Treatment-Related Prognostic Factors in Managing Osteosarcoma around the Knee with Limb Salvage Surgery: A Lesson from a Long-Term Follow-Up Study

Purpose The aim of this study was to assess the treatment-related factors associated with local recurrence and overall survival of patients with osteosarcoma treated with limb-salvage surgery. Patients and Methods Treatment-related factors were analyzed to evaluate their effects on local recurrence-free survival (LRFS) and overall survival (OS) in 182 patients from 2004 to 2013. Results The mean length of follow-up was 73.4 ± 34.7 months (median, 68 months; range, 12-173 months), and 63 patients died by the end of the follow-up. The 5-year and 10-year overall survival rates were 68.6 ± 6.6% and 59.4 ± 10.6%, respectively. Univariate analysis showed that treatment-related prognostic factors for overall survival were prolonged symptom intervals >=60 days, biopsy/tumor resection performed by different centers, previous medical history, incomplete preoperative chemotherapy (<8 weeks), and prolonged postoperative interval >21 days. In the multivariate analysis, biopsy/tumor resection performed by different centers, incomplete implementation of planned new adjuvant chemotherapy, and delayed resumption of postoperative chemotherapy (>21 days) were risk factors for poor prognosis; biopsy/tumor resection performed by different centers and tumor necrosis <90% were independent predictors of local recurrence. Conclusion For localized osteosarcoma treated with limb-salvage surgery, it is necessary to optimize timely standard chemotherapy and to resume postoperative chemotherapy to improve survival rates. Biopsies should be performed at experienced institutions in cases of developing local recurrence.


Introduction
Osteosarcoma is a very rare malignant bone tumor with a high predilection for the area surrounding the knee joint [1]. Since the 1980s, neoadjuvant chemotherapy with limbsalvage surgery (LSS) has become the current standard for the management of high-grade osteosarcomas. Amputation offers better local control, but it confers no clear survival benefit over LSS [2][3][4], whether in patients with osteosarcoma who respond poorly to chemotherapy or not. The goal of LSS is to preserve a functioning limb without increasing the risk of tumor recurrence. However, there is still concern about the adverse impact of limb-salvage surgery on local recurrence and overall survival due to closer resection margins [5][6][7][8].
For those patients with high-grade osteosarcoma around the knee without initial metastases, the surgical margin and the presence of tumor necrosis are generally accepted prognostic factors for local recurrence and overall survival [1,[9][10][11][12][13][14][15][16][17]. The prognostic relevance of other factors, such as tumor site and size, histologic subtype, neurovascular infiltration, elevated alkaline phosphatase (ALP), elevated lactate dehydrogenase (LDH), and age, is still controversial [18][19][20][21][22]. Other treatment-related factors, such as symptom intervals (SI) [23,24], incomplete preoperative chemotherapy [25,26], and delay of resumption of postoperation chemotherapy [27,28], are still not clear. Because of these contradictory reports, the relevance of treatment-related prognostic factors in patients with osteosarcoma needs further elucidation, especially for those patients who prefer limb sparing. Indeed, the identification of patients with a favorable prognosis and of those with a poor prognosis may be used to adjust treatment and surveillance schedules [29,30]. This is particularly important for patients who are at risk for local recurrence and inferior survival who have been treated with limb sparing. A reasonable accurate estimate of overall survival and development of local recurrence before treatment decisionmaking would be helpful to avoid upsetting patients and their parents or confusing them with impractical hope. Therefore, we want to assess the influence of several treatment-related prognostic factors in a large series of patients treated with limb-salvage surgery from East China in long-term followup, excluding those patients with primary metastases, which may lead to paradoxical conclusions. We wish to pay more attention to the timing of planned new adjuvant chemotherapy and resumption adjuvant chemotherapy and to try and identify the time intervals, including symptom interval, preoperative chemotherapy duration, and preoperative interval and postoperative interval, which can serve as prognostic factors.

