The AJCC staging system and post-operative nomograms use patient and tumor characteristics to provide prognostic estimates after resection of retroperitoneal sarcoma (RPS). While these variables help to predict survival at the time of diagnosis and resection, the applicability of these prognostic factors to survivors of RPS remains unknown. We hypothesized that the variables evaluated in the current staging system and post-operative nomograms would have limited ability to predict conditional survival in patients surgically treated for RPS.
A retrospective study was conducted using National Cancer Institute-sponsored tumor registries. We identified 1,199 patients who underwent surgical resection for non-metastatic RPS from 1988 to 2007. Conditional survival was defined as time-specific estimates conditioned on living a certain number of years post-diagnosis. Cox proportional hazards regression was used to assess the impact of various factors on sarcoma-specific survival (SSS) at baseline and up to 5 years after diagnosis.
Older age, male gender, histologic subtype, and high tumor grade predicted worse SSS at the time of diagnosis. After 1 year of survival, older age, male gender, and histologic subtype were no longer significant predictors of conditional survival. Only high grade tumors remained a significant predictor of worse prognosis after 5 years of survival (HR 1.95).
Retroperitoneal sarcomas are a rare group of tumors that frequently result in recurrence and death. Approximately 1000 new cases are diagnosed each year in the United States,1 and retroperitoneal sarcoma patients often present with large tumors that invade adjacent organs. Although the mainstay of treatment remains margin-negative surgical resection, the 5-year survival rate after surgical resection remains low, ranging from 36% to 63%.2-13 After treatment, patients are often provided an estimated prognosis that is largely determined by elements of the American Joint Committee on Cancer (AJCC) staging system for sarcoma.14 This staging system provides prognostic estimates for both extremity and retroperitoneal sarcoma based on a combination of tumor size, grade, depth, nodal involvement, and the presence of distant metastasis.
Several studies have called into question the applicability of a combined staging system for both extremity and retroperitoneal sarcoma, demonstrating its inability to adequately predict long-term survival.1, 10, 15, 16 These single-hospital experiences have been validated by a population-based study,16 in addition identifying tumor histology and involvement of surrounding structures as important predictors of survival; neither of these variables, however, is included in the current staging system. These limitations have significant clinical implications, as staging systems are typically designed to provide prognostic estimates to patients and guide clinicians to derive stage-specific oncologic treatment decisions. Therefore, sarcoma-specific nomograms have been developed to overcome these limitations in the staging systems.17-19 With growing interest in cancer survivorship, the relevance of the current prognostic systems to those who have survived years after retroperitoneal resections remains unknown. Conditional survival estimates may supply more meaningful data to those surgically treated for cancer, as they provide survival-specific prognostic information. Several studies have demonstrated that for various cancers, including extremity soft tissue sarcoma, melanoma, and ovarian, gastric, and rectal cancer, some predictors of mortality may diminish as survival time increases.20-28
We hypothesized that the variables evaluated in the current staging system and postoperative nomograms would have limited ability to predict conditional survival in patients surgically treated for abdominal and retroperitoneal sarcoma. Therefore, the aim of our study was to determine whether the patient and tumor factors contained within these prognostic tools would be able to predict future survival in patients who have already survived years after their initial diagnosis and surgical resection.
