The Nomogram Application to Predict the Overall and Disease-specic Survival in Synovial Sarcoma: A Population Study

Background: Synovial sarcoma is an uncommon soft sarcoma that lacks prognostic prediction models. The nomograms were employed to predict patients’ survival of synovial sarcoma. Methods: Materials collected from 1941 synovial sarcoma cases in the SEER database were analyzed. We employed univariate and multivariate cox analyses to identify the independent prognostic variables. Based on these outcomes, the nomograms were built for predicting 1-, 3-, and 5- year overall survival rate and disease-specic survival rate and then validated in external dataset. C-indices, calibration plots and ROC curves were applied to assess nomogram accuracy. Results: Patients were randomly classied into the training (n=1361) and testing (n=580) cohorts. Age, race, sex, primary anatomic site, chemotherapy, subtypes, surgery, SEER historic stage and tumor size were identied as independent prognostic variables (P<0.05) and then the nomogram of overall survival was constructed. Similarly, the nomogram of disease-specic survival was constructed. C-indices of training cohort for predicting overall and disease-specic survival were 0.779 and 0.779, respectively. Corresponding values of testing cohort were 0.769 and 0.765, respectively. The AUC values of prognostic model in the training cohort at 1-, 3-, 5- years of overall survival were 0.845, 0.785, and 0.788, respectively. Corresponding of disease-specic survival were 0.835, 0.783, and 0.787, respectively. The calibration plots illustrated excellent consistency between the survival rate of predicted and actual survival. Conclusion: we constructed the reliable nomograms for synovial sarcoma patients to predict overall and disease-specic survival, which can offer precise and personalized survival prediction. cell), anatomic site, tumor size (>9.5 cm), nonsurgical treatment,


Introduction
Synovial sarcoma is a rare subtype of soft sarcomas, which include roughly 5% to 10% of all soft sarcomas (1)(2)(3). It commonly affects adolescents and adults under 30 years of age with an incidence 800-1000 case per year in the USA (4). Lower limbs are the primary sites of the majority of synovial sarcoma, particularly the thigh area. Synovial sarcoma commonly appears to small round cell sarcoma in morphology, among which over 90% have the special chromosome translocation -t (X; 18) (p11. 2; q11. 2), producing SS18-SSX mixture gene (5,6). Currently, wide resection, chemotherapy or excision combine with chemotherapy and/or radiotherapy have been abroad employed to cure synovial sarcoma (7). The 5-and 10-year survival rate of most of synovial sarcoma are 60% and 50%, respectively (8). Although several prognosis factors for synovial sarcoma including age, tumor size, tumor localization, tumor stage, tumor depth and surgery have been explored (9)(10)(11). Most of previous studies only served these variables as a solitary quota to assess prognosis, which restricted the in uence for precisely predicting individualized survival of synovial sarcoma patients. Given this limitation, we were the rst to construct the innovative prognostic models. Nomogram is an effective and convenient prognostic tool to assess the patients' prognosis of tumor (12). In this study, the nomograms of overall and disease-speci c survival were built, which can assess the survival probability of personal more precisely than a single quota via combining all prognostic factors. Owing to its powerful robustness and more precisely predictive capacity, the nomogram can enhance the precise prediction of personal prognosis (13). The Surveillance, Epidemiology, and End Results (SEER) dataset between 1975 and 2016 collected clinical materials of synovial sarcoma patients from 18 national registries represents about 30% of overall population in the USA that permitted detailed analyses of survival of synovial sarcoma (14). The aim of this study was the rst to build powerful prognostic nomograms for predicting the 1-, 3-, and 5-year overall survival and disease-speci c survival rates for synovial sarcoma patients.

