The role of preoperative 18 F-uorodeoxyglucose positron emission tomography/computed tomography ( 18 F-FDG PET/CT) in retroperitoneal sarcoma

18 F-uorodeoxyglucose positron emission tomography/computed tomography ( 18 F-FDG PET/CT) scan was used to predict pathologic grade based on the maximum standardized uptake value (SUVmax) in soft tissue sarcoma and bone sarcoma. In retroperitoneal sarcoma (RPS), the effectiveness of PET scan was not well known. This study is designed to investigate the association of SUVmax with histopathologic grade and usefulness of 18 F-FDG PET/CT scan preoperatively. Patients undergoing primary surgery for retroperitoneal sarcoma with preoperative 18 F-FDG PET/CT imaging were investigated between January 2001 and February 2020 at Samsung Medical Center. The relationship between SUVmax and histologic features was assessed. The association of SUVmax with overall survival (OS), local recurrence (LR), and distant metastasis (DM) was studied. Of the total 129 patients, the most common histologic subtypes were liposarcoma (LPS, 68.2%) and leiomyosarcoma (LMS,15.5%). The median value of SUVmax was 4.5 (range, 1- 29). The value of SUVmax was correlated with higher tumor grade (p < 0.001, Spearman coecient 0.627) and mitosis (p < 0.001, Spearman coecient 0.564) and showed a higher value in LMS (12.04±6.73) than in dedifferentiated liposarcoma (DDLPS, 6.32±4.97, p = 0.0054). The optimal threshold to distinguish high tumor grade was 4.8. High SUVmax group based on the above threshold showed poor prognosis in OS, LR, and DM (p < 0.001). SUVmax was correlated with pathologic parameters (tumor grade, mitosis) in RPS and was higher in the LMS group than DDLPS group. In addition, prognosis (OS, LR, DM) was poor at high SUVmax values (p < 0.001). The value of SUVmax 4.8 is the optimal threshold to rule out high-grade tumors and predict prognosis.


Introduction
Retroperitoneal sarcomas (RPS) are rare neoplasm of mesenchymal origin derived from connective tissue. The incidence of RPS is 1% of all human malignancies 1 . The most common histologic types are liposarcoma and leiomyosarcoma, which account for 70% of all retroperitoneal sarcomas 2,3 .
Surgical resection including that of adjacent organs is the most important treatment for RPS [4][5][6] .
Research on preoperative radiotherapy and adjuvant chemotherapy is ongoing, but the effectiveness of these has not been determined. Therefore, most patients require only surgery with multi-organ resection. However, in cases of patients with locally advanced and high-risk primary sarcomas, neoadjuvant chemotherapy can improve the likelihood of negative resection margins, which are associated with reduced local recurrence 7,8 .
Preoperative diagnosis is becoming increasingly important as preoperative radiotherapy and adjuvant chemotherapy research continues. Percutaneous biopsy is a very good preoperative diagnosis method and can safely classify histologic subtype 9,10 . In con rming the metastasis of retroperitoneal sarcoma, computed tomography (CT) scans are su cient to screen for presence. Since su cient preoperative information can be obtained through CT and percutaneous biopsy 3 , the role of 18 F-uorodeoxyglucose positron emission tomography/computed tomography ( 18 F-FDG PET/CT) has not been de ned.

Page 3/15
There have been several studies on the use of 18 F-FDG PET/CT in sarcomas, but most of those studies included both bone sarcoma and soft tissue sarcoma 11,12 . Alternatively, whole soft tissue sarcomas not speci c to RPS were targeted 13,14 . Previously, our research team conducted a study on the association between maximum standardized uptake value (SUVmax) and retroperitoneal liposarcoma (LPS) 15 .
However, there was a limitation that only LPS was included.
In this study, we aimed to investigate prognostic signi cance of SUVmax in RPS and preoperative usefulness of 18 F-FDG PET/CT imaging.

Results
Clinicopathologic data. In total, 136 patients who underwent primary surgery for RPS from 2001 to 2016 and underwent preoperative 18 F-FDG PET/CT to determine the presence of metastasis were identi ed.
Three patients with Ewing's sarcoma were excluded. Four patients were excluded due to insu cient pathological data such as mitosis and necrosis. After excluding these patients, data from a total of 129 patients were investigated. The histologic subtypes were dominantly LPS (68.2%) and LMS (15.5%). DDLPS accounted for 68% of the LPS patients, followed by well-differentiated liposarcoma (WDLPS) and pleomorphic liposarcoma (PLS). The distribution of tumor grades was similar for the three grades.
Demographic and clinicopathological details are shown in Table 1.  The univariate analysis of prognostic factors for LR was performed on all RPS patients. The signi cantly associated factors were SUVmax (p < 0.001), high tumor grade (p < 0.001), mitosis (≥20/10 HPF, p = 0.024), and necrosis (≥ 50%, p < 0.001). On multivariate analysis, the only factors independently associated with LR were high tumor grade (p = 0.013) and necrosis (≥ 50%, p = 0.028). However, in the analysis conducted by histologic subtypes, SUVmax (p < 0.001) and high tumor grade (p = 0.002) were the main factors for LR in LPS groups (Table 3). Optimal threshold to distinguish high grade sarcoma. The ROC curve analysis demonstrated that the AUC for high tumor grade (Grade III) was maximal when the threshold SUVmax was 4.8. The AUC for high tumor grade at the cut-off SUVmax was 0.820 (p < 0.001). At this threshold, the values of sensitivity and speci city were 0.77 and 0.80, respectively (Fig. 2).
Prediction of outcome with optimal SUVmax threshold. The SUVmax threshold was divided into a high SUVmax group and a low SUVmax group based on a cut off of 4.8; and survival analysis was performed for OS, LR, and DM. In analysis of the entire RPS group, the high SUVmax group showed poor prognosis in OS, LR, and DM (p < 0.001). When analyzed by histologic subtype, the liposarcoma group's high SUVmax subgroup showed poor prognosis in OS (p < 0.001) and LR (p = 0.004). However, there was no difference in the LMS group (Fig. 3).

