Predicting Fracture Through Benign Skeletal Lesions with Quantitative Computed Tomography

Snyder, Brian D. MD, PhD; Hauser-Kara, Diana A. PhD; Hipp, John A. PhD; Zurakowski, David PhD; Hecht, Andrew C. MD; Gebhardt, Mark C. MD

Journal of Bone & Joint Surgery - American Volume:
doi: 10.2106/JBJS.D.02600
Scientific Articles
Abstract

Background: There are no proven radiographic guidelines for predicting fracture risk in children and young adults with a benign skeletal lesion. An in vivo diagnostic study was conducted to determine whether a reduction in the load-carrying capacity of a bone measured with quantitative computed tomography was more accurate than current radiographic guidelines for predicting pathologic fracture in patients with a benign skeletal lesion.

Methods: Eighteen patients who presented with a fracture through a benign skeletal lesion were compared with eighteen patients who had a benign skeletal lesion that had been thought to be at increased risk for fracture on the basis of currently used radiographic criteria but had not fractured over a two-year period. Structural analysis was performed to calculate the resistance of the affected bones to compressive, bending, and torsional loads with use of serial transaxial quantitative computed tomography data obtained along the length of the bone containing the lesion and from homologous cross sections through the contralateral, normal bone. At each cross section, the ratio of the structural rigidity of the affected bone divided by that of the normal, contralateral bone was determined. The cross section with the greatest reduction in compressive, bending, and torsional rigidity was identified as that most likely to fracture.

Results: The mean age (and standard deviation) of the thirty-six patients was 12.5 ± 3.6 years. Twenty lesions were located in the femur; eleven, in the tibia; three, in the humerus; one, in the ulna; and one, in the pelvis. A combination of the minimum bending and torsional rigidities calculated from the tomographic data provided optimal performance in differentiating between the fracture and non-fracture groups (100% sensitivity and 94% specificity). In contrast, plain radiographic criteria demonstrated 28% to 83% sensitivity and 6% to 78% specificity.

Conclusions: The combination of bending and torsional rigidity measured noninvasively with quantitative computed tomography was more accurate (97%) for predicting pathologic fracture through benign bone lesions in children than were standard radiographic criteria (42% to 61% accuracy). We believe that this method can provide accurate objective criteria for planning treatment of benign bone lesions and monitoring treatment response.

Level of Evidence: Therapeutic Level III. See Instructions to Authors for a complete description of levels of evidence.

Author Information

1 Departments of Orthopaedic Surgery (B.D.S., D.Z., and M.C.G.) and Biostatistics (D.Z.), Children's Hospital, Harvard Medical School, 300 Longwood Avenue, Boston, MA 02115. E-mail address for B.D. Snyder: brian.snyder@childrens.harvard.edu

2 Orthopedic Biomechanics Laboratory, Beth Israel Deaconess Medical Center, Harvard Medical School, 330 Brookline Avenue, Boston, MA 02215

3 Spine Research Laboratory, Baylor College of Medicine, 6620 Main Street, Houston, TX 77030

Article Outline

Benign skeletal neoplasms represent a diverse group of pathologic and clinical entities that vary greatly in aggressiveness and clinical behavior1,2. The true incidence is unknown, but benign fibrous lesions such as nonossifying fibromas and fibrous cortical defects have been found in up to 33% of asymptomatic children who were evaluated with radiographs of long bones for reasons other than surveillance of a lesion3,4. After confirming that the lesion is benign, the orthopaedist must decide whether the defect has weakened the bone sufficiently to cause a pathologic fracture and whether treatment is indicated. Pathologic fractures in the axial and appendicular skeleton are often associated with pain and loss of function5, so preventing these fractures is a primary goal of treatment. Optimal treatment for benign bone lesions remains controversial. Depending on the type of lesion, its anatomic location, and the suspected risk of fracture, treatment may include observation, restricted weight-bearing, bracing, intralesional injection of steroids or demineralized bone matrix with or without bone marrow aspirate, and curettage and packing of the defect with bone graft with or without stabilization of the affected bone with hardware to prevent fracture3,6-22. Two common lesions, unicameral bone cyst and nonossifying fibroma, can heal spontaneously, so the major clinical challenge is deciding whether a given lesion increases the risk that the involved bone will fracture. Reliable methods for assessing the fracture risk associated with a benign bone lesion could be used to select the most appropriate treatment for a patient and as a mechanical assay to monitor the bone response to treatment.

There are no proven clinical or radiographic guidelines for predicting which children are at risk for pathologic fracture and for objectively monitoring the response of the involved bone to treatment. Previous investigators have performed retrospective studies in an attempt to predict the risk of pathologic fracture on the basis of the patient's age, the stage and activity of the lesion, the anatomic site of the lesion, the size of the lesion, and/or the percentage of cortical destruction4,16,23-26. However, they failed to identify any single radiographic parameter that accurately predicted the occurrence of a pathologic fracture through any benign bone lesion.

