Current guidelines for estimating fracture risk in patients with lytic bone defects are based upon retrospective patient series and have not been validated prospectively or in the laboratory. The prevention of pathological fracture and spinal cord compression from metastatic bone tumors depends on better techniques and more objective criteria for deciding when prophylactic stabilization is necessary. To monitor fracture risk associated with lytic bone defects, we have been studying the use of non-invasive imaging modalities in conjunction with structural rigidity analysis.

The Problem with Current Clinical Guidelines for Predicting Pathologic Fracture

          Hayes quantified the factor of risk for bone fracture as the load-bearing requirement divided by the load bearing capacity of the bone.1 When this ratio exceeds unity, the bone is expected to fail. Defect geometry, the type of lesion, and the anatomic site influence the load bearing capacity of bone, whereas age, anatomic site, and activity level influence the load bearing requirements.
Ideally, methods for predicting fracture risk would incorporate measurements of both bone material behavior (by monitoring apparent bone density) and bone structural geometry (by monitoring cross-sectional area and cross-sectional moment of inertia). The geometry of a metastatic osteolytic defect is the simplest parameter to measure in the clinical setting. Consequently, most clinical guidelines for the management of lytic defects in bone are based solely upon analysis of the geometry of the lesion as depicted on radiographs.

Estimations of Fracture Risk Base on Radiographs

          Current clinical guidelines for managing lytic defects in bone arose from retrospective analysis of radiographs in several clinical series. Prophylactic stabilization is advised when a lytic defect either 1) is greater than 2.5 cm in diameter, or 2) destroys greater than 50% of the cortex.1-7 Unfortunately, these guidelines do not hold up to careful analysis.

2.5-Centimeter Diameter
           The 2.5-cm defect guideline was originally suggested based upon an initial review of 19 patients in whom femoral fractures were associated with well-defined, osteolytic defects 2.5 cm in diameter with cortical involvement.8 However, among these 19 pathologic fractures, 8 occurred in association with defects smaller than 2.5 cm. Their false-negative rate using the 2.5-cm criteria was therefore over 40%. In addition, defects with diffuse borders were considered unpredictable. Moreover, the 2.5-cm criterion was suggested based solely on the analysis of femoral defects, and has not been tested at other anatomical sites.

          A 2.5 cm defect in the shaft of an 70-year-old, 40 kg, post-menopausal woman would likely result in a substantially greater strength reduction compared to a 2.5 cm defect in a 30-year-old, 50 kg, athletic, fertile woman. The failure of radiographic criteria to account for these types of differences in load bearing capacity and load bearing requirements may explain why a study of patients receiving radiation therapy for metastatic cancer to bone found no difference in defect size between patients with and without subsequent fracture.4

50% Cortical Destruction
          review of four adult patients considered to have an impending fracture.9 The same guideline was subsequently suggested based on a review of clinical records of 66 patients with 100 osseous metastases.5 Patients were divided into four groups depending on the level of cortical involvement (0-25%, 25-50%, 50-75% or 75-100%). While the overlap between the defect size in the fracture and non-fracture group was large, only one fracture occurred for a cortical involvement of less than 50%.
The percent cortical involvement was measured from radiographs, in most cases by estimating the maximum width of the defect and dividing by the width of the bone. For lesions that were difficult to measure, a tube of paper was used to represent the diameter of the bone. The outline of the lesion was drawn on the tube as it appeared in the radiographs. The tube was then unrolled, and the cortical involvement expressed as the perimeter of bone compromised by tumor divided by the periosteal perimeter. No experiments were reported to address the inter- or intra-observer accuracy or precision of the method. Clearly, for two bones with the same periosteal diameter but different cortical wall thickness, the total cross sectional area of bone removed would be greater for thick-walled bone. In addition, since errors of up to 100% can occur measuring very simple diaphyseal defects from plane radiographs10, there are major limitations in the predictive capabilities of these radiographic methods.

