Abstract
Angiosarcoma (AS) is a kind of highly aggressive cancer with high occurrence and mortality rates. This study aimed to establish a comprehensive and validated prognostic nomogram with various clinical indicators in non-metastatic AS patients after surgery. Data of non-metastatic AS patients diagnosed after surgery between 2010 and 2015 was retrieved from the surveillance epidemiology and end results database. Univariate and multivariate Cox proportional hazards regression analysis were performed to identify the independent prognostic factors associated with survival to construct the predictive nomogram of 3- and 5-year overall survival (OS) and cancer-specific survival (CSS) rates. Concordance-index (C-index), calibration plots and receiver operating characteristic (ROC) curves were applied to evaluate the predictive ability of the nomograms. 251 patients in total were divided into the training group (N = 177) and the validation group (N = 74). After the multivariate Cox regression analysis, gender, AJCC stage group 7th ed, T, N stage 7th ed, histologic grade and primary site were statistically identified as independent factors with OS and CSS (P < 0.05). We incorporated the significant factors above and age into nomograms. The C-index of the nomograms for OS and CCS in the training cohort was 0.757 (95%CI 0.697–0.817) and 0.762 (95%CI 0.702–0.822), meanwhile, the C-index of those in the validation cohort was 0.749 (95%CI 0.668–0.830) and 0.756 (95%CI 0.676–0.836) respectively. The results of calibration plots and ROC curve showed the nomograms qualified to measure the risk and prognosis. Our study has developed novel and practical nomograms for predicting prognosis in patients with non-metastatic AS after surgery contributing to cancer management.
Similar content being viewed by others
Introduction
Soft tissue sarcoma is a highly aggressive human malignancy, among which the rare Angiosarcoma (AS) accounted for less than 1%1,2. AS, originating from blood or lymphatic vessels, can occur in any position of the body with high proportions of recurrence and distant metastasis after surgery. It has been reported that the median overall survival (OS) varied from 30 to 50 months for AS patients, resulting in a poor prognosis3. Previous study confirmed the absence of surgical excision was correlated with the poor prognosis of AS patients thus surgery is the preferred therapy for AS patients with no metastasis4, however, they would face the relapse and distant metastasis. It has been demonstrated that AS patients with elder age5, prior radiation therapy6, tumor size > 5 cm, primary site from Liver7 related to a worse prognosis. However, these factors can only serve as a single predictive indicator, which led to limited effect in comprehensively assessing the prognosis of AS. Therefore, a more integrated system combining multi-factors is needed.
Nomogram is widely applied as a predictive tool that incorporates multiple hazardous variables to forecast the disease prognosis in a visualized manner8. However, limited data on this issue in non-metastatic AS patients after surgery has been published. Under the circumstances of various clinical characteristics in each patient with AS, an applicable, accurate and fully-varied prognostic nomogram is imperatively warranted. Hence, by utilizing the data from surveillance epidemiology and end results (SEER) database, we present this study to predict the independent factors in non-metastatic AS after surgery and to construct a prognostic model for better guiding the clinical treatment and follow-up schedules for patients.
Materials and methods
Patients enrolled from SEER
The basic information of post-operative AS patients with no metastasis including clinical characteristics and survival data was extracted from the database which is supported by the National Cancer Institute named SEER using the SEER Stat software (version 8.3.6; National Cancer Institute, USA). In line with the International Classification of Disease for Oncology Third Edition (ICD-O-3: 9120/3), the eligible AS patients were screen out. The diagnosis year of patients was during 2010 and 2015. The inclusion criteria of this study were: (1) AS was the unique primary cancer (2) diagnostic confirmation via positive histology; (3) active follow-up; (4) with hospital inpatient/outpatient or clinic reporting source; (5) available data on gender, age, race, survival time, American Joint Committee on Cancer (AJCC) TNM stage, histologic grade, primary cancer site and other details; (6) surgery had been performed with no metastasis. Patients not appropriate for the inclusion criteria were excluded in this study.
