Abstract
Background
The extent of tumor necrosis after neoadjuvant chemotherapy is an important predictive factor of survival in osteosarcoma. However, the response to chemotherapy is not known until after the definitive resection and limits the utility of this information for operative planning. Our study questions include: 1) Are there clinical and radiographic factors following neoadjuvant chemotherapy, but prior to the tumor resection, that may aid in predicting response to treatment? 2) Can we combine these criteria into a predictive composite score that can identify good and poor responders to chemotherapy?
Methods
We identified consecutive patients diagnosed with osteosarcoma and managed with neoadjuvant chemotherapy prior to surgical resection. We assessed post-chemotherapy tumor ossification, tumor size and growth, and the presence of pain to devise a scoring criteria to predict the percent necrosis on the final histologic specimen. Bivariate analyses were done, and a receiver operating characteristic curve was constructed to determine predictive capacity.
Results
Out of the 40 patients included in this study, 15 (38%) had a good response (≥ 90% necrosis) to treatment and ten patients (25%) had a poor response with ≤ 50% necrosis. Tumor size, growth and increase in ossification were significantly associated with a good response to treatment. For good responders, a composite score of 6 was seen to attain the highest sensitivity and specificity, 100% and 84%, respectively. Tumor size, no change in ossification, and post-chemotherapy pain were significantly associated with a poor response to treatment. For poor responders, a composite score of 7 was seen to have the highest sensitivity and specificity, 100% and 63%, respectively.
Conclusion
Compared to the use of one single factor, our combined scoring criteria demonstrated a far improved accuracy in identifying good responders to neoadjuvant chemotherapy, where a score of 6 or less is predictive of a good response. However, the specificity of this scoring criteria to predict poor responders was low, indicating that this criterion may not be the most accurate method to identify poor responders. The utility of this score has implications regarding pre-operative counseling of the patient and operative planning.
Level of Evidence: III
Keywords: orthopaedic surgery, osteosarcoma, neoadjuvant chemotherapy, predictive factor
Introduction
In the past, osteosarcoma was treated with surgery alone, and more specifically, amputation. The five-year survival rate at that time was abysmally low at approximately 22%.1 The addition of chemotherapy significantly improved the prognosis of patients with osteosarcoma and also contributed to limb-salvage surgery being an effective treatment. Now, five-year survival rates have increased to 60%.1,2 Despite these initial advancements in treatment, local recurrence and metastatic spread of osteosarcoma continues to be a significant problem and subsequent attempts to improve survival rates have been disappointing. It is well accepted that the extent of tumor necrosis after neoadjuvant chemotherapy is an important predictive factor of local recurrence and survival in osteosarcoma.3-8 In most of these studies, tumor necrosis of 90% or greater was considered “good” and less than 90% was defined as “poor” response.3-10
Recent attempts have investigated the predictive value of radiological response to neoadjuvant chemotherapy in osteosarcoma patients.11-15 However, to date there is not an easily determined and accurate scoring system using clinical and radiograph factors to predict response to neoadjuvant chemotherapy. From the perspective of the surgeon, assurance of a good response to chemotherapy will comfortably allow for a closer margin (within one-two mm of important structures) and a limb sparing surgery, while a poor response may make a plan for wider margins, or possibly an alteration in plan to an amputation, more advisable. The difficulty in this clinically is that the neoadjuvant response to chemotherapy chemotherapy. is not known until days after the surgery has been completed, compromising the surgeon’s ability to make the most appropriate operative decision. Accurately predicting the response to chemotherapy would have the potential to improve outcomes by allowing the surgeon to make adjustments in the treatment plan, including surgery and surveillance, prior to the operative encounter.
Our goals are to (1) identify clinical and radiographic factors following neoadjuvant chemotherapy, but prior to the tumor resection, that may aid in predicting a good or poor response to treatment; and (2) create a composite score that will identify good and poor responders to chemotherapy.
