The Relationship Between Prevalent Spine Fractures and Future Fracture Risk
A number of studies have demonstrated that the presence of a spine fracture is predictive of future fractures, independent of bone density (33-39). The strongest association is between existing spine fractures and future spine fractures with estimates of the increase in risk from only one prevalent spine fracture of 3- to 11.1-fold. One of the first such studies was from Ross et al. (33). This study was performed in the same group of women from the Kuakini Osteoporosis Study who were described earlier in the discussion of the definition of a spine fracture threshold. In this study, the presence of one vertebral fracture at baseline resulted in a fivefold increase in the risk for new vertebral fractures. If two vertebral fractures were present at baseline, the risk for new vertebral fractures increased 12-fold.
In a second study, Ross et al. (35) evaluated 380 postmenopausal women with an average age of 65 who were participants in a multicenter trial of etidronate therapy for postmenopausal osteoporosis. In this study, the presence of one or two spine fractures at baseline increased the risk of future spine fractures 7.4-fold.
Nevitt et al. (34) evaluated the effect of the number and location of prevalent spine fractures on future fracture risk using data from the Fracture Intervention Trial (FIT), a placebo-controlled, randomized trial of alendronate therapy in postmenopausal osteoporosis. Data from 6082 women were included in this analysis, roughly half of whom were receiving a placebo. Vertebral fractures at baseline were found in 1950 women. Four hundred sixty-two new vertebral fractures occurred in 344 women during an aver age of 3.8 years of follow-up. Nevitt et al. found that the presence of just one vertebral fracture at baseline resulted in a threefold increase in the risk for new vertebral fractures compared to women without a vertebral fracture at baseline. This was true even after adjustment for age, total hip BMD, and weight. Prevalent spine fractures appeared to be stronger predictors of incident spine fractures in the upper spine than in the lower spine.
Data obtained during the Study of Osteoporotic Fractures, a prospective study of 9704 women aged 65 and older, were analyzed by Black et al. (36) to determine the effect of prevalent spine fracture on future fracture risk. Seven thousand two hundred thirty-eight women had technically adequate films for morphometric evaluation of spine fracture at baseline and after an average follow-up of 3.7 years. Prevalent spine fractures were present at baseline in 1915 women. Two or more deformities were seen in 797 and four or more, in 211. Over the follow-up period, 389 women developed new vertebral fractures. The risk for new vertebral fracture increased as the number of prevalent vertebral fractures increased. In women with one prevalent vertebral fracture, the relative risk for new vertebral fracture was 3.2. For women with three or more prevalent vertebral fractures, the relative risk of new vertebral fracture was 10.6 even after adjustment for age.
In a population-based epidemiologic study (37) of vertebral fracture incidence between 1985 and 1994 in Rochester, Minnesota, 820 residents (619 women, 201 men) were diagnosed as having one or more vertebral fractures. The residents were followed until death or the last clinical contact during the study period. The average age at the time of the first vertebral fracture was 67.3 years for women and 55.5 years for men. Compared to fracture incidence rates expected in the general population, the observed rate was almost three times higher in individuals age 35 or older with a prior vertebral fracture. The increase in risk for new vertebral fractures was 12.6-fold higher for both sexes combined and 11.1-fold higher for women alone.
Prevalent vertebral fractures have also been shown to be predictive of nonvertebral fractures. Two hundred fifty subjects (225 women, 25 men) with an average age of 74 years were evaluated for the presence of vertebral deformity at baseline and the subsequent development of nonvertebral fracture during a follow-up period of 3 years (38). During this period, 39 subjects suffered nonvertebral fractures of which 10 were hip fractures, 17 were forearm fractures, and 13 were at a variety of other skeletal sites. Of the 39 subjects who developed nonvertebral fractures, 27 had spine deformities at baseline. Spinal deformities were graded as mild or severe based on the number of vertebrae affected and/or the degree of deformity. After adjusting for age, sex, and BMD in the femoral neck (DXA), subjects with severe spinal deformities at baseline had a fourfold increase in the risk of nonvertebral fracture (RR 4.1; 95% CI 1.3 to 12.4). Subjects with mild spinal deformities had a relative risk of 1.5 for the development of nonvertebral fractures but this increase in relative risk was not significant at the 95% CI.
