Patients with different disease stages and treatment received. *, Two hundred fifty seven patients had no treatment and 33 had missing resection data; †, 287 patients had no treatment and 20 had missing resection data.
Kaplan-Meier curve of overall survival rate in patients with locoregional disease with or without surgery and/or radiotherapy.
Zhou Z, McDade TP, Simons JP, Ng SC, Lambert LA, Whalen GF, Shah SA, Tseng JF. Surgery and Radiotherapy for Retroperitoneal and Abdominal SarcomaBoth Necessary and Sufficient. Arch Surg. 2010;145(5):426–431. doi:10.1001/archsurg.2010.70
To evaluate the effect of surgical resection and radiotherapy (RT) in retroperitoneal or abdominal sarcoma.
Surveillance, Epidemiology, and End Results, 1988-2005.
Patients 18 years or older with initial diagnosis of primary retroperitoneal and nonvisceral abdominal sarcoma.
Main Outcome Measures
Survival for 2 years after diagnosis. Kaplan-Meier survival was stratified based on surgery and RT status. Cox proportional hazards model was used to assess adjusted effects of surgery and RT on survival in patients with locoregional disease.
Of 1901 patients with locoregional disease, 1547 (81.8%) underwent resection; 447 (23.5%) received RT. Overall, patients who received both surgery and RT demonstrated improved survival compared with patients who underwent either therapy alone; patients undergoing monotherapy in turn had more favorable survival compared with patients who received neither therapy (P < .001, log rank). Cox analysis demonstrated that surgical resection (hazard ratio [HR], 0.24; 95% confidence interval [CI], 0.21-0.29; P < .001) and RT (0.78; 0.63-0.95; P = .01) independently predicted improved survival in locoregional disease only. In adjusted analyses stratified for American Joint Commission on Cancer (AJCC) stage, for stage I disease (n = 694), RT provided an additional benefit (HR, 0.49; 95% CI, 0.25-0.96; P = .04) independent of that from resection (0.35; 0.21-0.58; P < .001). For stage II/III (n = 552), resection remained protective (HR, 0.24; 95% CI, 0.18-0.32; P < .001); however, RT was no longer associated with a significant benefit (0.78; 0.58-1.06; P = .11).
In a national cohort of retroperitoneal and abdominal sarcomas, surgical resection was associated with significant survival benefits for AJCC disease stages I to III. Radiotherapy provided additional benefit for patients with stage I disease. Resection should be offered to reasonable surgical candidates with nonmetastatic retroperitoneal/abdominal sarcomas; radiotherapy may most benefit patients with early-stage disease.
Retroperitoneal and abdominal soft tissue sarcoma accounts for approximately 15% to 20% of newly diagnosed soft tissue sarcoma reported annually in the United States.1,2 The overall 5-year survival is reported to be 40% to 60%, with disease mortality being mainly because of tumor recurrence.3- 6 Surgery and radiotherapy (RT) (particularly in the adjuvant setting) are the major treatment options for retroperitoneal and abdominal sarcoma.7,8 However, large-scale studies that evaluate the independent and combined benefits of these 2 main modalities are lacking. Furthermore, it is also unknown how such benefit might vary regarding different disease stages, characterized by tumor size, grade, and lymph node involvement. Although there have been reports9- 14 on several cohort studies or case series based on single-institution experiences with surgical treatment and RT, it is unclear to what extent these findings from small numbers of patients can be generalized to most patients with retroperitoneal and abdominal sarcoma. The goal of the current study was to evaluate the survival benefit in a population-based setting among patients with retroperitoneal and abdominal sarcoma after surgical resection and/or RT, using National Cancer Registry data from Surveillance, Epidemiology, and End Results (SEER), 1988-2005.
Data were obtained from the SEER program administered by National Cancer Institute. Seventeen tumor registries participate in the program (Table 1). Combined, these sites provide a representative survey of the United States, approximating 26.0% of the population. This sampling is comparable to the general US population for a number of characteristics, including those of potentially at-risk groups.15,16 Information within the registry includes patient demographics (age, sex, and race), disease-related data (cancer site, tumor characteristics, and lymph node involvement), and first course of treatment (cancer-directed surgery and RT). Survival rate data are available through the SEER database via its linkage with the National Death Index, a centralized index of death records obtained through state vital statistics offices.