Materials And Variables
We performed a retrospective review of synovial sarcoma patients utilizing the Surveillance, Epidemiology, and End Results (SEER) database between January 1975 and December 2016 published in November 2018. The SEER*Stat software (version 8.3.5) was applied to obtain patient information.
For present study, we employed the International Classi cation of Disease for Oncology morphology (ICD-O-3) codes to extract patients with synovial sarcoma from the database (9040/3, 9041/3, 9042/3, and 9043/3 for not otherwise speci ed (NOS), spindle cell, epithelioid cell, and biphasic). Clinicopathological features including age, sex, race (white, black, and others/unknown), subtypes of synovial sarcoma (spindle cell, epithelioid cell, biphasic, and NOS), anatomical site (upper limb, lower limb, pelvis, head and neck and other), grade (I, II, III, IV, and unknown), marital status (single/domestic partner, married, DWS (divorced, widowed and separated), and unknown), SEER historic stage (localized, regional, and distant), and tumor size. The cutoff value of age and tumor size were evaluated through X-tile software, which was formerly proved to identify optimal cut-points of variables (15). The best cutoffs of age were 19 and 59 years, so patients were divided into three groups (<=18, 19 59, or >=60 years) (Fig.1). The optimal tumor size cutoff was 9.5 cm, so patients also were classi ed as two clusters (<=9.5 cm, or >9.5 cm) (Fig.1). Besides, the information of radiation therapy, chemotherapy, and surgery was also obtained. Radiation therapy, and chemotherapy, and surgery were categorized as yes or no/unknown. Surgery was categorized as yes or no.

Statistical analysis
The variables of categorical are presented as frequencies and ratios and compared with the chi-squared tests. Xtile software was employed to evaluate the optimal cutoff values for age and tumor size on the foundation of overall survival. Univariate and multivariate analysis were employed to determine the variables correlated with overall survival and disease-speci c survival via using univariate and multivariate cox proportional hazards regression analyses. The nomograms were built for predicting the 1-, 3-, and 5-year overall and disease-speci c survival rates. The risk score of each patient was calculated according the following formula: risk score =∑ coe cient (prognostic variables) ×prognostic variables .
Validation and evaluation of nomogram: The training and testing cohorts were used to validate the nomograms. Receiver operating characteristic (ROC) curves were employed to assess the performance of nomograms using the R package "survivalROC". The calibration plots were applied to evaluate the consistency between the predicted survival rate and actual survival. Besides, concordance-index (C-index) was performed to assess the performance of nomograms, which was a practical assessment value close to counting the AUC value of ROC (16). The construction and veri cation of nomograms were applied utilizing the R package "rms". The univariate and multivariate cox proportional hazards regression analyses and chi-squared tests were employed with R software (version 4.1.0). P<0.05 of two-sided tests was regarded as statistically signi cant.

Results
Basic characteristics of patients 1941 synovial sarcoma patients between 1975 and 2016 were obtained from the SEER database. To construct and validate nomograms, we randomly divided 0.7 and 0.3 of the patients into the training cohort (n=1361) and testing cohort (n=580). Of these patients, most of them were 19~59 years old (71.12% and 72.76% in the training and testing cohorts, respectively), white (79.79% and 82.24%) and male (53.12% and 52.76%). The majority of primary anatomic location of these synovial sarcoma patients were lower limb (55.99% and 50.17%). With respect to SEER historic stage, localized tumors (62.97% and 63.10%) took the majority, followed by regional tumor (23.07% and 23.62%), distant tumor (10.51% and 9.48%) and unstaged tumor (3.45% and 3.79%). With regard to subtypes of synovial sarcoma, NOS (not otherwise speci ed) (42.47% and 41.38%) was most frequent followed by spindle cell (35.42% and 34.48%), biphasic cell (21.75% and 23.79%) and epithelioid cell (0.37% and 0.34%). Regarding marital status, single/domestic partner (46.29% and 43.79%) and married (42.98% and 45.34%) took the majority. In two cohorts, the majority of patients had undergone radiation therapy, chemotherapy and surgery. In regard to tumor size, the majority of tumors were less than 9.5cm (90.67% and 86.72%). All of synovial sarcoma patients' clinicopathological features are shown in Table 1.

