracture (Table 5). Prevalent fractures in the upper spine were significantly more strongly associated with new “upper” than with new lower spine fractures. Although prevalent fractures in the “lower” spine were also more strongly associated with new upper than Table 3. Odds ratios (95% CI) for new vertebral fractures in the placebo group, by location of prevalent fractures Prevalent fractures, number and locationa

Percent of all women

Odds ratiob for ⱖ1 new fracture, any location

0 ⱖ1 upper (T4–10) ⱖ1 lower (T11–L4)

71.2 16.3 17.6

1.0 (reference) 2.6 (1.9, 3.6) 2.6 (1.9, 3.6)

0 1 upper (T4–10) ⱖ2 upper 1 lower (T11–L4) ⱖ2 lower

71.2 12.4 4.0 13.5 4.1

1.0 (reference) 2.0 (1.3, 3.0) 3.9 (2.4, 6.4) 2.1 (1.5, 3.1) 3.4 (2.1, 5.5)

a

Categories in the first (left) column indicate the number and location of prevalent fractures. For example, women with a single prevalent fracture in the upper spine had a 2.0-times-greater odds of having a new vertebral fracture during follow-up, compared to women without a prevalent fracture. Women with multiple prevalent fractures may have them in more than one location. b Adjusted for age, weight, and total hip bone mineral density (BMD).

Bone Vol. 25, No. 5 November 1999:613– 619

M. C. Nevitt et al. Vertebral fracture risk by spinal region

Table 4. Odds ratios (95% CI) for new vertebral fractures in specific locations in the placebo group, by number of prevalent vertebral fractures Number of prevalent fractures 0 ⱖ1

Odds ratioa for new fractureb at: “Upper” (T4–10)c

“Lower” (T11–L4)c

1.0 (reference)

1.0 (reference)

p value for comparison of odds ratiosd

5.2 (3.2, 8.3)

2.3 (1.6, 3.3)

0

1.0 (reference)

1.0 (reference)

1

2.3 (1.3, 4.1)

1.8 (1.2, 2.7)

0.3470

13.1 (7.8, 22.1)

3.4 (2.1, 5.3)

0.0003

ⱖ2

0 ⱖ1

Midthoracic (T7–8)c

T-L junction (T12–L1)c

1.0 (reference) 5.6 (2.7, 11.6)

1.0 (reference) 2.1 (1.3, 3.4)

0.0051

with lower spine fractures, this difference was not significant. Results were similar using new midthoracic vs. new T-L junction fractures as the outcome (data not shown). Association of BMD With Incident Fractures, by Location of Fractures, in the Placebo Group All three BMD measurements (lumbar spine, femoral neck, total hip) were significantly associated with new vertebral fractures at locations in both the upper and lower spine (Table 6). For both spine and femoral neck BMD, the association with new fractures in the upper spine was significantly stronger than the association with new fractures in the lower spine. This difference was not significant for total hip BMD. All three BMD measurements Table 5. Odds ratios (95% CI) for new vertebral fractures in specific locations in the placebo group, by location of prevalent fractures

Location of prevalent fractures None ⱖ1 Upper (T4–10) ⱖ1 Lower (T11–L4) a

Upper (T4–10)b

Lower (T11–L4)b

p value for comparison of odds ratiosc

1.0 (reference) 4.6 (3.0, 7.2) 3.8 (2.5, 5.9)

1.0 (reference) 1.8 (1.2, 2.6) 2.2 (1.5, 3.2)

0.0011 0.1625

Adjusted for age, weight, and total hip BMD. The incidence of new fractures among all women was 3.6% at T4 –10, 4.8% at T11–L4, 1.6% at T7– 8, and 2.9% at T12–L1. Women with multiple new fractures may have them in more than one location. c p value is for comparison of odds ratios by location of new fractures. b

Table 6. Odds ratios (95% CI) for associations of incident vertebral fractures with bone density in the placebo group, adjusted for age and weight Bone mineral density (BMD) measure Spine Femoral neck Total hip

Spine Femoral neck Total hip

Odds ratioa (95% CI) for new fracture at: Upper (T4–10)

Lower (T11–L4)

p value for comparison of odds ratiosb

2.1 (1.7–2.6) 3.0 (2.1–4.4) 3.4 (2.4–4.9)