The risk of a pathologic fracture depends both on the strength of the bone and on the loads applied to it27. A fracture is probable when the load applied to the bone during a specific activity exceeds the load-bearing capacity of the bone. The load-bearing requirement of the bone depends on the patient's size, weight, activity level, and loading regimen. The loads applied to the bone may be dramatically reduced if a patient's activities are limited by pain. The load-bearing capacity depends on the structural properties of the bone, which are determined by the material properties of the bone tissue, the anatomic site of the lesion, the geometry of both the lesion and the host bone, and the aggressiveness of the lesion. The increased fragility associated with these radiolucent lesions suggests that the strength of the bone tissue surrounding the lesion is degraded and/or the stresses generated within the bone during loading are increased because of changes in bone geometry.

Any method for predicting fracture risk must be able to measure both changes in the material properties of bone tissue and changes in bone geometry induced by the neoplasm. Rigidity, the product of the bone-tissue modulus of elasticity and the cross-sectional geometry of the bone, describes the structural behavior of a bone and its resistance to deformation when subjected to axial, bending, or twisting loads28. The bone-tissue modulus is a function of the bone-mineral density29,30. The bone geometry is represented by the cross-sectional area and the moment of inertia31. The moment of inertia quantifies how the bone tissue is distributed in space; it varies as the fourth power of the distance of the bone tissue relative to a specific bending axis so that the resistance of the bone to bending dramatically increases as bone tissue is distributed away from that bending axis. Therefore, the cortical expansion induced by a bone cyst may partly compensate for the mechanical effect of the lesion itself.

Structural mechanics analysis and concepts from composite beam theory can be used to predict the load-bearing capacity of a bone with a lytic lesion since bone fails at a constant strain independent of bone density32,33. We previously demonstrated ex vivo that the force required to fracture a bone with a simulated lytic lesion was proportional to the structural rigidity of the weakest cross section through the bone by calculating the structural rigidities for serial transaxial cross sections through the bone at the site of the simulated lesion34,35. In the present study, we extended our structural mechanics analysis to the in vivo prediction of fracture risk in children and young adults with a variety of benign skeletal neoplasms. The specific aims of this in vivo diagnostic study were to (1) derive guidelines for accurately predicting a fracture through any benign osteolytic lesion involving the skeleton on the basis of the reduction in structural rigidity calculated from serial transaxial quantitative computed tomography images acquired through the lesion and the adjacent host bone, and (2) compare the sensitivity, specificity, and accuracy of current fracture-risk guidelines based on measurements made on plain radiographs with those of the new fracture-risk guidelines based on a measurement of the reduction in structural rigidity of the affected bone with quantitative computed tomography.

Back to Top | Article Outline

Materials and Methods

Study Cohort

After we obtained approval from the committee for clinical investigations of our institution, 150 children and young adults with a benign lesion of the appendicular skeleton underwent clinical, radiographic, and quantitative computed tomography evaluation between 1994 and 2001. Eighteen patients had a pathologic fracture through the benign osteolytic lesion. These patients were compared with patients who did not have a fracture and who met the following inclusion criteria: (1) the affected bone was at increased risk for fracture on the basis of clinical and radiographic criteria; (2) no treatment had been provided other than observation by the treating physician; and (3) the patient had been followed for a minimum of two years since identification of the benign bone lesion. Patients were excluded from the study if (1) they had received surgical treatment in the absence of a fracture, (2) they had used crutches or were advised to substantially decrease their activities for more than a few weeks (since this modification in activity alone would have decreased the fracture risk), (3) they had been followed for less than two years, or (4) quantitative computed tomography imaging of the affected and contralateral limbs had been technically unsatisfactory (e.g., serial computed tomography images of the same anatomic region in both limbs had not been made, a calcium hydroxyapatite bone phantom had not been included with the computed tomography images of the bones for converting Hounsfield units to equivalent bone mineral density, or serial transaxial computed tomography images had not been made 2.5 cm proximal and distal to the margins of the lesion). Eighteen patients without a fracture met the study inclusion criteria. The size, type, and anatomic location of the osteolytic lesions in the patients who were excluded from the study were not significantly different from those in the patients who were included in the study.

Back to Top | Article Outline
Structural Analysis with Quantitative Computed Tomography

The defect and the adjacent bone in the affected limb and homologous regions in the contralateral, normal limb of each patient were scanned with a high-speed helical quantitative computed tomography scanner (GE Highlight Advantage; General Electric, Waukesha, Wisconsin) with use of sequential, transaxial, cross-sectional slices (3.0 mm thick and spaced 3.0 mm apart) from 2.5 cm proximal to 2.5 cm distal to the margins of the defect. The typical field of view resulted in an inplane resolution of approximately 0.5 to 0.7 mm/pixel. The modulus of elasticity (in megapascals)—i.e., the intrinsic stiffness of the bone—for each pixel in the image was calculated from the apparent density with use of empirical relationships29,30 (see Appendix). Although bone is composed of mineral, organic matrix, and water, computed tomography primarily reflects the attenuation signature of the mineral phase. Therefore, the density measured with quantitative computed tomography approximates the density of the mineral phase, or the ash density (ρash, in grams per cubic centimeter). Six solid hydroxyapatite phantoms (Computerized Imaging Reference Systems, Norfolk, Virginia) of known mineral densities (0.003, 0.078, 0.178, 0.538, 1.048, and 1.597 gm/cm3) were placed in the same image field of view (Fig. 1). Linear regression between the attenuation coefficient and the corresponding ash densities for the phantom chambers was performed for each examination. The regression equations were used to convert the gray level of the quantitative computed tomography image (Hounsfield unit) into an equivalent bone density for each voxel element forming the image36. The axial rigidity, bending rigidity, and torsional rigidity for each transaxial cross-sectional image through the bone containing the lesion and the homologous cross section through the contralateral bone were calculated by summing the modulus-weighted area of each pixel comprising the bone section by its position relative to the centroid of the bone (Fig. 2 and Appendix). The rigidity of the affected bone was divided by the corresponding rigidity of the intact bone to calculate a ratio that expressed the reduction in the load-carrying capacity of the affected bone relative to the intact bone (Fig. 3). To account for a “worst-case” scenario, the minimum principal moments of inertia were used when these ratios were calculated.

