SEARCH

SEARCH BY CITATION

Keywords:

  • femur;
  • fracture prediction;
  • soft tissue sarcoma;
  • radiation;
  • nomogram

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONFLICT OF INTEREST DISCLOSURES
  7. REFERENCES

BACKGROUND:

The objective of the current study was to formulate a scoring system to enable decision making for prophylactic stabilization of the femur after surgical resection of a soft tissue sarcoma (STS) of the thigh.

METHODS:

A logistic regression model was developed using patient variables collected from a prospectively collected database. The study group included 22 patients who developed a radiation-related pathological fracture of the femur after surgery and radiotherapy for an STS of the thigh. The control group of 79 patients received similar treatment but did not sustain a fracture. No patients received chemotherapy. The mean follow-up was 8.6 years. The variables examined were age, gender, tumor size, radiation dose (low [50 grays (Gy)] vs high [≥60 Gy]), extent of periosteal stripping (<10 cm, 10-20 cm, and >20 cm), and thigh compartment involvement (posterior, adductor, anterior or other [ie, abductors and groin]).

RESULTS:

On the basis of an optimal regression model, the ability to predict radiation–associated fracture risk was 91% sensitive and 81% specific. The area under the receiver operating characteristic curve was 0.9, which supports this model as a very accurate predictor of fracture risk.

CONCLUSIONS:

Radiation-related fractures of the femur after combined surgery and radiotherapy for STS are uncommon, but are difficult to manage and their nonunion rate is extremely high. The results of the current study suggest that it is possible to predict radiation-associated pathological fracture risk using patient and treatment variables with high sensitivity and specificity. This would allow for the identification of high-risk patients and treatment with either close follow-up or prophylactic intramedullary nail stabilization. The presentation of this model as a nomogram will facilitate its clinical use. Cancer 2010. © 2010 American Cancer Society.

Radiation-associated pathologic fractures are rare but serious complications of combined modality treatment of soft tissue sarcomas (STS). The reported fracture rates vary between 1.2% and 6.4%. Although any bone can be affected, the femur is involved most frequently.1-7

Factors implicated in determining the risk for a radiation-associated pathologic fracture include age, gender, radiation dose, periosteal stripping, anatomical location, and chemotherapy. The relative impact of each of these factors is difficult to interpret because of diverging conclusions that have been drawn in different studies and the rarity of these fractures.5, 6

The management of radiation-associated pathologic fractures of the femur is controversial. These fractures tend to have a high nonunion rate and are a major cause of patient morbidity. Several treatments have been suggested to assist union in addition to surgical stabilization, including the liberal use of bone grafting or vascularized fibular grafting; however, the early use of endoprosthetic replacement is a reasonable alternative given the high risk of nonunion.8

Prophylactic femoral intramedullary (IM) nailing is an attractive alternative to the more common wait–and–see approach after combined surgery and radiotherapy for an STS of the thigh. This should minimize the risk of subsequent fracture, nonunion, and additional surgery. However, it has been suggested that prophylactic nailing should not be contemplated on a regular basis because the fracture risk is low.5, 6 Furthermore, routine IM nailing adds additional time to often lengthy procedures, as well as potential morbidity. For example, large, deep tumors in the thigh that are likely to be at high fracture risk are also at risk of wound healing complications, particularly after preoperative radiation.9

The rationale behind the current study was to develop a prognostic model to enable the accurate prediction of the risk of fracture of the femur based on easily defined clinical and treatment parameters. This will subsequently help the surgeon decide whether to prophylactically stabilize the femur in different clinical situations. The presentation of the model as a clinical nomogram that incorporates the different risk factors will facilitate the surgeon's decision making in treating patients with an STS of the thigh.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONFLICT OF INTEREST DISCLOSURES
  7. REFERENCES

Patient Groups

After obtaining Institutional Review Board approval, our prospectively collected database was reviewed and 101 patients were identified who had received combined modality treatment using surgery and radiotherapy for an STS of the thigh between 1986 and 2006 and for whom complete clinical information was available. A total of 22 patients sustained a radiation-associated fracture of the femur, defined as a fracture occurring in the previous radiation field and associated with minimal energy. The patient and treatment variables of this group were compared with the 79 patients who had received the same treatment of a thigh STS, but had not sustained a fracture. A review of the literature identified the following risk factors for analysis of prediction of fractures of the femur after combined modality treatment of an STS of the thigh: age, gender, tumor size, site of the tumor (adductor, posterior, anterior compartment, or other anatomic areas around the hip that are not included in the thigh compartments, such as the hip abductor and groin areas), radiation dose (low [50 grays (Gy)] vs high [≥60 Gy]), and the amount of periosteal stripping (<10 cm, 10-20 cm, and >20 cm). The extent of periosteal stripping was adapted from the criteria proposed by Helmstedter et al6 using information from surgical notes. When the length of periosteal stripping was not specifically cited in the surgical note, it was concluded from the tumor size and muscles stripped as summarized in the pathology report and the surgical note, as well as from preoperative cross-sectional imaging. Chemotherapy, although cited as a possible risk factor for fracture in some studies,1, 5 was not included in our model because none of the patients in this cohort received chemotherapy as part of their treatment.