All of the patients enrolled were randomly assigned to the training cohort and validation cohort in the ratio of 7:3. All the data in SEER can be downloaded freely with a publicly available ethics approval. All methods were carried out in accordance with relevant guidelines and regulations.
Clinical variables
Data of demographic and clinical variables such as age, race, gender, primary site, histologic grade, TNM stage, survival data, cause of death, and survival status was acquired from the SEER program. Overall survival (OS) was the primary interest in this study calculated from the diagnosis to the final follow-up or all-cause death. Cancer-specific survival (CSS), the second endpoint defined as the period from the diagnosis to the death due to cancer progression. By using X-tile bioinformatics software (Yale University, USA, Version 3.6.1), the optimal cut-off value of age was 50- and 83-years.
Statistical analysis
The baseline of categorical variables was compared by the Chi-square test and the Kaplan–Meier method was utilized for survival analysis. Univariate and multivariable Cox regression analysis were presented to identify factors significantly related to survival. Nomograms to predict the 3- and 5-OS and CSS rates were constructed by integrating those independent risk factors from multivariable Cox results. The nomograms were validated both internally (based on the training group) and externally (based on the validation group). The C-index (Harrell’s concordance index) was applied to estimate the predictive ability of nomograms. In general, if a C‐index value is larger than 0.7, the predictive model is regarded as a good prediction9. In addition, a receiver operating characteristic (ROC) curve was built to calculate the corresponding areas under the curve (AUC) value. It was universally acknowledged that the larger the AUC value was, the more significance the parameter accounted for. The calibration plot was undertaken to compare the nomogram‐predictive survival probabilities and the actual one with a 45° reference line. All of data analysis was conducted by R Software (Version 4.0.1, Vienna, Austria) with relevant packages such as rma, survival, stdca, cmprsk. Statistical significance in Univariate analysis was set with P < 0.2 while P < 0.05 was limited for multivariate analysis.
Results
Input data from SEER
Based on the inclusion and exclusion criteria, 251 of the 5071 patients registered in the SEER database between 2010 and 2015 were enrolled. The patients’ exclusion criteria were as follows (Fig. 1): (a) AS was not the primary cancer; (b) without positive histology; (c) unspecific gender, age, race, histology grade, survival months, AJCC TNM stage, tumor primary site and other details; (d) no active follow-up; (e)without surgery. Finally, 251 patients were brought into this study randomly divided into the training group (N = 177) and the validation group (N = 74). The demographic and clinicopathologic information of the patients in training group was displayed in Table 1. The number of patients aged < 50, 50–82, ≥ 83 years were 33 (13.1%), 181 (72.2%), 37 (14.7%). The majority population were White people (219, 87.3%) and the rest were Black people (17, 6.8%) with only a minority of other races (15, 5.9%). The number of people with primary site soft tissue, heart mediastinum, peritoneum and bone were 219 (87.3%), 17 (6.7%), 5 (2%) and 10 (4%), respectively. The numbers of people with histologic grades I (well-differentiated), II (Moderately differentiated), III (Poorly differentiated), and IV (Undifferentiated) were 25 (10%), 34 (13.5%), 98 (38.7%), and 94 (37.8). The ratio of people with clinical stages IA, IB, IIA, IIB, and III were 30 (12%), 23 (9.2%) and 79 (31.5%), 15 (6%), 104 (41.3%), respectively. (Table 1).
Construction of nomogram
According to the COX analysis and threshold value, the variables, including gender, histologic grade, AJCC stage group 7th ed, T and M stage 7th ed and primary site were statistically identified with OS in univariate COX regression as well as in the multivariate COX analysis (Table 2). Age was found no significant in univariate COX but was significant in multivariate COX analysis. Nonetheless, age was an acknowledged influencing factor in diverse kinds of diseases and lots of studies have arrived at the agreement that age was significantly associated with survival in AS patients so we included age in the nomogram5,10,11,12. Aiming to predict the 3- and 5-year OS rates, the nomograms were created with these independent factors (Fig. 2). Additionally, the independent variables mentioned above also contributed significantly to the CSS (Table 3), which were utilized to draw the nomogram for CSS prediction (Fig. 3).