Univariate and multivariate cox regression analyses results for overall survival and disease-speci c survival
In the training cohort, univariate and multivariate analyses were developed to determine independent prognostic variables for overall survival and disease-speci c survival. As is listed in Table 2, race, age, sex, primary anatomic site, subtypes, radiation therapy, chemotherapy, surgery, marital status, SEER historic stage and tumor size were notably correlated with overall survival in univariate analyses. Thus, these eleven factors were further subject to multivariate cox analysis. The results of multivariate cox analysis showed that nine variables (age, race, sex, primary anatomic site, subtypes, chemotherapy, surgery, SEER historic stage and tumor size) were independent prognostic variables for overall survival. Similarly, as is listed in Table 3, age, race, sex, primary anatomic site, radiation therapy, grade, subtypes, chemotherapy, surgery, marital status, SEER historic stage and tumor size were notably correlated with disease-speci c survival in univariate analysis. Thus, these twelve variables were further subject to multivariate cox analysis. The results of multivariate cox analysis illustrated that nine variables (sex, age, race, primary anatomic site, chemotherapy, subtypes, surgery, SEER historic stage and tumor size) were independent prognostic variables for disease-speci c survival.

Construction of the nomograms
The nine signi cant independent variables (age, race, sex, subtypes, primary anatomic site, chemotherapy, surgery, SEER historic stage and tumor size) from multivariate cox analysis were applied to construct the nomogram for predicting the 1-, 3-, and 5-year overall survival rates ( Fig.2A). Similarly, the nine signi cant independent variables (age, sex, race, primary anatomic site, subtypes, chemotherapy, surgery, SEER historic stage and tumor size) from multivariate cox analysis were applied to construct the nomogram for predicting the 1-, 3-, and 5-year disease-speci c survival rates (Fig.2B). The total points of all variables on the nomogram were summed and then exchanged into the probability rate of 1-, 3-, and 5-year overall survival or disease-speci c survival with guideline of the linear parallel lines. The nomogram of overall survival illustrated that race (being black), having the largest absolute values, was the strongest factor to the negative prognosis, followed by distant tumor (metastasis), age, subtypes (epithelioid cell), anatomic site, tumor size (>9.5 cm), nonsurgical treatment, chemotherapy, and being male. The nomogram of disease-speci c survival indicated that being black, having the largest absolute values, was the strongest factor to the negative prognosis, followed by distant tumor (metastasis), subtypes (epithelioid cell), age, anatomic site, tumor size (>9.5 cm), nonsurgical treatment, chemotherapy, and being male. The calibration plots illustrated wonderful agreement between the survival of predicted probability and actual survival (Fig.4).