1.5 (1.3–1.8) 1.8 (1.3–2.5) 2.7 (2.0–3.6)

0.0215 0.0363 0.4610

Midthoracic (T7–8)

T-L junction (T12–L1)

2.8 (2.0–3.9) 3.6 (2.2–6.0) 3.5 (2.2–5.6)

1.4 (1.1–1.7) 1.4 (1.0–2.1) 1.8 (1.3–2.6)

0.0001 0.0022 0.0125

a

0.0056

Values indicate the estimated increase in odds of a new fracture in a given spine region, shown in the columm heading, relative to women without prevalent fractures (the reference group). The leftmost column indicates the number of prevalent fractures (without regard to spinal location. a Adjusted for age, weight, and total hip BMD. b Women with multiple new fractures may have them in more than one location. c The incidence of new fractures among all women was 3.6% at T4 –10, 4.8% at T11–L4, 1.6% at T7– 8, and 2.9% at T12–L1. d p value is for the comparison of the odds ratio by location of new fracture.

Odds ratioa (95% CI) for new fracture at:

617

The odds ratios represent the increase in odds of fracture for a 1 SD decrease in baseline BMD. b p values are for comparison of odds ratios by location of new fracture.

were significantly more strongly associated with new midthoracic compared with new T-L junction fractures. Effect of Alendronate on New Fractures in the Upper and Lower Spine The incidence of new vertebral fractures among women randomized to alendronate was lower than that of the women assigned to placebo by approximately one half at nearly all individual vertebral levels (Figure 1b). The reduction in incidence among women on alendronate was similar for fractures in the upper spine (odds ratio: 0.43; 95% CI: 0.31, 0.66) and lower spine (0.49; 0.37, 0.65). There was a tendency for a greater reduction in risk of new fractures at T7– 8 (0.32; 0.18, 0.57) compared with new fracture at T12–L1 (0.45; 0.31, 0.66), but these odds ratios were not significantly different. Among women with a preexisting vertebral fracture, the reduction in risk of new vertebral fractures in the alendronate group was similar regardless of the location of the preexisting fracture (data not shown). Discussion Previous studies have shown that preexisting vertebral fractures are associated with an increased risk of additional vertebral fractures.3,12,35,37,41 The association is independent of BMD. Our study confirms this previous finding in a large sample of postmenopausal women with low bone mass, and extends it by showing that the risk of new vertebral fractures increases progressively with an increasing number, up to five or more, of prevalent fractures. Whether the strong relationship between preexisting and new fractures reflects an impairment in bone quality, or abnormal spine biomechanics associated with preexisting fractures or some factor associated with physical frailty, is an important question, but one that is not addressed in the present analyses. The association between prevalent and new vertebral fractures reported here probably underestimates the true increase in risk associated with prevalent vertebral fractures, because women with prevalent fractures in our study had, on average, only 2.9 years of follow-up, compared to 4.3 years for women without preexisting vertebral fractures at baseline. Some osteoporosis treatment trials have excluded women with more than four preexisting vertebral fractures.32,40,42 Published analyses of the FIT cohort have shown a comparable