MATERIALS AND METHODS
A retrospective cohort study was conducted using the Surveillance, Epidemiology, and End Results (SEER) database to identify patients who were diagnosed with abdominal or retroperitoneal sarcoma and underwent surgical resection from 1988 to 2007. The SEER cancer registries provide population-based cancer surveillance for 17 regions that represent approximately 28% of the United States.29 SEER currently collects information on patient demographics, tumor characteristics, and treatment types including age, race, marital status, tumor grade, stage, histology, radiotherapy administration, and types of sarcoma surgery. However, resection margin status, receipt of adjuvant chemotherapy, and disease recurrence status are not currently collected by SEER, and therefore are not evaluated in this study.29
This study evaluated men and women older than 18 years who underwent surgical resection for primary nonmetastatic abdominal and retroperitoneal sarcoma. Only patients who received surgical resection were included. Surgical resection was defined as simple/partial surgical removal of the primary site, total surgical removal, or radical surgery. By using the International Classification of Disease for Oncology third edition, we further restricted our analysis to patients diagnosed with malignant fibrous histiocytoma (MFH), leiomyosarcoma (LMS), liposarcoma, or sarcoma of other histologies (excluding gastrointestinal stromal tumor, Ewing sarcoma, and Kaposi sarcoma). During the study period we evaluated, regional nodal disease was not routinely evaluated; however, the seventh edition of the AJCC staging system now designates patients with lymph node–positive disease as stage III, and thus we included patients with positive nodal disease. We in addition excluded patients with distant metastatic disease (n = 437) or a history of cancer (n = 2072), so that previous surgery or treatment with chemotherapy or radiation therapy (RT) would not confound our current findings. Patients from rural Georgia and Alaska were excluded from analysis because of small sample size (n = 13), as well as those diagnosed in the Louisiana registry during 2005 (n = 17) because of the negative impact Hurricane Katrina had on data collection.
Variables were collected and analyzed in conformance with the AJCC staging system.14 Tumors were further classified as low (grade 1 or 2) or high (grade 3 or 4) grade. At the time of presentation, many patients will have extension of their retroperitoneal sarcoma into surrounding organs, and therefore patients with direct extension of tumor into abdominal organs or structures were included in the analysis.
Conditional Survival Concept
Conditional survival was defined as time-specific estimates conditioned on surviving a certain number of years postdiagnosis and has been well described in the literature.21-23, 30 Patients are provided an initial prognostic estimate based on tumor and personal characteristics after surgical resection. Conditional survival estimates are recalculated with each additional year of survival by incorporating the tumor and patient characteristics as well as survival time (Fig. 1).
Chi-square tests were conducted to detect differences in patient demographics and tumor characteristics between patients at baseline, and conditional on surviving 1, 3, or 5 years after diagnosis and surgical resection. Kaplan-Meier methods were used to assess unadjusted conditional survival analysis. Cox proportional hazards regression was used to assess the effects of current AJCC measures as well as relevant patient-, tumor-, and treatment-related factors on sarcoma-specific survival while adjusting for covariates. The association between these factors was evaluated at baseline and conditional on surviving 1, 3, and 5 years after diagnosis and surgical resection. We performed extensive multivariate modeling, including interaction analyses.
All statistical analyses were completed using SAS software (version 9.2; SAS Institute, Cary, NC). This study was exempt from review by the Human Subjects Committee of the University of Minnesota Institutional Review Board because we used pre-existing data with all personal identifiers removed.
We identified 1199 patients who underwent surgical resection for nonmetastatic primary abdominal and retroperitoneal sarcoma from 1988 to 2007. At the time of diagnosis, the mean patient age was 59.6 years (range, 18-95 years). At baseline, the cohort was comprised of mostly females (57.3%) and non-Hispanic whites (71.9%). At the time of diagnosis, a large percentage of patients had tumors of the retroperitoneum (91.1%) that were >5 cm (84.9%), low grade (43.6%), and of liposarcoma histology (50.9%).
Bivariate analysis showed tumor grade as the only factor that varied between the cohort of patients at time of diagnosis (43.6% low grade) compared with those who survived 1 (47.9% low grade), 3 (54.4% low grade), or 5 years (58.6% low grade; Table 1).
Table 1. Cohort Demographic Factors Across Years of Survival
Baseline, n=1199, %
1 Year, n=964, %
3 Years, n=568, %
5 Years, n=382, %
Abbreviations: AJCC, American Joint Committee on Cancer; LMS, leiomyosarcoma; LPS, liposarcoma; MFH, malignant fibrous histiocytoma; n, sample size of each cohort.