Methods
This study was an institutional review board-approved single-institution retrospective chart review of patients who had osteosarcoma from an ongoing cohort of extremity sarcoma patients from September 2010 to February 2020. Our population of interest included patients diagnosed with osteosarcoma and managed with multi-agent neoadjuvant chemotherapy, followed by tumor excision. We included all histologic subtypes of high-grade osteosarcoma, as well as localized or metastatic disease at diagnosis. We excluded cases of recurrent osteosarcoma, patients who were not treated with neoadjuvant chemotherapy or did not undergo tumor removal and patients with incomplete medical records.
Patient records, including clinic notes, pathology reports, and radiology reports, were reviewed to determine underlying patient and tumor characteristics. From the medical records, we identified potential predictive factors including presence of pain, tumor size and growth on MRI, and tumor ossification on plain radiographs.
Presence of pain was measured using the 10-point Visual Analog Scale (VAS). VAS score ranges of 1-3, 4-6, and 7-10 were minimal, moderate, and severe pain, respectively. The VAS score at the pre-chemotherapy visit was recorded as the baseline pain while the VAS score after completion of chemotherapy, just prior to surgery, was recorded as the post-chemotherapy pain.
Tumor size in the greatest dimension and growth was recorded using the MRI radiology reports before and after chemotherapy. The term “No change” in tumor size was reported if there was ≤ five mm size change. MRI imaging review by the authors was used if the radiology report was unclear in its description (Figure 1). Post-chemotherapy tumor size and overall change in size was utilized in the composite score.
Figure 1.
Measurements of tumor size in greatest dimension on MRI of femur of two patients. (A) MRI shows an increase in the greatest dimension of the tumor after neoadjuvant chemotherapy. (B) MRI shows a decrease in the greatest dimension of the tumor after neoadjuvant chemotherapy.
Tumor ossification was assessed on plain radiographs by two separate observers. Pre-chemotherapy and post-chemotherapy radiographs were viewed side by side, where ossification was graded as none, minimal, moderate, or extensive. There were no pre-set criteria to classify the ossification, hence this was done subjectively by both observers prior to independent review (Figure 2). We determined a patient to have an overall increase in ossification if the post-chemotherapy ossification was graded higher than the pre-chemotherapy ossification. A second assessment of radiographs were then done by both observers together at a later date. Kappa statistics were utilized to measure the degree of agreement between each observer.
Figure 2.
Examples of tumor ossification on plain radiographs AP view of distal femur of three different patients. (A) Minimal ossification of the tumor was demonstrated in this plain radiograph. (B) Moderate ossification of the tumor was demonstrated in this plain radiograph. (C) Extensive ossification was demonstrated in this plain radiograph.
Our primary outcome was percent tumor necrosis on final histopathology report, which was categorized based off the Huvos four-grade system.16 For the analysis, we defined a good response as necrosis ≥ 90% and a poor response when necrosis is ≤ 50% on the final histologic specimen.
Bivariate methods (Fisher’s exact testing) were used to investigate the association of identified clinical and radiographic factors with percent tumor necrosis. With these identified clinical and radiographic factors, we devised scoring criteria to predict good (≥ 90%) or poor (≤ 50%) responders to neoadjuvant chemotherapy (Table 1). With the devised scoring criteria, we scored all 40 patients and conducted sensitivity and specificity analyses to test the statistical validity of the scoring criteria. We also constructed a receiver operating characteristic (ROC) curve to determine predictive capacity of the composite score we created.
Table 1.