Patients and Methods
We conducted a retrospective, single-center review of patients with osteosarcoma around the knee after limbsalvage surgery between January 2004 and December 2013. During that time, 254 patients were treated for osteosarcoma around the knee. The study was approved by the Institutional Review Board. The exclusion criteria included metastatic disease at initial presentation (28 patients), limb amputation as the primary procedure (17 patients), initial age >=60 years (14 patients), and incomplete follow-up records (13 patients). 182 patients were included in the final analysis.

Assessment of Patient, Tumor, and Treatment-Related
Variables. Age, sex, tumor site, symptom interval, pathological fracture status, previous biopsy history, previous medical history, serum alkaline phosphatase level, preoperative chemotherapy duration, preoperative interval, bone margin length, histologic response (tumor necrosis), tumor size, neurovascular infiltration, postoperative interval, presence of local recurrence, metastatic disease, and overall survival were recorded for each patient (  Symptom interval (SI) was defined as the time from the first onset of symptoms or signs to a definitive diagnosis and initiation of treatment [24].
Previous medical history was defined as the patients having accepted inadequate treatment due to an erroneous diagnosis, including erroneous drug and surgical treatment, before they were referred to professional musculoskeletal oncologists. Patients diagnosed with secondary osteosarcoma were also assessed and included in this group.
At diagnosis, the serum ALP levels were measured in international units (IU), and the activity of ALP was estimated by the p-nitrophenyl phosphate method. ALP ranges of 60.0-300.0 IU/L for patients ≤14 years and 38.0-115.5 IU/L for patients >15 years were considered normal.
Responses to neoadjuvant chemotherapy (tumor necrosis) were graded on the basis of the amount of tumor necrosis in the resected specimen. More than 90% tumor necrosis was regarded as a good response; a cut-off of 90% tumor necrosis is usually used to distinguish good and poor responders [31].
Standard chemotherapy followed the protocols of IOR/OS-4 [32]. The standard length of preoperative chemotherapy duration was 10 weeks of 2 cycles of multidrug therapy. The preoperative interval was defined as the time from the end of preoperative chemotherapy to the primary surgery.
Neurovascular infiltration was defined as tumor growth outside of the normal compartment and infiltration of soft tissue, such as the nerve and vasculature, which can be inspected by enhanced MRI preoperatively and confirmed during surgery and by pathological examination. Tumors measuring at least 10 cm in length of the involved bone or more were defined as large and all others as small.
The postoperative interval was defined as the time from primary surgery to the resumption of postoperative chemotherapy.

Diagnosis, Treatment
, and Surveillance. Diagnosis of osteosarcoma, established by clinical and radiological findings, was confirmed on core needle biopsy or open biopsy instead of fine biopsy. Thereafter, 2 cycles of neoadjuvant chemotherapy were given to patients. The chemotherapy protocol was IOR/OS-4 [32]. Histologic response to preoperative chemotherapy was evaluated and graded according to previously reported criteria [31].
Surgical excision of the primary tumor involved en bloc resection of the overlying biopsy tract with the tumor. Resection of the tumor and reconstruction of the bone defect were performed according to the operating instructions provided by Enneking et al. [33]. Proximal and distal bony margins were defined by the respective treatment protocol according to Enneking's classification. Detailed bone and soft tissue margin widths were guided by Kawaguchi et al. [34,35]. The surgical margin was determined by 3 techniques: macroscopic inspection of the specimen, careful dissection, and histologic examination. Tumor size, neurovascular infiltration, and tumor necrosis were also recorded via pathological examination by macroscopic inspection.
Postoperative chemotherapy was given once the wound healed adequately. The chemotherapy protocol was modified if the tumor response was poor according to IOR/OS-4. Postoperative follow-up was performed every 2-3 months in the first two years, every 4 months until 5 years, and every 6 months in the following years. Imaging modalities included bone scans, chest computed tomography (CT) scans, abdominal ultrasounds, and positron emission tomography (PET) scans.
Seventy patients showed good response to neoadjuvant chemotherapy with tumor necrosis >=90%. The average preoperative chemotherapy duration was 10.8 ± 2.1 weeks, and patients were grouped according to their length of preoperative chemotherapy: 110 patients (60.4%) with a duration of 8-12 weeks, 38 patients (20.9%) with a duration greater than 12 weeks, and 34 patients (18.7%) with a duration less than 8 weeks due to noncompliance or severe drug toxicity. Seventyeight patients (42.9%) underwent primary surgery after 14 days from the end of preoperative chemotherapy, while 77 patients (42.3%) had a delay of greater than twenty-one days to the resumption of chemotherapy (Table 1).

