The Effects of Metastatic Lesion on the Structural Determinants of Bone: Current Clinical and Experimental Approaches

Abstract

Metastatic bone disease is incurable with an associated increase in skeletal-related events, particularly a 17–50% risk of pathologic fractures. Current surgical and oncological treatments are palliative, do not reduce overall mortality, and therefore optimal management of adults at risk of pathologic fractures presents an unmet medical need. Plain radiography lacks specificity and may result in unnecessary prophylactic fixation. Radionuclide imaging techniques primarily supply information on the metabolic activity of the tumor or the bone itself. Magnetic resonance imaging and computed tomography provide excellent anatomical and structural information but do not quantitatively assess bone matrix. Research has now shifted to developing unbiased data-driven tools that can predict risk of impending fractures and guide individualized treatment decisions. This review discusses the state-of-the-art in clinical and experimental approaches for prediction of pathologic fractures with bone metastases. Alterations in bone matrix quality are associated with an age-related increase in skeletal fragility but the impact of metastases on the intrinsic material properties of bone is unclear. Engineering-based analyses are non-invasive with the capability to evaluate oncological treatments and predict failure due to the progression of metastasis. The combination of these approaches may improve our understanding of the underlying deterioration in mechanical performance.

1. Introduction

Metastatic bone disease (MBD), resulting from primary tumor invasion to bone presents a major clinical concern [1–3]. MBD leads to several complications [4,5], including hypercalcemia, bone marrow failure, anemia, and more commonly, skeletal-related events (SREs). SREs are defined as bone pain, a pathological fracture, spinal cord compression, and necessity for radiation or surgery to bone [6,7]. Cancer-induced bone pain afflicts 28–45% of patients with bone metastasis [8], and often requires palliative radiation therapy (RT), opioid analgesia, nonsteroidal anti-inflammatory drugs, or bisphosphonates [9–11]. Breast and prostate carcinomas have the greatest tendency to metastasize to bone (65–75%), followed by thyroid (60%), lung (30–40%), and renal (20–25%) carcinomas [8,12,13]. The spine and the pelvis are most frequently sites affected by metastasis [14–17] with up to 73% of subjects with breast, 68% of prostate, and 35% of kidney and lung cancers were found to have vertebral metastases at autopsy [8]. In the appendicular Skelton, the humerus and femur are also common sites for metastases (4). In the spine, 15–20% of prostate, breast, and lung cancer patients suffer an SRE [18], and 9–39% of these patients are at high risk for SRE-based complications for which pathologic fractures are an integral part [19–22]. Metastatic epidural spinal cord compression, which can cause paralysis, is the most dreaded spinal related complication.

Surgery and other oncological treatments for SREs pose significant health and economic burden to this frail patient population, their immediate families, and society as a whole [23,24]. MBD patients report clinically significant decreases in quality of life parameters, including physical (e.g., pain, fatigue), emotional (e.g., depression, anxiety), and functional well-being (e.g., mobility, independence) [15,7,25,24]. Depending on the type of primary tumor and site of metastasis within the skeleton, the risk of pathological fractures ranges from 17–50% [26,1,27]. Aging or rapid degradation in bone properties associated with endocrine therapy for cancer expose patients to additional morbidity due to increased risk of osteoporotic vertebral and hip fractures [28]. The assessment of fracture risk is thus critical both at the time of diagnosis and during treatment. Although several systems exist for predicting overall survival in patients with spinal bone metastases [29], predictive accuracy for fracture risk is too low to be actionable. This poor prediction is due to the limited knowledge of the effect of metastases on the material, mechanical, and architectural properties of trabecular and cortical bone. Combined, these properties determine the ability of the skeletal structure to carry loads of daily living. Surgical intervention is only attempted in response to severe symptoms of pain or neurological deficits. Once the fracture has occurred, treatment may require an emergency surgical repair of the affected skeletal region with the prospects for recovery depending on the severity of the structural deficit and the time elapsed between the collapse and treatment. In this cohort with limited life expectancy and quality of life, further deterioration and surgery must be avoided if possible [30–33]. Hence, optimal management of adults with the metastatic bone disease who are at risk of fracture represents an important unmet medical need.

1.1. Radiological classification of bone metastases

Bone is a composite hierarchically structured material built from organic (proteins) and mineral (hydroxyapatite) components. Bone metastasis disrupts skeletal homeostasis [34] and demands an understanding of the underlying mechanisms that contribute to the deterioration of bone’s mechanical performance. Radiologically, bone metastases appear as osteoblastic (bone-forming), osteolytic (bone-destroying), or mixed lesions containing both types. Figure 1 demonstrates clinical CT, MR and radionuclide bone scan images of mixed spinal metastases. The lesions present as high density on CT, reflecting the net increased in bone formation compared to normal tissue. The MR images demonstrate markedly hypointense marrow due to loss of normal fat best illustrated on the T1 image. The STIR image demonstrates dramatic low signal due to the loss of hematopoietic marrow and replacement of marrow space with expanded calcified trabecular bone. The radionuclide bone scan is sensitive neither the calcium density in the bone, nor to the composition of the marrow space. Instead, the high intensity on this study reflects the high level of calcium bone formation in the diffuse metastases. Osteoblastic lesions, typical of prostate and many breast primary cancers, appear as sclerotic (i.e., increasing bone density on X-ray imaging) [35]. These lesions are generally accepted as evidence of a less aggressive tumor with relatively slower growth and has been associated with prolonged survival [36,37]. Osteolytic lesions, typical of breast, lung, urinary, and myeloma, appear as three types: 1) geographic: large, solitary, well defined, radiolucent areas with sharply demarcated edges; 2) moth-eaten: multiple smaller radiolucent areas (2–5 mm) that may coalesce to form larger areas and have ill-defined margins; and 3) Permeative: usually occurring in the most aggressive lesions, manifest as multiple tiny radiolucent areas (<1mm) seen principally in cortical bone [38]. Mixed lesions, often observed in patients with breast and lung cancers [39], are thought to occur due to secondary bone formation in an attempt to repair bone loss. Figure 2 presents high magnification microCT- image of human vertebrae with lytic, mixed and blastic lesions and their effect on the architecture and density of the vertebral bone.