Statistical Analysis

Statistical analysis was performed using SAS statistical software (version 9.1; SAS Institute Inc, Cary, NC). Simple logistic regression analysis was performed to examine the significance of each variable for predicting the probability of a fracture. Multiple logistic regression analysis was used to derive the predictive equation for the probability of a fracture. To minimize overfitting, all variables were used in the model regardless of statistical significance. The utility of the model was evaluated using sensitivity, specificity, and area under the receiver operating characteristic (ROC) curve. A nomogram was developed based on the multiple logistic regression model.10

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONFLICT OF INTEREST DISCLOSURES
  7. REFERENCES

A total of 22 of 101 patients treated with surgery and radiotherapy for an STS of the thigh developed a radiation–associated pathologic fracture of the femur. The median time to fracture after combined modality treatment was 4 years (range, 1-13 years). The majority (68%) of fractures occurred within 5 years of treatment (Fig. 1). The 79 patients who did not sustain a fracture had a minimum follow-up of 4.5 years (median, 8 years; range, 4.5-19 years).

thumbnail image

Figure 1. A histogram depicting fracture occurrence over time in 22 patients who underwent combined modality treatment for soft tissue sarcoma of the thigh is shown.

Download figure to PowerPoint

Patient demographic, oncologic, and surgical and radiation treatment variables were compared between the fracture and nonfracture patient groups (Table 1). Univariate analysis revealed that age (P = .005), tumor size (P = .005), length of periosteal stripping (P = .007), and the muscle compartment involved (P = .02) were significantly different between the 2 groups. Gender was found to be marginally different between the groups (P = .08), and no difference was identified with regard to radiation dose (P = .36).

Table 1. Univariate Analysis of Clinical and Treatment Variables of Radiation-Related Pathological Fractures of the Femur in 101 Patients Treated for a Soft Tissue Sarcoma of the Thigh
VariablesNo Fracture (n=79)Fracture (n=22)OR (95% CI)aPa
  • OR indicates odds ratio; 95% CI, 95% confidence interval.

  • a

    The ORs and P values were derived from single variable logistic regression models.

Median age at surgery (range), y55 (20-84)65 (39-89)1.05 (1.02-1.09).005
Gender   .08
 Male (reference)42 (53.2%)7 (31.8%)1.0 
 Female37 (46.8%)15 (68.2%)2.4 (0.9-6.6) 
Median tumor size (range), cm8 (1-28)15 (4-28)1.2 (1.1-1.3).005
Radiation dose   .36
 Low dose (reference)30 (38.0%)6 (27.3%)1.0 
 High dose49 (62.0%)16 (72.7%)1.6 (0.6-4.6) 
Periosteal stripping, cm   .007
 <10 (reference)68 (86.1%)12 (54.5%)1.0 
 10-2010 (12.7%)6 (27.3%)3.5 (1.0-11.1) 
 >201 (1.3%)4 (18.2%)22.7 (2.3-220.6) 
Compartment/tumor location   .02
 Other (abductor, groin) (reference)23 (29.1%)1 (4.6%)1.0 
 Adductor16 (20.3%)4 (18.2%)5.7 (0.6 −56.3) 
 Posterior11 (13.9%)2 (9.1%)4.2 (0.3-51.2) 
 Anterior29 (36.7%)15 (68.2%)11.9 (1.5-96.8) 

On the basis of multiple logistic regression analysis using all potential prognostic variables (Table 2), age (P = .003) and tumor size (P = .008) remained statistically significant predictors of fracture risk. In the presence of all the variables, gender reached conventional levels of significance (P = .02), whereas muscle compartment was found to be only marginally significant (P = .07) and periosteal stripping ceased to remain significant (P = .54). Radiation dose remained nonsignificant (P = .35).

Table 2. Multivariate Logistic Regression Analysis of Clinical and Treatment Variables of Radiation-Related Pathological Fractures of the Femur in 101 Patients Treated for a Soft Tissue Sarcoma of the Thigh
VariableOdds Ratio95% CIP
  1. 95% CI indicates 95% confidence interval.