Statistics.
Kaplan and Meier estimates were used to assess overall survival (OS) and local recurrence-free overall survival (LRFS) [36]. For survival analysis, the primary end points were time to death, so overall survival was measured from the date of diagnosis to death or the endpoint of followup. LRFS was defined as the time from the date of primary surgery to the day when local recurrence was confirmed by biopsy or surgery. Univariate and multivariate Cox proportional hazards regression analyses were performed to determine which parameters were significant. A 95% confidence interval level was used; p < 0.05 was considered as significant to identify factors predictive of local recurrence and overall survival. The multivariate logistic regression model included all the potential predictors. Predictors were excluded (one by one) if the p value of the log likelihood ratio test was greater than 0.10. Predictor exclusion was continued until all the remaining predictors had p values less than 0.10, which was then defined as the final prediction model. SPSS Version 20.0 (SPSS Inc., Chicago, IL) was used for all statistical analyses.

Prognostic Factors for Overall
Survival. The 5-year and 10year overall survival rates were 68.6 ± 6.6% and 59.4 ± 10.6%, respectively ( Figure 1). The mean length of follow-up was 73.4 ± 34.7 months (median, 68 months; range, 12-173 months), and 63 patients had died by the end of follow-up.
Univariate analysis showed that the prognostic factors for overall survival were young age at diagnosis, prolonged symptom intervals >=60 days, biopsy/tumor resection performed by different centers, previous medical history, elevated ALP at diagnosis, neurovascular infiltration, tumor size >=10 cm, tumor necrosis <90%, incomplete preoperative chemotherapy (<8 weeks), and prolonged postoperative interval >21 days. In the multivariate analysis, the risk factors were young age at diagnosis, biopsy/tumor resection performed by different centers, elevated ALP at diagnosis, tumor size >=10 cm, tumor necrosis <90%, prolonged postoperative interval >21 days, and shorter preoperative chemotherapy (<8 weeks) ( Table 2).
Univariate analysis showed that prognostic factors for LR-free survival were biopsy/tumor resection performed by different centers, neurovascular infiltration, tumor size >=10 cm, and tumor necrosis <90%. Patients with a previous medical history or a larger tumor size were inclined to develop local recurrence, although it was not significant. However, in the multivariate analysis, only biopsy/tumor resection performed by different centers and tumor necrosis <90% were independent predictors of local recurrence. There was no significant difference in the probability regarding sex, age, bone margin width, elevated ALP level at diagnosis, preoperative chemotherapy duration, preoperative interval, and postoperative interval (Table 3).
For patients who underwent biopsy by different centers, their 5-year LR-free survival was 58.1±19.6% and was much worse than that of those whose biopsy was performed in the same center (p<0.001) ( Figure 5). The 5-year LR-free survival for patients with necrosis <90% was 80.2±7.6%, significantly worse than those with necrosis >=90% (p=0.008) ( Figure 5).

