Abstract
Background
Stratification of the fracture risk is an important treatment component for patients with multiple myeloma, which is associated with up to an 80% risk of pathologic fracture. The Mirels score, which is commonly used to estimate the fracture risk for patients with osseous lesions, was evaluated in a cohort in which fewer than 15% of lesions were caused by multiple myeloma. The behavior of multiple myeloma lesions often differs from that of lesions caused by metastatic disease, and accurate risk stratification is critical for effective care. To our knowledge, the Mirels score has not been validated specifically for multiple myeloma.
Questions/purposes
Our purpose was: (1) To develop a novel scoring system for the prediction of pathologic fracture in patients with long-bone lesions from multiple myeloma; and (2) to compare the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and receiver operating characteristic (ROC) area under curve (AUC) between the novel scoring system and the Mirels system.
Methods
Between 2003 and 2017, 763 patients at one center with the diagnosis of multiple myeloma were reviewed, of whom 174 presented with long-bone disease involvement. Of those, 5% (nine of 174) were missing data or radiographs at a minimum of 1 year and had not reached an endpoint (fracture or surgery) before that time and were therefore excluded. Many patients have more than one lesion; consequently, we used the largest lesion in each patient, resulting in 163 lesions in as many patients. Ten percent (16 of 163) of these patients eventually developed a fracture and 4% (six of 163) underwent prophylactic stabilization (excluded from analysis because of outcome uncertainty). During the study period, prophylactic stabilization was performed at the discretion of the orthopaedic oncologist. Fifty-one percent (83 of 163) of patients were female, and the mean (± SD) age was 60 ± 10 years at radiographic lesion identification. All lesions were characterized before determining whether the patient underwent pathologic fracture. We identified variables associated with pathologic fracture on univariate analysis. Variables independently significant on logistic regression analysis were used to generate scoring algorithms at varying weights and scoring cutoffs for comparison via ROC curves. We then selected a novel score based on ROC performance, and compared the sensitivity, specificity, PPV, and NPV of that scoring system to that of Mirels score. ROC AUCs were compared after bootstrapping 100,000 iterations. Alpha was set at 0.05.
Results
After controlling for potential confounders, such as age, sex, and duration of myeloma diagnosis, we found the following factors were independently associated with the occurrence of pathologic fracture: larger lesion size (area, cm2) (log odds 0.17; p = 0.03), longer lesion latency (years from diagnosis to lesion identification) (log odds 0.25; p = 0.03), presence of pain (relative risk [RR] 2.9; p = 0.04), and metaphyseal location (RR 3.2, compared with epiphyseal or diaphyseal; p = 0.003). These variables were used to formulate a novel scoring system. Compared with the Mirels system, the novel system was more sensitive (69% [95% CI 61 to 76] versus 38% [95% CI 30 to 46]; p < 0.05) but not different in terms of specificity (87% [95% CI 80 to 91] versus 87% [95% CI 81 to 92]; p > 0.05), PPV (37% [95% CI 29 to 45] versus 25% [95% CI 19 to 33]; p > 0.05), NPV (96% [95% CI 91 to 99] versus 92% [95% CI 87 to 96]; p > 0.05), or AUC (0.85 [95% CI 0.74 to 0.92] versus 0.67 [95% CI 0.51 to 0.81]; p > 0.05).
Conclusion
The novel scoring system was found to be more sensitive than the Mirels system for predicting pathologic fracture in our retrospective cohort of patients with multiple myeloma-related bone disease. Specificity, PPV, NPV, and ROC AUC were not different with the numbers available. Thus, the novel scoring system may serve as a more effective screening tool to determine which patients with multiple myeloma would benefit from further radiologic or orthopaedic evaluation based on a skeletal survey.
Level of Evidence
Level III, diagnostic study.
Introduction
Cancer deaths have decreased by approximately 25% during the past 25 years [15], causing some cancers to become more like chronic diseases [10]; nevertheless, they are associated with substantial morbidity, including pathologic fractures [3, 19]. Multiple myeloma, which was diagnosed in approximately 32,000 people in the United States in 2019 [1], is associated with up to an 80% risk of pathologic fracture [11]. Although many of these fractures occur in the spine, appendicular lesions are also associated with a substantial fracture risk. Patients with multiple myeloma who have a pathologic fracture have a 20% higher risk of death within 2 years than those who do not [17], making stratification of the fracture risk critical.