In the fracture group, virtual reduction of the fracture fragments was performed with use of segmented three-dimensional reconstructed computed tomography images (AVS 5.0; Advanced Visual Systems, Waltham, Massachusetts) in order to represent the bone structure just prior to the occurrence of the fracture (Fig. 4).

Back to Top | Article Outline
Radiographic Analysis

Plain radiographic criteria for increased fracture risk associated with an osteolytic lesion included a defect length of 3.3 cm, a defect width of ≥2.5 cm, or involvement of ≥50% of the cortex3,4,23,31,37-44. Two observers not involved in the treatment of the patients independently evaluated the radiographs. Each observer used Vernier calipers to measure the defect length, defect width, and percentage of cross-sectional involvement of the cortex by the lesion on anteroposterior and lateral radiographs.

Back to Top | Article Outline
Statistical Analysis

Univariate analysis, with the Student t test, Fisher exact test, and chi-square test as appropriate, was used to compare demographic data, including age, gender, affected side, diagnosis, site of the defect, and affected bone, between the fracture and non-fracture groups. Radiographic and quantitative computed tomography variables were assessed for normality with use of the Kolmogorov-Smirnov test, and no significant skewness was detected. Therefore, the two-sample Student t test was used to compare the fracture and non-fracture groups. Sensitivity, specificity, and accuracy were calculated, with use of standard formulas45, to determine the diagnostic performance of radiographic and quantitative computed tomography parameters, and 95% confidence intervals were calculated with use of the Pratt approximation46. A power analysis with nQuery Advisor software (version 4.0; Statistical Solutions, Boston, Massachusetts) indicated that a sample size of thirty-six (eighteen cases and eighteen controls) would provide 80% power (α = 0.05, β = 0.2) to detect an effect size (mean difference/standard deviation) of 1.0 for each of the radiographic and quantitative computed tomography variables.

Multiple stepwise logistic regression analysis was performed to identify the optimal combination of quantitative computed tomography predictors and cutoff values to differentiate between fracture cases and controls. A likelihood ratio chi-square test was used to evaluate the significance of the variables in the final model, and the predicted probabilities of fracture were derived by maximum likelihood estimation46. Age, gender, diagnosis, pain at presentation, and defect location were also tested for significance as possible predictors of fracture. For all comparisons, a two-tailed p < 0.05 was considered significant. The data were analyzed with use of the SPSS software package (version 12.0; SPSS, Chicago, Illinois).

Back to Top | Article Outline

Results

Thirty-six patients with a benign bone lesion met the inclusion criteria for the study. Twenty patients were male and sixteen were female, and their ages ranged from two to nineteen years (mean [and standard deviation], 12.5 ± 3.6 years). Eighteen patients had a nonossifying fibroma; fourteen, a unicameral bone cyst; two, a fibrous dysplasia; and two, an aneurysmal bone cyst. Twenty lesions were located in the femur; eleven, in the tibia; three, in the humerus; one, in the ulna; and one, in the pelvis. The results of the univariate comparisons of the demographic data between the fracture and non-fracture groups are shown in Table I. The only significant difference was in the mean age of the patients at the time of the computed tomography, with the patients who had a fracture being younger than those who did not have a fracture (p < 0.001). In addition, clinical features, including age, gender, diagnosis, pain at presentation, and defect location, were evaluated with use of multiple logistic regression analysis to determine whether any of these characteristics were associated with fracture. Younger age was independently correlated with an increased risk of fracture (p < 0.01), whereas gender (p = 0.63), diagnosis (p = 0.48), pain (p = 0.27), and the site of the defect (p = 0.08) were not predictive of a fracture.

Standard radiographic criteria based on the size of the defect were neither sensitive nor specific for predicting fracture risk. There was no significant difference between the fracture and non-fracture groups with regard to any of the radiographic parameters for prediction of fracture risk on the basis of the lesion size (Table II). The accuracy of a defect length of ≥3.3 cm, a defect width of ≥2.5 cm, or destruction of ≥50% of the cortex as measured on anteroposterior and lateral radiographs was only fair (range, 42% to 61%) (Table III).

Reconstructed quantitative computed tomography images were available for seventeen patients who had a fracture. One set of quantitative computed tomography data concerning a proximal humeral fracture through a unicameral bone cyst was excluded because the “intact” bone could not be reconstructed with confidence as a result of the complexity of the fracture fragments. All biomechanical parameters derived with quantitative computed tomography differed significantly (p < 0.001) between the fracture and non-fracture groups (Table IV). All such parameters were 100% sensitive for predicting that a fracture would occur and were more specific than the radiographic criteria for predicting that a fracture would not occur through the lesion (Tables III and V). When the ratio of the minimum bending rigidity (EI) of the affected limb to that of the contralateral normal limb (EIlesion/EInorm) was <67% (i.e., the lesion reduced the bending rigidity of the bone by >33%), then bending rigidity was the most accurate (94% accurate) single structural parameter for predicting the occurrence of a fracture through the lesion.