Our unadjusted Kaplan-Meier analyses showed that sarcoma-specific median survival increased with each additional year of survival (Fig. 2). At time of diagnosis, the sarcoma-specific median survival was 83 months. Patients who had already survived 1 year had a sarcoma-specific median survival of 98 months. After 3 years of survival, sarcoma-specific median survival was 177 months. We could not calculate the sarcoma-specific median survival at 5 years because >50% of the cohort survived the duration of follow-up. These findings indicate that with each additional year of survival a patient experiences, his or her chances for continued survival increase.
At the time of diagnosis, older age, male sex, histologic subtype, and high tumor grade predicted worse sarcoma-specific survival. Patients older than 81+ years had the worst survival compared with the youngest patients (81+ years vs 18-35 years, hazard ratio [HR], 3.32; P < .0001). Men had a worse prognosis compared with women (HR, 1.47; P < .001). A diagnosis of MFH, LMS, or other sarcoma histology was associated with a higher risk of mortality at baseline compared with liposarcoma. Patients with high-grade tumors had worse sarcoma-specific survival than those with low-grade tumors (HR, 2.48; P < .0001). Race, nodal disease, receipt of adjuvant RT, and tumor size were nonsignificant predictors of survival.
In patients who survived 1 year after diagnosis, older age, male sex, high tumor grade, and the histologic subtype of LMS or other sarcoma histology remained significant predictors of sarcoma-specific survival, whereas MFH histology lost prognostic significance.
In patients who survived 3 years after diagnosis, however, older age, male sex, and histology no longer predicted future sarcoma-specific survival. High-grade tumors were the only predictor of worse survival (HR, 1.70; P = .004). At 5 years of survival, high tumor grade remained a significant predictor of sarcoma-specific survival (high grade vs low grade, HR, 1.95; P < .02; Table 2).
Table 2. Impact of AJCC Staging Factors and Others on Conditional Sarcoma-Specific Survival After Resections of Retroperitoneal Sarcoma
Baseline HR, n=1199, D=497
1 Year HR, n=954, D=366
3 Year HR, n=568, D=181
5 Year HR, n=382, D=87
Abbreviations: AJCC, American Joint Committee on Cancer; D, death; HR, hazard ratio; LMS, leiomyosarcoma; LPS, liposarcoma; MFH, malignant fibrous histiocytoma; n, sample size of each cohort; Ref, referent; RT, radiotherapy.
P < .05. We also adjusted for race and tumor registry location.
Because of the close relationship between tumor histology, size, and grade, we conducted several interaction analyses and found no significant interactions between tumor size and histology (P = .10) or tumor size and grade (P = .40). However, we identified a significant interaction between tumor histology and tumor grade (P = .046). High tumor grade (vs low) predicted worse outcomes in LMS and liposarcoma histologies. For liposarcoma tumors, the association between grade and survival was significant at baseline and at 1 and 3 years of survival. Grade was significant in LMS tumors at baseline and 1 year of survival, lost significance at 3 years, but regained significance after 5 years of survival. This fluctuation in significance is likely a reflection of sample size. There was no association detected in patients with MFH histology, and significance was only detected at baseline in other sarcoma histology (Table 3). We further performed Kaplan-Meier analysis at baseline by histology and grade. There was not a significant survival difference for patients with MFH histology, but a significant survival difference at baseline for patients with LMS, liposarcoma, and other sarcoma histology was found (Fig. 3).
Table 3. Interaction Analyses of Tumor Histology and Grade
Years of Survival
MFH, HR (95% CI)
LMS, HR (95% CI)
LPS, HR (95% CI)
Other, HR (95% CI)
Abbreviations: MFH, malignant fibrous histiocytoma; HR, hazard ratio; CI, confidence interval; LMS, leiomyosarcoma; LPS, liposarcoma; n, sample size of each cohort.
