Fax: (713) 794-5573
Prognostic factors for patients with localized soft-tissue sarcoma treated with conservation surgery and radiation therapy
An analysis of 1225 patients
Article first published online: 30 APR 2003
Copyright © 2003 American Cancer Society
Volume 97, Issue 10, pages 2530–2543, 15 May 2003
How to Cite
Zagars, G. K., Ballo, M. T., Pisters, P. W. T., Pollock, R. E., Patel, S. R., Benjamin, R. S. and Evans, H. L. (2003), Prognostic factors for patients with localized soft-tissue sarcoma treated with conservation surgery and radiation therapy. Cancer, 97: 2530–2543. doi: 10.1002/cncr.11365
- Issue published online: 30 APR 2003
- Article first published online: 30 APR 2003
- Manuscript Accepted: 28 JAN 2003
- Manuscript Revised: 21 JAN 2003
- Manuscript Received: 29 AUG 2002
- National Cancer Institute, U.S. Department of Health and Human Services. Grant Number: CA 06294
- radiation therapy;
- prognostic factors
Prognostic factors for patients with soft-tissue sarcoma who are treated with conservative surgery and radiation are documented poorly.
The clinicopathologic features and disease outcome for 1225 patients with localized sarcoma who were treated with conservative surgery and radiation were reviewed retrospectively. Actuarial univariate and multivariate statistical methods were used to determine significant prognostic factors for local control, metastatic recurrence, and disease specific survival.
The median follow-up of surviving patients was 9.5 years. The respective local control rates at 5 years, 10 years, and 15 years were 83%, 80%, and 79%. Factors predictive of local recurrence were positive or uncertain resection margins; tumors located in the head and neck and the deep trunk; presentation with local recurrence; patient age > 64 years; malignant fibrous histiocytoma, neurogenic sarcoma. or epithelioid sarcoma histopathology; tumor measuring > 10 cm in greatest dimension; and high pathologic grade. Freedom from metastasis at 5 years, 10 years, and 15 years was 71%, 68%, and 66%, respectively. Factors that were predictive of metastatic recurrence were high tumor grade; large tumor size (> 5 cm); and leiomyosarcoma, rhabdomyosarcoma, synovial sarcoma, or epithelioid sarcoma. The respective disease specific survival rates at 5 years, 10 years, and 15 years were 73%, 68%, and 65%. Adverse factors for disease specific survival were high tumor grade; large tumor size (> 5 cm); tumors located in the head and neck and deep trunk; rhabdomyosarcoma, epithelioid sarcoma, or clear cell sarcoma; patient age > 64 years; and positive or uncertain resection margins.
Soft-tissue sarcoma comprises a heterogeneous group of diseases. Prognostic factors for local recurrence, metastatic recurrence, lymph node recurrence, disease free survival, and disease specific survival are different, and optimal treatment strategies need to take this complexity into account. Cancer 2003;10:2530–43. © 2003 American Cancer Society.
Although the combination of conservation surgery (S) and radiation therapy (XRT) has been a mainstay in the treatment of patients with soft-tissue sarcoma (STS) for several decades,1–4 few studies have reported comprehensive analyses of prognostic factors after this treatment strategy. Most data on prognostic factors for patients with these tumors come from analyses of patient groups treated with and without XRT and often are restricted to patients with tumors arising in specific anatomic regions, especially the extremities. These studies have defined a number of relevant outcome determinants.5–12 It is recognized widely that different disease endpoints, particularly metastatic and local recurrence, are influenced by different factors.13 With respect to metastatic recurrence, virtually all reports have found that tumor grade, tumor size, and tumor depth relative to the investing fascia are crucial prognostic factors,5, 7–9, 14 and these parameters are incorporated into the American Joint Committee on Cancer (AJCC) staging system.15 The roles of other factors, such as histopathologic subtype and site of origin, are controversial. Factors that are significant for local control are less well defined. It has been found generally (although not universally9) that microscopic resection margins are important.2, 5, 10, 16, 17 Less consistently, it has been found that prior local recurrence,5, 18, 19 older age,5, 9 retroperitoneal or head and neck location,7–9 high grade,9, 10 and selected histopathologic subtypes5 are significant. In many reports, however, these factors either were not analyzed or were identified as insignificant. Much of this uncertainty is related to the rarity of STS, and few studies have had enough patients to perform adequately powerful statistical analyses of numerous potentially significant factors for a variety of outcome endpoints. To address these issues, we performed a retrospective review of the large experience with S and XRT at the University of Texas M. D. Anderson Cancer Center (MDACC). Although certain treatment factors were included in this analysis, its primary objective was to elucidate significant tumor and patient determinants of disease outcome.