Numerous risk factors for pathologic fracture have been identified, including patient age, sex, cancer type, lesion size, cortical involvement, bone matrix quality, and symptoms [5, 6, 13, 16, 20, 22]. The Mirels system, a commonly used tool for stratifying the risk of osseous lesions [12], uses several of these factors to calculate a fracture risk score. However, the study of 78 lesions that established the Mirels system included only 11 lesions in patients with multiple myeloma [12]. The mechanical behavior of multiple myeloma lesions often differs from that of lesions from metastatic disease, in that multiple myeloma lesions may consolidate or heal by forming a sclerotic rim if controlled by medical or radiation therapy. This healing pattern has meaningful implications in predicting fracture risk and may justify a scoring system unique to multiple myeloma and distinct from the Mirels score. In addition, the Mirels score was evaluated in patients who had received radiation therapy. Not all patients with bone lesions receive radiation therapy, but they may benefit from observation and an accurate assessment of the fracture risk. Therefore, a scoring system that accounts for the presence or absence of radiation therapy may be better suited for patients with multiple myeloma. Other methods of assessing pathologic fracture risk, including CT-based structural rigidity analyses, have improved the prediction of fractures compared with the Mirels system but require advanced imaging and analytical tools that are not widely available [4].
In patients at a high risk of having a pathologic fracture, prophylactic stabilization should be considered. Compared with treatment of a completed pathologic fracture, prophylactic stabilization may involve a shorter operative time, a shorter hospital stay, less perioperative pain, and better function [2, 21]. Because many multiple myeloma lesions are asymptomatic, an accurate prediction of the fracture risk may help medical and radiation oncologists determine which patients may be observed and which may benefit from further orthopaedic evaluation.
We therefore sought to (1) develop a novel scoring system for the prediction of pathologic fracture in patients with long-bone lesions from multiple myeloma, and (2) compare the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and receiver operating characteristic (ROC) area under curve (AUC) between the novel scoring system and the Mirels system.
Patients and Methods
Patient Selection
The study methods were approved by our university’s institutional review board before data gathering (IRB 00170713). Patients with a diagnosis of multiple myeloma between 2003 and 2017 were identified in the cancer registry at our high-volume comprehensive cancer center. We identified 763 patients with the diagnosis of multiple myeloma (Fig. 1). Inclusion criteria were the availability of medical records with radiographic and clinical data permitting an evaluation of long-bone lesions, clinical records with symptom and disease course information, and follow-up of at least 1 year or until a pathologic fracture occurred or prophylactic surgical stabilization was performed. Seventy-seven percent (589 of 763) of patients had no identifiable long bone disease involvement. Of the 23% (174 of 763) with long bone disease involvement, 94% (163 of 174) had follow-up radiographs for 1 year or until pathologic fracture or stabilization. Overall, 5% (eight of 163) were excluded for inadequate follow-up, and 2% (three of 163) were excluded for inadequate radiographs. Exclusion criteria were the presence of concurrent metastatic cancer, history of connective tissue disease, or history of fracture of the bone of interest, though none of the 163 patients met any of these criteria.
Fig. 1.
A flow diagram for patients included in study is shown here.
Radiographic Evaluation
Extremity radiographs were evaluated for long-bone lesions. The largest lesion in each patient was characterized according to anatomic location, lesion dimensions on orthogonal radiographs, presence and number of lesions surrounding the lesion being examined, presence of endosteal scalloping, presence of a sclerotic rim on the initial radiograph, and fraction of the lesion diameter to bone diameter (width fraction). All measurements were made by two trained researchers (GRT, JAB) affiliated with our orthopaedic oncology division. Discrepancies were adjudicated by a third evaluator (ASL), who is a fellowship-trained orthopaedic oncologist. All target lesions were characterized before the fracture outcome was determined to prevent bias regarding lesion selection and characterization (Fig. 2A-C).
Fig. 2 A-C.
These AP radiographs are from a 67-year-old man who presented with a midshaft humeral lesion (A) 15 years after being diagnosed with multiple myeloma. He had moderate pain not exacerbated by function and no history of radiation to the site. The lesion measured 2.2 cm × 4.5 cm, for a lesion size of 9.8 cm2. (B) There was an eccentric lytic lesion associated with endosteal scalloping, distal to the lesion of interest. (C) Approximately 1 week after this lesion was identified, he had a fracture through the lesion and was promptly treated with open reduction and internal fixation of the completed fracture, followed by postoperative radiation of the site.