Nomogram validation
Internal and external methods were used to validate the performance of our nomograms. The predictive accuracy of the nomograms was assessed by exhibiting C-index. The C-indexes were 0.757 (95%CI 0.697–0.817) and 0.762 (95%CI 0.702–0.822) for OS and CSS subjecting to the internal validation. In the external validation, the C-index for the OS nomogram was 0.749 (95%CI 0.668–0.830), while for the CSS nomogram 0.756 (95%CI 0.676–0.836). The calibration curves for the likelihood of 3 and 5-year OS were demonstrated no obviously deviations from the reference line showing a relatively fair agreement between the model prediction and the observed situations (Fig. 4). The validation results showed that the predicted values of both nomograms were in good agreement. In the ROC curves, the AUC values were also applied to evaluate the predictive performance in training (Fig. 5).
Discussion
The TNM classification system has been universally used to the clinical prediction of prognosis for patients with Soft tissue sarcomas. Nonetheless, it focused mainly on tumor size, lymph nodes and distant metastasis. Some particular clinical factors closely correlated with the prognosis of AS. Besides, Soft tissue sarcoma included a diverse of histological subtypes among them each is heterogeneous from one another. For these reasons, a more highly-specific and sensitive predictive system for AS is necessary.
As a medical system, the nomogram can both predict the independent risk factors associated with diseases among individuals and assist clinicians to discern those patients with high-risk and then adopt optimal therapeutic regimens for them. X-tile software was used to optimize the cut-off value of the variable, age. It selected 50 and 83 as the cut-off point for age which can help better figure out its potential effects on survival rate. The predicting system ability can be greatly enhanced by adding this factor.
In this study, 251 eligible AS patients from the SEER database were all diagnosed in 2010–2015. People with White race, aged 50–82, diagnosed at stage III, and primary site from soft tissues made up the majority of the investigated population. Based on the univariate and multivariate regression, gender, age, AJCC stage group 7th ed, T and N stage 7th ed, histologic grade and primary site were recognized as the independent predictors of prognosis which were utilized to establish nomograms. The C-index of the model in the internal and external validation were all greater than 0.7 for OS (0.757, 0.749) and CSS (0.762, 0.756). The following ROC curves showed it was qualified to predict the 3-year and 5-year survival rates.
Our study has found a significant relationship between gender and the prognosis of non-metastatic AS patients. Biing Luen Lee reported male sex was the possible predictor of poor prognosis in post-operative AS which was consistent with us. However, some research suggested gender appeared no obvious correlation with OS probably due to the limited sample size4,13. As displayed in the nomogram, age was also observed as an independent factor. Retrospective studies by Therese Dettenborn and Clothilde Lindet have reached an agreement on this issue that the elder age (> 70) influenced the AS prognosis adversely5,10. The nomograms uncovered the stage T and N were also the risk variables. Richard J. Cassidy et.al illustrated that tumor size ≥ 5 cm functioned as a negative role in Scalp angiosarcomas. Of note, they also found age ≥ 65 related to worse survival11. Inna Schott found the negative lymph node status corresponded to the better OS14. As for the histologic grade, a recent study concluded AS with grade III histology tended to experience a shorter survival which was consistent with our result15. The median survival time of AS was various in different primary sites16,17,18, but which site with the highest mortality has not been well documented. Our study elaborated AS from heart mediastinum harbored the highest risk compared with soft tissue, bone and peritoneum. These findings indicated AS patients with risk factors stated above had better take relevant examinations more frequently.