Discussion
Synovial sarcoma is a malignant mesenchymal neoplasm with incomplete epithelial differentiation that comprise about 5 % ~ 10% of all soft tissue sarcomas(3). It is highly correlated with recurrence and metastases, which occur in about 50% to 70% of cases (17). The 5-and 10-year survival rate of synovial sarcoma are about 60% and 50%, respectively (8). Because of the low incidence of synovial sarcoma amid the population, it was a big challenge for doctors to research this soft sarcoma among large patient populations (18). In this study, utilizing a large synovial sarcoma patient cohort from the SEER dataset, we were rst to construct two novel, e cient and convenient nomograms for assessing individual overall survival and disease-speci c survival rates for patients with a rare cancer-synovial sarcoma. The nomograms illustrated successful accuracy and robustness when applied to both training and testing cohorts, which demonstrated good clinical capacity of the nomograms for this uncommon soft sarcoma.
In order to precisely lter the optimal prognostic factors, we employed univariate and multivariate cox proportional hazards regression analyses to determine independent prognostic variables. The results indicated that age, sex, race, primary anatomic site, chemotherapy, subtypes, surgery, SEER historic stage and tumor size are independent prognostic factors for overall survival and disease-speci c survival of synovial sarcoma patients. In past studies, growing age of patient was correlated with a statistically notable decline in the survival rate of synovial sarcoma patients (4,19,20). Pan et al found that synovial sarcoma patients over 50 years old had poorer survival results (10). Similarly, we con rmed growing age of patient is an independent risk variable for synovial sarcoma patients. X-tile software was applied to cluster the data of patient age on the foundation of survival status and time. The optimal cut-points of variables were identi ed and it was originally used in breast cancer. We identi ed that the best values of age of synovial sarcoma patients were 19 and 59 years. Similarly, tumor size is also an independent risk factor of synovial sarcoma patients. Some studies found that patients with larger tumor sizes had a worse survival and poorer prognosis (11,21,22). Similarly, we con rmed increasing tumor size of patient is a negative factor for synovial sarcoma patients. X-tile software was also employed to achieve the optimal cut-point of tumor size. We showed that the best tumor size cut-point of synovial sarcoma patients was 9.5 cm. Aytekin et al showed that male and being colored individual of synovial sarcoma patients had poorer prognosis (23). We also noticed that sex had a signi cant in uence on synovial sarcoma patient survival rate and race was correlated with overall survival. Besides, we also identi ed that surgery had a positive in uence on synovial sarcoma patient survival rate. similar results have been reported by previous studies (24,25).
In our study, we determined that SEER historic stage and anatomic site were the most important prognostic variables for synovial sarcoma patients. As for tumor stage at diagnosis, previous studies reported that synovial sarcoma patients with metastases own a remarkably poorer survival prognosis (10,26,27). It was reported that patients with lung metastases only and treated with curative metastasectomy have better relative survival (24).
Similarly, we found that synovial sarcoma patients with distant metastases had the highest risk of death.
Synovial sarcoma appear mostly in the lower extremities, while it is less common in the upper limbs, trunk and head and neck (28). Sultan et al and Xiong et al indicated that tumors located in the extremities of synovial sarcoma patients had better prognosis (20,29). Similarly, compared with lower limb and central lesions, tumors located in upper limb associated with increased overall survival (11). Our study also illustrated that tumor site was associated with the survival of synovial sarcoma patients. Besides, we identi ed chemotherapy as an independent risk prognostic negative variable for patients of synovial sarcoma. Despite wide application, the bene ts of chemotherapy for synovial sarcoma patients remain controversial. Some previous studies reported that chemotherapy was not associated with better survival (30,31). However, some studies reported that chemotherapy had a positive effect on survival rate and reducing distant metastasis (32,33). Patients who received no chemotherapy and patients who were unknown for having been received chemotherapy or not had the same code. Therefore, there was some certain deviation in the accuracy of chemotherapy data. More robust evidence about the bene ts of chemotherapy are needed from future studies. Besides, we found that different subtypes of synovial sarcoma have different overall survival and disease-speci c survival and subtype is an independent prognostic variable. And pervious study has proved that among different subtypes of synovial sarcoma, the subtype of biphasic owned the best survival, but the subtype of epithelioid cell favoured the worst survival (34).
Based on the abovementioned independent prognostic variables, we constructed prognostic nomograms that provide an e cient and convenient method to assess1-, 3-, and 5-year overall survival and disease-speci c survival for synovial sarcoma patients. The nomograms can enhance the accuracy of predicting individual survival results of synovial sarcoma patients at a speci c time point.
Although the nomograms in our study demonstrated favorable predictive capability, there are some limitations should be admitted. First, the data information about chemotherapy were limited in the SEER database, which may lead to some relevant bias. Second, the database contains a certain number of missing data, which might have reduced the amounts of eligible patients. Finally, the performance of our prognostic nomograms should be externally validated in another independent, large-scale database.

Conclusion
The present study identi ed age, sex, race, anatomic site, chemotherapy, subtypes, surgery, SEER historic stage and tumor size as independent prognostic factors for both the overall survival and disease-speci c survival of synovial sarcoma patients. These independent prognostic factors were combined to construct the nomograms for predicting prognosis of synovial sarcoma patients. Using our nomogram, the 1-, 3-and 5-year overall survival and disease-speci c survival rates for synovial sarcoma patients can be assessed, which enable surgeons to evaluate personalized survival rate more reliable and accurate and identify mortality risk.

Declarations
Ethics approval and consent to participate: Not applicable.

Consent for publication:
Not applicable.
Availability of data and materials: The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests:
The authors declare that they have no competing interests.