Figure 1,

Sagittal reformatted CT, Sagittal T1 MR, Sagittal STIR MR, and coronal radionuclide bone scan of mixed spinal metastases. The striking low signal on MR images is due to the replacement of normal marrow adipose tissue with tumor cells and sclerosis that replaces marrow space with dense bone.

Figure 2,

Radiographic appearance of the effect of lytic, mixed, and osteoblastic metastases on vertebral bone architecture. The osteolytic lesion leads to rarefication of bone trabecula within the vertebra, which, on larger magnification, can be observed to be driven by both the physical loss of bone trabecula and thinning of the remaining trabeculae. For both the vertebra with mixed metastases and with osteoblastic metastases, the changes in bone architecture can be observed to be driven by the thickening of the trabeculae, growth of secondary trabeculae in between the trabeculae and unorganized remodeling leading to large deposition of bone material.

1.2. Clinical Assessment of bone metastases

About 75% of patients with bone metastases present with pain, requiring further workup [40]. Imaging forms a critical component in identifying the extent of anatomical involvement and monitoring the development and progression of metastatic tumors. Plain radiography is sometimes the first imaging test performed. It is relatively inexpensive and accessible, providing high-resolution information on the mineral aspect and simple two-dimensional geometric measurements of cortical bone. It has been employed to assess the risk of pathological fractures in human femurs based on measures of tumor size, location, and type of lesion [41]. However, it has been estimated that by the time a lesion becomes radiographically detectable in long bone, around 25–75% loss of bone mineral has occurred [42], resulting in the bone involved to have weakened significantly [43,42]. Lack of standardize criteria for patient positioning [44] and in the inability predict fracture based on radiographs in case of permeative or diffuse lesions without clear boundaries [45–47] further limits the application of this modality. Critically, although radiographic based studies have identified possible risk factors for fracture risk of femurs with bone metastases, their positive predictive value (PPV) is very low [48], resulting in up to two thirds of the patients without an impending fracture being refereed to preventive surgery [45]. Section 2 presents a detailed discussion on the limitation of the prediction of the fracture risk based on radiographs.

In the spine, plain radiography based classifications offer only modest specificity [49], resulting in patients exposed to unnecessary prophylactic fixation with attendant operative risks and complications, while missing many patients who have weakened vertebral bodies due to metastases not detected with this imaging modality [50]. Plain radiography has low sensitivity to density differences as compared to X-ray computed tomography (CT) and is not quantitative. Therefore, only large alterations of density are detected with this technique.

Although similar to plain radiography in that both primarily assess the distribution of calcium in bone, CT is far more sensitive than plain radiography to calcium content. CT provides quantitative volumetric information and displays the architecture of cortical and trabecular bone. Thus, CT is superior in identifying the degree of destruction of these osseous components by the tumor [48]. Magnetic resonance imaging (MRI) provides excellent soft-tissue resolution, does not expose patients to ionizing radiation, and can help discriminate between metastatic and osteoporotic vertebral compression fractures [51].

Magnetic resonance imaging has long been known to be far more sensitive than CT for the detection of osseous metastases [52,53]. This higher sensitivity is due to the CT’s ability to detect changes only in the calcified matrix of the cortical and trabecular bone whilst being remarkably insensitive to changes in the composition of the marrow space. Thus, tumors can completely replace the normal mixture of adipose tissue and hematopeotic cells in adult marrow with the CT images remaining normal (see Figure 3). By contrast, due to the MR exquisite sensitivity to the composition of marrow space with the replacement of adipose in marrow with tumor cells produces dramatic changes in MR signal intensity. Other MR characteristics, such as blood flow or diffusion properties, have less well-established value for detecting the tumor. Metastases have been reported to have slow diffusion and high blood flow.

Figure 3.

On the sagittal MR image of the spine, [A], the replacement of normal mixture of adipose tissue and hematopeotic cells in the adult marrow with the tumor tissue results in dramatic changes in MR signal intensity (highlight by the yellow and red arrows). By contrast, on the corresponding sagittal spine CT image, [B], the osseous tissue appears normal. Note the difference in the appearance of the failed level (marked by the white star symbol).

MR is intermediate in its ability to further characterize the marrow space. Tumor cells typically demonstrate high glucose utilization and thus high uptake of FDG. Detecting this does not depend on bone formation (osteoblastic) or bone resorption (osteoclastic) activity. Nuclear medicine techniques, including positron emission tomography (PET) and single-photon emission computed tomography (SPECT), determine the uptake of radionuclides by tumor or bone itself. Radionuclide bone scanning most often uses ⁹⁹mTc with methylene diphosphonate (MDP). This adsorbs to the mineralization front of bone, thus reflecting the activity of osteoblasts. Note that almost all osteolytic tumors also have elevated osteoblastic activity and are thus well displayed on bone scans. Multiple myelomas is a well-known exception to this, with lesions often demonstrating little or no uptake on bone scan.