Logistic regression modeling revealed that the ratio of bending rigidities (EIlesion/EInorm) (p < 0.0001) and the ratio of torsional rigidities (GIlesion/GInorm) (p < 0.0001) were each highly informative quantitative computed tomography-derived biomechanical parameters for predicting fracture occurrence. This finding is consistent with the clinically observed fracture patterns that implicated bending (Figs. 5-A and 5-B) and/or torsion (Figs. 5-C and 5-D) as the mechanisms of failure. The relationship between the ratio of bending rigidities and the probability of fracture and the relationship between the ratio of torsional rigidities and the probability of fracture are presented in Figures 6-A and 6-B. An EIlesion/EInorm ratio of 50% is associated with an estimated probability of fracture of 98%, whereas an EIlesion/EInorm ratio of 75% is associated with an estimated probability of fracture of <10%. Similarly, the theoretical fracture-probability curve indicates that a GJlesion/GJnorm ratio of 50% is associated with an estimated fracture risk of nearly 100%, a GJlesion/GJnorm ratio of 75% indicates an estimated fracture risk of approximately 50%, and a GJlesion/GJnorm ratio of 85% indicates an estimated fracture risk of only 5%.

Multiple stepwise logistic regression analysis indicated that bending rigidity and torsional rigidity were each highly significant independent predictors of fracture occurrence and, together, provided the optimal combination of quantitative computed tomography-derived biomechanical parameters for differentiating patients with a fracture from those without a fracture. Although a univariate analysis based on Student t tests indicated that total bone mineral content (BMC) and axial rigidity (EA) were significantly lower (p < 0.001) in the fracture group (Table IV), regression analysis indicated that neither the BMClesion/BMCnorm ratio (p = 0.91) nor the EAlesion/EAnorm ratio (p = 0.41) provided additional predictive information beyond that provided by the combination of EIlesion/EInorm and GJlesion/GJnorm ratios. A risk-of-fracture (ROF) equation was derived from the multiple stepwise logistic regression model as a function of the minimum values for the EIlesion/EInorm and GJlesion/GJnorm ratios calculated for the transaxial cross sections through the lesion (but not necessarily on the same cross section) with use of the maximum likelihood estimation:

where exp denotes the base of the natural logarithm (approximately 2.71), β indicates the coefficients of the regression model, EI is the minimum EIlesion/EInorm ratio, and GJ is the minimum GJlesion/GJnorm ratio over all cross sections through the lesion. The β estimates were determined with use of EI and GJ as continuous variables. With use of a fracture-risk probability threshold of >33%, the resultant combination of bending and torsional rigidities was 97% accurate and 100% sensitive for detecting patients with a fracture and 94% specific for detecting patients without a fracture.

Back to Top | Article Outline

Discussion

Figure. No caption a...
Figure. No caption a...

One problem with predicting a pathologic fracture in a child with a benign bone lesion is that the natural history of many benign bone disorders is poorly understood2,10,11,25. Since skeletal lesions are often initially diagnosed on the basis of plain radiographs, several investigators have attempted to estimate the load-bearing capacity of a bone with a lytic lesion by measuring the geometry of the defect on the radiograph. Previous investigators have considered the defect geometry, pain, and the anatomic site, type, and activity of the lesion to be predictors of the risk of a fracture through an osteolytic lesion of the appendicular skeleton4,7,16,23-27,31,38-44,47-50. Those authors developed guidelines for specific lesions on the basis of retrospective reviews of clinical cases. In general, it was hypothesized that absolute size parameters correlated directly with the risk of pathologic fracture on the basis of the assumption that the larger the lesion, the greater the mechanical defect and thus the greater the chance for a fracture. However, these investigators failed to derive any method based on clinical and/or radiographic findings that accurately predicted the occurrence of a pathologic fracture prospectively16,37,51. While patient age and lesion activity were frequently identified as important factors, no investigator took into account the remodeling potential of the host bone in response to these lesions or considered the structural mechanics that ultimately determine the load-bearing capacity of a bone with a lesion. Furthermore, each of the methods reported for fracture prediction pertained to only a specific type of lesion; none of the methods were globally applicable to any type of benign osteolytic neoplasm. In the present study, we used structural engineering parameters derived from transaxial quantitative computed tomography images of the involved bone in an attempt to overcome these weaknesses.

Nonossifying fibroma was the most common diagnosis in our series. Arata et al.4 proposed that lower-extremity nonossifying fibromas that were >50% of the transverse cortical diameter as seen on both anteroposterior and lateral radiographs and that were >33 mm in length were predictive of a pathologic fracture. Our results suggest that these absolute lesion-size parameters do not accurately predict a fracture.

Figure. No caption a...
Figure. No caption a...