A retrospective cohort of patients with retroperitoneal and abdominal sarcoma was assembled based on SEER using the 1988-2005 data submission from the National Cancer Institute. Patients 18 years or older who had an initial diagnosis of primary retroperitoneal and nonvisceral abdominal sarcoma from January 1988 through December 2005 were included in the study. Disease diagnosis was established by using International Classification of Diseases for Oncology, Third Edition17 histology codes for primary site sarcoma coded as retroperitoneal or abdominal, including peritoneum, mesentery, or omentum (codes available on request from the corresponding author). For abdominal sarcoma, patients with nonvisceral sarcomas were included because of their similar histologic subtypes and a growth pattern of their sarcomas similar to those of retroperitoneal sarcomas.1 The heterogeneous groups of visceral, intestinal, or gastrointestinal stromal tumor subtypes, Kaposi sarcoma, and multiple hemorrhagic sarcomas were excluded. Subsequently, a locoregional disease cohort was generated by identifying patients coded in SEER as having local or regional disease, using the SEER_historic_stage_A variable.
For each patient, information was collected on nonidentifying demographics (age at diagnosis, sex, race, and marital status18- 20), year of diagnosis, location of tumor (abdominal vs retroperitoneal), disease status (localized, regional, or distant or metastatic), TNM classification and tumor grade, and type of treatment (surgical resection and/or RT). American Joint Commission on Cancer (AJCC) stage information in SEER is frequently missing or unknown. Therefore, for the purposes of our analyses of locoregional disease, a new variable (AJCC stage) was derived based on TNM classification and tumor grade information, as defined by the sixth edition of the AJCC Staging Manual:21,22 AJCC stage I (T1a-2b, N0, M0, grade I or II), stage II (T1a, 1b, 2a, N0, M0, grade III or IV), and stage III (T2b, N0, M0, grade III or IV).
Patients who were coded as receiving cancer-directed surgery with curative intent were defined as having surgical resection. Those who received only non–cancer-directed operations (eg, debulking, procedures directed at regional and/or distant sites or nodes, biopsy, exploration laparotomy, bypass, or ostomy) were considered not to have had surgical resection. For RT, patients who received preoperative, intraoperative or postoperative external beam radiation, radioisotopes, radioactive implants, or a combination of beam radiation with implants or isotopes were all classified as having had RT. SEER registries capture RT data for the first course of treatment or within 4 months of diagnosis, whichever is longer. Information is not available regarding number of RT treatments.
A multivariate logistic regression model was used to identify independent predictors for receiving surgical resection or RT among patients with locoregional disease. Significance of individual variables, including patient information (age at diagnosis, sex, race, and marital status), tumor location (abdominal vs retroperitoneal), and tumor stage (AJCC stage I vs II/III), were examined in the model first through univariate analysis and then by a multivariate logistic regression model using backward selection. Odds ratios and 95% confidence intervals (CIs) of significant predictors for treatment type were reported from the final multivariate model.
The primary study outcome was defined as death owing to all causes with total length of follow-up to 2 years from the initial diagnosis of sarcoma. If applicable, the date of death was ascertained. Data from patients who were alive by the end of 2 years were censored.
Kaplan-Meier curves were first constructed to compare survival rates based on overall patient and disease characteristics (age category, sex, tumor location, and disease stage). Next, Kaplan-Meier survival rate analysis was performed for the subset of patients with local or regional disease, who were thus eligible for potentially curative therapy, including surgical resection and/or RT. Kaplan-Meier curves of patients who received surgical resection or RT alone, both treatments, or neither were compared. Median survival time in each group and the P value from a log-rank test were reported.
A Cox proportional hazards model was used to assess the effect of surgery and/or RT on patient survival rate. Univariate analysis was performed to evaluate the association of each treatment type (surgical resection and RT) on survival rate, followed by a multivariate Cox model to include patient information (age at diagnosis, sex, race, and marital status), tumor location (abdominal or retroperitoneal), and tumor characteristics (AJCC stage I, II, or III) for adjustment. We also adjusted for year of diagnosis (grouped by 5-year intervals) to take into account any possible temporal changes in patient care during the study period. A backward selection algorithm was used to select significant variables. Age, sex, and year of diagnosis as clinically important variables were kept in the model regardless of their significance. Treatment effect was further examined by stratification based on AJCC tumor stage (II/III vs I). The hazard ratios (HRs) and 95% CIs were reported from these survival rate analyses. All data management and statistical analyses were performed using SAS statistical software, version 9.2 (SAS Institute Inc, Cary, North Carolina), and P < .05 was considered statistically significant.
A total of 2504 patients with retroperitoneal and abdominal sarcomas were identified from the SEER database. Mean age was 62 years. A total of 1118 (47.4%) patients were male and 2055 (82.1%) were white. Retroperitoneal sarcoma was diagnosed in 2230 (89.1%) patients. Of the 2504 patients, 1901 (75.9%) had locoregional disease (Table 1). The remaining 603 patients had distant disease at diagnosis. Among patients with locoregional disease, 1547 (81.8%) had surgical resection, and 23.5% of these surgical patients also received RT. Sixty-four patients (3.4%) received only RT (Figure 1).