In the study from Black et al. (36) utilizing data from the Study of Osteoporotic Fractures, prevalent spine fractures were also predictive of nonspine fractures in general and hip fractures, specifically. The risk of any nonspine fracture was increased 1.9-fold, whereas the risk of hip fracture was increased 3.8-fold. The authors could not show a statistically significant increase in the risk for wrist fracture based on the finding of a prevalent spine fracture. In the Rochester, Minnesota population-based study from Melton et al. (37), the finding of a prevalent spine fracture resulted in a 2.3-fold increase in the risk for hip fracture. Unlike the study from Black et al., Melton and colleagues did find a significant increase in the risk of distal forearm fracture of 1.6.
Klotzbuecher et al. (39) reviewed the available literature in 2000 to summarize the known associations between prevalent fracture and future fracture risk of all types. They performed a literature search that spanned 1966 through 1999, identifying 15 publications that reported associations between prevalent spine fractures and subsequent fractures. Based on this review, Klotzbuecher et al. concluded that prevalent spine fracture increases the risk for future spine fracture 4.4-fold (95% CI 3.6 to 5.4). The risk of subsequent hip fracture was increased 2.3-fold (95% CI 2.0 to 2.8) and the risk of subsequent wrist fracture was increased 1.4-fold (95% CI 1.2 to 1.7). The authors noted that in 5 of the 15 studies reviewed, the associations between prevalent spine fracture and subsequent fracture were reduced by only 20% or less when adjustments were made for BMD. Nevertheless, BMD was also a strong predictor of future fractures, independent of prior fractures. Klotzbuecher et al. concluded that BMD and prevalent fractures were complementary in the prediction of future fracture risk.
The assessment of fracture risk is clearly incomplete without an assessment for vertebral fracture. It is estimated that only 33% of vertebral fractures are symptomatic (40). Of those fractures that are not clinically symptomatic, 78% remain unrecognized (41). A more aggressive effort to evaluate patients for vertebral fracture is clearly indicated. The lack of a clear "gold standard" for defining the types of vertebral deformities that are the result of bone fragility and thus fractures, remains controversial. Semiquantitative and quantitative approaches for defining vertebral fractures are used clinically. Either can be applied to plain radiographs as well as densitometric spine images.
Vertebral Fracture Assessment with Genant's Semiquantitative Technique
Genant's semiquantitative technique relies on the expertise of the observer rather than direct measurements of the physical dimensions of the vertebrae (42). Based on the physical appearance, vertebrae are characterized as being normal or deformed. The types of deformation are mild (grade 1), moderate (grade 2), and severe (grade 3). Deformed vertebrae are also described based on the shape of the deformation as wedged (anterior fracture), biconcave (middle fracture), or crushed (posterior fracture). These deformities are illustrated in Fig. 10-5. Although physical measurements are not made with this technique, a grade 1 deformity roughly corresponds to a 20 to 25% reduction in the anterior, middle, or posterior height of the vertebra and a 10 to 20% reduction in vertebral area. A grade 2 deformity is the result of a 25 to 40% reduction in any of the three heights and a reduction in vertebral area of 20 to 40%. A grade 3 deformity occurs when there is a 40% reduction in any of the three heights and a 40% reduction in vertebral area. An expert in the Genant semiquantitative technique also brings to bear knowledge of vertebral deformities caused by various disease processes other than fracture. This adds an invaluable qualitative aspect to the evaluation of vertebral fracture with this technique. This technique has been traditionally used with plain radiographs of the spine but semiquantitative vertebral fracture assessment can be performed with fan-array DXA spine images as well.
Quantitative techniques rely on physical measurements to diagnose vertebral fracture. Reference points are placed on each vertebral body. A common method is the placement of six points, one point at each corner of the vertebral body and one point at the midpoint
of each of the endplates. Using these points, the anterior, mid-, and posterior heights (ha, hm, and hp, respectively) of the vertebra are measured. The vertebral area is calculated as the polygon area defined by the six points. In addition to the heights themselves, the anterior-posterior height ratio (ha/hp) and the midposterior height ratio (hm/hp) are calculated. Other ratios include the wedge index (Iw = hp/ha) (43) and the biconcavity index (hm/ha) (44). These measurements were originally made from plain radiographs. In recent years, measurements were made from digitized films. With the advent of fan-array DXA spine imaging and morphometry software, quantitative vertebral morphometry can be performed by the densitometrist as well.