Clinical and Demographic Data
Using patient medical records, we extracted data on age, sex, BMI, medical comorbidities, laboratory values (albumin, alkaline phosphatase, beta globulins, calcium, creatinine, gamma globulins, and hemoglobin), fracture history, systemic therapy (such as antiproteasomal therapy, antiresorptive therapy, corticosteroids, lenalidomide, and bone marrow transplant), date of multiple myeloma diagnosis, date of lesion identification, and presence of pain and pain severity in the extremity as described by Mirels (mild, moderate, or functional).
Fifty-one percent (83 of 163) of patients were female. The mean ± SD duration of patient follow-up was 4.7 ± 3 years after identification of the lesion. Mean (± SD) patient age was 60 ± 10 years; mean BMI value was 28 ± 6 kg/m2; and 39% of patients (64 of 163) were noted to be alive at the time of data collection. Twenty-six percent (43 of 163) of patients had been diagnosed with diabetes, and 6% (nine of 163) were current smokers. Sixty percent (97 of 163) of lesions were femoral, 39% (64 of 163) were humeral, and 1% (two of 163) were tibial. During the study period, 10% of patients (16 of 163) were diagnosed with pathologic fractures (nine femoral and seven humeral), and 4% (six of 163) underwent prophylactic stabilization before fracture. Those who underwent prophylactic stabilization before fracture were excluded from analyses for determining factors associated with pathologic fracture.
Lesion Latency
We defined lesion latency as the time from the diagnosis of multiple myeloma to identification of the lesion of interest on radiographs. Lesion latency indicated the chronicity of the lesion with respect to the overall disease course if the lesion was not present on an initial skeletal survey. Patients with a previous diagnosis of multiple myeloma were considered to have lesion latency only if the lesion of interest was identified more than 3 months after presentation to our center. If the lesion was identified within 3 months of presentation, lesion latency was considered absent.
Radiation History
Radiation history was defined as having undergone radiation therapy involving the extremity of interest before the occurrence of a skeleton-related event, defined as pathologic fracture, need for radiation to the site for pain or fracture, or surgery involving the bone. Radiation history was considered absent for patients who did not undergo radiation, underwent radiation of a different area of the body, or received low-dose radiation concurrent with bone marrow transplantation.
Statistical Analysis
Data were imported into Matlab, version R2018a (Mathworks, Natick, MA, USA) for analysis. Mirels scores for each lesion were calculated by two independent reviewers (GRT, JAB). In the Mirels system, categories of site, pain severity, lesion, and size are scored from 1 to 3 points. An aggregate score of ≥ 9 points indicates surgical stabilization, a score of 8 warrants clinical judgement regarding surgery, and a score of ≤ 7 indicates a lesion that should be clinically observed. Scoring discrepancies were adjudicated by a third evaluator (ASL), who is a fellowship-trained orthopaedic oncologist.
We performed an exploratory analysis using Student t-tests and chi-square tests to determine which variables would be included in the logistic regression analysis. Factors associated with pathologic fracture between groups (lesions that did fracture and lesions that did not fracture) were lesion size (area) (mean difference 5 cm2 [95% CI -1 to 10 cm2]; p < 0.001), lesion latency (years from diagnosis of multiple myeloma to identification of lesion) (mean difference 2 years [95% CI 0 to 4 years ]; p = 0.002), Mirels score (mean difference 1 [95% CI 0.1 to 1.2]; p = 0.02), presence of pain (relative risk 2.6 [95% CI 1.1 to 6.5]; p = 0.04), and width fraction (lesion diameter divided by bone diameter) (mean difference 13% [95% CI 0 to 26]; p = 0.004) (Table 1). P values less than 0.05 were considered statistically significant.
Table 1.