For patients with locoregional disease, an analysis of predictors of surgical resection showed that patient age, AJCC stage, and race were significant determinants. Older patients and patients with more advanced disease stage (AJCC stages II/III) were less likely to receive surgical resection, whereas white race was associated with a higher likelihood of receiving surgery.
Advanced stage (AJCC stages II/III) was a positive predictor for RT, as was retroperitoneal location. Older age, however, was associated with a lower likelihood of receiving RT. For both treatments, sex nor diagnosis year appeared to be significant predictors. Neither treatment type independently predicted the other (Table 2).
Kaplan-Meier survival rate analysis based on overall patient and disease characteristics showed significantly worse survival outcome in patients older than 80 years (P < .001, log-rank test) and with nonretroperitoneal disease (P = .002, log-rank test) (data not shown). Among the 1901 patients with locoregional disease, 589 deaths (31.0%) occurred during the 2-year follow-up period. Crude analysis based on Kaplan-Meier curves showed distinct differences in survival outcomes for patients who received both surgical treatment and RT vs surgery or RT alone vs neither, where the RT alone and neither treatment groups frequently represented patients with stage II and III disease and had the worst outcome. An overall significant difference was found in terms of median survival (P < .001, log-rank test) (Figure 2).
Univariate analysis based on Cox proportional hazards modeling showed significant survival benefit associated with either surgical resection (HR, 0.24; 95% CI, 0.21-0.29) or RT (0.78; 0.63-0.95) in patients with locoregional diseases. Older age, male sex, and advanced AJCC stage (II/III) were individually associated with higher risk of mortality (Table 3).
Multivariate analysis, adjusting for patient characteristics and year of diagnosis and stratified by AJCC stage (I vs II/III), showed survival benefit provided by surgical resection in the stage I and II/III groups, providing 65.0% to 76.0% risk reduction for mortality in locoregional disease (stage I: HR, 0.35; 95% CI, 0.21-0.58; stage II to III: 0.24; 0.18-0.32). The benefit of RT in addition to surgical resection was most pronounced in patients with AJCC stage I disease who had an additional 50.8% risk reduction (HR, 0.49; 95% CI, 0.25-0.96). No statistically significant independent benefit of RT in patients with stage II to III disease was found (HR, 0.78; 95% CI, 0.58-1.06; P = .11). Patient age but not tumor location was a significant risk factor for mortality in both the stage I and stage II to III groups. Furthermore, male sex and earlier year of diagnosis (1988-1993) appeared to be risk factors for poor outcomes in patients with stage II to III disease (Table 3).
Results from this large national cohort of patients with retroperitoneal and abdominal sarcomas showed that surgical resection was associated with a significant survival advantage in locoregional disease and that the benefit persisted across both AJCC stage I and II/III disease groups. Radiotherapy provided a significant additional benefit for patients with AJCC stage I disease more than 2 years after diagnosis.
There is a paucity of literature evaluating the survival benefit provided by surgical resection and/or RT in treating retroperitoneal and abdominal sarcoma.23 Among patients with locoregional disease, surgical resection provides the only means of a potentially curative treatment. Given the high late recurrence rate of the disease, however, there has been a long controversy regarding the benefit of perioperative RT to treat residual disease after surgery. The results shown by different studies12,24- 26 were varied. A previous study performed by Porter et al18 showed an overall relatively low use of adjuvant RT (25.9%) after retroperitoneal sarcoma surgery. A recent phase 3 randomized controlled trial to compare preoperative RT plus surgery vs surgery alone, conducted by the American College of Surgeons Oncology Group (ACOSOG Z9031), was prompted by a similar question, and the study is currently under way.
By using the national SEER database, which contains relatively comprehensive information regarding patient and tumor characteristics and survival rate data, we studied 2504 patients with retroperitoneal and abdominal sarcoma. Survival benefit of surgical resection and RT was evaluated among 1901 patients with locoregional disease who were thus eligible for potentially curative therapy to reduce confounding by disease severity. Of note, RT was primarily used in addition to surgical resection rather than as single-modality treatment. A 65.0% to 76.0% risk reduction in mortality was found to be associated with surgical resection in locoregional sarcoma, and a further 50.8% reduction was offered by the addition of RT to patients with AJCC stage I disease. Reduced benefit by RT (22.1% risk reduction) was found with AJCC stage II to III diseases and was no longer significant after adjusting for age, sex, race, and temporal trends. Although RT has been routinely recommended for patients with more advanced-stage disease to preoperatively or postoperatively address the issues of margin status and residual disease, our results suggest that RT may most benefit those patients with operable earlier-stage sarcoma.