The Proposed Scoring System (The Larger the Score, The Worse the Outcome)
| Score | |||
|---|---|---|---|
| Variable | 1 | 2 | 3 |
| Change in Size | Smaller | No Change | Larger |
| Post-Chemo Size (cm) | ≤ 10 cm | > 10 cm | |
| Post-Chemo Pain | No | Yes | |
| Ossification Change | Increase | No Increase | |
Results
Out of the 630 extremity sarcoma patients, a total of 40 patients met the study criteria with a median age of 20 years (range, 8-70 years) (Figure 3). There were 22 male patients and 18 female patients. Seven patients presented with metastases at diagnosis, while 33 did not. Lower extremity (32) osteosarcomas were the most frequent, followed by upper extremity (7) and pelvic (1) osteosarcomas. The most common histologic subtype of osteosarcoma was conventional (34), followed by telangiectatic (3) and then one each of giant cell rich, periosteal, and parosteal with a high-grade component. Of the conventional subtypes of osteosarcoma, 16 cases were unspecified, 10 were chondroblastic, 6 were osteoblastic and 2 were of the fibroblastic variant (Table 2).
Figure 3.
Patients’ participation status flowchart.
Table 2.
Patients’ Characteristics and Tumors and Treatment Factors Between the 2 Groups which are Based on Tumor Necrosis at 90% and 50% Cutoff
| Demographics | Total patients (n=40) | ≥ 90% Necrosis (%) (n=15) | < 90% Necrosis (%) (n=25) | > 50% Necrosis (%) (n=30) | ≤ 50% Necrosis (%) (n=10) |
|---|---|---|---|---|---|
| Age | |||||
| < 20 | 20 | 11 (74) | 9 (36) | 18 (60) | 2 (20) |
| 20-40 | 8 | 2 (13) | 6 (24) | 4 (13) | 4 (40) |
| > 40 | 12 | 2 (13) | 10 (40) | 8 (27) | 4 (40) |
| Sex | |||||
| Male | 22 | 9 (60) | 13 (52) | 17 (57) | 5 (50) |
| Female | 18 | 6 (40) | 12 (48) | 13 (43) | 5 (50) |
| Location | |||||
| Lower Limb | 32 | 10 (67) | 22 (88) | 23 (77) | 9 (90) |
| Upper Limb | 7 | 4 (27) | 3 (12) | 6 (20) | 1 (10) |
| Pelvis | 1 | 1 (6) | 0 (0) | 1 (3) | 0 (0) |
| Histological subtype | |||||
| Conventional | 34 | 13 (87) | 21 (84) | 24 (80) | 10 (100) |
| Telangiectatic | 3 | 2 (13) | 1 (4) | 3 (10) | 0 (0) |
| Parosteal | 1 | 0 (0) | 1 (4) | 1 (3) | 0 (0) |
| Periosteal | 1 | 0 (0) | 1 (4) | 1 (3) | 0 (0) |
| Giant cell rich | 1 | 0 (0) | 1 (4) | 1 (3) | 0 (0) |
| Metastasis at Diagnosis | |||||
| Yes | 7 | 3 (20) | 4 (16) | 5 (17) | 2 (20) |
| No | 33 | 12 (80) | 21 (84) | 25 (83) | 8 (80) |
Overall, of the 40 patients included in this study, 15 (38%) had a good response to treatment, with four of those patients having complete tumor necrosis. Ten patients (25%) had a poor response to treatment with ≤ 50% necrosis. All 40 patients were scored using the criteria, with all patient scores ranging from 4 to 11.
In the good responders’ group, scores ranged from 4 to 6, while in those without a good response, scores ranged from 4 to 11. The data indicated that 86% of patients (6 of 7) with a score of 4, 80% of patients (4 of 5) with a score of 5, and 71% of patients (5 of 7) with a score of 6 were good responders. There were no good responders with a score greater than 6. (Table 3). The sensitivity and specificity of each score were plotted on the ROC curve, where a score of 6 is seen to attain the highest sensitivity and specificity: 100% and 84%, respectively. Thus, a score of 6 reflects a cutoff point for predicting a good response to neoadjuvant chemotherapy response to treatment. (Figure 4). Our criterion was able to attain a positive predictive value of 79%, a negative predictive value of 100%, and accuracy of 90%.