Age at index surgery2.3 (per 10-y change)1.3-3.9.003
Gender  .02
 Male (reference)1.0  
 Female7.01.4-35.4 
Tumor size, cm2.3 (per 5-cm change)1.2-4.3.008
Radiation dose   
 Low dose (reference)1.0  
 High dose2.00.5 - 8.9.35
Periosteal stripping, cm  .54
 <10 (reference)1.0  
 10-202.30.5-10.2 
 >201.70.1-24.7 
Compartment/tumor location  .07
 Other (reference)1.0  
 Adductor3.40.3-43.1 
 Posterior0.80.04-14.9 
 Anterior12.21.1-133.6 

The overall fit of the model as represented by the area under the ROC curve was 0.90 (Fig. 2). Two cutpoints were suggested. Using a cutpoint with a predicted probability of fracture of 0.20 from the model provided a sensitivity of 91% and a specificity of 81%, whereas a cutpoint of 0.35 corresponded to a sensitivity of 82% and a specificity of 91%. A nomogram was developed from this multivariate logistic regression model (Fig. 3), in which a patient's probability of fracture can be calculated by finding the number of points on the top axis that corresponds to each variable's value for that individual, summing these points, and then drawing a vertical line from the resulting total points axis to the probability of fracture axis.

thumbnail image

Figure 2. The receiver operating characteristic curve depicts the trade-off between the sensitivity (sens) and specificity (spec) of a chosen cutpoint for fracture probability as obtained from the multiple logistic regression model. Increasing the probability for fracture as calculated by the nomogram decreases the sensitivity and increases the specificity and positive predictive value for the chosen cutpoint.

Download figure to PowerPoint

thumbnail image

Figure 3. A nomogram to predict a radiation-associated fracture of the femur after combined modality treatment of a soft tissue sarcoma of the thigh is shown. A patient's probability of fracture can be calculated by finding the number of points on the top axis (which corresponds to each variable's value for that individual), summing these points, and then drawing a vertical line from the resulting total points axis to the probability of fracture axis.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONFLICT OF INTEREST DISCLOSURES
  7. REFERENCES

The outcome of patients with radiation-associated pathologic fractures of the femur after combined management of STS of the thigh is dismal. Lin et al7 described union in only 4 of 12 such fractures, which united after open reduction and internal fixation at a range of 16 to 28 months despite bone grafting being performed in each case. Five patients required additional surgical procedures, and 17 complications occurred in 10 patients. Helmstedter et al6 reported on a series of 20 radiation-associated fractures, 12 of which occurred in the femur. The nonunion rate was 60% in those fractures treated surgically. Furthermore, the authors identified a significant decline in functional outcome after surgical treatment of fractures involving the femur. Livi et al3 reported on 7 fractures occurring in patients who had received high–dose radiotherapy for STS of the extremity. Four of the fractures healed after minor surgery, whereas 3 patients underwent amputation of the affected limb. However, no information was provided with regard to the site of the fracture or the surgical treatment necessary to achieve union.

Given the extremely poor outcome of the conventional treatment of radiation-associated fractures of the femur in patients with STS, better treatment options are necessary. Although different techniques of internal fixation, bone grafting, or endoprosthetic replacement have all been advocated to manage the fracture once it occurs, fracture prevention might be the most important intervention.8 Our goal was prediction of the risk of fracture based on well–understood clinical parameters. If accurate prediction of the risk of subsequent fracture was possible, then prophylactic stabilization could potentially be used selectively at the time of STS resection. We realize that prophylactic stabilization does not entirely preclude subsequent fracture of the femur, because anecdotal cases have been described even after the insertion of an IM nail. Conversely, to the best of our knowledge, no widely applicable biological solution is yet available to treat these fractures in a reliable fashion.8

Several patient and treatment variables have been proposed as risk factors for radiation-associated pathologic fractures of the femur in patients undergoing combined surgical resection and radiotherapy for STS of the thigh. Lin et al found 6 factors to be significant on univariate analysis: amount of periosteal stripping, anterior compartment location, age, gender, receipt of chemotherapy, and external beam radiotherapy. However, in a subset of their patients who had all undergone periosteal stripping, only female gender and chemotherapy were found to have prognostic value.5 Helmstedter et al6 concurred with the finding of the prior study that periosteal stripping is a risk factor for fracture because 32% of the patients in their study who underwent moderate or extensive periosteal stripping went on to fracture their femur. In addition, anterior compartment involvement and intralesional resection increased the risk of fracture whereas age and gender were not considered to be risk factors for fracture in their study. Holt et al4 found that high–dose radiation, as is often used in postoperative regimens, was associated with a higher risk of fracture than lower dose radiation, which is more typical of preoperative treatment. In addition, although moderate or extensive periosteal stripping was found to be associated with a higher rate of fracture, multivariate analysis failed to confirm this as a significant variable. Livi et al3 noted that all fractures in their study group occurred in postmenopausal women. Cannon et al1 found that circumferential exposure of bone to radiation and surgical exposure of bone by periosteal stripping were associated with a higher incidence of fractures, but the small number of patients in their study did not allow for further analysis.1