PET imaging (18F FDG PET) is most commonly employs fluorodeoxyglucose (FDG) to detect energy metabolism in tumors. Neoplastic cells typically employ anaerobic glycolysis (Warburg effect), leading to high rates of glucose consumption. By imaging FDG uptake, one can detect cancer in bones and elsewhere. Although both ⁹⁹mTc bone scanning and PET scanning are valuable in evaluating metastatic disease, they reflect different processes and provide different information. ⁹⁹mTc bone scanning is far more sensitive than plain radiography for detecting skeletal metastases but offers quite limited anatomic detail. Data consistently indicate that tumor metabolism (FDG PET) is more sensitive than osteoblast activity (99mTc methods).

Studies employing meta-analysis to compare the diagnostic value of these imaging modalities in detecting vertebral metastases reported the sensitivity of CT, MR, 18F-FDG PET, and 99mTc MDP bone scintigraphy (BS) to be; 72.9 –, 92.3%, 90.6 – 94.1%, 89.7 – 89.8 %, 80.0 – 86.0% [52,53], and for BS with single-photon emission computed tomography respectively BS-SPECT at 90.3% [53]. For a review of these modalities and their ability to highlight the morphological and metabolic changes that occur in the skeleton due to bone metastases, see Vassiliou et al. [54].

The continual technical development and advancement of clinical imaging modalities provide increases in these modalities specificity and sensitivity to evaluate the lesion mediated degradation of the bone as well as measures of its regeneration with treatment. However, it is important to note that a reduction in tumor growth and burden does not indicate a corresponding increase in bone structural integrity and strength. Hence most of these imaging modalities may offer limited ability to predict the risk of pathologic fractures most, commonly afflicting the vertebrae and femurs [49].

2. Assessment of fracture risk in patients with metastatic bone disease

About 10% of all skeletal metastasis affects the long bones, with 70% affecting the axial skeleton [5,55]. In the appendicular skeleton, the proximal femur and proximal humerus are the two most common sites [56], with almost 30% of patients with femoral metastases experiencing a pathologic fracture [57]. In femurs, pathological fractures predominantly occur in the diaphysis and the subtrochanteric region, followed by the neck and trochanteric regions with lesions varying greatly in size and shape [46,57]. These fractures often require surgical interventions with instrumentation due to poor healing of the diseased bone [37]. In a retrospective study by Fidler et al. [58], radiographic measures of defect size involving at least 50–75% of the cortex were prognostic of high risk for pathologic fractures in a cohort of 66 patients with femoral metastasis. Menck et al. [46], suggested a more granular classification with axial expansion of cortical destruction zone of: ≥13 mm in the neck region, ≥30 mm in other parts of the femur, a ratio between width of the metastasis and bone ≥0.60, and cortical destruction of the circumference ≥50% as critical for risk for pathologic fractures of the femur. However, in a study of 203 female patients with 220 measurable lesions on the femoral shaft and proximal femur, neither a specific percent involvement of the bone nor a critical diameter of the lesion was found to predictive of fracture [45]. Mirel et al, [41] extended this classification to include 4 basic characteristics including 1) site of the lesion; 2) nature of lesion; 3) size of the lesion; and 4) pain. All the features were assigned progressive scores ranging from 1 to 3 (Table 1) to establish a score based system for predicting the risk of femoral pathologic fractures.

Table 1.

Mirels’ scoring system, adapted from Mirels [41]

Score	Site of lesion	Size of lesion	Nature of lesion	Pain
1	Upper limb	< 1/3 of the cortex	Blastic	Mild
2	Lower limb	1/3–2/3 of the cortex	Mixed	Moderate
3	Trochanteric region	> 2/3 of the cortex	Lytic	Functional

Evaluation of this protocol’s prognostic utility showed only modest specificity [50]. In part, these findings arise due to the absence of standardized criteria for patient positioning during radiological evaluations, resulting in significant errors in the measurement of lesion size on the radiographs [44]. Furthermore, the measurement of lesion size remains highly subjective [48] and cannot be applied in the case of permeated or diffuse lesions without clear boundaries [46,47]. Although radiological parameters such as cortical involvement showed a high negative predictive value (NPV), their positive predictive value (PPV) was very low [48], suggesting that up to two-thirds of the patients without an impending fracture were selected for preventive surgery [45].

During weight bearing and walking based daily actives, the femur undergoes a high degree of forces with a magnitude of up to 4 times body weight. Combined with the lesion’s effect on bone’s structural strength and the biological process underlying bone homeostasis and remodeling, the lesioned femur may suffer mechanical instability that is often associated with the presentation of severe pain [59]. Due to the high risk of fracture through areas of metastasis [60], surgical intervention is aimed at 1) prophylactic fixation to prevent impending pathological fractures, 2) stabilization of the failed femur due to pathological fracture, 3) segmental resection of tumors, and 4) arthroplasty for replacing joints that have been destroyed by tumor [61,37,62]. This group of patients is, however, complex and is highly susceptible to surgical complications, including: infection, deep vein thrombosis and pulmonary embolism intraoperative or postoperative death [63,64]. Uniquely to this patient population concern remains as to the effects of the introduction of surgical instrumentation on the local and systematic spreading of cancer [65], Due to the complexity of pathologic fractures, the general lack of bone healing and the risk associated with the previous radiotherapy, implant failure remains a common complication [66]. A detailed review of the factors associated with the failure of intramedullary instrumentation for femoral metastases is presented by Willeumier et al. [66].