618

M. C. Nevitt et al. Vertebral fracture risk by spinal region

reduction in vertebral fracture risk with alendronate treatment in women having just one, and those with two or more, prevalent vertebral fractures.12 We found no attenuation of this antifracture effect even among women with five or more baseline fractures, a group with the highest rate of new fractures during follow-up. Our results provide further evidence that women with advanced vertebral osteoporosis can benefit from antiresorptive therapy and contribute useful information to therapeutic trials. We found that the elevated risk of future fractures signaled by preexisting fractures is similar regardless of the location of prevalent fractures in the spine. Moreover, we observed a reduction in new vertebral fractures at locations throughout the spine among women taking alendronate, and regardless of the location of prevalent fractures. Thus, our findings suggest that the location of preexisting vertebral fractures should not be a factor in identifying high-risk patients or in the fracture efficacy of alendronate therapy. Our study confirms previous community-based cross-sectional studies,8,15,16,18,31,36,43 which show an uneven distribution of fractures in the spine with a peak in fracture prevalence and incidence around T7– 8 and another peak around T12–L1. Factors that preferentially compromise the biomechanics of the spine in these two locations are thought to contribute to the bimodal distribution of fracture by location.8 Fractures at different locations in the skeleton may have distinct pathogeneses that are reflected in a variable association with indicators of osteoporosis and other risk factors.30 For example, several recent studies find that cervical and intertrochanteric hip fractures, despite their very close proximity in the skeleton, differ substantially in their relationship to several important risk factors, including age, previous vertebral and other fractures, bone density of the spine and trochanter, and ultrasound measurements of bone mass.28 Differences in femoral and pelvic dimensions between the two types of fracture17 strongly suggest that biomechanical differences related to anatomy influence the type of hip fracture. Our findings suggest that in elderly women, new fractures in the upper spine (T4 –10) are more strongly associated with a previous vertebral fracture, and with low BMD, than are new fractures in the lower spine (T11–L4). These differences were similar, although somewhat more pronounced, for new midthoracic (T7– 8) compared with new T-L junction (T12–L1) fractures. This pattern is consistent with a different relationship of fractures in these locations to osteoporosis, and may reflect variations in the biomechanical and material properties of vertebral bodies, as well as different patterns of spinal loading, in different segments of the spine. A recent ex vivo study of vertebral bodies from T1 to L5 found that the density of cancellous bone measured by qCT decreased progressively, whereas areal measures of BMD, vertebral cross-sectional area, and failure loads increased progressively from the upper thoracic to the lower lumbar spine.38 The impact of osteoporosis on bone strength may be influenced by these systematic differences across segments. Patterns of spinal loading also vary across segments and may influence fracture risk.29 Vertebrae in the lower spine may be more exposed to large forces generated by falls, heavy lifting, or activities that accentuate bending and twisting,7 which might result in a weaker association with bone density. Analysis of vertebral fracture risk taking spinal location into account is complex, and there are several methodological factors that could influence our findings. Our grouping of vertebral levels to define location is somewhat arbitrary and results may differ using other groupings. However, our results were consistent across several alternative groupings including T6 –9 compared with T12–L3, and T7– 8 compared with T12–L1. Our morphometric methods for assessing fracture may be subject to greater error in certain regions of the spine, which could weaken the ability to predict fractures in these regions. For example,