Patient and lytic lesion characteristics, laboratory values, and treatment factors for patients with multiple myeloma who had a pathologic fracture versus those who did not.
| Variable | Non-fracture group (n = 141) | Fracture group (n = 16) | p valuea | ||||
| Mean ± SD | Median (Q1 to Q3) | N (%) | Mean ± SD | Median (Q1 to Q3) | N (%) | ||
| Patient characteristics | |||||||
| Age (years) | 60 ± 11 | 61 ± 8 | 0.69 | ||||
| BMI (kg/m2) | 28 ± 6 | 29 ± 7 | 0.37 | ||||
| Current smoker | 8 (5.7) | 1 (6.2) | 0.93 | ||||
| Diabetes | 39 (28) | 6 (38) | 0.41 | ||||
| Disease duration (years)b | 6 (4 to 8) | 6 ± 4 | 5 (2 to 9) | 0.75 | |||
| Female sex | 72 (51) | 8 (50) | 0.94 | ||||
| Pain scorec | 1 (1 to 1) | 1 (1 to 2) | 0.20 | ||||
| Presence of pain | 29 (21) | 7 (44) | 0.04 | ||||
| Lytic lesion characteristics | |||||||
| Metaphyseal location | 46 (33) | 11 (69) | < 0.01 | ||||
| Mirels score | 7 ± 1 | 8 ± 1 | 0.02 | ||||
| Number of lesions | 3 (1 to 5) | 3 (2 to 5) | 0.61 | ||||
| Lesion duration (years)d | 4 (2 to 7) | 2 (0 to 3) | < 0.01 | ||||
| Lesion latency (years)e | 0 (0 to 1) | 1 (0 to 3) | < 0.01 | ||||
| Lesion size (cm2) | 1 (1 to 3) | 3 (1 to 7) | < 0.01 | ||||
| Scallopingf | 48 (34) | 5 (31) | 0.82 | ||||
| Sclerotic rim on initial radiograph | 5 (3.5) | 2 (13) | 0.10 | ||||
| Site scoreg | 2 (1 to 2) | 2 (1 to 3) | 0.97 | ||||
| Width fraction (%)h | 30 ± 16 | 43 ± 26 | < 0.01 | ||||
| Laboratory values | |||||||
| Albumin (g/dL) | 4 (4 to 4) | 4 (4 to 4) | 0.96 | ||||
| Alkaline phosphatase (U/L) | 69 (53 to 94) | 79 (59 to 97) | 0.98 | ||||
| Beta globulins (g/dL) | 1 (1 to 1) | 1 ± 0 | 1 (1 to 1) | 0.79 | |||
| Calcium (mg/dL) | 10 ± 1 | 10 ± 1 | 0.69 | ||||
| Creatinine (mg/dL) | 1 (1 to 1) | 1 (1 to 1) | 0.38 | ||||
| Gamma globulins (g/dL) | 1 (1 to 2) | 1 (1 to 1) | 0.92 | ||||
| Hemoglobin (g/dL) | 12 (10 to 13) | 12 (10 to 13) | 0.66 | ||||
| Treatment factors | |||||||
| Therapy | |||||||
| Antiproteasomali | 113 (80) | 14 (88) | 0.48 | ||||
| Antiresorptivej | 127 (90) | 13 (81) | 0.28 | ||||
| Bone marrow transplant | 73 (52) | 8 (50) | 0.89 | ||||
| Corticosteroid | 137 (97) | 16 (100) | 0.49 | ||||
| Lenalidomide | 115 (82) | 13 (81) | 0.98 | ||||
| Radiationk | 10 (7.1) | 3 (19) | 0.11 | ||||
| Follow-up (years)l | 4 (2 to 7) | 2 (1 to 5) | 0.03 | ||||
P values from univariate analysis.
From diagnosis of multiple myeloma to the last visit at our health network.
As in Mirels scoring system: mild, 1 point; moderate, 2 points; functional (worsened by activity of loading the bone), 3 points.
From lesion identification to the latest follow-up or fracture.
From diagnosis of multiple myeloma to identification of the lesion of interest.
Invasion of the lesion into the bone cortex.
As in Mirels scoring system: upper limb, lower limb, or peritrochanteric.
Width of the lesion divided by width of the bone.
Consisting of bortezomib and carfilzomib.
Consisting of bisphosphonates and denosumab.
History of radiation to the anatomic compartment containing the lesion of interest.
From lesion identification to the last visit at our health network.
Of these variables, those that were independently significant on logistic regression analysis were used to generate scoring algorithms at varying weights and scoring cutoffs for comparison using ROC curves. We then selected a novel system based on ROC performance and compared the sensitivity, specificity, PPV, and NPV of that scoring system with that of the Mirels system. We compared ROC AUCs after bootstrapping 100,000 iterations.