Accurate risk fracture prediction is paramount for managing these lesions. For example, misclassification as high risk for fracture may lead to unnecessary nailing of the proximal femur with its attended morbidity and reduced life expectancy. With the radiographic definition of impending pathological fracture remaining ambiguous, there is an essential need to establish image based methodology for prediction of risk of pathological fracture in long bone afflicted with metastatic bone diseases that will enable most effective of current surgical and minimally invasive approaches to improve patient’s prognosis.

In the spine, retrospective clinical studies [67–69] reported defect geometry, destruction of the pedicles, pain, age, anatomic site, lesion type, and activity levels to be risk factors for pathological vertebral fractures (PVF). Relative lesion size remains the most often used radiographic predictor of pathological vertebral fractures [70,71]. However, similar to radiographic measures in the femur, defect size was not found to be prognostic for PVF in patients with osteolytic lesions [72,73], a finding supported by laboratory experiments that showed defect size accounted for < 50% of the variation in vertebral strength [74–77]. This lack of specificity might be attributed to limited contrast in trabecular bone, encompassing up to 90% of the vertebral body, resulting in difficulty in detecting and measuring lesions on radiographs compared to cortical bone [78]. Indeed, it has been estimated that 30–75% of normal bone mineral content must be lost, corresponding to a 50–90% reduction in bone strength [79], before osteolytic lesions in the lumbar vertebrae become apparent on radiographs [80]. Taneichi et al. [81] proposed a radiographic based probability risk model incorporating lesion size, destruction of vertebral structures, and regional specific risk factors. Although 100% specific, the risk model was only 20% sensitive in predicting the occurrence of pathologic vertebral fractures in a cohort of 100 breast cancer patients with spinal bone metastases. Combining age, vertebral height, prior fracture and bone mineral density (BMD), application of the Fracture Risk Assessment (FRAX) algorithm [82] showed limited prediction of observed fractures in breast [83] and prostate cancer [84,85] patients.

The spinal instability neoplastic score (SINS) [86], which categorizes risk based on vertebral level, mechanical pain, “ bone lesion quality “, radiographic alignment, vertebral body collapse, and posterolateral spinal element involvement, is being increasingly used to assess risk of impending instability of the spine. Of its components, mechanical pain often manifested as severe movement-related pain [87,88], is thought to be the result of structural instability of the lesioned spine [89]. Mechanical pain forms a critical criterion for selecting patients for surgical stabilization [90] and is strongly associated with risk for PVF [91,92]. However, its prognostic value within SINS for the occurrence of PVF remains controversial [93]. Of its components, SINS ordinal-based classification of “bone lesion quality”, based on characterization of the lesion radiographic appearance (i.e., osteolytic, mixed or osteoblastic), is highly subjective with poor reproducibility between different clinicians [86,94]. Critically, although aimed as a surrogate measure of vertebral strength, this classification cannot account for the remarkable variation in the metastases effect on vertebral strength [95]. Thus, prognostic utility of SINS for predicting PVF [93,96,97] remains controversial [98,99]. As a result, the management of these infirm patients is often reactive, with patients only offered therapy after complications occur. Patients considered at high risk for fractures, but who have not suffered increasing pain are rarely offered prophylactic treatment intended to reduce the risk of PVF [100,101,86,21].

3. Effect of metastases on bone strength and fragility.

3.1. The determinants of bone strength and fragility.

The mechanical strength of bone is strongly related to its apparent density (the mass of bone present in the volume of interest) [102–106], architecture (the geometric distribution of the mass) [107,108], and intrinsic bone material properties (bone matrix and mineral composition). Clinically, the quality of metastatic bone is described by its radiologic appearance (i.e., osteolytic, osteoblastic, or mixed) and is associated with bone mineral density (BMD). However, BMD derived from dual x-ray absorptiometry (DXA), or quantitative computed tomography (QCT) is only a surrogate measure of bone strength and does not fully explain fracture incidences. Bone quality encompasses all factors that contribute to bone fracture independently of BMD. In contrast to research on the effects of aging on bone fragility, the mechanisms underlying the effect of metastases on bone mechanical and structural properties remain unclear.

In trabecular bone, the size and shape of the trabeculae, as well as anisotropy (trabeculae connectivity and orientation), contribute to bone strength while in cortical bone, cross-sectional area, cortical thickness and porosity are the main determinants [109]. Hydroxyapatite mineral is significantly associated with bone material stiffness [110–112], whereas type-I collagen content and maturity of collagen crosslinking [113] contributes primarily to its toughness and post-yield deformation [114]. Non-collagenous proteins (NCPs) constitute the remaining fraction of bone extracellular matrix (ECM). NCPs are involved in coordinating cellular activity, bone matrix organization, and, more recently, have been shown to play a structural role in bone facilitating microdamage formation and energy dissipation [115,116]. An understanding of the relationship between bone quantity, quality, and mechanical properties can improve the prediction of fracture occurrences due to metastases and reduce associated morbidity. This can be achieved with the use of human and animal models.