The raw data was acquired from the SEER database containing abundant demographic characteristics, tumor properties, and large population of survival data from different races and countries which is of good representative population and can better reflect population experience than those studies only from a single center19. The results of the nomograms arrived at favorable consistency between predictive survival probabilities and the actual conditions, indicating that this model enabled us to distinguish hazard indicators and predict prognosis precisely.
This nomogram provides important clinical use in AS which are listed as follows. Firstly, Clinicians can more accurately and faster to estimate the survival chances and recognize the personalized risk of AS patients so as to adopt corresponding interventions. For post-operative patients, Individual treatment strategies and the follow-up plans can be optimized based on the different prognosis of patients. For example, a 45-year-old female patient was diagnosed with AS originated from soft tissues. Her tumor size was found T1b with N1 and M0 metastasis. The postoperative pathology was confirmed as grade III with stage IIA. This patient got 175 scores totally in OS nomogram and the corresponding 3- and 5-year predicted OS were 26% and 16%. Following the same method, the predicted CSS can be easily obtained. In this way those at high-risk had better follow up more frequently and detailly. The heterogeneity among patients should be considered when clinicians formulate clinical decision-making, thus, early preventive interventions may be executed to prolong the survival for those high-risk population. Secondly, the soft tissue sarcomas, a variety of huge admixture, contains more than 50 subtypes which are diverse from each other. Our study has specifically targeted AS as a separate category to help its better management. Moreover, some patients with AS cannot be clearly staged for some inevitable reasons which are often staged as Tx. It is usually more difficult for clinicians to predict their prognosis and often required multidisciplinary teamwork. Our study has focused on this special vague cohort to guide better predict prognosis and treatment. In addition, our nomograms have incorporated AJCC stage T, N, and showed great consistency with the training group, supporting this credible model into clinical practice. In short, this nomogram can be utilized as a useful stratification tool for further clinical AS researches.
Still, there were some important clinical variables not available in the SEER database such as chronic lymphedema history, smoking or alcohol drinking habits. Research has reported people having smoking habit or suffering from chronic lymphedema were more susceptible to AS12,20. Additionally, specific genes mutation acted oncogenic roles in AS21,22 while the SEER database did not contain genetic testing information so we are unable to evaluate these effects. Moreover, the data in this study was derived from the USA population, whether it is applicable to Asians or Europeans needs further evidenced-based study.
Conclusion
Taken together, our study constructed a novel predictive model to identify the risk in post-operative AS patients with no metastasis. The nomograms were capable to predict 3- and 5-year OS and CSS. These findings can help clinicians in decision-making on personalized treatment strategy.
References
Young, R. J. et al. Angiosarcoma. Lancet Oncol. 11(10), 983–991 (2010).
Gaballah, A. H. et al. Angiosarcoma: Clinical and imaging features from head to toe. Br. J. Radiol. 90(1075), 20170039 (2017).
Florou, V. & Wilky, B. A. Current and future directions for angiosarcoma therapy. Curr. Treat. Options Oncol. 19(3), 14 (2018).
Buehler, D. et al. Angiosarcoma outcomes and prognostic factors: A 25-year single institution experience. Am. J. Clin. Oncol. 37(5), 473–479 (2014).
Dettenborn, T. et al. Prognostic features in angiosarcoma of the head and neck: A retrospective monocenter study. J. Craniomaxillofac. Surg. 42(8), 1623–1628 (2014).
Merfeld, E. et al. Clinical outcomes and prognostic features of angiosarcoma: Significance of prior radiation therapy. Clin. Oncol. (R. Coll. Radiol.) 31(4), 232–241 (2019).
Fayette, J. et al. Angiosarcomas, a heterogeneous group of sarcomas with specific behavior depending on primary site: A retrospective study of 161 cases. Ann. Oncol. 18(12), 2030–2036 (2007).
Balachandran, V. P. et al. Nomograms in oncology: More than meets the eye. Lancet Oncol. 16(4), e173–e180 (2015).