Different criteria have been proposed for the diagnosis of prevalent or incident fracture based on quantitative morphometry. Several authorities have proposed that a prevalent fracture is present if there is a 15% or greater reduction in the ha/hp or hm/hp ratio or the ratio of the posterior height of one vertebra to the posterior height of an adjacent vertebra (hp/hpa) when compared to the mean value for a normal population (45-47). A more stringent 20% reduction in these ratios has been proposed as well. A reduction of 3 SD in the ha/hp or hm/hp compared to normative data to define vertebral fracture was proposed by Ross et al. and Eastell et al. (48,49). McCloskey and Kanis (50,51) also proposed utilizing a 3 SD reduction in any of the ratios combined with reductions in ratios calculated using a predicted posterior height. These morphometric definitions of vertebral fracture require comparisons to normative reference data for a population. Heights may also be adjusted for body size using the dimensions of the fourth thoracic vertebra (T4). In other words, the hp for T12 can be adjusted or normalized for size by dividing it by the hp at T4 in the individual. The resulting posterior dimension for T12 is then abbreviated nhp, reflecting the normalization for size. Minne et al. (52) proposed defining vertebral fracture as being present when any of the three normalized heights was below the third percentile of the normal range. Because vertebrae are expected to have slightly different shapes depending on the vertebral level, individual heights must be compared to normal values that are specific for that vertebral level.
The definition of incident fractures tends to be more straightforward. When plain radiography is used to capture images of the vertebrae it is imperative that the same radiographic technique be used to avoid artifactual changes in the vertebral shapes. When this is done, a decrease of 15% in the ha, hm, or hp from the baseline film is indicative of an incident fracture. A reduction of 20 to 25% has also been proposed for the definition of an incident fracture (53). An even more stringent criterion is a reduction of 20 to 25% in any of the three heights with a minimum absolute reduction of 4 mm (54).
Performance Comparisons of Semiquantitative and Quantitative Techniques
Quantitative techniques rely heavily on the accuracy of point placements as well as comparisons to reference databases. Point placement can be subjective and affected by the deformity itself or patient positioning. Differences of opinion exist regarding the validity and design of reference databases for vertebral morphometry, just as they do for bone densitometry. Genant's semiquantitative technique is based on the visual recognition of quantitative changes in vertebral shape as well as knowledge of disease processes that cause deformities other than fracture. The performance of these techniques in identifying vertebral fractures have been compared in several studies (55-60). In order to compare the techniques, a gold standard for the diagnosis of vertebral fracture was generally created by a consensus reading of radiographs by experts. When done in this manner, the semiquantitative and quantitative approaches generally perform equally well but the threshold chosen for identifying fracture by quantitative morphometry profoundly effects the agreement between the two techniques. The combination of a semiquantitative and quantitative technique may be better than either alone. Spine imaging with fan-array densitometry combined with precise, computerized measurements enable the densitometrist to utilize both.
Fan-array DXA spine imaging is one of the newer applications for DXA. The spine can be imaged from T4 to L5 in the lateral or PA projection. DXA spine imaging can be performed in seconds to minutes, depending on the scan mode, but always at a fraction of the radiation exposure of conventional spine radiographs.
Lateral spine images, such as the image shown in Fig. 10-6, can be evaluated using Genant's semiquantitative method. Morphometric software can also be used to define and measure vertebral heights, as shown in Fig. 10-7. This technique is called morpho-metric X-ray absorptiometry (MXA) in contrast to the use of conventional radiographs for morphometric measurements, which is called morphometric radiography (MXR). In 1998, Rea et al. (61) evaluated 161 postmenopausal women for fracture using conventional lateral spine radiographs and fan-array lateral spine imaging. Both image types were evaluated using Genant's semiquantitative method. The sensitivity for moderate and severe (grade 2 and 3, in the Genant method) vertebral fractures for analyzable vertebrae on the DXA spine images was 91.9%. When mild (grade 1) fractures were included, the sensitivity dropped to 77.4%. The specificity was extremely high, however,
Fig. 10-6. A lateral spine DVA™ image from a Lunar Prodigy showing a grade 2 fracture at T12. Case courtesy of GE Medical Systems, Madison, WI.