3.2. The effects of metastases on bone quantity and architecture.

Consistent with its radiographic presentation, human vertebral bone with osteoblastic lesions exhibits higher bone volume to trabecular volume fraction (BV/TV) and trabecular number compared to normal bone [117,118]. Trabeculae in osteoblastic bone were also shorter and thicker for a given BV/TV than trabeculae in osteolytic bone [119]. The effective length to width ratio (slenderness) of the trabeculae is important in determining stability to buckling and bending. In proximal femoral biopsies with osteolytic lesions from breast cancer patients, increased bone resorption (bone volume loss) and trabecular thinning were associated with increased femoral fragility [120]. In vitro studies of cortical and trabecular bone samples containing osteolytic metastases [119,121,122] and in cadaveric vertebral specimens obtained from donors with breast, lung and urinary cancer [123], affirmed the role of lesion mediated changes in bone quantity and architecture.

A few animal models have recapitulated the clinical metastatic spread of primary cancers with evidence of different lesion types, and structural and architectural changes have been assessed. Similar to human studies, cortical femoral bone destruction, decreased trabecular BV/TV, and BMD were observed in tumor-bearing mice of breast cancer metastasis [124]. The tumor-bearing limbs had significantly lower stiffness and energy to failure compared to control limbs. BMD and lysis score (a scoring system that distinguishes between lesion size, location, and perforation through the femoral cortex) were modest predictors of mechanical strength. A significant reduction in elastic modulus of tibial metaphyseal trabecular bone was associated with osteolytic lesions bones [125]. Prostate cancer colonization in the femurs of mice disrupts the normal unidirectional, layer-by-layer, compact lamellar structure of cortical bone [126]. More recently, reduced bone mineral density and trabecular microarchitectural indices were observed in osteolytic and mixed metastatic rat vertebral bone [127].

3.3. The effects of metastases on bone quality and material properties.

In bone biopsies from patients with osteoblastic lesions, the degree of mineralization (the mineral density at the level of calcified bone tissue) was lower with a broader range than comparable sites in patients without metastases [117]. The broadening indicates that a higher percentage of newly formed bone is poorly mineralized even though formation is accelerated. Other measures of bone matrix quality, such as mineral crystallinity, matrix phosphate/amide ratio, and carbonate substitutions, and enzymatic collagen crosslinks, were lower in the newly-formed bone compared to the surrounding pre-existing bone [128]. These findings are further supported by micro-CT imaging [118] demonstrating osteoblastic bone to exhibit lower bone mineral content at the trabecular level.

In osteolytic lesions, mineral crystallinity and collagen mineralization were also lower. However, carbonate substitution was higher [129]. This suggests that osteolytic lesions caused weaker bones with imperfect crystal structure. Nazarian et al. [119] reported metastatic vertebral bone to show 50% lower modulus and tissue hardness compared to normal and osteoporotic bone. However, with nearly half of the specimens obtained from prostate cancer donors and with no separate values reported for tissue osteolytic vs. osteoblastic metastases, it is unclear how these measures would differ among lesion type.

Similarly to human studies, osteolytic rat vertebral bone had decreased crystallinity, mineral to matrix ratio, and mineral crystal width compared to healthy controls [130,131]. Other material properties, such as increased carbonation level [132], decreased collagen mineralization, and mineral maturity was also associated with the presence of osteolytic tumors. Risedronate treatment of bone with osteolytic tumors significantly recovered the nanomechanical stiffness with intermediate values between healthy and metastatic bones [125]. Osteoblastic rat vertebral bone showed a significant decrease in tissue mineral homogeneity with diminished modulus and hardness compared to osteolytic and healthy bone [133]. The collagen fibrils of newly synthesized bone in areas previously involved in osteoblastic metastasis were highly disordered and lacked the typical parallel packing arrangement seen in healthy lamellar bone [127]. The preferential alignment of collagen fibril and HA crystals was also disrupted and contributed to the impaired mechanical function (stiffness and toughness) of these bones [126]. It is important to note that the mineral quality of non-metastatic breast cancer tibial metaphyseal bone is profoundly modified [134] by evidence of smaller, less perfect, and less oriented hydroxyapatite crystals. This suggests that mammary tumors could remotely alter bone mineral properties through circulating tumor-derived factors and increased local osteogenesis.

Post-translational modifications of type-I collagen affect the quality and structural integrity of the bone matrix. While enzymatic modifications contribute to collagen stability, nonenzymatic modifications, in particular, glycation, forms advanced glycation end-products (AGEs). The accumulation of AGEs has been shown to be detrimental to bone matrix quality and is associated with increased skeletal fragility [135–137]. Pentosidine (PEN), a clinically relevant AGE, was elevated in osteolytic rat vertebral bone while enzymatic crosslinks (deoxypyridinoline, DPD) decreased [130,131]. Whole bone stiffness and maximum load were also lower and correlated negatively with PEN. Carboxymethyl-lysine (CML), a major end-stage AGE, is significantly elevated in breast carcinomas and plasma of prostate cancer patients [138,139]. The high CML content in breast cancer patients was also correlated with relapse-free survival after chemotherapy [138].

Lastly, non-collagenous proteins (NCPs) constitute the remaining component of the bone matrix, but their role in the prediction of pathologic fractures in bone metastasis remains unknown. Osteopontin (OPN), osteocalcin (OC), and bone sialoprotein (BSP) are major NCPs in bone matrix that show a divergent pattern in their expression based on the type of metastasis. OPN is preferentially expressed in osteolytic lesions (83% of breast cancer tumors analyzed), while BSP is elevated in osteoblastic lesions (89% of prostate cancer tumors analyzed). OC is elevated in metastatic tumors of both prostate and breast cancer. These NCPs have been shown to play a role in bone biology and mechanical properties through the use of animal models (Morgan, Poundarik, and Vashishth 2015). Taken together, the aforementioned studies show that metastases alter bone matrix quality and increase fragility.