MATERIALS AND METHODS
The medical records of all patients with STS who underwent S and received XRT at MDACC between 1960 and 1999, inclusive, were reviewed. Patients were identified from a registry that has been maintained prospectively in the Department of Radiation Oncology since the early 1950s. One thousand two hundred twenty five consecutive patients, selected according to the criteria detailed below, were the focus of this report. No patient had metastatic disease (lymph node or distant) at the time of presentation, and each patient underwent conservative, macroscopic, total tumor resection (either before or after XRT). Patients who were treated with neutron or tourniquet-hypoxia XRT were excluded. Desmoids, cystosarcomas, angiosarcomas, dermatofibrosarcoma protuberans, and Kaposi sarcomas—tumor types that are excluded from the AJCC staging system15—also were excluded from this analysis.
Each patient had been evaluated with a full history, physical examination, routine blood tests, chest radiography, and other studies as appropriate and available during the years encompassed by this review. The histopathologic diagnosis was established by a review of slides at our institution. Our pathologists render a diagnosis of unclassified sarcoma when they judge that a sarcoma does not fulfill the criteria for any of the recognized subcategories. Grading of STS at MDACC uses histotype as the definition of grade20, 21 for the majority of tumors. Thus, pleomorphic malignant fibrous histiocytoma (MFH), synovial sarcoma, pleomorphic liposarcoma, dedifferentiated liposarcoma, neurogenic sarcoma (malignant peripheral nerve sheath tumor), rhabdomyosarcoma, epithelioid sarcoma, clear cell sarcoma, alveolar soft part sarcoma, Ewing sarcoma, primitive neuroectodermal tumor, osteosarcoma, and mesenchymal chondrosarcoma, by definition, are high grade. Myxoid MFH, myxoid liposarcoma, extraskeletal myxoid chondrosarcoma, and (for this analysis) hemangiopericytoma, by definition, are intermediate grade; and atypical lipomatous tumor, by definition, is low grade. However, grading on an individual basis is done for leiomyosarcoma, fibrosarcoma, and unclassified sarcomas. Thus, each tumor was assigned to a low, intermediate, or high grade. Tumor size was documented as the greatest tumor dimension at the onset of treatment. The influence of tumor size on outcome was evaluated using the AJCC criterion of 5 cm (≤ 5 cm is T1, and > 5 cm is T2) as well as other cut-off points. Tumor depth relative to the superficial fascia was not documented uniformly and could not be used as a variable in this study. Tumor site was classified into head and neck (above the clavicles), upper extremity (at and distal to the shoulder), lower extremity (at and distal to the groin), and trunk (all other sites, including buttock). Trunk sites were subclassified further into superficial (muscular and subcutaneous parietes) and deep (intrathoracic, intraabdominal [including retroperitoneal], and intrapelvic).