Table 3.
Probability of a Good Response to Treatment “Tumor Necrosis ≥ 90%”
| Score | Good Response ≥ 90% | Not Response Good <90% | % Good | % Not Good |
|---|---|---|---|---|
| 4 | 6 | 1 | 86% | 14% |
| 5 | 4 | 1 | 80% | 20% |
| 6 | 5 | 2 | 71% | 29% |
| 7 | 0 | 7 | 0% | 100% |
| 8 | 0 | 5 | 0% | 100% |
| 9 | 0 | 3 | 0% | 100% |
| 10 | 0 | 3 | 0% | 100% |
| 11 | 0 | 3 | 0% | 100% |
Figure 4.
ROC Curve for good responders. A score of 6 is seen to attain the highest sensitivity and specificity, 100% and 84% respectively; this reflects the cutoff score for predicting a good response to treatment.
We used this same scoring criteria to predict the poor responders to treatment, ≤ 50% tumor necrosis after neoadjuvant chemotherapy. Patient scores in this group ranged from 7 to 11, while all other patients with a tumor necrosis of > 50% had scores ranging from 4 to 11. The data indicated that 57% of patients (4 of 7) with a score of 7, 40% of patients (2 of 5) with a score of 8, 0 of 3 patients with a score of 9, 33% of patients (1 of 3) with a score of 10, and 100% of patients (3 of 3) with a score of 11 were considered to be poor responders to treatment (Table 4). Again, the sensitivity and specificity of each score were plotted on the ROC curve, where it is shown that a cutoff score of 7 is seen to have the highest sensitivity and specificity: 100% and 63%, respectively (Figure 5).
Table 4.
Probability of a Poor Response to Treatment “Tumor Necrosis ≤ 50%”
| Score | Poor Response ≤ 50% | Not Poor Response >50% | % Poor | % Not Poor |
|---|---|---|---|---|
| 4 | 0 | 7 | 0% | 100% |
| 5 | 0 | 5 | 0% | 100% |
| 6 | 0 | 7 | 0% | 100% |
| 7 | 4 | 3 | 57% | 43% |
| 8 | 2 | 3 | 40% | 60% |
| 9 | 0 | 3 | 0% | 100% |
| 10 | 1 | 2 | 33% | 67% |
| 11 | 3 | 0 | 100% | O% |
Figure 5.
ROC Curve for poor responders. A score of 7 is seen to attain the highest sensitivity and specificity, 100% and 63% respectively; this reflects the cutoff score for predicting a poor response to treatment.
Bivariate analyses of each predictive factor utilized in our proposed scoring criteria was done to assess whether any individual factor was as effective as the scoring criteria to predict treatment response, and to justify its inclusion in the overall scoring criteria. There was a strong association between the tumor’s growth response to treatment and tumor necrosis. Among the 15 good responders; the tumor decreased in size in 11 (73%) patients and had no change in size in four (27%) patients (p < 0.01) (Table 5.) In addition, 14 (93%) of the good responders had a post-treatment tumor size of ≤10 cm, compared to only one patient who had a tumor size of >10 cm (p < 0.01). There was also an association between change in tumor size and ≤ 50% tumor necrosis. Among the poor responders; the tumor grew larger in three (30%) patients and had no change in size in seven (70%) patients (p = 0.03) (Table 6).
Table 5.
Predictive Factors for Tumor Necrosis in Osteosarcoma at 90% Cutoff
| Predictive Factors | ≥90% Necrosis (%) (n=15) | < 90% Necrosis (%) (n=25) | p value |
|---|---|---|---|
| Tumor change in size | p < 0.01* | ||
| Larger | 0 (0) | 9 (36) | |
| Smaller | 11 (73) | 2 (8) | |
| No change | 4 (27) | 14 (56) | |
| Post-chemo tumor size | p < 0.01* | ||
| > 10 cm | 1 (7) | 13 (52) | |
| ≤ 10 cm | 14 (93) | 12 (48) | |
| Post-chemo pain | p = 0.32 | ||
| Yes | 4 (27) | 12 (48) | |
| No | 11 (73) | 13 (52) | |
| Post-chemo ossification | p = 0.02* | ||
| Increase | 15 (100) | 16(64) | |
| No Increase | 0 (0) | 9 (36) |
*Proportions between the two groups are compared using Chi-square test or Fischer exact, p value of ≤ 0.05 is significant.