Bone Vol. 25, No. 5 November 1999:613– 619

vertebrae at the T-L junction are further from the central ray in our X-ray protocol which increases radiographic image distortion, compared with the midthoracic region. However, we attempted to reduce these errors through expert radiologist review of incident morphometric fractures. In addition, our results for T12–L3 were nearly identical to those for T12–L1. Our definition of new fracture requires an absolute height decrease of at least 4 mm, which may bias toward detection of more severe fractures in the smaller vertebrae of the upper spine. However, in practice, this criterion only altered the classification of new fracture in a very small number of previously fractured vertebrae. Nevertheless, it is possible that the location of fractures may be correlated with the severity of fractures in a way that could influence our findings. This requires further exploration. Although we excluded levels fractured by morphometry or by visual inspection from the calculation of lumbar spine BMD, undetected fractures at L2– 4 could bias this measurement and weaken its association with lower spine fractures. Analysis of fractures defined by location increases the number of comparisons and hence the likelihood of chance findings. Several of our findings are not consistent with a stronger relationship to osteoporosis for fractures in the upper compared with the lower thoracolumbar spine. For example, the location of prevalent fractures did not influence their association with new fractures, and the effect of alendronate treatment did not appear to vary by location of new fractures. We used somewhat different approaches to defining prevalent and new fractures, and it is possible that the rate of misclassification differs between prevalent and new fractures in a way that could influence our findings. The method for defining new fractures (changes in vertebral heights over time) may be more specific for osteoporotic fractures, and less subject to misclassification errors related to natural variations in vertebral shape or other diseases affecting vertebral shape than the method for defining prevalent fractures (ratios of heights of vertebral bodies). Associations with risk factors would tend to be biased toward the null for fractures based on the method with greater random error. Because of the many complex methodological issues involved in assessing vertebral fractures, our findings for the prediction of new fractures in different locations need confirmation in other studies and with different methods. We conclude that the risk of future vertebral fractures increases progressively with the number of prevalent fractures, up to five or more, independently of age and BMD. The magnitude of association with an increased risk of future vertebral fractures is approximately equal for prevalent fractures located in either the lower or upper spine. Furthermore, we found no evidence that the effectiveness of alendronate in reducing the risk of future vertebral fracture was attenuated in women with multiple prevalent fractures, or that it varied by the location of prevalent fractures; alendronate treatment reduced the risk by approximately half. Women with five or more vertebral fractures may benefit from treatment and contribute useful information in clinical trials, and there appears to be no evidence for selection of women for treatment based on the location of a prevalent fracture. On the other hand, in our population of older women, low BMD and prevalent fractures were associated more strongly with the risk of new fractures in the upper (T4 –10) spine than with new fractures in the lower (T11–L4) spine. This finding suggests that, in older women, osteoporosis is a stronger risk factor for fractures in the upper thoracolumbar spine and that there may be differences in the pathophysiology of fractures at different locations in the spine. References 1. Black, D. M., Cummings, S. R., Karpf, D. B., Cauley, J. A., Thompson, D. E., Nevitt, M. C., Bauer, D. C., Genant, H. K., Haskell, W. L., Marcus, R., Ott,