4. Prediction of Skeletal Strength with metastatic disease

Fracture of a skeletal structure occurs when it is no longer able to carry the loads imposed by daily tasks or activity. Structural rigidity, a product of the geometry and mechanical strength of the bone tissue, quantifies the bone’s capacity to resist applied force or moment-based loading. Conceptually, as illustrated in Figure 4, the risk of a skeletal structure to undergo failure can be estimated via a factor of risk of fracture (ψ), defined as the ratio of the sum of the in-vivo loading imposed to the failure strength of the structure. When this ratio is near or larger than 1, the skeletal structure is expected to be at risk for failure.

Figure 4.

A graphical illustrating demonstrating the concept of factor of fracture risk. The graphical portion of the image was from a previous publication. Experimental validation of finite element analysis of human vertebral collapse under large compressive strains. Hosseini HS, Clouthier AL, Zysset PK. J. Biomech Eng. 2014 Apr;136(4). doi: 10.1115/1.4026409. PMID: 24384581 The three vertebrae pictures were part of Figure 5 in that publication. We would like to acknowledge the fact that part of our image is composed from this image.

4.1. Quantitative protocols for image-based prediction of pathologic vertebral strength

a. Bone mineral-based predictions:

Radiographic measure of bone mineral density (BMD) is the current standard for estimation of vertebral fracture risk in osteoporosis patients [140]. However, being region-specific [102] and structurally simplistic due to the two-dimensional projection nature of this assessment, it was reported to be a weak predictor of strength for vertebrae with osteolytic lesions in vitro [76,77] and in vivo [141]. In recent years, the application of engineering analysis to predict the lesions’ effect on the strength of skeletal structures containing bone metastases from routine clinical imaging has been an active area of research. The most common approaches are the mechanics of solid (MOS) and finite element analysis (FEA).

b. Mechanics of solid approach (MOS):

Known as computed tomography-based structural rigidity analysis (CTRA) [27,142,77] or CT structural analysis protocol (CT-SAP) [76], this approach applies classical engineering methodology to compute the structural rigidity of skeletal structures allowing the estimation of the failure load. Structural rigidity represents the capacity to carry, either single or combined, axial, bending, or torsional loads that are applied on the body under the constraint of a pre-specified degree of strain.

• Engineering strain is the amount that a material deforms in the direction of the applied stress per unit length.

• Engineering stress is the magnitude of applied load (axial, bending, torsional) divided by the actual cross-sectional area of the specimen at that load.

As illustrated in Figure 5, the volumetric geometry of the skeletal structure is segmented from the medical imaging data (CT, MRI). As preparatory stage for computing structural rigidity, image pixels identified as either trabecular or cortex bone tissue within the segmented structure are assigned bone mineral density value based on the pixel’s gray value (CT Hounsfield unit). Empirical laws are then applied to convert the density value to bone elastic modulus based on the specific bone tissue (trabecular or cortex). For each axial section (image) within the segment volume, a series of computational steps computes the section’s axial, bending, and torsional structural rigidities based on the assumption that the mechanical behavior of the bone material follows Hooke’s law for linear elastic materials. This law state that for small deformation the proportionality between applied stress (σ) and resulting elastic strain (ε) is defined by the material Young’s (elastic) modulus, i.e., E= σ / ε

• Elastic modulus is a measure of the material resistance to being deformed elastically (i.e., ability to return to its original shape when stress applied to it is removed).

Setting the constraint that the highest degree of strain allowed within the bone tissue to be its yield strain (a value denoting transition from elastic (fully recoverable) to plastic (imposing a permanent) deformation value), the computed rigidity parameters are employed to estimate the section’s ability to sustain applied axial and moment-based loading. This constraint ensures that the resultant estimate of loading corresponds to the initiation of failure within the bone. As illustrated in Figure 5, the prediction of the overall strength of the skeletal structure is based on the assumption that failure will initiate at the structural section demonstrating the minimum value of rigidity (“weakest link”). In practice, this analytical based estimation represents threshold values, not necessarily the exact failure load or location of the failure.

Figure 5. A graphical illustrating of the mechanics of solids based approach (MOS).

From the routine clinical CT data, vertebral anatomy and bone tissue are segmented [A] and a calibration curve, providing conversion of the CT grey values to bone density values, computed from standard density phantom [B]. Based on this calibration, each pixel corresponding to the bone is assigned a bone density value (r), computed from the CT Hounsfield units. Empirical formulae [C] are employed to convert the assigned density to tissue-specific (trabecular or cortex) elastic modulus. For each CT cross-section within the segmented anatomy, a density-weighted centroid is computed followed by the computation of the section’s axial (EA), bending (EI_ii) and torsional (Gk) structural rigidities [76,144,77] representing the section’s ability to sustain axial, bending, or torsional loads respectively [D]. (E is the modulus of elasticity (Young’s modulus) of the bone, A: is the cross-sectional area of the segmented CT geometry, I: is the second moment of area, G: is the torsional modulus for the trabecular or cortical bone and K: is the polar moment of area). Depending on the applied load, these structural rigidities are incorporated within force and moment equation [E] to estimate the CT cross-section failure (Fz) strength. Performed for the entire segmented volume [F], the CT section showing the least value for Fz is identified and used to provide the structure’s estimated failure strength [G].