Table 6.
Predictive Factors for Tumor Necrosis in Osteosarcoma at 50% Cutoff
| Predictive Factors | > 50% Necrosis (%) (n=30) |
≤ 50% Necrosis (%) (n=10) |
p value |
|---|---|---|---|
| Tumor change in size | p = 0.03* | ||
| Larger | 6 (20) | 3 (30) | |
| Smaller | 13 (43) | 0 (0) | |
| No change | 11 (37) | 7 (70) | |
| Post-chemo tumor size | p = 0.72 | ||
| > 10 cm | 10 (33) | 4 (40) | |
| ≤ 10 cm | 20 (67) | 6 (60) | |
| Post-chemo pain | p < 0.01* | ||
| Yes | 8 (27) | 8 (80) | |
| No | 22 (73) | 2 (20) | |
| Post-chemo ossification | p = 0.03* | ||
| Increase | 26 (87) | 5 (50) | |
| No Increase | 4 (13) | 5 (50) |
*Proportions between the two groups are compared using Chi-square test or Fischer exact, p value of ≤ 0.05 is significant.
After chemotherapy, 16 patients had residual pain while 24 patients reported being pain free. While there was no strong association among the good responders, we did find that eight (80%) of the poor responders had residual post-chemotherapy pain, compared to only two (20%) who did not have any pain (p < 0.01).
After evaluating the data, a high degree of agreement on evaluation of ossification existed between each observer, with a kappa score of .73, indicating 90% agreement. In context, a kappa score of 1 implies perfect agreement, while a kappa score of .61-.80 corresponds to substantial agreement (Table 7). There was a significant association with an overall increase in tumor ossification and a good response to treatment. All 15 good responders had an increase in post-chemotherapy tumor ossification compared to pre-chemotherapy (p = 0.02).
Table 7.
Inter-observer Reliability of Evaluation of Tumor Ossification
| Observer | Total (n =40) | ≥ 90% Necrosis (n=15) | ≤ 50% Necrosis (n=10) |
|---|---|---|---|
| A | |||
| Increase in ossification | 29 | 14 | 4 |
| No increase | 11 | 1 | 6 |
| B | |||
| Increase in ossification | 31 | 15 | 5 |
| No increase | 9 | 0 | 5 |
Inter-observer reliability measured by Cohen’s Kappa, with k=0.73 indicating 90% agreement.
We found a significant association with an overall no change in tumor ossification and a poor response to treatment, where 55% of patients with no change (5 of 9) in ossification, compared to 16% of patients who had an increase (5 of 31) were poor responders to treatment (p = 0.03).
Discussion
Patients diagnosed with osteosarcoma will typically be managed with a course of neoadjuvant chemotherapy, followed by tumor resection. Many studies have shown that the final histologic response to neoadjuvant chemotherapy is an important predictive factor for overall outcome in osteosarcoma. Unfortunately, this knowledge is not known until days after the surgery has been completed. Based on the patient’s progression of pain, tumor growth, or other factors, a physician may have some idea of how well a patient is responding to chemotherapy during their treatment. However, to date, there has been no definitive clinical tool utilizing the association of clinical findings with response to chemotherapy. Therefore, the main purpose of our study was to create a scoring criterion that will accurately predict how a patient has responded to neoadjuvant chemotherapy based upon predictive clinical and radiographic factors prior to the tumor resection.