This large population-based study of >1100 primary nonmetastatic abdominal and retroperitoneal sarcoma patients demonstrates that the variables contained within the AJCC staging system and current postoperative nomograms do not adequately predict continued survival in persons who have survived years after their initial retroperitoneal sarcoma diagnosis. Despite the presence of several patient and tumor characteristics indicative of poor prognosis at baseline, many of these variables become nonsignificant predictors after only 1 year of survival. Only tumor grade continued to predict survival in long-term survivors of abdominal and retroperitoneal sarcoma. To our knowledge, the current study is among the first population-based studies of conditional survival in persons who survived years after resection of primary nonmetastatic abdominal and retroperitoneal sarcoma.
In agreement with our findings, previous literature supports that tumor grade, histologic type, sex, and age are all important predictors of survival at the time of diagnosis.1, 3, 9, 10, 12, 15, 16 Canter et al evaluated the impact of histologic subtype according to varying histologic grade for soft tissue sarcoma. In this study, which included all soft tissue sarcomas, it was found that 5-year survival rates varied within categories of low-, intermediate-, and high-grade disease based on histologic subtype.10 A recently published nomogram analyzing 343 patients with retroperitoneal sarcoma found that older age, larger tumor size, multifocal disease, incomplete surgical resection, and poorly differentiated liposarcoma were independent predictors of worse sarcoma survival at baseline.17
The importance of tumor grade overall and within specific histologic categories was demonstrated in our work. These findings are especially important when certain histologic subtypes exhibit worse behavior regardless of their grade (eg, LMS) or when they are diagnosed with higher grade (atypical lipomatous tumors vs pleomorphic liposarcoma).12, 13 For instance, we found that higher tumor grade in liposarcoma had a poorer prognostic significance at baseline and after years of survival, likely reflecting the varying histology spectrum of liposarcoma. Although less conclusive because of sample size, similar trends were also observed for LMS histologic subtypes. Other studies confirm the importance of tumor grade and histology in determining prognosis at baseline; our work highlights that these variables, most notably tumor grade, remain the only significant predictor of continued survival even after years of survival. Nevertheless, these findings should not be viewed as a deterrent to the impact of sarcoma histology on oncologic decision making or disease prognosis.
The lack of association between tumor size and long-term conditional survival in those surgically treated for primary nonmetastatic retroperitoneal sarcoma is noteworthy. Tumor size is an integral part of the current AJCC staging system and published prognostic nomograms.14, 17-19 Although its validity in predicting prognosis in extremity soft tissue sarcoma has been established, its relevance for patients with abdominal and retroperitoneal sarcoma has been disputed in some single-hospital and population-based experiences.5, 6, 8, 9, 15, 16 One reason for this may be that at the time of diagnosis, most patients present with tumors >15 cm.11 Alternatively, other factors such as low grade and variations in histologic subtypes (eg, differentiation of lipomatous tumors) may represent significant predictors of conditional survival that override the impact of size itself. This may be manifested by certain subtypes of lipomatous tumors or sarcomas that exhibit a more indolent course in the context of larger tumor size. Despite several modifications to our multivariate model, larger tumor size was not found to impact conditional survival. In light of our findings, our study adds further support to the current body of literature questioning the relevance of tumor size as a prognostic indicator for retroperitoneal sarcoma at the time of diagnosis, as well as its role in predicting continued survival.5, 6, 8, 9, 15, 16
The current study in addition sheds light on the limited impact of adjuvant RT on survival after surgical treatment of primary nonmetastatic retroperitoneal sarcoma. After adjusting for covariates, we found no significant survival benefit for patients treated with adjuvant external beam RT compared with surgical resection alone. On the basis of 2 randomized clinical trials, the benefit of RT is well established in patients with extremity soft tissue sarcoma.31, 32 Improvements in local control and a decrease in local recurrence have been demonstrated by recent single- and multi-institution and population-based studies evaluating the use of neoadjuvant, adjuvant, or combination RT for retroperitoneal sarcoma.6, 33, 34 Feng et al demonstrated an improvement in local control with high-dose RT in a retrospective evaluation of patients who received neoadjuvant, adjuvant, or combination RT. There was no benefit in overall survival, however, and this study did not compare local control in patients who received radiation versus patients who did not.33 Stoeckle et al also demonstrated a decrease in local recurrence in patients treated with adjuvant RT; however, in this study of 165 patients, not all patients received complete surgical resection.6 In a population-based study using SEER, Zhou et al demonstrated that patients with stage I disease benefitted from RT in addition to resection (HR, 0.49; 95% confidence interval [CI], 0.25-0.96; P = .04), but patients with stage II or stage III disease did not see any additional benefit with RT (HR, 0.78; 95% CI, 0.58-1.06; P = .11).34 This study evaluated several different types of RT and was not limited to adjuvant therapy only. These findings contrast with those of another recently published SEER study by Tseng et al, which found that postoperative RT did not improve disease-free or overall survival for patients with retroperitoneal sarcoma, except for those patients with MFH.35 These differences, in addition to other variations in inclusion and exclusion criteria, may explain why our study did not demonstrate a benefit with RT.