MOS based approaches have been applied to predict the effect of critical osteolytic defects on the strength of human core bone samples [143], human femurs [144–146], and human spine motion segments [76,77] as well as the strength of rat femoral bone containing osteolytic metastases [147]. Results have shown that MOS strength predictions are more accurate than those based on BMD alone. In a prospective cross-sectional clinical study, an MOS approach provided higher sensitivity, specificity, positive predictive, and negative predictive values compared to the Mirel’s scoring system (a description of Mirel’s scoring system is provided in section 2) [144]. In a cohort of 124 patients with metastatic disease of the femur, a multicenter study by Nazarian et al. [148] showed the application of CTRA improved fracture risk prediction compared to the Mirel based criterion. On follow up period of four months the MOS approach showed higher sensitivity and specificity for predicting both the risk of new fractures as well as NPV value, highlighting the potential advantage of the MOS approach in preventing unnecessary surgical interventions in these infirmed patients. Permitting rapid analysis and prediction of strength with modest expertise and computing facilities, MOS approaches can advance the accuracy of predicting impending pathological fractures over current clinical criteria [146].

c. Finite Element Modeling (FEM):

Finite element modeling is the most widely used computational technique for analyzing the effect of functional loading and disease conditions on the mechanical response of tissues to loading. In addition to statics, FEM is well suited to studying dynamic and time-varying mechanical responses of hard and soft tissues and the resulting change in the distribution of strain and stress within them. This computational approach employs a discretization strategy (of which the graphical representation is the mesh model) to subdivide the physical model (global domain) to small and finite domains (elements) with differential equations [149]. As illustrated in Figure 6, the initial process of segmentation of the anatomy, creation of calibration curves associated with image intensity values and, assignment of material properties of specific tissues is largely similar to the MOS approach.

Figure 6.

FEA based approach: Similar to the MOS approach, the anatomy of the skeletal structure is segmented from the CT [A] and a calibration curve associating bone density value (r) with the CT Hounsfield units is established [B]. For each segmented anatomy, mesh models are created to match the tissue structure of this anatomy (osseous [D] or soft tissue [F]). Based on the established calibration model (CT for bone elements) or known association between the MR metric and the properties of the soft tissue, empirical formula for bone [E] and tissue-based laws for MR [I] are used to assign material models [G] to the appropriate mesh elements. Depending on the complexity of the modeled anatomy, the simulated physical model is completed by assembling the individual mesh models and assigning boundary and loading conditions (required for solving the system of numerical equations) as well as prescribing the kinematic constraints between each component of the model [J]. Depending on the type analysis required linear or non-linear, static vs. dynamic), the appropriate finite element numerical solver is used to compute the structural response of the physical model [K]. Based on the set of outputs (strain, stress) requested, the state of stress and strain estimated for the structure of interest can be visualized.

However, at this point, the two approaches diverge in both requirement of expertise and resources The next stage is the application of the numerical mesh model (Fig 4), represents a numerical-based discretization of the continuous geometric shape into geometric and topological cells. Based on the physicality of the problem and the complexity of the structure to be modeled, the mesh can be composed of a regular (structured) or irregular (adaptive) lattice, the latter allowing to capture complicated geometry with increased geometrical accuracy and computational efficiency. As part of the meshing process, the type of mesh element is selected based on the element’s numerical formulation designed to model the physical laws that govern the balance of forces and the constitutive relations that relate the stresses to strains in the tissue or structure constituent materials. This stage is followed by the application of material laws appropriate to the structure. These can be Isotropic (mechanical and thermal properties that are the same in all directions), Anisotropic (mechanical and thermal characteristics that differ with respect to two of the material geometrical axes under the same loading conditions) or Orthotropic (subgroup of the anisotropic materials with the different material properties along three mutually orthogonal axes). In bone, these material laws are related to the internal structure of both cortical and trabecular bone and to the bone density. For soft highly hydrated tissues, such as ligaments, cartilage, and discs, these material laws may describe more complex time-varying behavior, such as creep or stress relaxation response and poroelasticity (describing the effect of the interaction between fluid flow and the deformation of the solid matrix of the material).

In a unique difference with respect to the MOS approach, in order to solve this system of differential equations, one prescribes boundary or initial conditions. For skeletal modeling, conditions include;

1. The physical constraints at the boundary of the geometry that will affect its deformation (for example, the constraints applied on the intervertebral disc via its connection to the vertebral body).

2. Contact conditions that affect the interaction of the geometry with external bodies (for example, the contact between the acetabulum and the head of the femur via the cartilage tissue) and

3. The loading applied on the structure under examination (for example, the pattern of compressive bending and torsional loads estimated to act on the femoral head at toe-off).

The workflow for assembling an FE model is illustrated in Figure 6. For a detailed review of the method and its application in orthopedics, see Taylor et al. [150].