Studies by Miwa et al. evaluated various radiologic factors, including sclerotic change on plain radiographs and percent reduction of the maximal diameter of the mass on MRI. They used these imaging modalities and findings to create a combined radiological scoring system and found a high correlation with histologic response to neoadjuvant chemotherapy.12,13 Our study included similar radiologic factors of tumor ossification and tumor change in size, along with the final tumor size itself in our scoring criteria to evaluate the likelihood of having a good response to treatment. We found a significant association between these individual factors and a good response to treatment. Several other studies have also shown that increased ossification of the tumor could be a sign of chemotherapeutic response.17-20 While these factors individually demonstrated high sensitivity in predicting a good response, we found that specificity was low. Only 54% (14 of 26) of patients with a final tumor size of ≤ 10 cm actually had a good response to treatment. In addition, 48% (15 of 31) of patients with an increase in post-chemo ossification had a good response to treatment. This suggests that these factors, while sensitive, are not adequate by themselves to predict a good response to neoadjuvant chemotherapy.
A similar study by Amit et al. looked at the association of pain with histologic response to chemotherapy and found that resolved pain with a great VAS score reduction was predictive of a good histologic response with a positive predictive value of 66.67%. They also reported that persistence of pain with minimal reduction predicted a poor response, with a negative predictive value of 87.5%.21 Our cohort did not show the same association between pain and a good response to treatment. However, our data did show that presence of pain does seem to predict the likelihood of being classified as a poor responder to treatment, where 50% (8 of 16) of patients with persistent pain had a poor response. Since the clinical evaluation of pain can be subjective, our scoring criteria simplifies this issue by classifying pain as either absent or present after treatment.
Our approach to this study was similar to Mirels in intent, where he evaluated the predictive value of known risk factors of an impending pathologic fracture in a long bone affected by metastases and then compared the accuracy with a proposed scoring system comprised of those risk factors. In his study, the weighted scoring system demonstrated the highest accuracy, where a score of 9 or greater highly indicated prophylactic fixation of the lesion.22 Our study was able to demonstrate that our proposed scoring criteria was more accurate in predicting a good histologic response to treatment compared to any single identified clinical or radiographic factor. The results of our ROC curve were promising, as it demonstrated that our proposed criterion was able to achieve a high sensitivity and specificity of predicting a good response to neoadjuvant chemotherapy at the cutoff score of 6. At the end of a patient’s course of neoadjuvant chemotherapy, the attending physician can assess the patient’s response to treatment by utilizing this criterion. With a patient having a score of 6 or less, the attending physician can conclude that there is a high likelihood of the patient responding well to treatment and can carry on with their typical surgical management. On the other hand, a score of greater than 6 suggests that the patient has not had an optimal response to treatment and may warrant heightened awareness of the intended surgical margins in the preoperative plan to ensure they will be widely free of tumor. To take a hypothetical example, in osteosarcoma involving the proximal fibula, the surgeon must decide whether to preserve the peroneal nerve. In cases with a composite score ≤ 6, the decision may be justified as the tumor is likely diffusely necrotic and a periosteal margin will be adequate and negative. However, if there are clinical or radiographic features that are concerning and result in a score of 7 or greater, the surgeon may consider sacrificing the nerve to ensure margins clear of viable tumor and decrease the risk of local recurrence. The utility of this score has implications regarding pre-operative counseling of the patient, surgical margins, borderline limb salvageable presentations, and functional preservation. In our study, all the poor responders of ≤ 50% necrosis had a score of 7 or more. However, the specificity of this scoring criteria to predict the poor responders was very low at 63%, indicating that this criterion may not be the most accurate method to identify poor responders.