On the basis of the findings of the above literature, we incorporated 6 candidate variables into our model: patient age, gender, tumor compartment location, tumor size, radiation dose, and length of periosteal stripping. Univariate analysis demonstrated that age at the time of surgery, tumor size, periosteal stripping, and compartment involvement were statistically significant risk factors for fracture. Gender was found to be only marginally significant and radiation dose was not found to be a significant factor. On multivariate analysis, age, tumor size, and gender continued to be significant risk factors.

Incorporation of all patient variables into the predictive model regardless of statistical significance minimizes the chance of overfitting and increases the possibility of good external validation of the model.11 Even in the absence of being statistically significant, all commonly accepted risk factors believed to contribute to fracture risk (eg, radiation dose) were included in the final model. To facilitate using these data in the clinical setting, a multiple regression model was developed that would be the basis for a clinical nomogram. Evaluation of this model using a ROC curve revealed a large area of 0.9 under the curve. The area under the ROC curve represents the accuracy of the model and its ability to correctly stratify the patient according to his fracture risk. Its presentation as a nomogram allows the surgeon to easily translate an individual patient's risk factors into the predicted probability of a fracture provided by the statistical model. The choice of what cutpoint of predicted probability should be used to decide to ultimately stabilize the femur during surgery remains in the hands of the individual surgeon; however, based on data from the current study, the ROC curve suggests 2 possible decision points: a probability of fracture of 20%, which results in 91% sensitivity and 81% specificity, or a probability of fracture of 35%, which provides 82% sensitivity and 91% specificity. Preference for either probability depends on whether the surgeon wishes to maximize the sensitivity or the specificity. Increasing the fracture probability threshold will decrease the sensitivity and increase the specificity for the chosen cutpoint. The positive predictive value of the model also increases in this situation. However, external validation of the model and cutpoints are needed before the nomogram can be recommended for use in clinical practice.

Periosteal stripping has been quantified by Helmstedter et al into 4 different categories (none, <10 cm, 10-20 cm, and >20 cm), and is an accepted risk factor for radiation-associated pathologic fractures.6 In the current study, it was not found to be an independent risk factor on multivariate analysis. One possible reason is that accurate documentation of the actual extent of periosteal stripping was not regularly listed in the surgical note.1 In the current study, we applied the system of Helmstedter et al retrospectively by checking patient surgical notes, pathology reports, and preoperative magnetic resonance imaging scans to attempt to quantify the amount of periosteal stripping that was performed. Despite the finding that periosteal stripping did not appear to be an independent risk factor on multivariate analysis, it was incorporated into the statistical model combined with the other proposed risk factors, based on its demonstrated influence in other studies.

Tumor involvement of the anterior compartment of the thigh has been implicated as a significant fracture risk. Approximately 88% to 92% of fractures have been reported to occur in patients with anterior compartment involvement.5, 6 Our study group demonstrated a similar, but less pronounced, trend, with 15 of 22 fractures occurring after tumor resection from the anterior compartment. Anterior compartment involvement obtained statistical significance on univariate analysis but not on multivariate analysis, suggesting that other factors such as large tumor size played a more important role in leading to fractures in these patients.

To the best of our knowledge, tumor size has not been extensively examined as a risk factor for radiation-associated fracture. Intuitively, it would appear that a larger tumor would require a larger amount of periosteal stripping to obtain adequate margins, thereby devascularizing a greater amount of bone. Furthermore, a larger radiation volume would also be necessary, potentially exposing a greater volume of bone to the effects of radiation. The results of the current study confirmed that tumor size was an independent risk factor on multivariate analysis.

We were initially surprised that radiation dose was not found to be an independent risk factor for fracture in this study. On the basis of our previous investigation,4 which included many of the same patients from the current study, high–dose radiation was found to be a predictor of subsequent fractures. The statistical model used in the current study appears more robust, most likely accounting for at least part of this difference. In addition, the study by Holt et al4 included radiation-associated fractures from locations other than the femur. However, we believe the final multivariate model in the current study still predicts fracture risk more accurately when radiation dose is included as a predictive factor.