Pain is a common presenting symptom and is assumed to indicate progressive accumulation of damage in the bone prior to fracture; however, this has never been demonstrated conclusively42. Drennan et al. suggested that a bone with a large nonossifying fibroma causing pain might be predisposed to fracture15. Easley and Kneisl noted that, whereas three of their thirteen patients who did not have a fracture had pain, none of their nine patients with a fracture had prodromal pain and all had been playing sports (soccer, basketball, or jumping on a trampoline) at the time of injury16. Pain was not independently predictive of fracture risk in our study.

Nine of our eighteen patients with a fracture and seven of our eighteen patients without a fracture had a cystic lesion (an aneurysmal or unicameral bone cyst). In a retrospective review of the cases of seventy-five children with a unicameral bone cyst, fifty-two of whom had sustained a pathologic fracture, Ahn and Park noted that every cyst that was associated with a fracture had destroyed >85% of the cortex in the transverse plane as seen on both the anteroposterior and the lateral radiographs23. They did not report the specificity of this threshold for predicting fracture. In our series, the percentage of cortical destruction measured on either the anteroposterior or the lateral radiograph did not differ significantly between the fracture and non-fracture groups.

Nakamura et al. reported that the possibility of a fracture through a unicameral bone cyst was “very high” if the cyst wall thickness was <0.5 mm but that no fracture would occur if the cyst wall was thicker than 10 mm24. However, measurement of the cyst wall thickness relative to the bone circumference is highly dependent on the angle of the beam projected onto the plane of the radiograph, so the true thickness of the cyst wall cannot be reliably measured on plain radiographs unless this angle is known. Nakamura et al. also determined the average mineral content of the cyst from the gray-level attenuation measured on plain radiographs normalized by the width of the bone, and they proposed that this “microdensity” value be used to predict the risk of subsequent fracture after intracyst steroid injection. The authors did not validate this assertion or correlate this measurement to bone strength.

Kaelin and MacEwen performed a retrospective study of fifty-seven patients with a unicameral bone cyst and derived a “cyst index” (the area of the cyst on a plain radiograph divided by the square of the diameter of the bone diaphysis at its tubular part) to predict the risk of recurrent pathologic fracture and to assess the response of the affected bone to treatment after intracyst steroid injection26. A cyst index of >4.0 in the humerus and >3.5 in the femur was proposed as a predictor of high fracture risk, and indices below those thresholds were reported to be predictive of low fracture risk. Vasconcellos et al. conducted a retrospective study of thirty-two patients with a unicameral bone cyst of the humerus or femur to evaluate the reproducibility and utility of the cyst index as a predictor of initial bone fracture51. Although the diaphyseal diameter could be measured reliably, measurement of the cyst area was too subjective to produce an acceptable repeatability index. A cyst index of >4.0 yielded a sensitivity of 100%, a specificity of only 13%, a positive predictive value of 24%, and a negative predictive value of 100% for prediction of a fracture in the humerus at an average of 1.3 years after the radiograph was analyzed. With regard to prediction of a fracture in the femur, a cyst index of >3.5 was 50% sensitive and 33% specific, with a positive predictive value of 33% and a negative predictive value of 50%, at an average of 1.4 years after the radiograph was analyzed. Lowering the threshold value below that at which no fractures were observed gave the cyst index high sensitivity but low specificity and a high negative predictive value but a low positive predictive value. Therefore, the cyst index could not reliably discriminate between patients who would sustain a fracture and those who would not.

We also evaluated radiographic parameters that have been frequently reported to be predictive of a pathologic fracture through an osteolytic skeletal metastasis. These include (1) a lesion of ≥2.5 cm in diameter as a predictor of fracture, and (2) ≥50% cortical destruction as an indication for prophylactic stabilization31,38-41. These guidelines arise from several retrospective clinical studies of adults with skeletal metastases; however, neither guideline has been confirmed experimentally ex vivo or evaluated prospectively in vivo. The 2.5-cm-defect guideline was originally based on an initial review of the cases of nineteen patients and a subsequent review of the cases of 338 patients with a femoral fracture associated with a well-defined radiolucent defect of at least 2.5 cm in diameter involving the cortex42. This criterion was not verified at anatomical sites other than the femur. Critical review of the initial series of nineteen pathologic fractures reveals that eight of the fractures occurred in patients who did not have a defect of at least 2.5 cm in diameter. This finding corresponds to a false-negative rate of 42%. Furthermore, the guideline is difficult to apply to permeative lesions, which have indistinct borders.

Parrish and Murray proposed considering 50% cortical destruction as an indication for prophylactic stabilization on the basis of a retrospective study of four adult patients thought to be at risk for impending fracture. The same guideline was then used to appraise ninety-six patients with a total of 100 skeletal metastases40. The percentage of cortical involvement was measured on radiographs by estimating the maximum width of the defect and dividing it by the width of the bone. The patients were divided into four groups depending on whether 0% to 25%, 25% to 50%, 50% to 75%, or 75% to 100% of the cortex was involved. While there was considerable overlap in defect size between the fracture and non-fracture groups, only one fracture occurred in a bone with <50% cortical destruction. Neither the interobserver nor the intraobserver accuracy was evaluated. In our study, the percentage of cortical destruction was relatively sensitive (83%) but nonspecific (39%). The flaw with an analysis of fracture risk based on cortical destruction is that it fails to account for the cortical thickness: e.g., if the same size defect occurred in two bones with the same outer diameter but different cortical thicknesses, then the total cross-sectional area of bone destroyed would be greater in the thicker-walled bone, although the percentage of cortical destruction based on the ratio of the projected width of the defect to the bone diameter would be the same.