Failed Attempts To Identify Threshholds For Fracture Risk
          Other investigators have been unable to determine a radiographic measure that identifies patients at risk for pathologic fracture. Keene and colleagues reviewed the clinical histories of 203 patients with a total of 516 metastatic defects to the proximal femur.8 Defect size was measured from radiographs as the maximum defect dimension, and was normalized to the exterior dimensions of the bone. In this large series, the authors were unable to determine a defect size that discriminated between those that fractured from those that did not. Three reasons were cited. First, 57% of the lesions were permeative and did not have clear boundaries and were deemed unmeasurable. Second, 54% of the 26 fractures observed occurred through such unmeasurable lesions. Third, the 12 measurable lesions that fractured had defect sizes that overlapped with those that did not fracture. Zickel and Mouradian were also unable to determine a threshhold geometric measurement predictive of fracture based on radiographs of 50 patients with lytic bone defects associated with fracture or impending fracture.11

Scoring Systems
          By combining four risk factors: site (upper, lower, peritrochanteric); pain (mild, moderate, functional); lesion (blastic, mixed, lytic); and size (less than one-third, between one and two thirds, and greater than two-thirds of the diameter of the bone) into a single score Mirels12 derived a weighted scoring system in an attempt to quantify the risk of sustaining a pathologic fracture through a metastatic defect in a long bone. Summation of these factors into a single score provided greater accuracy than any single factor for determining fracture risk. Seventy-eight metastatic long bone lesions that were irradiated without prophylactic stabilization were analyzed retrospectively: 27 lesions fractured and 51 lesions did not fracture during the subsequent 6-month follow-up. The fracture risk percentage of a lesion could be predicted for any given score. As the score increased above seven, so did the percent of fracture risk. A score of nine attained the highest sensitivity and specificity. Lesions with a score of less than seven could safely be irradiated with only a 5% probability of fracture, while a score of nine had a 33% probability of fracture which might warrant prophylactic stabilization before radiotherapy. However, 67% of patients would receive potentially unnecessary surgery demonstrating that the scoring system was not accurate (i.e. percent test results that are correct or the number of true positive and true negatives divided by the total number of results).

A New Approach for Predicting Pathologic Fracture

          The elastic structural behavior of whole bones with or without lytic defects depends on both the material properties and the cross-sectional geometry of the bone. Rigidity, the product of the material modulus (a measure of the bone stiffness) and the cross-sectional moment of inertia (a measure of how the bone mass is distributed about a bending axis), describes the elastic behavior of a beam. Using composite beam analysis it should be possible to calculate failure loads and thereby predict fracture risk.

          Our underlying assumption in predicting bone fracture is that rigidity measured mechanically (by the slope of the linear portion of the load-deflection curve) is related to the failure load of the bone. If this assumption is true, we can then predict the failure load of a bone by calculating rigidity from the modulus of the bone tissue and cross-sectional moment of inertia measured using non-invasive imaging methods such as quantitative computed tomography (QCT), dual energy x-ray absorptiometry (DXA) and magnetic resonance imaging (MRI). To test this assumption more rigorously, we performed a series of experiments to examine whether rigidity was related to the yield and ultimate loads of bones with simulated osteolytic defects, then applied our techniques in the evaluation of benign bone tumors in children.

Calculations of Structural Rigidity Correlate with Failure Load

          We used a defect model consisting of regularly shaped defects of various sizes created in cylinders of homogeneous, transversely isotropic trabecular bone cored from whale vertebrae. Bone density was measured using computed tomography by comparison with solid hydroxapatite phantoms of known density (QCT). Specimens were tested to failure in four-point bending, uniaxial tension, or torsion. These failure loads were then compared to the minimum axial, bending, or torsional rigidity at the defect that was calculated based on the defect geometry and the bone density measured by QCT.
Correlation with failure load was better for mechanically measured rigidity and calculations of structural rigidity based on QCT than for structural or material properties alone.13 The structural rigidity correlated best with failure load, implying that trabecular bone fails at a constant strain rate, even in the presence of a defect. This conclusion was further supported by the observation that our test specimens failed at fairly constant strain.

Non-invasive Imaging Predicts Failure Load of Spines with Simulated Osteolytic Defects.

          We then applied this technique to the more clinically representative and geometrically complex human vertebra. Specifically, we investigated the use of dual x-ray absorptiometry (DXA), QCT, and Magnetic Resonance Imaging (MRI) as non-invasive tools for measuring the structural properties and for predicting the failure load of vertebrae with simulated lytic defects of intermediate size. The MRI was performed using porosity phantoms of known fractions of heavy water and saline.

           Human fresh frozen cadaver spines from elderly persons were segmented into 3-body functional spinal units (FSU) at the thoracic and lumbar levels. A uniform lytic defect was created in the middle vertebra at one of three different locations.(Fig. 1) We then used composite beam theory to calculate predicted failure loads based on data from the three non-invasive imaging modalities and making the assumption-supported by preliminary data and other reports14Ðthat bone fails at a strain of 1%.