Discussion
In this study, we presented a long-term assessment of overall survival and local recurrence in relation to prognostic factors among patients with nonmetastatic high-grade osteosarcoma around the knee. We found an overall cumulative 5-and 10-year survival rate of 68.6 ± 6.6% and 59.4 ± 10.6%, respectively, with a mean follow-up of 73.4 months. The overall survival in our study is on par with most data [1,9,20,22]. The incidence of local recurrence (LR) in this study was 13.7% (25/182), and the 5-year LR-free survival was 85.3 ± 5.3%. Our local recurrence rate of 13.7% is slightly higher than BioMed Research International 5   that presented in some reported data [6,12,37,38], but these data are equal to data from a larger series report [2,8,[39][40][41].
Although there is a tendency towards worse overall survival for patients with symptom delays of longer than 60 days, we are unable to find a positive correlation (p=0.193, HR=1.443 (0.830-2.509)) in the multivariate analysis between a prolonged SI and overall survival; that is, latency of diagnosis and new adjuvant chemotherapy is not associated with poor prognosis. However, when adjusted for lower tumor necrosis (<90%), patients with an SI longer than 60 days had adverse 5-year overall survival of 47.4±12.7%, which was much worse than that (71.7±12.2%) in patients with a shorter SI (p=0.014). This indicates that we should adjust adjuvant therapy and intense surveillance for those patients with a longer SI and a poor chemotherapy response. Similar results that no positive correlation exists between the SI and the outcome of osteosarcoma patients have been shown in other reports [1,23,24]. Some studies even show that the SI is shorter in patients with metastatic disease than in patients with localized disease, and patients with a longer duration     of symptoms appear to have a better prognosis [47,48]. Interestingly, Bielack et al. [1] found that longer symptom history and treatment delay do significantly correlate with poor chemotherapy response. Jin et al. [49] concluded that adolescents with longer patient delays may incur inferior prognoses. In this cohort, patients with primary metastases that showed aggressive tumor biology were excluded, but we still found a slight tendency towards a poor prognosis for patients with symptom delays (p=0.066, 1.622 (0.969-2.715)) in the univariate analysis. We recommend that we should pay more attention to patients with long symptom intervals. We recommend early diagnosis and treatment to shorten the SI in cases of poor survival because prolonged SI leads to poor chemotherapy response. For patients with prolonged SI at diagnosis, standard chemotherapy should be vigorously executed to obtain good tumor necrosis, avoiding decreasing chemotherapy intensity.
Incomplete preoperative chemotherapy and prolonged preoperative chemotherapy may incur poorer prognosis compared to standard preoperative chemotherapy in our study. Shorter or incomplete preoperative chemotherapy correlates with poor chemotherapy response [50,51] due to the heterogeneity of patient compliance with chemotherapy or inferior socioeconomic status [44,45,52]. Patients with shorter or incomplete preoperative chemotherapy were 2.809 (95% CI, 1.169-6.750) times at risk for poor prognosis in our study. Toxicity, patient choice, and financial inadequacy also resulted in prolonged preoperative chemotherapy duration in our study. Both incomplete preoperative chemotherapy and prolonged preoperative chemotherapy would reduce the overall dose intensity according to recent investigators [53,54]. The results of dose-intensity analyses performed by other investigators [25,26,55,56] support the hypothesis that the actual dose intensity delivered determines the outcome of treatment of osteosarcoma. On the other hand, some researchers [57,58] suggest that as preoperative therapy becomes more prolonged, the correlation of tumor necrosis with DFS (disease-free survival) decreases. Standard chemotherapy, if possible, should be rigorous to achieve a superior prognosis.
Similar to the results of Imran et al. [27], the mean surgery to postoperative chemotherapy interval was 23 ± 10.1 days (7-60 days) in our study, and the median length was 21 days. In the study by Imran et al. [27], overall survival was poorer for patients who had a delay of greater than 21 days before the resumption of chemotherapy compared with that of patients who had a shorter delay. Patients with a good response to the preoperative chemotherapy fared worse when the postoperative chemotherapy was delayed for more than sixteen days after the definitive surgery, whereas the impact of delayed resumption of chemotherapy on the patients with a poor response was not significant. The reason may be that a lengthy delay before the resumption of chemotherapy after definitive surgery could compromise the overall dose intensity. Patients with a poor histological response to preoperative chemotherapy have worse survival when they have a delay in resumption of chemotherapy of more than twenty-four days [57]. Our study confirmed these findings in that patients with a postoperative interval >21 days are inclined to incur poor outcomes (HR=2.844 (1.068-7.572), p=0.036). In this study, it seems that patients with a postoperative interval longer than 14 days face a negative ending, and the later they resume chemotherapy, the poorer their outcomes are. Whether intensive chemotherapy would impact the chemotherapy response and the survival thereafter is still controversial [53][54][55][56]; thus further study on delayed resumption of postoperative chemotherapy and its impact is still needed.