Based on our model, patient age, gender, and tumor size are the most important risk factors for radiation-associated fractures of the femur. This suggests that underlying osteoporosis may be an important contributing factor in these patients. Elderly females are by far the most likely patients to develop radiation-associated fractures, just as they are osteoporosis-related fractures in the general population. It is likely that physiologic changes in the femur that occur after the combined management of STS of the thigh, including decreased vascularity and decreased bone remodeling, but increased resorption, result in further decreased bone density, thereby rendering the femur even more susceptible to fracture with minimal trauma.12, 13

Statistical predictive models are available for many types of cancer that supply reliable prognostic information for individual patients.14-16 In the recent musculoskeletal oncology literature, nomograms have become a clinically accepted tool with which to predict survival in patients with STS.17-19 The model developed herein and presented as a clinical nomogram was based on multiple logistic regression analysis of clinical factors within our study group. The nomogram provides the clinician with the ability to stratify patients and decide on an appropriate treatment plan. The guidelines proposed in the literature with regard to whether it is necessary to prophylactically stabilize a femur after resection of an STS in combination with radiotherapy are simple algorithms that incorporate different risk factors without addressing their relative importance, or use terms such as “high–risk” patients and yet rely on the clinical impression of the surgeon.1, 5, 6 The nomogram presented in the current study enables the surgeon to quantify a patient's risk of developing a femur fracture by taking into account the relative importance of each risk factor.

The nomogram should help the surgeon to decide which patients might benefit the most from prophylactic IM nailing and assist as a patient counseling tool before treatment. It should be validated in other patient populations and is not meant to replace the periodic clinical and radiographic evaluation of patients after treatment. Fractures of the irradiated femur do occur despite IM nailing and the monitoring of high-risk patients should not be changed despite prophylactic fixation. Furthermore, the nomogram can help guide the physician and the patient to be more vigilant in the case of those patients who have not been stabilized despite having a high probability of fracture.

The nomogram does not provide an answer as to when one should perform prophylactic stabilization. IM stabilization at the time of index surgery in a femur that has been exposed to radiation and possible periosteal stripping could have deleterious effects on the vascularity of the involved bone by also disrupting the endosteal blood supply.20-23 In addition, wound healing problems are common after combined modality treatment of STS of the thigh, especially when the medial compartment is resected after preoperative radiotherapy,9 which may further justify a decision not to insert hardware at the time of the index procedure. Conversely, in our patient population, approximately 30% of fractures occurred within 2 years of surgery, which suggests that prophylactic stabilization should not be delayed in patients who are at highest risk based on their calculated fracture probability.

Ultimately, however, the surgeon will still have to make a clinical decision based on his experience: for example, whether a 50% probability of fracture is a sufficient reason to prophylactically stabilize a femur during the initial resection of a sarcoma or whether to observe the patient closely, including letting them know that new symptoms such as mild activity–related thigh pain might signify the early development of a stress fracture in the radiated field and to seek early reassessment.

One shortcoming of the current study was our inability to evaluate the role of chemotherapy as a risk factor for fracture. At the study institution, chemotherapy is typically only used in the curative adjuvant setting for patients with certain histological subtypes of STS such as rhabdomyosarcoma, soft tissue Ewing sarcoma, osteosarcoma, and synovial sarcoma. Other groups that identified chemotherapy as being a risk factor for femur fracture used neoadjuvant chemotherapy more frequently in patients with intermediate and/or high–grade tumors.1, 3, 5 Involvement of multiple compartments in the thigh and combined preoperative and postoperative radiotherapy could also influence the risk of fracture, but these factors were not included in our model because of a lack of patients who presented with these parameters in the current study. To validate our model, it will have to be tested in another independent patient population.

In summary, our statistical model, using the commonly accepted risk factors of patient age, gender, tumor compartment location, tumor size, radiation dose, and degree of periosteal stripping, allows for the prediction of radiation-associated fractures of the femur in patients with STS, with a high sensitivity and specificity. The nomogram presented herein can assist the surgeon in preoperative decision making when contemplating prophylactic stabilization of the femur after resection of an STS of the thigh. Accurate prediction of the risk of femur fracture and utilization of prophylactic IM nailing may reduce the incidence of this serious complication of sarcoma management and prevent additional patient morbidity.

CONFLICT OF INTEREST DISCLOSURES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONFLICT OF INTEREST DISCLOSURES
  7. REFERENCES

Yair Gortzak is a Joey and Toby Tanenbaum fellow in Surgical Oncology, 2008-2009.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONFLICT OF INTEREST DISCLOSURES
  7. REFERENCES