at 98.4% in analyzable vertebrae. As a result, the negative predictive value of DXA spine imaging for vertebral fractures was excellent. Even when mild fractures were included, the negative predictive value for spine fractures from the DXA spine images was 97.5%, suggesting that DXA spine imaging was an excellent means of excluding the diagnosis of vertebral fracture. The agreement in fracture diagnoses between the visual assessment of DXA spine images and the semiquantitative assessment of standard spine radiographs for vertebrae that could be evaluated with both techniques was 96.3%.
Although DXA spine imaging spans T4 to L4, it is not uncommon for the uppermost thoracic vertebrae to be poorly visualized. In the study by Rea et al. (61), 94.9% of the vertebrae could be evaluated. T4 and T5 were the most common vertebrae that were too poorly visualized to be evaluated. In a study by Schousboe et al. (62) in which 342 women underwent DXA lateral spine imaging, 92.1%, or 4096 of the 4446 vertebrae studied, could be evaluated. In this study, T4 to T6 were less likely to be adequately visualized. The inability to consistently evaluate T4 to T6 on lateral DXA spine images does not present a significant problem in osteoporotic fracture identification. Several studies have demonstrated that the majority of fractures occur below these levels. The most common locations for vertebral fractures would appear to be T11 to L1, followed by T7 and T8 (34,59,62,63). In studies that have utilized DXA spine imaging for spine fracture diagnosis, a surprising percentage of women with nonosteoporotic bone densities have been found to have fractures. In the study by Schousboe et al. (62), 27.4% of the patients age 60 and older with osteopenic bone densities according to WHO Criteria were found to have vertebral deformities consistent with a diagnosis of fracture on DXA spine images. Of the patients age 60 and older with osteoporotic bone densities, 42% were found to have vertebral deformities as well. In this study, the diagnosis of fracture on the DXA image was based primarily on Genant's semiquantitative method. Faulkner et al. (64) evaluated 231 women with a mean age of 65 with DXA spine imaging utilizing proprietary mor-phometric software in which a diagnosis of spine fracture was based on a reduction in vertebral height or height ratio of 3 SD or more from the expected mean value. Using this definition of prevalent fracture, more than 50% of the women were found to have vertebral fractures. Based on bone density at the PA lumbar spine or proximal femur, 46.7% of the women had osteopenia and 26.4% had osteoporosis. Of the women with osteopenia, 49.1% were found to have spine fractures. More than 72% of the women with osteoporosis were also found to have spine fractures based on MXA measurements.
MXA, like densitometry, is a quantitative measurement technique. Like densitometry, the utility of MXA can be assessed in part, by its reproducibility or precision. The long-term precision of MXA was evaluated by Ferrar et al. (65) in a population-based sample of 48 postmenopausal women (average age 65 years) and in 50 postmenopausal women with osteoporosis (average age 67 years). Spine imaging was performed with a Hologic QDR 4500 A on two occasions. Precision was calculated for the population-based sample over a 2-year period. One-year precision was calculated for the osteoporotic women. In both groups, use of the "compare" option resulted in better precision. In the population-based sample, the RMS-SD was 0.60 mm with the "compare" option and 0.87 mm without the "compare" option for vertebral heights. The RMS-SD for height ratios in this group was 0.03 and 0.18, with and without "compare." Precision was poorer in the osteoporotic group with RMS-SD of 0.77 mm and 1.59 mm for vertebral heights, again with and without the "compare" option, respectively. Ferrar et al. concluded that the long-term precision of MXA was comparable to that of MXR. Rea et al. (66) also evaluated
the long-term precision of MXA and compared it to that of MXR. Although these authors concluded that the precision of MXA was not as good as that of MXR, they noted, as did Ferrer et al. (65), that the precision errors for both techniques were substantially smaller than the 20 to 25% reduction in vertebral height often used as a criteria for the diagnosis of incident vertebral fracture.
Was this article helpful?