Predicting vertebral fractures from density heterogeneity
Our results from laboratory experiments led us to believe that this range of uncertainty in prediction of fracture risk may be largely accounted for by the heterogeneity of bone density within a vertebra. The heterogeneity cannot be accounted by the current BMD protocols since they are average measures for an individual. The existing data led us to also hypothesize that variability in BMD is directly related to bone toughness and inversely related to strength and that this concept can be used clinically to define new methods to better estimate vertebral fracture risk and to suggest new strategies to restore bone structure.
Computed tomography (CT) images that were acquired as a part of a large and NIH-sponsored study (National Lung Screening Trial, NLST) are available for testing our hypothesis. The subjects of NLST are individuals with an increased risk for a vertebral fracture as they all have a significant history of smoking. Three CT scans acquired at one year intervals are available for each individual in the CT arm of the study. This is in essence the real life equivalent of a mechanical fatigue experiment done in a laboratory.
Therefore, our overall goal is to test the utility of bone density heterogeneity for prediction of a future vertebral fracture and progression of an existing one using CT images from our Institution’s arm of the NLST. The specific goals of the proposed period are:
1) To determine the differences in the distribution of BMD heterogeneity (BMD.SD) and average BMD in the
same population and to determine the extent to which men and women are different in their BMD.SD.
2) Using semi-quantitative and quantitative morphometric vertebral fracture assessment methods and the CT
images of the same individual from three time points taken one year apart, we will determine the extent to
which BMD.SD can predict occurrence of new and monitor the progress of existing vertebral fractures.
3) Using cross-sectional area of the vertebra, average BMD and BMD.SD measured from CT images, we will
determine the extent to which fracture prediction can be unified over sexes and vertebral levels.
Accuracy in the prediction of osteoporotic fracture risk using bone mineral density (BMD) alone is limited requiring additional measures to improve prediction accuracy. From our previous observations, we suggest that variability of bone quality within an individual bone is an important contributor to bone quality. We have shown that BV/TV within a bone varies and we believe this variability when used in combination with the mean density (BMD) will be able to increase the accuracy of fracture risk prediction.
Our data show clear differences in cancellous bone microstructural variability between groups with different propensity to vertebral fracture. We found such a difference between men and women even though they have the same average microstructure in T12-L1 with women exhibiting greater variation. If one further considers that T12-L1 has the highest incidence of vertebral collapse of the T3 to L5 levels, these results suggest that microstructural variability within a bone may be a key quality in determining vertebral fracture risk.
Direct assessment of vertebral properties by mechanical testing also showed that decreasing vertebral strength (maximum load it can carry during fracture) is strongly associated with increasing variability of bone mass (coefficient of variation of bone volume fraction) within the same vertebra. This association was independent of average bone mass, thus reinforcing the importance of variability.
Interestingly, we also found that increasing variability is associated with increased displacement at maximum load (i.e., reduced brittleness). Our co-investigator demonstrated that regional variations of bone density in a vertebra can predict vertebral strength and fatigue life but that the predictive regions were different for each. While both increased strength and reduced brittleness are clearly desirable properties for a vertebra, the requirements on the variation of bone quality necessary for increased strength appear different from, perhaps even competing with, those necessary for reduced brittleness. We do not yet know the relative contribution of intravertebral variability to vertebral fracture risk, since extrapolating from in vitro studies is unreliable. The few in vivo studies examining similar concepts were limited cross-sectional studies which separated groups with or without existing fractures limiting their predictive value.
Together, we feel that there is reasonable evidence that imaging biomarkers based on microstructural variability (either calculated from statistical or spatial distribution of bone mass) may be predictive of vertebral bone fracture risk. The overall objective of the proposed research is to determine the extent to which bone mass variability parameters can predict vertebral fractures. Our overall hypothesis that the statistical variation of bone density within a vertebra when used in combination with the average bone mineral density (BMD) is a better predictor of vertebral fracture compared to the average BMD used alone. We propose to use the computed tomography (CT) image archive constructed by the National Lung Screening Trial (NLST). This represents a unique opportunity as the study cohort is a large group of smokers, a high-risk group for vertebral fractures, who underwent multiple chest CT scans that included the thoracic spine at yearly intervals.
Aim 1: Using CT images from the NLST, determine the extent of the variability of BMD (measured as the standard deviation of BMD: BMD.SD) within vertebra that do not have a fracture.
H1a: There will be a greater variability of BMD.SD than average BMD between randomly selected populations.
H1b: BMD.SD will be greater in women than in men.
Aim 2: Compare CT images of the same individual from three time points taken one year apart to determine the extent to which BMD.SD could predict incident vertebral fracture and in vivo fatigue fracture progress.
H2a: Standard deviation of bone density at time 2 shows a greater deviance from the standard deviation of bone density at time zero among incident fracture cases than non-fracture controls.
H2b: Standard deviation of bone density is a predictor of thoracic vertebral fracture independent of average BMD.
H2c: Standard deviation of bone density is a predictor of fracture progress (in vivo fatigue damage rate).
H2d: Standard deviation of bone density is correlated with fracture grade.
Aim 3: Using crossectional area (CSA), BMD and BMD.SD measured from CT images from the NLST, determine the extent to which fracture prediction can be unified over sexes and vertebral levels.
H3: Relationship between fracture and BMD parameters will be unified by accounting for CSA differences.
Since the current methods for estimation of bone quality and fracture risk aim at bone strength and use bone-average measurements, the knowledge gained from this research may substantially change our clinical approach for predicting risk of fracture. Further, a better understanding of how strength- and fatigue-related predictors contribute to vertebral fracture risk will promote the development of interventions such as exercise, surgery and drugs aimed at restoring bone structure optimized for fatigue resistance as well as strength.