Results
Developing a Novel Scoring System for Fracture Prediction in Multiple Myeloma
We found the following factors were independently associated with the occurrence of pathologic fracture: larger lesion size (area, cm2) (log odds, 0.17; p = 0.03), longer lesion latency (years) (log odds 0.25; p = 0.03), presence of pain (relative risk 2.9; p = 0.04), and metaphyseal location (relative risk 3.2, compared with epiphyseal or diaphyseal; p = 0.003). The novel scoring system selected ranges from 0 to 11 using these variables (Table 2).
Table 2.
Novel scoring system to predict the risk of pathologic fracture in patients with multiple myeloma-related bone lesions
| Category | Point value |
| Lesion latencya | |
| < 6 months | 0 |
| 6 months to 1 year | 2 |
| > 1 year | 4 |
| Lesion size (cm2) | |
| < 5 | 0 |
| ≥ 5 | 2 |
| Lesion location | |
| Diaphysis or epiphysis | 0 |
| Metaphysis | 3 |
| Painb | |
| No | 0 |
| Yes | 2 |
From diagnosis of multiple myeloma to identification of the lesion of interest.
As in Mirels scoring system; functional means worsened by activity or loading of the bone.
Comparing the Novel System With the Mirels System
The novel scoring system, using a cutoff score of 6 or greater, was more sensitive (69% [95% CI 61 to 76] versus 38% [95% CI 30 to 46]; p < 0.05) but not different in terms of specificity (87% [95% CI 81 to 92] versus 87% [95% CI 80 to 91]; p > 0.05), PPV (37% [95% CI 29 to 45] versus 25% [95% CI 19 to 33]; p > 0.05), and NPV (96% [95% CI 91 to 99] versus 92% [95% CI 87 to 96]; p > 0.05) compared with the Mirels system (Table 3). The novel system ROC AUC was not different compared with that of Mirels (0.85 [95% CI 0.74 to 0.92] versus 0.67 [95% CI 0.51 to 0.81]; p > 0.05) (Fig. 3).
Table 3.
Performance of a novel scoring system and the Mirels system in the multiple myeloma cohort
| Scoring system | % (95% confidence interval) | Area under the curve | |||
| Sensitivity | Specificity | PPV | NPV | ||
| Novel system | 69 (61 to 76) | 87 (80 to 91) | 37 (29 to 45) | 96 (91 to 99) | 0.85 (0.74 to 0.92) |
| Mirels | 38 (30 to 46) | 87 (81 to 92) | 25 (19 to 33) | 92 (87 to 96) | 0.67 (0.51 to 0.81) |
Fig. 3.
These receiver operating characteristic curves represent the Mirels scoring system and novel scoring system in patients with multiple myeloma. *Indicates the performance of the scoring system at a discrete score suggesting prophylactic stabilization.
Discussion
Understanding the risk of pathologic fracture in patients with multiple myeloma bone disease and the ways in which that risk is influenced by treatments is critical to preserving skeletal integrity and quality of life. Tools for estimating the fracture risk, such as the Mirels scoring system, have typically been applied to cohorts consisting largely of patients with metastatic carcinoma [4, 12]. Multiple myeloma commonly demonstrates a clinical course distinct from many osseous metastases from carcinoma, which may justify a unique scoring system. To our knowledge, a unique scoring system for patients with multiple myeloma does not exist. We therefore developed a novel scoring system after analysis of lesion characteristics and behavior. We identified lesion size (area), lesion latency, presence of pain, and metaphyseal location as independent predictors of risk of pathologic fracture, leading to the formulation of a novel scoring system that had improved sensitivity compared with the Mirels system in this retrospective analysis of multiple myeloma patients.
Limitations
This study has several limitations. The limited number of events in the cohort (16 fractures) may have resulted in sparse-data bias [9]; on some of our comparisons, the confidence intervals were quite wide, and it is conceivable that even a few more fractures may have changed our findings. Therefore, readers should interpret our findings with caution, and we hope that future studies will seek to validate this scoring system, which we find useful. However, given the limited research in this vulnerable patient population and that this is the largest study of pathologic multiple myeloma fractures to our knowledge, these results may be useful to clinicians and promote further research with larger sample sizes and multiple care centers.