The local recurrence rate in this study is 13.7%, which is slightly higher than the current international standard (4-10%). Neurovascular infiltration [6,8], large tumor size, and inferior tumor necrosis [38,59,60] are commonly regarded as risk factors for local recurrence. Under the promotion of limb-salvage surgery, these factors have been widely and intensely debated [2,12,15,19,61,62]. However, in our study, we attribute the higher incidence of local recurrence to inadequate/close margins and poor tumor necrosis. Reasons for inadequate/close margins include unprofessional biopsy and erroneous/unplanned surgery due to an erroneous diagnosis. In this study, 31 patients (17%) received a biopsy in other medical centers and then asked for limb-salvage surgery in our hospital; 17 patients (9%) received prior surgery, such as intralesional/marginal curettage. These unplanned surgeries may lead to problems in managing wide soft tissue margins due to unplanned approaches. That is why all of these recurrent tumors were located in the soft tissue around the wound or near the neurovascular bundle, although frozensection histology examination by intraoperative biopsies certified that there were no residual tumor cells. A previous medical history due to malignant transformation or erroneous diagnosis and treatment has recently been accepted by some researchers as a risk factor [6,[63][64][65]. Biopsy/tumor resection performed by different centers in this study is a significant predictor of the development of local recurrence and poorer overall survival. This is similar to the reports by Picci et al. [66], Andreou et al. [6], and Poudel et al. [39]. The biopsy approach must be planned so that it can be completely and widely excised at the moment of the resection. The procedure should be performed by an experienced surgical team who will also be handling the definitive surgery. Poor necrosis due to nonstandard chemotherapy (incomplete chemotherapy) is widely accepted as a risk factor for local recurrence. In this study, 72 patients (39.6%) did not achieve complete or standard chemotherapy due to noncompliance with chemotherapy and low socioeconomic status. Thus, only 70 patients (38.5%) had good tumor necrosis (>=90%). This is one of the reasons for the high incidence of local recurrence in this study.
Due to the lack of uniformity in patient analyses and methods and statistics that were calculated on study populations whose minimum follow-up was often less than 3 years, a number of clinical and pathologic features show bias in different studies and present with contradictory prognostic significance. However, our present analysis evaluated a large number of patients according to the previously mentioned prognostic variables, followed up for at least 5 years. In our study, data about the variables evaluated were available for almost all patients. The main shortcoming is that data were not derived from a randomized study, but we collected information from each patient prospectively. The tumors of the patients included in this study were somehow heterogeneous in terms of biological behavior and stage, but we rigorously set and executed inclusion and exclusion criteria. We believe that our data provide a meaningful contribution to the previous literature.

Conclusion
For localized osteosarcoma treated with limb-salvage surgery, it is necessary to optimize timely standard chemotherapy and to resume postoperative chemotherapy to improve survival rates. Biopsies should be performed at experienced institutions in cases of developing local recurrence.

Data Availability
The data used to support the findings of this study and the datasets used and/or analyzed during the current study are available from the corresponding author upon request (Chunlin Zhang, shzhangchunlin123@163.com).

Ethical Approval
The study received ethical approval from the Institutional Review Board of Shanghai Tenth People's Hospital affiliated to Tongji University.

Consent
Written informed consent was obtained from all patients before surgical treatment including the fact that the data may be included in future publications.

Disclosure
Jianping Hu, Chunlin Zhang, and Kunpeng Zhu are co-first authors. The IRB and ethical committee of the Shanghai Tenth Peoples' Hospital would review the requests because of the patients' information. After the approval of the committee with confirmation of the reasonable requests, the dataset would be freely available.

Conflicts of Interest
The authors declare that they have no conflicts of interest.