There are limitations to this study that merit mention. First, our numbers are small, and this is preliminary data from which no clinical decisions should be drawn at this stage. Validation studies of different populations of many more patients will be required to confirm or improve upon these suggested values and predictive variables. Next, we lacked objective criteria supported by literature that could be used to measure the degree of tumor ossification. However, our study did demonstrate significant inter-observer reliability in independently reviewing ossification with our determined subjective criteria. This indicates that our methodology in grading the change in ossification is reliable and accurate. Still, with the lack of objective criteria, other observers may have difficulty in reproducing identical results and this will be important to note on validation studies. Further, our study did not investigate overall survival and notable events such as local recurrence. These outcomes are very important in the management of osteosarcoma, and further investigation must still be done to demonstrate the utility and importance of a predictive criteria. Necrosis of the specimen is well known as an important factor that predicts survival, so we believe we are justified to use this as our surrogate measure for later oncologic events.
In conclusion, our study supports the accuracy of our proposed criteria and has the potential to be an effective tool in assessing and managing cases of osteosarcoma, in particular by predicting good responders to neoadjuvant chemotherapy. In our evaluation of identified factors, we found that a decrease in tumor size, the tumor size itself, and an increase in tumor ossification is significantly associated with a good response to multi-agent neoadjuvant chemotherapy. While we also did not find an association between pain and a good response to treatment, we still found some relation regarding a poor response to chemotherapy. The inclusion of all four of these factors in a combined scoring criteria shows a far improved accuracy in identifying good responders to neoadjuvant chemotherapy compared to the use of one single factor, or any other combination of the four factors. However, since our study is small, we propose further investigation of our criteria by assessing and scoring a larger sample size to ensure similar levels of sensitivity and specificity.
References
- 1.Li X., et al. “A comparative study between limb-salvage and amputation for treating osteosarcoma.”. J Bone Oncol. 2016;5(1):15–21. doi: 10.1016/j.jbo.2016.01.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Misaghi A., et al. “Osteosarcoma: A comprehensive review.”. SICOT-J. 2018;4:12–12. doi: 10.1051/sicotj/2017028. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Bacci G., et al. “Predictive factors for local recurrence in osteosarcoma: 540 patients with extremity tumors followed for minimum 2.5 years after neoadjuvant chemotherapy.”. Acta Orthop Scand. 1998;69(3):230–236. doi: 10.3109/17453679809000921. [DOI] [PubMed] [Google Scholar]
- 4.Bacci G., et al. “Prognostic factors for osteosarcoma of the extremity treated with neoadjuvant chemotherapy: 15-year experience in 789 patients treated at a single institution.”. Cancer. 2006;106(5):1154–1161. doi: 10.1002/cncr.21724. [DOI] [PubMed] [Google Scholar]
- 5.Bacci G., et al. “Grade of chemotherapy-induced necrosis as a predictor of local and systemic control in 881 patients with non-metastatic osteosarcoma of the extremities treated with neoadjuvant chemotherapy in a single institution.”. Eur J Cancer. 2005;41(14):2079–2085. doi: 10.1016/j.ejca.2005.03.036. [DOI] [PubMed] [Google Scholar]
- 6.Davis A. M., et al. “Prognostic factors in osteosarcoma: A critical review.”. J Clin Oncol. 1994;12(2):423–431. doi: 10.1200/JCO.1994.12.2.423. [DOI] [PubMed] [Google Scholar]