Discussion
This study analyzed the relationship between SUVmax and the pathologic details of RPS and the usefulness of SUVmax for prediction of prognosis. We showed that SUVmax is associated with the highgrade portion of RPS. In addition, we demonstrated that the range of SUVmax varies according to histologic subtype.
Distinction between DDLPS and LMS. Our key nding was that LMS (12.04±6.73) showed the higher range of SUVmax value than DDLPS (6.32±4.97). DDLPS and LMS have potential for neoadjuvant chemotherapy; micro-metastasis potential is lowered and unresectable tumor can be reduced in size before surgery 7 . In adjuvant chemotherapy, anthracycline-based chemotherapy is the cornerstone of rstline treatment in localized soft tissue sarcoma 16 . However, based on many retrospective studies, different histology-driven-chemotherapy can be applied to DDLPS and LMS. In addition, multi-center prospective study (STRASS-2) is ongoing to determine whether these treatments affect the prognosis 17 .
The distinction between high grade LPS and LMS is becoming increasingly important through these studies. Our nding suggests that 18 F-FDG PET/CT can be useful in distinguishing these two histology subtypes preoperatively.
Detecting the high-grade portion through 18 F-FDG PET/CT imaging. Due to its multifocal nature and large size, RPS can be di cult to target accurately during biopsy at the time of detection. In addition, preoperative biopsies tend to underestimate the nal grade probably due to sampling error 18 . For example, in LPS, when solid portion and fatty portion exist together, the high-grade portion is likely to be the solid portion. However, when there are several solid portions, prediction of only the high-grade portion is di cult with CT. Because of these di culties, the TARPSWG guideline suggests that 18 F-FDG PET/CT be available for de ning biopsy target areas 3 . The current study demonstrated that SUVmax of tumors was correlated with higher tumor grade (p < 0.001, Spearman coe cient 0.627) and mitosis (p < 0.001 Spearman coe cient 0.564). This result is similar to other studies showing the association between pathologic details and SUVmax 18, 19 . These results support the suggestion of using the TARPSWG guideline to set SUVmax as the biopsy target areas.
Prognosis prediction with SUVmax. A previous study conducted by our research team demonstrated an SUVmax cutoff > 4.5 to be associated with a higher grade and worse prognosis in RPS. In this study, only LPS was used, and there was a limitation in that SUVmax was performed on heterogenous populations including metastatic and recurrent tumors 15 . Subramaniam et al. also reported that when the SUVmax value was higher than 5.0, the prognosis was poor, and the high SUVmax value and grade were related. This study investigated the homogenous population ; only the DDLPS and LMS groups were studied.
However, the small number of patients was mentioned as a limitation 18 . In both studies referenced above, OS and relapse free survival (RFS) were mentioned in the analysis of SUVmax and prognosis.
The current study investigated a relatively large number of patients given the low prevalence of RPS, excluding metastatic and recurrent tumors. In addition, the study showed a correlation between SUVmax and DM, which was not shown in other studies. The cut-off value of SUVmax (4.8) was a good measure for predicting prognosis but showed relatively low sensitivity (0.77) for predicting tumor grade and was not particularly useful in the LMS group. Therefore, 18 F-FDG PET/CT is considered to be useful as a measure of prognosis of LPS patients or to rule out high-grade tumors by utilizing relatively high speci city (0.8).
Limitations. The current study is limited by its retrospective nature and small number of LMS patients. A large-volume study is needed to nd the SUVmax value that can differentiate between DDLPS and LMS and to study the role of 18 F-FDG PET/CT in recurrent and metastatic tumors.

Conclusion
Tumor SUVmax was correlated with pathologic parameters (tumor grade, mitosis) in RPS and was higher in LMS than DDLPS. In addition, the prognosis (OS, LR, DM) was poor at high SUVmax values (p < 0.001).
The value of SUVmax 4.8 is the optimal threshold to rule out high grade tumors, and the prognosis can be predicted through this value. Statistical analysis. Factors affecting the prognosis of RPS were analyzed through univariate and multivariate Cox regression models. The Cox proportional hazard model was used to evaluate prognostic variables, and an estimated hazard ratio (HR) with 95% con dence interval (95% CI) was presented. P < 0.05 was considered statistically signi cant.

Methods
The receiver-operating characteristic (ROC) methodology was used to calculate the ideal threshold to distinguish high-grade sarcoma. The area under the curve (AUC) was calculated for each parameter using the non-parametric method to represent the overall predictive or prognostic performance.
For survival analysis, Kaplan-Meier estimates and the log-rank test were used. OS, LR, and DM were analyzed using time-to-event regression. OS was calculated from the date of surgery to the date of death. LR was de ned a tumor was found on CT scan, and the duration was calculated based on the CT scan date. When late identi cation of recurrence occurred because of progression at follow-up, not initial, testing, the date was calculated as the date of the rst discovery. DM was de ned as a tumor found in organs such as liver, lung, brain, and bone, and the date was calculated as when the tumor was detected by clinical symptoms or imaging tests. All analyses were performed using R 4.0.4 software (The R Core Team, Vienna, Austria). Comparison of median SUVmax with histologic subtypes.