In previous studies of the inter-rater reliability of the Mirels score, the overall score has demonstrated moderate reproducibility, but pain severity has performed poorly [4, 8]. The pain component we used in the novel scoring system is simplified compared with the Mirels pain severity score because it is binary (presence or absence of pain in the affected location), whereas the Mirels system uses three grades of pain. We did not evaluate the inter-rater reliability of the scoring systems because our team approach may have inflated agreeability because of our consistent collaboration and involvement in interpreting the Mirels score and in the development of the novel score. An independent study should evaluate the inter-rater reliability of the Mirels score and this novel score. We theorize that our novel system would perform similarly, if not better, because of its simplified pain component.
We excluded lesions that had not fractured or had undergone surgical treatment with follow-up less than 1 year. Lesions with follow-up of less than 1 year may behave differently than lesions with longer follow-up. Therefore, our cohort may have been biased toward lesions with unique behavior. However, fractures that occurred less than 1 year after lesion identification were included, regardless of duration of follow-up. In addition, 13% of patients (22 of 163) may have been lost to follow-up (not known to have died and did not present for care at our institutions within 6 months before data collection). These patients may have transferred their care, had an indolent disease course that needed less frequent follow-up, or received treatment at other care centers. However, these patients who were lost to follow-up did not have different baseline characteristics and had mean follow-up of 3.9 years, which is similar to duration of follow-up of the entire cohort and approximately four times the follow-up needed for inclusion. In addition, the nature of the practice patterns of the orthopaedic oncology team at our high-volume referral center suggests that few patients would likely be treated for pathologic or impending pathologic fractures at other facilities.
Although most patients had multiple long-bone lesions that required evaluation, we selected only the largest lesion in each patient. The Mirels score included multiple lesions per patient. We limited our cohort to one lesion per patient to avoid analyzing lesions that were not independent events. Therefore, the novel score was developed to evaluate the patient’s largest lesion at presentation and should be used accordingly in clinical practice. We performed additional analyses of patients with multiple lesions, using up to three lesions per patient, resulting in similar conclusions (data not shown). Selecting the largest lesion per patient may have biased the sample toward the lower extremity. However, our additional analyses of patients with multiple lesions had similar numbers of upper and lower extremity lesions. Furthermore, lower extremity lesions, in addition to bearing more weight compared with those in the upper extremity, are of particular importance to oncology patients because of their effects on ambulation. Further studies may be conducted to evaluate performance of the novel system in smaller lesions in other long-bones, either in different anatomic sites or within the same bone.
We excluded prophylactically stabilized lesions because of the uncertainty of their clinical outcomes. We believe that such lesions would have artificially increased sensitivity and PPV through inclusion of false positives. However, because these lesions accounted for less than 4% of the patient cohort, we determined that the exclusion was an acceptable limitation of our analysis.
Fracture history was not used in the novel system because most patients already had other skeletal manifestations of their disease, such as rib or vertebral fractures. Furthermore, laboratory measurements were not different between the groups, which may be attributed to the variability of these measures in the setting of multiple myeloma and the influence of systemic and antiresorptive therapies. These variables were not predictors of pathologic fracture in the current study, but they may be useful for patients with extensive fracture histories and if they are assessed at standardized time points.
Developing a Novel Scoring System for Fracture Prediction in Multiple Myeloma
Variables that were independently associated with pathologic fracture were lesion size (area), lesion latency, metaphyseal location, and presence of pain. Several of these factors incorporated into the novel scoring system align with previous research. The metaphysis may be more susceptible to pathologic fractures because of its different biomechanical properties compared with other portions of bone [14]. Larger lesions likely compromise the structural integrity of the bone to a greater degree than do smaller lesions as a result of removal of structural bone stock. Pain may be a marker of mechanical insufficiency and has been identified as a marker of impending fracture in patients with osseous metastatic disease [3]. The novel scoring system was constructed using these variables, with pain appropriately represented.