Ex vivo studies of the behavior of bones with simulated defects have indicated that the load-carrying capacity of a bone with a defect depends on many factors that are not represented by radiographic criteria based on lesion size alone. The critical parameters that determine the load-bearing capacity of a bone containing a lytic lesion include the amount of bone loss, the cross-sectional structural geometry of the bone, the material properties of the remaining bone tissue, the location of the lesion with respect to the applied loads, and the loading mode34,37. The advantage of our analysis of structural rigidity with quantitative computed tomography is that it reflects changes in both the material properties of the bone tissue and the bone cross-sectional geometry for any bone or lesion. Bone mineral density alone, which is often used to predict the risk of an osteoporotic fracture in the elderly, did not prove to be as discriminatory as the structural rigidity parameters, since mineral density does not reflect how the bone mineral is distributed in space, which determines the load-carrying capacity of the bone.

Fracture-risk indices based on lesion size alone fail to account for the compensatory remodeling of the host bone that occurs in response to the lesion. Fibrous lesions tend to induce cortical thickening around their margins, and septa form within the lesion and buttress it against collapse. Cystic lesions induce periosteal expansion of the host bone, which serves to increase the bone's moment of inertia. These remodeling strategies partially compensate for the structural consequences of the lesion itself. Accurate predictions of fracture risk can be made only when the structural properties of the entire bone containing the lesion are taken into account. We demonstrated that the minimum bending and torsional rigidities measured on serial transaxial cross-sectional computed tomography images through the lesion and host bone were more sensitive and specific for predicting pathologic fracture in children with a benign bone lesion than were standard clinical and radiographic criteria based on the size of the lesion alone. A clinical example would be two patients who each have a large unicameral bone cyst in the proximal part of the femur of a size that exceeds all radiographic criteria for increased fracture risk; the structural rigidity analysis could indicate that the fracture risk was increased for the first patient (Fig. 7-A) but not for the other patient because compensatory remodeling counterbalanced the reduction in bending and torsional rigidity caused by the osteolytic lesion (Fig. 7-B).

One limitation of this analysis is that it required a normal contralateral bone for comparison. Because of the diverse range of patients' activities, the actual loads applied to the bones were unknown. Therefore, we compared the structural properties of the affected bone with those of the contralateral, normal bone at homologous cross sections to calculate the reduction in the load-carrying capacity of the affected bone. We previously evaluated the variation from bilateral symmetry for bending and axial rigidity of the femora of middle-aged adults52. The average differences in axial and bending rigidity between the left and right femora at the subcapital, base-of-the-neck, intertrochanteric, subtrochanteric, and mid-diaphyseal regions ranged from 2% to 6%. While individual differences of up to 13% were observed, these differences are smaller than the 33% reduction in bending rigidity that we found to be the threshold for fracture occurrence. If the other limb is also pathologically involved, as seen with polyostotic fibrous dysplasia, osteogenesis imperfecta, metabolic bone disease, and metastatic cancer, one cannot assume that the contralateral limb provides an appropriate internal control. In that case, the fracture risk must be calculated by comparing the estimated load applied to the bone for a specific activity to the calculated failure load of the bone.

Another limitation of our computed tomography-based technique is that it exposes the patient to relatively large doses of radiation, so that repeated measurements are less desirable in children. We are currently developing a quantitative magnetic resonance imaging-based method of analysis53 to eliminate any exposure to radiation, and newer multidetector helical computed tomography scanners expose patients to much lower doses of radiation. Other limitations of our computed tomography-based analysis are that computed tomography scanners are not as readily available as standard radiographic units, the application of our algorithm requires sophisticated knowledge of image-analysis software to properly align virtual images of right and left bone pairs, and the interpretation of the results requires a background in structural mechanics. Some of these issues can be addressed by allowing clinicians to send computed tomography images by means of the Internet, over secure lines, to be analyzed by our research group.

An important advantage of our computed tomography-based analysis of structural rigidity is that it provides objective data that can be used to serially evaluate the progression or regression of a lesion in response to treatment. One of the most difficult aspects of managing a child with a benign bone lesion is deciding on the proper course of treatment. Most of these children are active and participate in sports or other activities associated with a high fracture risk. The clinician is often frustrated by the lack of data to help decide whether activity modification or surgical intervention is necessary. Our computed tomography-based structural analysis could be used to choose among various treatment options on the basis of the relative probability of fracture and the risks associated with each treatment. Unlike measurements of bone mineral content and density with dual energy x-ray absorptiometry (DEXA), our computed tomography-based rigidity analysis accounts for changes in bone size related to growth and remodeling. As with DEXA, anatomic landmarks can be used to uniformly reference relevant cross sections for computed tomography analysis to facilitate repeated measures over time.

In summary, the combination of the ratios of bending and torsional rigidities as determined with computed tomography-based structural analysis of homologous regions of affected and normal bones provides a sensitive and specific method for accurately predicting fracture risk in children with a variety of benign neoplasms involving the skeleton. Several ex vivo studies34,35,37 as well as the present in vivo study have shown that the often-cited radiographic guidelines of a 2.5-cm lesion width, a 3.3-cm lesion length, and/or destruction of 50% of the cortex are associated with large errors in estimating the load-bearing capacity of the affected bone. Computed tomography-based structural analysis enables the orthopaedic surgeon to quantify the mechanical properties of the bone at a particular site and to determine if there is a sufficient reduction in the load-carrying capacity of the bone to merit intervention.