Several concerns and limitations are associated with this study. First, as in any retrospective, nonrandomized database study, the current study is subject to possible confounding from unmeasured factors and missing clinical details, such as specific RT plan, surgical reports, margin status, and tumor histologic type. With regard to the RT treatments, we have information for the first course, thus mostly perioperative treatment, and no information regarding subsequent RT and treatment compliance. Therefore, our analysis of RT is best considered as an intent-to-treat analysis. Despite the likely changes in subsequent RT treatments, our results still suggested that initial surgery plus RT produced a better overall survival rate compared with surgery alone during a 2-year follow-up in stage I disease. We considered different radiotherapies, including time sequence of perioperative RT (eg, preoperative, postoperative and/or intraoperative), as 1 combined group of treatment in addition to surgery compared with surgery alone, and the current study did not try to discern treatment effect among different RT methods.
Second, there is a concern of lack of detailed operative reports. Our primary analysis was based on a relatively homogeneous patient population with locoregional diseases that are potentially surgically resectable. Moreover, we do not anticipate the extent of surgery to differ substantially between our comparison groups of surgery vs surgery plus RT in the initial treatment of locoregional disease, minimizing the concern of confounding in this case. As to the lack of knowledge on margin status in surgery, one could reasonably assume that patients with positive margins likely underwent more extensive surgical resection and were more likely to receive RT in addition to surgery. Although there might be an increased risk of recurrence and mortality associated with patients with positive margins in the surgery plus RT group, we were still able to show that combined treatment was superior to surgery alone in terms of overall survival rate during the study period (Figure 2). Therefore, the lack of knowledge of margin status is unlikely to change our conclusion. However, missing margin status limits our ability to assess disease-free survival rate as an outcome, if positive margin is associated with higher risk of recurrence, in which case margin status if available needs to be adjusted in the comparison to reduce potential bias. A similar argument pertains to the missing information about tumor histologic type. Despite the fact that different tumor histologic types may influence the choice of RT, and, hence, potentially heterogeneous RT responsiveness and outcomes in the surgery plus RT group, we were able to show that this group performed better than surgery alone in stage I disease.
Third, the SEER database does not contain information regarding chemotherapy that patients may have received. However, there is no obvious reason to suspect that chemotherapy would be broadly indicated outside a clinical trial among patients with locoregional disease,1,27 and their use, if any, would be substantially different across our 2 comparison groups.
Finally, we have shown the treatment effects on overall survival rate rather than progression-free survival rate during 2-year follow-up because of the lack of information regarding disease recurrence. Two-year follow-up was chosen a priori as a reasonable cutoff point because of concerns of greater loss to follow-up and compromised data quality associated with longer study duration.
Our results suggest that resection should be offered to all reasonable surgical candidates given the magnitude of survival benefit and that RT may most benefit those patients with operable stage I sarcoma. This finding suggests a need for improved tailoring of treatments for stage I retroperitoneal and abdominal sarcomas. Careful analyses of the results of currently ongoing and future randomized controlled trials are warranted to confirm and expand these findings.
Correspondence: Jennifer F. Tseng, MD, MPH, Surgical Outcomes Analysis and Research (SOAR), University of Massachusetts Medical School, Worcester, MA 01655 (firstname.lastname@example.org).
Accepted for Publication: October 27, 2009.
Author Contributions:Study concept and design: Zhou, McDade, Lambert, Shah, and Tseng. Acquisition of data: Zhou, Simons, Ng, Shah, and Tseng. Analysis and interpretation of data: Zhou, McDade, Ng, Lambert, Whalen, Shah, and Tseng. Drafting of the manuscript: Zhou. Critical revision of the manuscript for important intellectual content: Zhou, McDade, Simons, Ng, Lambert, Whalen, Shah, and Tseng. Statistical analysis: Zhou, Ng, and Tseng. Obtained funding: Tseng. Administrative, technical, and material support: McDade, Simons, Whalen, Shah, and Tseng. Study supervision: Whalen, Shah, and Tseng.
Financial Disclosure: None reported.
Funding/Support: Dr Tseng is funded by a Howard Hughes Medical Institute Early Career Award and the American Surgical Association Foundation.
Previous Presentation: This study was presented at the 90th Annual Meeting of the New England Surgical Society; September 11, 2009; Newport, Rhode Island; and is published after peer review and revision.
Additional Information: Drs Zhou, McDade, and Simons contributed equally to this article.