In our clinical practice, we have noted that lesions for which Mirels scores predict fracture are often present for a long period without clinical consequences and without undergoing stabilization, indicating that the risk assessment was incorrect. Our findings suggest that a prolonged interval between the diagnosis of multiple myeloma and the development of a given lesion (lesion latency) correlated strongly with pathologic fracture risk. It is unclear why this was the case. Chronic lesions, particularly those treated with chemotherapy or systemic antiresorptive therapy, are more likely to develop a sclerotic rim than are newly formed lesions [7, 18]. This tendency was evident in the current cohort as well. Lesion latency may be a surrogate for a relatively new lesion that has not had the opportunity to remodel the surrounding bone to improve structural integrity. There was no difference in the mean time after multiple myeloma diagnosis between the fracture and non-fracture groups. This finding may suggest that lesion latency is not only a surrogate for a new lesion with respect to the disease course of multiple myeloma but also for refractory osseous disease. Furthermore, lesion latency may reflect the overall progression of disease late in the management of multiple myeloma rather than a local osseous phenomenon alone. In the current analysis, anatomic location did not predict pathologic fracture. Although the anatomic location was included in the scoring system proposed by Mirels, it was not found to be a predictor of fracture risk in the Mirels study [12]. In addition, sclerotic rim was not associated with fracture. This may be attributable to our methods, which analyzed only the initial radiograph. Subsequent radiographs of these lesions may have demonstrated formation of a sclerotic rim, but this was not investigated because of our intention to use only one radiograph for the novel scoring system.
Comparing the Novel System to the Mirels System
Our novel scoring system was more sensitive than Mirels in the prediction of fractures but did not differ in terms of specificity, PPV, NPV, or ROC AUC. Analysis of a greater number of fractures may improve statistical comparisons for these metrics. Given the advent of CT-based structural rigidity analysis and the continued progression of fracture risk stratification technologies, a more sensitive scoring system, such as the one described, may be useful for determining which patients would benefit from these more advanced techniques. In addition to its ability to limit further radiation exposure in patients being treated for cancer, the novel scoring system may be valuable in communities with limited healthcare resources by offering simplicity and relatively low cost. Clinicians may also use the novel scoring system to decide which patients would benefit from referral to an orthopaedic oncologist for further evaluation or transfer of their care to a more resource-rich health system with advanced technologies. Our results indicate that the Mirels system is limited in determining the risk of pathologic fracture in patients with multiple myeloma because of lower sensitivity, and thus would not be useful as a radiographic tool to determine which patients are at risk of pathologic fracture.
Conclusion
The novel scoring system presented herein may be more effective than the Mirels system for estimating the fracture risk in patients with multiple myeloma, as evidenced by its better sensitivity and seemingly comparable specificity, PPV, NPV, and ROC AUC. This novel system may help estimate fracture risk on the basis of skeletal surveys, which are performed routinely for patients with multiple myeloma. Future studies should include more fractures to reduce the impact of sparse-data bias, validate the system in other care centers, grade the system’s inter-rater reliability, and ultimately evaluate the novel system in a prospective manner. Because many multiple myeloma lesions are asymptomatic and are found on a skeletal survey, this novel scoring system may be especially helpful for medical and radiation oncologists, who may be able to determine which patients may be observed and which may benefit from an advanced radiographic examination or orthopaedic evaluation.
Acknowledgments
We thank Rachel Box MS, Jenni Weems MS, and Kerry Kennedy BA, for their editorial assistance. We also thank Jad El Abiad MD, and Vaishali Laljani for their research coordination assistance.
Footnotes
Each author certifies that neither he nor she, nor any member of his or her immediate family, has funding or commercial associations (consultancies, stock ownership, equity interest, patent/licensing arrangements, etc) that might pose a conflict of interest in connection with the submitted article.
All ICMJE Conflict of Interest Forms for authors and Clinical Orthopaedics and Related Research® editors and board members are on file with the publication and can be viewed on request.
Each author certifies that his or her institution approved the human protocol for this investigation and that all investigations were conducted in conformity with ethical principles of research.
References
- 1.American Cancer Society. Key Statistics About Multiple Myeloma. Available at: https://www.cancer.org/cancer/multiple-myeloma/about/key-statistics.html. Accessed on March 12, 2019.