Back to Top | Article Outline

Appendix

The details of the calculations used in this study are available with the electronic versions of this article, on our web site at jbjs.org (go to the article citation and click on “Supplementary Material”) and on our quarterly CD-ROM (call our subscription department, at 781-449-9780, to order the CD-ROM). ▪

NOTE: The authors greatly appreciate the assistance of Betsy Meyers, PhD, who performed the initial statistical analysis.

A commentary is available with the electronic versions of this article, on our web site () and on our quarterly CD-ROM (call our subscription department, at 781-449-9780, to order the CD-ROM).

In support of their research for preparation of this manuscript, one or more of the authors received grants or outside funding from the Whitaker Foundation, Rosslyn, Virginia, and National Cancer Institute Extramural Activities Grant EPN/636 2 RO1 CA40211-13. None of the authors received payments or other benefits or a commitment or agreement to provide such benefits from a commercial entity. No commercial entity paid or directed, or agreed to pay or direct, any benefits to any research fund, foundation, educational institution, or other charitable or nonprofit organization with which the authors are affiliated or associated.

Investigation performed at Children's Hospital and the Orthopaedic Biomechanics Laboratory, Beth Israel Deaconess Medical Center, Boston, Massachusetts

1. , Kennon RE, Jelinek JE. Benign bone tumors of childhood. J Am Acad Orthop Surg. 1999;7: 377-88.
2. , Flynn JM. Pathologic fractures associated with tumors and unique conditions of the musculoskeletal system. In: Beaty JH, Kasser JR, editors. Rockwood and Wilkins' fractures in children. 5 th ed. Philadelphia: Lippincott, Williams and Wilkins; 2001. p 139-240.
3. , Wilkins R, Conrad EU 2nd. Benign bone tumors. Instr Course Lect. 1996;45: 425-46.
4. , Peterson HA, Dahlin DC. Pathological fractures through non-ossifying fibromas. Review of the Mayo Clinic experience. J Bone Joint Surg Am. 1981;63: 980-8.
5. . Solid aneurysmal bone cyst with pathologic bone fracture. Skeletal Radiol. 1995;24: 214-6.
6. , Windhager R, Lang S, Richling B, Kotz R. [Incidence or recurrence of aneurysmal bone cysts following surgical treatment and adjuvant therapy with phenol]. Z Orthop Ihre Grenzgeb. 1995;133: 422-8. German.
7. , Lamont-Havers W, Schweon D, Kliman A. Pathologic fracture risk in rehabilitation of patients with bony metastases. Clin Orthop Relat Res. 1985;192: 222-7.
8. , Taminiau AH. Simple bone cysts treated by multiple drill-holes. Acta Orthop Scan. 1997;68: 86.
9. , Marcove RC, Huvos AG, Mike V. Aneurysmal bone cysts. A clinicopathologic study of 66 cases. Cancer. 1970;26: 615-25.
10. . On fibrous defects in cortical walls of growing tubular bones: their radiographic appearance, structure, prevalence, natural course, and diagnostic significance. Adv Pedatr. 1955;7: 13-51.
11. , Capanna R. Unicameral bone cysts. Round table discussion. Treatment comparison of steroid injection versus surgery. In: Uhthoff HK, editor. Current concepts of diagnosis and treatment of bone and soft tissue tumors. New York: Springer; 1984. p 321-7.
12. , Capanna R, Picci P. Unicameral and aneurysmal bone cysts. Clin Orthop Relat Res. 1986;204: 25-36.
13. , Dal Monte A, Gitelis S, Campanacci M. The natural history of unicameral bone cyst after steroid injection. Clin Orthop Relat Res. 1982;166: 204-11.
14. , De Nayer P, Malghem J, Noel H. Induced healing of aneurysmal bone cysts by demineralized bone particles. A report of two cases. Arch Orthop Trauma Surg. 1996;115: 141-5.
15. , Maylahn DJ, Fahey JJ. Fractures through large non-ossifying fibromas. Clin Orthop Relat Res. 1974;103: 82-8.
16. , Kneisl JS. Pathologic fractures through nonossifying fibromas: is prophylactic treatment warranted? J Pediatr Orthop. 1997;17: 808-13.
17. . Prophylactic internal fixation of secondary neoplastic deposits in long bones. Br Med J. 1973;1: 341-3.
18. , Dos Santos JB, Korukian M, Laredo Filho J. Conservative treatment of solitary bone cysts—a study of 55 patients. Rev Paul Med. 1992;110: 131-7.
19. , Reinus WR, Wilson AJ. Quantitative analysis of the plain radiographic appearance of aneurysmal bone cysts. Invest Radiol. 1995;30: 433-9.
20. . Letter: the fate of simple bone cysts which fracture. Clin Orthop Relat Res. 1974;101: 302-4.
21. , Gaur S, Sharma S. Percutaneous autogenous bone marrow grafting in 20 cases of ununited fracture. Acta Orthop Scand. 1993;64: 671-2.