Bone Vol. 25, No. 5 November 1999:613– 619

2.

3.

4.

5.

6.

7.

8.

9.

10. 11.

12.

13. 14.

15.

16.

17.

18. 19.

20.

21. 22. 23.

S. M., Torner, J. C., Quandt, S. A., Reiss, T. R., and Ensrud, K. Randomised trial of effect of alendronate on risk of fracture in women with existing vertebral fractures. Lancet 348:1535–1541; 1996. Black, D. M., Arden, N. K., Palermo, L., Pearson, J., and Cummings, S. R. Prevalent vertebral deformities predict hip fractures and new vertebral deformities but not wrist fractures. Study of Osteoporotic Fractures Research Group. J Bone Miner Res 14:821– 828; 1999. Black, D. M., Nevitt, M. C., Palermo, L., Ensrud, K., Genant, H. K., Cummings, S. R., and the Study of Osteoporotic Fractures Research Group. Prediction of new vertebral deformities. J Bone Miner Res 6(Suppl.):S135; 1993. Black, D. M., Palermo, L., Nevitt, M. C., Genant, H. K., Epstein, R., San Valentin, R., and Cummings, S. R. Comparison of methods for defining prevalent vertebral deformities: The Study of Osteoporotic Fractures. J Bone Miner Res 10:890 –902; 1995. Black, D. M., Reiss, T. F., Nevitt, M. C., Cauley, J., Karpf, D., and Cummings, S. R. for the Fracture Intervention Trial Research Group. Design of the Fracture Intervention Trial. Osteopor Int 3(Suppl.):S29 –S39; 1993. Burger, H., van Daele, P. L., Algra, D., Hofman, A., Grobbee, D. E., Schutte, H. E., Birkenha¨ger, J. C., and Pols, H. A. P. Vertebral deformities as predictors of non-vertebral fractures. Br Med J 309:991–992; 1994. Cooper, C., Atkinson, E. J., O’Fallon, W. M., and Melton, L. J. Incidence of clinically diagnosed vertebral fractures: A population-based study in Rochester, Minnesota, 1985–1989. J Bone Miner Res 7:221–227; 1992. Cooper, C., O’Neill, T., and Silman, A. The epidemiology of vertebral fractures. European Vertebral Osteoporosis Study Group. Bone 14(Suppl.): S89 –S97; 1993. Cummings, S. R., Block, G., McHenry, K., and Baron, R. B. Evaluation of two food frequency methods of measuring dietary calcium intake. Am J Epidemiol 126:796 – 802; 1987. National Osteoporosis Foundation Working Group on Vertebral Fractures. Report assessing vertebral fractures. J Bone Miner Res 10:518 –523; 1995. De Smet, A. A., Robinson, R. G., Johnson, B. E., and Lukert, B. P. Spinal compression fractures in osteoporotic women: Patterns and relationship to hyperkyphosis. Radiol 166:497–500; 1988. Ensrud, K. E., Black, D. M., Palermo, L., Bauer, D. C., Barrett-Connor, E., Quandt, S. A., Thompson, D. E., and Karpf, D. B., for the Fracture Intervention Trial Research Group. Treatment with alendronate prevents fractures in women at highest risk. Arch Intern med 157:2617–2624; 1997. Ga¨rdsell, P., Johnell, O., and Nilsson, B. E. The predictive value of forearm bone mineral content measurements in men. Bone 11:229 –232; 1990. Ga¨rdsell, P., Johnell, O., Nilsson, B. E., and Nilsson, J. Å. The predictive value of fracture, disease, and falling tendency for fragility fractures in women. Calcif Tissue Int 45:327–330; 1989. Genant, H. K., Jergas, M., Palermo, L., Nevitt, M., Valentin, R. S., Black, D., and Cummings, S. R. 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 11:984 –996; 1996. Genant, H. K., Wu, C. Y., van Kuijk, C., and Nevitt, M. C. Vertebral fracture assessment using a semiquantitative technique. J Bone Miner Res 8:1137– 1148; 1993. Gluer, C. C., Cummings, S. R., Pressman, A., Li, J., Gluer, K., Faulkner, K. G., Grampp, S., and Genant, H. K. Prediction of hip fractures from pelvic radiographs: The study of osteoporotic fractures. The Study of Osteoporotic Fractures Research Group. J Bone Miner Res 9:671– 677; 1994. Hedlund, L. R., Gallagher, J. C., Meeger, C., and Stoner, S. Change in vertebral shape in spinal osteoporosis. Calcif Tissue Int 44:168 –172; 1989. Jergas, M. and San Valentin, R. Techniques for the assessment of vertebral dimensions in quantitative morphometry. In: Genant, H. K., Jergas, M., and van Kuijk, C., Eds. Vertebral Fracture in Osteoporosis. San Francisco, CA: Radiology Research and Education Foundation, University of California; 1995. Kotowicz, M. A., Melton, J., III, Cooper, C., Atkinson, E. J., O’Fallon, M. W., and Riggs, B. L. Risk of hip fracture in women with vertebral fracture. J Bone Miner 9:599 – 605; 1994. Lauritzen, J. B. and Lund, B. Risk of hip fracture after osteoporosis fractures. Acta Orthop Scand 64:297–300; 1993. Liang, K. Y. and Zeger, S. L. Longitudinal data analysis using generalized linear models. Biometrika 73:13–22; 1986. Liberman, U., Weiss, S. R., Broll, J., Minne, H. W., Quan, H., Bell, N. H., Rodriguez-Portales, J., Downs, R. W., Jr., Dequeker, J., Favus, M., Seeman, E., Recker, R. R., Capizzi, T., Santora, A. C., Lombardi, A., Shah, R. V., Hirsch, L. J., and Karpf, D. B., for the Alendronate Phase III Osteoporosis Treatment