- 2.Bickels J, Dadia S, Lidar Z. Surgical management of metastatic bone disease. J Bone Joint Surg Am. 2009;91:1503-1516. [DOI] [PubMed] [Google Scholar]
- 3.Coleman RE. Clinical features of metastatic bone disease and risk of skeletal morbidity. Clin Cancer Res. 2006;12:6243s-6249s. [DOI] [PubMed] [Google Scholar]
- 4.Evans AR, Bottros J, Grant W, Chen BY, Damron TA. Mirels' rating for humerus lesions is both reproducible and valid. Clin Orthop Relat Res. 2008;466:1279-1284. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Fidler M. Prophylactic internal fixation of secondary neoplastic deposits in long bones. Br Med J. 1973;1:341-343. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Fidler M. Incidence of fracture through metastases in long bones. Acta Orthop Scand. 1981;52:623-627. [DOI] [PubMed] [Google Scholar]
- 7.Hinge M, Andersen KT, Lund T, Jorgensen HB, Holdgaard PC, Ormstrup TE, Ostergaard LL, Plesner T. Bone healing in multiple myeloma: a prospective evaluation of the impact of first-line anti-myeloma treatment. Haematologica. 2016;101:e419-e422. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Howard EL, Shepherd KL, Cribb G, Cool P. The validity of the Mirels score for predicting impending pathological fractures of the lower limb. Bone Joint J. 2018;100-B:1100-1105. [DOI] [PubMed] [Google Scholar]
- 9.Leopold SS, Porcher R. Editorial: Sparse-data Bias-What the Savvy Reader Needs to Know. Clin Orthop Relat Res. 2018;476:657-659. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Li S, Peng Y, Weinhandl ED, Blaes AH, Cetin K, Chia VM, Stryker S, Pinzone JJ, Acquavella JF, Arneson TJ. Estimated number of prevalent cases of metastatic bone disease in the US adult population. Clin Epidemiol. 2012;4:87-93. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Melton LJ, 3rd, Kyle RA, Achenbach SJ, Oberg AL, Rajkumar SV. Fracture risk with multiple myeloma: a population-based study. J Bone Miner Res. 2005;20:487-493. [DOI] [PubMed] [Google Scholar]
- 12.Mirels H. Metastatic disease in long bones. A proposed scoring system for diagnosing impending pathologic fractures. Clin Orthop Relat Res. 1989;249:256-264. [PubMed] [Google Scholar]
- 13.Murray JA, Bruels MC, Lindberg RD. Irradiation of polymethylmethacrylate. In vitro gamma radiation effect. J Bone Joint Surg Am. 1974;56:311-312. [PubMed] [Google Scholar]
- 14.Nobakhti S, Katsamenis OL, Zaarour N, Limbert G, Thurner PJ. Elastic modulus varies along the bovine femur. J Mech Behav Biomed Mater. 2017;71:279-285. [DOI] [PubMed] [Google Scholar]
- 15.Noone AM, Howlader N, Krapcho M, Miller D, Brest A, Yu M, Ruhl J, Tatalovich Z, Mariotto A, Lewis DR, Chen HS, Feuer EJ, Cronin KA. SEER Cancer Statistics Review, 1975-2015, National Cancer Institute, based on November 2017 SEER data submission. Available at: https://seer.cancer.gov/csr/1975_2015/. Accessed on January 30, 2019.
- 16.Parrish FF, Murray JA. Surgical treatment for secondary neoplastic fractures. A retrospective study of ninety-six patients. J Bone Joint Surg Am. 1970;52:665-686. [PubMed] [Google Scholar]
- 17.Saad F, Lipton A, Cook R, Chen YM, Smith M, Coleman R. Pathologic fractures correlate with reduced survival in patients with malignant bone disease. Cancer. 2007;110:1860-1867. [DOI] [PubMed] [Google Scholar]
- 18.Schulze M, Weisel K, Grandjean C, Oehrlein K, Zago M, Spira D, Horger M. Increasing bone sclerosis during bortezomib therapy in multiple myeloma patients: results of a reduced-dose whole-body MDCT study. AJR Am J Roentgenol. 2014;202:170-179. [DOI] [PubMed] [Google Scholar]
- 19.Selvaggi G, Scagliotti GV. Management of bone metastases in cancer: a review. Crit Rev Oncol Hematol. 2005;56:365-378. [DOI] [PubMed] [Google Scholar]
- 20.Snell W, Beals RK. Femoral metastases and fractures from breast cancer. Surg Gynecol Obstet. 1964;119:22-24. [PubMed] [Google Scholar]
- 21.Ward WG, Holsenbeck S, Dorey FJ, Spang J, Howe D. Metastatic disease of the femur: surgical treatment. Clin Orthop Relat Res. 2003;S230-244. [DOI] [PubMed] [Google Scholar]
- 22.Zickel RE, Mouradian WH. Intramedullary fixation of pathological fractures and lesions of the subtrochanteric region of the femur. J Bone Joint Surg Am. 1976;58:1061-1066. [PubMed] [Google Scholar]