          The spines were mechanically tested to failure using a custom-designed spine-testing machine to load the spine asymmetrically with axial compression and forward flexion. The failure load was quite variable in spite of a consistent relative defect size, further emphasizing that geometric properties alone cannot reliably predict fracture risk. For DXA measurements, the measured failure load correlated better with material than structural properties and was not well predicted by composite beam theory. For QCT measurements, the measured failure load correlated equally well with either material or cross-sectional structural properties, and was well predicted by the calculations based on composite beam theory. These data suggest that QCT can help predict the fracture load of vertebrae compromised by lytic tumors. Such calculations might help guide treatment recommendations.

          Quantitative MRI measurements did not perform as well in this study as in previous studies using porosity phantoms made of foam filled with corn oil, suggesting that further work is needed to refine our MRI methods.

Structural Rgidity Measured by QCT Predicts Fracture in Long Bones with Benign Tumors.

          Children with benign bone tumors represent an ideal group for establishing the in vivo sensitivity and specificity of our ability to predict fracture risk since these patients are not receiving any other therapies (such as chemotherapy or radiation therapy) that would otherwise bias our results or introduce confounding variables. Benign bone tumors in children represent a diverse group of pathological and clinical entities that vary greatly in aggressiveness and clinical behavior. After the orthopedist has established the correct diagnosis of the tumor, the dilemma is to decide whether the defect has weakened the bone enough to lead to pathological fracture. There are few clinical or radiographic guidelines for predicting fracture risk in children with osteolytic defects.

          Our clinical protocol uses QCT to characterize benign lytic bone lesions in children. In addition to calculations based upon comparison with hydroxapatite phantoms, the reduction in the load carrying capacity of the affected bone was estimated by calculating the ratio of the structural rigidity of the bone containing the tumor to the rigidity of the uninvolved contralateral side. The fracture risk was also characterized based upon the following radiographic criteria: length of the lesion greater than 33mm; width greater than 25mm; or cortical destruction greater than 50%.

          A number of patients have been followed for two years or until fracture. Based upon this preliminary data, a receiver operating characteristic curve suggests that a 35% reduction in bending rigidity compared to the contralateral limb provides the most sensitive and specific threshold for discriminating fracture cases. The radiographic criteria appear to be sensitive, but not specific for fracture. These preliminary patient data suggest that minimum bending rigidity measured non-invasively by QCT is sensitive and more specific than standard clinical and radiographic criteria for predicting pathologic fracture through benign bone tumors.

Conclusions

          We are rigorously testing the hypothesis that structural analysis using composite beam theory can accurately predict the failure load of bones with simulated osteolytic defects in regularly shaped trabecular bone specimens. Preliminary data from experiments in human cadaver spines and a small clinical subset of children with benign bone tumors suggest that non-invasive estimations of bone density and geometry can be used for accurate prediction of failure load and fracture risk. To establish this as a new clinical tool for predicting fracture risk in patients with skeletal metastases, the next step is to demonstrate that structural rigidity measured by QCT can predict fracture prospectively in a clinical population of breast and prostate cancer patients with skeletal metastases.

Acknowledgments

This work was funded in part by NIH Grant CA40211-11.

Brian D. Snyder, MD, PhD is Acting Director of the Orthopaedic Biomechanics Laboratory at Beth Israel Deaconess Medical Center and Instructor of Orthopaedic Surgery at Harvard Medical School

Greg D. Cabe, MS, James Hong, MS, John R. Tedrow, MS and Kelli M. Whealan, MS are graduate students from Massachusetts Institute of Technology

John A. Hipp, PhD is Director of Research at the Spine Laboratory of the Baylor College of Medicine.

S. Daniel Kwak, PhD is a Researcher at the Orthopaedic Biomechanics Laboratory at the Beth Israel Deaconess Medical Center and Instructor of Orthopaedic Surgery at Harvard Medical School

Andrew Hecht, MD is a Resident in the Harvard Combined Orthopaedic Residency Program

Address correspondence to:
Brian D. Snyder, MD, PhD; Orthopedic Biomechanics Laboratory; Beth Israel Deaconess Medical Center, 330 Brookline Ave, RN 115 Boston, MA 02215.
e-mail: bds@obl.caregroup.harvard.edu