Royston, P. & Altman, D. G. External validation of a Cox prognostic model: Principles and methods. BMC Med. Res. Methodol. 13, 33 (2013).
Lindet, C. et al. Localised angiosarcomas: The identification of prognostic factors and analysis of treatment impact. A retrospective analysis from the French Sarcoma Group (GSF/GETO). Eur. J. Cancer 49(2), 369–376 (2013).
Cassidy, R. J. et al. The importance of surgery in scalp angiosarcomas. Surg. Oncol. 27(4), A3-a8 (2018).
Lee, B. L. et al. Investigation of prognostic features in primary cutaneous and soft tissue angiosarcoma after surgical resection: A retrospective study. Ann. Plast. Surg. 78(3 Suppl 2), S41-s46 (2017).
Li, R. et al. Clinical characteristics and prognostic factors of primary splenic angiosarcoma: A retrospective clinical analysis from China. Cell. Physiol. Biochem. 49(5), 1959–1969 (2018).
Schott, I. et al. Localized angiosarcoma, not one disease: A retrospective single-center study on prognosis depending on the primary site and etiology. Sarcoma 2021, 9960085 (2021).
Sinnamon, A. J. et al. A prognostic model for resectable soft tissue and cutaneous angiosarcoma. J. Surg. Oncol. 114(5), 557–563 (2016).
Jiang, S. et al. Surgery and chemotherapy improve the prognosis of primary hepatic angiosarcoma: A retrospective study based on propensity score matched survival analysis. Eur. J. Surg. Oncol. 47(3 Pt B), 690–698 (2021).
Zhang, C. et al. Clinical characteristics associated with primary cardiac angiosarcoma outcomes: A surveillance, epidemiology and end result analysis. Eur. J. Med. Res. 24(1), 29 (2019).
Wang, B., Chen, L. J. & Wang, X. Y. A clinical model of bone angiosarcoma patients: A population-based analysis of epidemiology, prognosis, and treatment. Orthop. Surg. 12(6), 1652–1662 (2020).
Doll, K. M., Rademaker, A. & Sosa, J. A. Practical guide to surgical data sets: Surveillance, epidemiology, and end results (SEER) database. JAMA Surg. 153(6), 588–589 (2018).
Dyroff, S., Layfield, L. J. & Crim, J. Angiosarcoma arising in massive localized lymphedema. Skelet. Radiol. 49(5), 815–818 (2020).
Huang, S. C. et al. Recurrent CIC gene abnormalities in angiosarcomas: A molecular study of 120 cases with concurrent investigation of PLCG1, KDR, MYC, and FLT4 gene alterations. Am. J. Surg. Pathol. 40(5), 645–655 (2016).
Painter, C. A. et al. The angiosarcoma project: Enabling genomic and clinical discoveries in a rare cancer through patient-partnered research. Nat. Med. 26(2), 181–187 (2020).
Acknowledgements
We would like to appreciate SEER database for downloading the free and public data.
Funding
This study was supported by Zhejiang Provincial Project for the key discipline of traditional Chinese Medicine (Yong GUO, no.2017-XK-A09, http://www.zjwjw.gov.cn/).
Author information
Authors and Affiliations
Contributions
B.B.W., T.J. and Z.X.Y. wrote the main manuscript text and T.Y.S. and Y.Y.L. were in charge of the architecture of figures and table, and revision of this article. All authors contributed to the article and approved the submitted version.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Jiang, T., Ye, Z., Shao, T. et al. Prognostic nomograms for predicting overall survival and cancer-specific survival in patients with angiosarcoma, a SEER population-based study. Sci Rep 12, 3479 (2022). https://doi.org/10.1038/s41598-022-07444-5
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-022-07444-5
- Springer Nature Limited
This article is cited by
-
Competing-risks model for predicting the prognosis of patients with angiosarcoma based on the SEER database of 3905 cases
Holistic Integrative Oncology (2024)