FEM allows estimates of traditional mechanical properties such as stiffness, ultimate strength, and strain energy density distribution. This approach has produced unique insights into the effects of metastases on the failure process of the femur [61,145,151] and vertebra [152–156] with osteolytic defects. In cadaveric femurs with and without simulated metastatic defects, FEM based failure predictions strongly correlated with experimental failure force (r² = 0.82, p<0.001) [157]. Based on a dataset of ten human cadaver femurs, for which osteolytic defects ranging from 20–45mm were simulated at the anterior or medial aspect of the greater trochanter, FEM produced higher performance for predicting femoral strength compared to clinical assessment based on radiographic analysis [157]. In a study of thirty-nine patients with osteolytic tumors within the femur, Eggermont et al. [158] reported patient-specific FEM to show 89% sensitivity, 79% specificity for predicting femoral fractures and 97% NPV for predicting femurs at low risk of fracture. By comparison, clinical-based prediction for the same set of femurs showed up to 33% sensitivity, 84–95% specificity for predicting femoral fractures, and 0.78–0.84% NPV for predicting femurs at low risk of fracture [158]. These findings demonstrate the utility of FEM for predicting the failure of skeletal structures containing bone metastasis and that the application of FEM may help identify patients who may suffer fracture and those who are at low risk of fracture leading to better selection of patients for prophylactic therapy.

c. Comparative advantage and future directions:

Both approaches share similar work-flow for segmenting the anatomy and tissue components of the skeletal structure under investigation. Direct comparison of MOS – and FE-estimation of failure loads for simulated lesions in rat tibias [159] and in cadaveric femurs [160] found no differences in accuracy for predicting strength.

• The MOS approach offers the advantage of direct application for standard clinical CT data without the need for specialized software packages and hardware resources, resulting in order of magnitude lower time for computation and solution compared to that required for preparation and computation of FE models [160]. This potentially translates to higher throughput for analysis of patient data. However, at present, this approach is limited to a computation based on assumption of linear elasticity, can account only for the contribution of osseous tissue to structure strength and cannot model the effect of soft tissue on kinematic constraints and the transmission and distribution of loading which affect the magnitude and state of stress within the structure.

• The FE approach allows the application of both linear and non-linear material laws with the increased complexity of axis-dependent properties (anisotropic, orthotropic) and can incorporate the contribution of soft tissues such as the cartilage in the femur and the intervertebral disc and spinal ligaments in the spine. Possessing time-dependent material properties [161–163], these soft tissues are important modulators of both the magnitude and duration of loads that act on bone and thus determine its response to loading [161,162] and its strength [164–166]. However, creation and analysis of FE models require costly software and hardware and involves a more complex preparation processes for model creation. It requires significant expertise in assuring model quality and interpreting the results. These demands limit the use of FEM in clinical practice.

The use of the MOS and FEM forms an active research area, and we foresee both methods benefiting from the development of machine- and deep-learning approaches for many of the steps required. An example includes non-supervised segmentation of skeletal structures from routine clinical imaging and identification of location and volume of tumor tissue within the complete volume (either single or multiple levels) of the skeletal structure. FEM based approaches continue to benefit from improvements in technology for script-based preparation of mesh models the rapid increase in computational capability and utilization of hardware-based accelerator units for the faster solution of more complex models. FEM further offers potential increased accuracy due to recent advances in estimating the mechanical stiffness of the intervertebral disc [167,168] and mechanical properties of tumor tissue [169,170] from magnetic resonance imaging. As the computational challenges give way to improved methods, we expect the FEA approach to gain an advantage for prediction of bone strength in the clinical setting.

Continual development of these engineering approaches may offer unique and exciting possibilities in several areas. 1) Assessing the risk of fracture progression in cancer patients with an existing pathologic spinal fracture, a common clinical presentation. 2) assessing the trajectory of risk due to the dynamic nature of the metastatic disease and 3) as part of pre-surgical planning for augmentation based approaches and 4) as a non-invasive assay to evaluate the efficacy and potential detrimental effects of radiological and pharmaceutical. Accomplishing these goals will require validation in large-scale retrospective and prospective clinical studies.

5. Conclusions

Metastatic bone disease is incurable and exposes patients to the risk of skeletal-related events. Optimal treatment is required since intervention is only attempted in response to consequential pain, neurological deficit, or pathologic fractures. Recent experimental methodologies focus on image- and biological-based markers as new tools for addressing the risk of skeletal complications with bone metastasis. Bone structure, architecture, and material composition are significant contributors to fracture resistance. These measures of bone quality have been shown to be altered by bone metastasis in pre-clinical animal models to a larger extent than changes in mass or density alone, but they are not quantifiable with current clinical modalities.

Mechanics of solids (MOS) and finite element analysis (FEA) approaches offer the ability to directly compute the effects of the metastasis mediated damage to skeletal anatomy on the structural rigidity of the affected skeletal structure. Being an active area of research both methods were shown to provide a highly accurate prediction of the strength of lesioned skeletal structures from routine CT scanning protocols that is independent of the radiological appearance of the lesion. Likely to further benefit from the development of machine- and deep-learning approaches and from continual development in both CT and MRI clinical imaging protocols, these approaches may allow the development of data-driven prediction of risk of fracture that is individualized to the patient.

Routinely collected as part of clinical care for metastatic patients, the use of bone turnover biomarkers are increasingly being advocated as prognostic indicators for the risk of skeletal events, and patient’s survival. Combined with imaging-based approaches, utilization of bone turnover markers values may provide improved prediction of the effect of disease trajectory on the risk of fracture, thus improving clinical management of patients with bone metastasis. Taken together, advancement in current and development of new methodologies will elucidate the differential effect of lesion type on bone quality, improve the prediction of pathologic fractures, and help guide treatment.

Highlights.

• Metastatic bone disease dramatically affects the skeletal bone architectural and structural competence resulting in a host of debilitating skeletal related events (SREs).

• This interdisciplinary based review:

  • Discusses the current challenges faced by clinical based prognostic protocols for predicting pathologic bone fractures.
  • Presents the state-of-the-art clinical and experimental evidence on the effect of bone metastases on the bone’s strength and fragility.
  • Highlights current engineering- and computational-based analyses approach for predicting pathologic fractures with bone metastases.