22. , Kriger J, Bronfman R, Lauf E. Unicameral bone cysts: treatment with methylprednisone acetate injections. J Foot Ankle Surg. 1994;33: 6-15.
23. , Park JS. Pathological fractures secondary to unicameral bone cysts. Int Orthop. 1994;18: 20-2.
24. , Takagi K, Kitagawa T, Harada M. Microdensity of solitary bone cyst after steroid injection. J Pediatr Orthop. 1988;8: 566-8.
25. , Francis KC, Johnson AD, Kiernan HA Jr. Current concepts on the treatment of solitary unicameral bone cyst. Clin Orthop Relat Res. 1973;97: 40-51.
26. , MacEwen GD. Unicameral bone cysts. Natural history and the risk of fracture. Int Orthop. 1989;13: 275-82.
27. . Biomechanics of cortical and trabecular bone: implications for assessment of fracture risk. In: Mow VC, Hayes WC, editors. Basic orthopaedic biomechanics. New York: Raven Press; 1991. p 93-142.
28. , Rubin D, Krempl E. Introduction to continuum mechanics. 3rd ed. Boston: Butterworth-Heinemann; 1993.
29. , Cowin SC, Bowman JA. On the dependence of elasticity and strength of cancellous bone on apparent density. J Biomech. 1988;21: 155-68.
30. , Schneider E. Estimation of mechanical properties of cortical bone by computed tomography. J Orthop Res. 1991;9: 422-31.
31. . Impending fracture associated with bone destruction. Orthopedics. 1992;15: 547-50.
32. , Keaveny TM. Yield strain behavior of trabecular bone. J Biomech. 1998;31: 601-8.
33. . Yield behavior of bovine cancellous bone. J Biomech Eng. 1989;111: 256-60.
34. , Kwak SD, Tedrow JR, Inoue K, Snyder BD. Noninvasive imaging predicts failure load of the spine with simulated osteolytic defects. J Bone Joint Surg Am. 2000;82: 1240-51.
35. , Cabe GD, Tedrow JR, Hipp JA, Snyder BD. Failure of trabecular bone with simulated lytic defects can be predicted non-invasively by structural analysis. J Orthop Res. 2004;22: 479-86.
36. , Rosenthal DI. Quantitative computed tomography scanning for measurement of bone and bone marrow fat content. A comparison of single- and dual-energy techniques using a solid synthetic phantom. Invest Radiol. 1987;22: 799-810.
37. , Springfield DS, Hayes WC. Predicting pathologic fracture risk in the management of metastatic bone defects. Clin Orthop Relat Res. 1995;312: 120-35.
38. , Lawton GD, Snell WE. Prophylactic internal fixation of the femur in metastatic breast cancer. Cancer. 1971;28: 1350-4.
39. . Incidence of fracture through metastases in long bones. Acta Orthop Scand. 1981;52: 623-7.
40. , Murray JA. Surgical treatment for secondary neoplastic fractures. A retrospective study of ninety-six patients. J Bone Joint Surg Am. 1970;52: 665-86.
41. , Sim FH, Springfield DS. Metastatic disease of the femur. Orthopedics. 1992;15: 621-30.
42. , Sellinger DS, McBeath AA, Engber WD. Metastatic breast cancer in the femur. A search for the lesion at risk of fracture. Clin Orthop Relat Res. 1986;203: 282-8.
43. , Mouradian WH. Intramedullary fixation of pathological fractures and lesions of the subtrochanteric region of the femur. J Bone Joint Surg Am. 1976;58: 1061-6.
44. . Metastatic disease in long bones. A proposed scoring system for diagnosing impending pathologic fractures. Clin Orthop Relat Res. 1989;249: 256-64.
45. . Approximate binomial confidence limits. J Am Stat Assoc. 1986;81: 843-55.
46. , Day NE. Statistical methods in cancer research. Volume I - The analysis of case-control studies. IARC Sci Publ. 1980;32: 5-338.
47. , Keaveny TM. The dependence of shear failure properties of trabecular bone on apparent density and trabecular orientation. J Biomech. 1996;29: 1309-17.
48. , Seitz CB, Eyre HJ. Nonoperative management of femoral, humeral, and acetabular metastases in patients with breast carcinoma. Cancer. 1980;45: 1533-7.
49. . New trends in the management of lower extremity metastases. Clin Orthop Relat Res. 1982;169: 53-61.
50. , Cummings S, Browner WS, Newman TB, Hearst N. Designing clinical research: an epidemiologic approach. Baltimore: Williams and Wilkins; 1988.
51. , Grace A, Lurie F, Fillman R, Moritz B, Yandow S. Cyst index as a predictor of simple bone cyst fracture. Read at the Annual Meeting of the Pediatric Orthopaedic Society of North America. 2003 May 2-4; Amelia Island, FL.
52. , Pierre MA, Snyder BD. Bilateral variation of bone structure of the femur in normal adults measured by ex-vivo orthogonal radiography, DEXA, and QCT. Trans Orthop Res Soc. 2001;26: 313.
53. , Hipp JA, Mulkern RV, Jaramillo D, Snyder BD. Magnetic resonance imaging measurements of bone density and cross-sectional geometry. Calcif Tissue Int. 2000;661: 74-8.
Copyright 2006 by The Journal of Bone and Joint Surgery, Incorporated