M. C. Nevitt et al. Vertebral fracture risk by spinal region

24.

25.

26.

27.

28. 29.

30. 31.

32.

33.

34. 35.

36.

37.

38.

39.

40.

41. 42.

43.

619

Study Group. Effect of oral alendronate on bone mineral density and the incidence of fractures in postmenopausal osteoporosis. N Engl J Med 333: 1437–1443; 1995. Looker, A. C., Wahner, H. W., Dunn, W. L., Calvo, M. S., Harris, T. B., Heyse, S. P., Johnston, C. C., Jr., and Lindsay, R. L. Proximal femur bone mineral levels of U.S. adults. Osteopor Int 5:389 – 409; 1995. Looker, A. C., Wahner, H. W., Dunn, W. L., Calvo, M. S., Harris, T. B., Heyse, S. P., Johnston, C. C., Jr., and Lindsay, R. Updated data on proximal femur bone mineral levels of US adults. Osteopor Int 8:468 – 489; 1998. Lufkin, E. G., Wahner, H. W., O’Fallon, W. M., Hodgson, S. F., Kotowicz, M. A., Lane, A. W., Judd, H. L., Caplan, R. H., and Riggs, B. L. Treatment of postmenopausal osteoporosis with transdermal estrogen. Ann Intern Med 117:1–9; 1992. Marshall, D., Johnell, O., and Wedel, H. Meta-analysis of how well measures of bone mineral density predict occurrence of osteoporotic fractures. Br Med J 312:1254 –1259; 1996. Mautalen, C. A., Vega, E. M., and Einhorn, T. A. Are the etiologies of cervical and trochanteric hip fractures different? Bone 18(Suppl.):133S–137S; 1996. Melton, L. J., III, Chao, E. Y. S., and Lane, J. Biomechanical aspects of fractures. In: Riggs, B. L. and Melton, L. J., III, Eds. Osteoporosis: Etiology, Diagnosis, and Management. New York: Raven; 1988. Melton, L. J., III, and Cummings, S. R. Heterogeneity of age-related fractures: Implications for epidemiology. Bone Miner 2:321–331; 1987. Melton, L. J., III, Lane, A. W., Cooper, C., Eastell, R., O’Fallon, W. M., and Riggs, B. L. Prevalence and incidence of vertebral deformities. Osteopor Int 3:113–119; 1993. Meunier, P. J., Sebert, J. L., Reginster, J. Y., Briancon, D., Appelboom, T., Netter, P., Loeb, G., Rouillon, A., Barry, S., Evreux, J. C., Avouac, B., Marchandise, X., and the FAVO Study Group. Fluoride salts are no better at preventing new vertebral fractures than calcium-vitamin D in postmenopausal osteoporosis: The FAVOS study. Osteopor Int 8:4 –12; 1998. Nevitt, M. C., Ettinger, B., Black, D. M., Stone, K., Jamal, S. A., Ensrud, K., Segal, M., Genant, H. K., and Cummings, S. R. The association of radiographically detected vertebral fractures with back pain and function: A prospective study. Ann Intern Med 128:793– 800; 1998. Ross, P. D. Clinical consequences of vertebral fractures. Am J Med 103(Suppl.):30S– 43S; 1997. Ross, P. D., Davis, J. W., Epstein, R. S., and Wasnich, R. D. Pre-existing fractures and bone mass predict vertebral fracture incidence in women. Ann Intern Med 114:919 –923; 1991. Ross, P. D., Fujiwara, S., Huang, C., Davis, J. W., Epstein, R. S., Wasnich, R. D., Kodama, K., and Melton, L. J., III. Japanese women in Hiroshima have greater vertebral fracture prevalence than Caucasians or Japanese-Americans in the US. Int J Epidemiol 24:1171–1177; 1995. Ross, P. D., Genant, H. K., Davis, J. W., Miller, P. D., and Wasnich, R. D. Predicting vertebral fracture incidence from prevalent fractures and bone density among non-black, osteoporotic women. Osteopor Int 3:120 –126; 1993. Singer, K., Edmondston, S., Day, R., Breidahl, P., and Price, R. Prediction of thoracic and lumbar vertebral body compressive strength: Correlations with bone mineral density and vertebral region. Bone 17:167–174; 1995. Spector, T. D., McCloskey, E. V., Doyle, D. V., and Kanis, J. A. Prevalence of vertebral fracture in women and the relationship with bone density and symptoms: The Chingford Study. J Bone Miner Res 8:817– 822; 1993. Storm, T., Thamsborg, G., Steiniche, T., Genant, H. K., and Sorensen, O. H. Effect of intermittent cyclical etidronate therapy on bone mass and fracture rate in women with postmenopausal osteoporosis. N Engl J Med 322:1265–1271; 1990. Wasnich, R. D., Davis, J. W., and Ross, P. D. Spine fracture risk is predicted by non-spine fractures. Osteopor Int 4:1–5; 1994. Watts, N. B., Harris, S. T., Genant, H. K., Wasnich, R. D., Miller, P. D., Jackson, R. D., Licata, A. A., Ross, P. D., Woodson, G. C., Yanover, M., Mysiw, W. J., Kohse, L., Rao, M. B., Steiger, P., Richmond, B., and Chesnut, C. H. Intermittent cyclical etidronate treatment of postmenopausal osteoporosis. N Engl J Med 323:73–79; 1990. Wu, C. Y., Li, J., Jergas, M., and Genant, H. K. Comparison of semiquantitative and quantitative techniques for the assessment of prevalent and incident vertebral fractures. Osteopor Int 5:354 –370; 1995.

Date Received: January 26, 1999 Date Revised: June 23, 1999 Date Accepted: June 23, 1999