- 7.Pakos E. E., et al. “Prognostic factors and outcomes for osteosarcoma: An international collaboration.”. Eur J Cancer. 2009;45(13):2367–2375. doi: 10.1016/j.ejca.2009.03.005. [DOI] [PubMed] [Google Scholar]
- 8.Raymond A. K., et al. “Osteosarcoma chemotherapy effect: A prognostic factor.”. Semin Diagn Pathol. 1987;4(3):212–236. [PubMed] [Google Scholar]
- 9.Bacci G., et al. “Neoadjuvant chemotherapy for high-grade central osteosarcoma of the extremity. Histologic response to preoperative chemotherapy correlates with histologic subtype of the tumor.”. Cancer. 2003;97(12):3068–3075. doi: 10.1002/cncr.11456. [DOI] [PubMed] [Google Scholar]
- 10.Li X., et al. “The relation of tumour necrosis and survival in patients with osteosarcoma.”. Int Orthop. 2011;35(12):1847–1853. doi: 10.1007/s00264-011-1209-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Cheon G. J., et al. “Prediction model of chemotherapy response in osteosarcoma by 18F-FDG PET and MRI.”. J Nucl Med. 2009;50(9):1435–1440. doi: 10.2967/jnumed.109.063602. [DOI] [PubMed] [Google Scholar]
- 12.Miwa S., et al. “Prognostic value of radiological response to chemotherapy in patients with osteosarcoma.”. PloS One. 2013;8(7):e70015–e70015. doi: 10.1371/journal.pone.0070015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Miwa S., et al. “A novel combined radiological method for evaluation of the response to chemotherapy for primary bone sarcoma.”. J Surg Oncol. 2012;106(3):273–279. doi: 10.1002/jso.23074. [DOI] [PubMed] [Google Scholar]
- 14.Taki J., et al. “Prediction of final tumor response to preoperative chemotherapy by Tc-99m MIBI imaging at the middle of chemotherapy in malignant bone and soft tissue tumors: Comparison with Tl-201 imaging.”. J Orthop Res. 2008;26(3):411–418. doi: 10.1002/jor.20522. [DOI] [PubMed] [Google Scholar]
- 15.Ueda T., et al. “Validation of radiographic response evaluation criteria of preoperative chemotherapy for bone and soft tissue sarcomas: Japanese Orthopaedic Association Committee on Musculoskeletal Tumors Cooperative Study.”. J Orthop Sci. 2008;13(4):304–312. doi: 10.1007/s00776-008-1235-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Huvos A. G., et al. “Primary osteogenic sarcoma: Pathologic aspects in 20 patients after treatment with chemotherapy en bloc resection, and prosthetic bone replacement.”. Arch Pathol Lab Med. 1977;101(1):14–18. [PubMed] [Google Scholar]
- 17.Fletcher BD. Response of osteosarcoma and Ewing sarcoma to chemotherapy: Imaging evaluation. AJR Am J Roentgenol. 1991 Oct;157(4):825–833. doi: 10.2214/ajr.157.4.1892044. [DOI] [PubMed] [Google Scholar]
- 18.Chuang V. P., et al. “Radiographic and angiographic changes in osteosarcoma after intraarterial chemotherapy.”. AJR Am J Roentgenol. 1982;139(6):1065–1069. doi: 10.2214/ajr.139.6.1065. [DOI] [PubMed] [Google Scholar]
- 19.den Heeten G.J., Thijn C.J.P., Kamps W.A., et al. The effect of chemotherapy on osteosarcoma of the extremities as apparent from conventional roentgenograms. Pediatr Radiol. 1986;16:407–411. doi: 10.1007/BF02386820. [DOI] [PubMed] [Google Scholar]
- 20.Smith J., et al. “Radiographic changes in primary osteogenic sarcoma following intensive chemotherapy. Radiological-pathological correlation in 63 patients.” Radiology. 1982;143(2):355–360. doi: 10.1148/radiology.143.2.6978499. [DOI] [PubMed] [Google Scholar]
- 21.Amit P., et al. “Role of Dynamic MRI and Clinical Assessment in Predicting Histologic Response to Neoadjuvant Chemotherapy in Bone Sarcomas.”. Am J Clin Oncol. 2014;37(4) doi: 10.1097/COC.0b013e31827b4f6f. [DOI] [PubMed] [Google Scholar]
- 22.Mirels H. Metastatic disease in long bones: A proposed scoring system for diagnosing impending pathologic fractures. 1989. Clin Orthop Relat Res. 2003 Oct;415 Suppl:S4–13. doi: 10.1097/01.blo.0000093045.56370.dd. [DOI] [PubMed] [Google Scholar]