Naturally, these retrospective and single-institution studies impart some limitations and interpretation bias.36-38 To date, no randomized controlled trial evaluating the role of adjuvant RT for retroperitoneal sarcoma has been conducted that supports these studies or demonstrates improvement in overall survival. Attempts at a randomized controlled trial in North America failed because of poor accrual to the American College of Surgeons Oncology Group trial Z9031, thus leaving the medical community without a definitive answer as to the clinical benefit of RT for persons with retroperitoneal sarcoma.
We acknowledge several SEER-related limitations in the present study. First, SEER does not collect information on resection margin status. Incomplete resection margins (R1/R2) have been repeatedly shown to decrease relapse-free and overall survival,2-9 and to control for this limitation we included only the SEER surgery codes that specified removal of the primary tumor site; however, we are unable to account for potential resections with positive resection margins. Therefore, SEER codes that were likely to leave macroscopic disease (eg, debulking procedures and local excision) were excluded from our analysis. Second, SEER does not capture information on disease recurrence. Patients with retroperitoneal sarcoma are known to have a high risk for disease recurrence within 2 to 3 years of their initial diagnosis and resection. In this regard, the current study may have underestimated the effects of local retroperitoneal recurrence in those who continued to survive. Finally, SEER does not collect information on systemic chemotherapy. To date, systemic chemotherapy remains investigational in the setting of ostensibly resectable retroperitoneal sarcoma except for chemosensitive histologic subtypes. Despite these mportant limitations, the current appraisal stems from a large and nationally representative study from over 2 decades, thus minimizing the referral and institutional biases frequently observed in single-hospital studies.
The current study presents an important set of future implications. First, our work complements published nomograms that have built upon the AJCC staging system's TNGM (TNM + group) designation to find additional variables, which more accurately predict survival at the time of diagnosis. Our conditional survival estimates add benefit to these nomograms by demonstrating how the significance of these variables changes with years of survival, as depicted in Figure 1. Second, the current work provides the first framework to help clinicians counsel survivors of retroperitoneal sarcoma. By demonstrating how demographic and tumor factors vary in significance over time, we highlight which variables remain prognostic at specific time points in survival. This information will allow clinicians to develop patient-specific surveillance schedules and aid in the development of resource-conserving cancer survivorship programs that may improve healthcare utilization and enhance the quality of life for patients. Third, as our knowledge of biomarkers and genotypes evolve, the present work can be built upon to reflect the addition of new prognostic variables.
The current AJCC guidelines and postoperative nomograms are limited in their ability to assist health care professionals in predicting continued survival for patients who have survived years after diagnosis and resection for primary nonmetastatic abdominal and retroperitoneal sarcoma. Tumor grade is the only variable that remains a significant predictor of survival. Conditional survival estimates allow clinicians to provide survivors with more meaningful prognostic estimates that may impact surveillance schedules and streamline adjuvant therapy decisions and design of future clinical trials.