At MDACC, the preferred treatment strategy for patients with STS has been a combination of conservative excision and XRT. The goal of surgery was to obtain a tumor free margin of 1–3 cm; when this was not feasible because of adjoining significant structures, an attempt was made to obtain tumor free, albeit closer, margins. Prior to referral to MDACC, virtually all patients had undergone at least a biopsy of the lesion, and management was dictated to some extent by the immediately antecedent surgical procedure. When tumors appeared to have been excised completely, examination of the patient, review of the operative report, communication with the referring surgeon, evaluation of the pathology report, and radiologic investigation were used to estimate the completeness of resection and to decide whether reexcision was necessary. Resection margins were defined as positive if the pathology report on the final resection stated that tumor was present at the resection margin; margins were defined as negative if the report stated that tumor was not present at the margin, regardless of the magnitude of the margin; and margins were uncertain if the report did not comment explicitly on the margin. XRT was delivered with megavoltage beams (60Co or greater energy) using techniques appropriate to each site, as described previously.22 Small numbers of patients received brachytherapy either as the sole form of XRT or as a boost after external beam therapy; a few patients received a single-dose intraoperative XRT boost. The sequencing of XRT and S was determined by the surgical procedure immediately preceding referral and by preferences among the sarcoma treatment team. During the 1960s and 1970s, the general preference was for postoperative XRT;1 later, in the 1980s and 1990s, the general preference was for preoperative XRT.23 Three S and XRT treatment sequences were identified: 1) S XRT (postoperative XRT), in which all definitive surgery, including reresection, was completed before any XRT (typically, 60–70 grays [Gy]); 2) XRT S (preoperative XRT), in which macroscopic tumor was irradiated (typically, 50 Gy) before final definitive resection; and 3) S XRT S (sandwich XRT), in which preoperative XRT (typically, 50 Gy) was delivered after limited excision of macroscopic tumor (prior to referral) before final, definitive resection of the tumor bed. Chemotherapy was used increasingly after 1980 and was doxorubicin-based in virtually all patients who received this modality. In general, chemotherapy was recommended for patients with high-grade, large tumors and was given mostly after completion of local-regional treatment.
After treatment, patients were followed at 3-month to 6-month intervals for the first few years and yearly thereafter. Follow-up was calculated from the time of completion of XRT. At the time of this analysis, 678 patients were alive, and 547 patients had died; the duration of follow-up for surviving patients was 1.3–37.7 years, with a mean of 11.0 years and a median of 9.5 years. The major endpoints of this study were local recurrence (defined as any recurrence at or adjacent to the initial primary site), lymph node recurrence (any recurrence in the regional lymph nodes), metastatic recurrence (any hematogenous recurrence), freedom from any recurrence (i.e., disease free survival, for which the first recurrence at any site is an event), and disease specific survival (with death due to sarcoma or treatment as an event; all other deaths were censored when they occurred). Overall survival (uncorrected) also was analyzed but was less meaningful compared with disease specific survival. Actuarial curves were constructed using the Kaplan–Meier method, and tests of significance between actuarial data were done with the log-rank statistic.24 Multivariate proportional hazards regression was performed using standard techniques.24 Preliminary data exploration to elucidate optimal cut-off points for continuous variables used recursive partitioning.25 However, because this technique uses categorical endpoints, it is not optimal for time-to-failure analyses and was not used as a major determinant of significant prognostic factors. Differences between means, proportions, and distributions were evaluated with the Mann–Whitney, Fisher exact, or chi-square tests, as appropriate.26 The method recommended by Rothman27 was used to calculate 95% confidence intervals (95%CIs) for actuarial data. The expected survival curve was an age-matched and gender-matched curve calculated from United States vital statistics data.28
Patient and Tumor Characteristics and Treatment
Table 1 summarizes a variety of patient, tumor, and treatment features. Patient age ranged from 1 year to 88 years, with a mean of 47 years and a median of 48 years. Seventy-seven patients (6%) were age ≤ 18 years. Female patients were significantly more common than male patients in this series (binomial test; P = 0.005). Among the 194 patients who presented with locally recurrent disease, the current presentation was the first recurrence in 137 patients, the second recurrence in 40 patients, and the third or greater recurrence in 17 patients. The 79 deep trunk tumors were abdominal or pelvic retroperitoneal in 73 patients and intrathoracic in 6 patients. The relative paucity of deep trunk tumors in this series reflects the fact that all patients included had XRT, and the sarcoma management team generally was reluctant to recommend XRT for patients with retroperitoneal tumors on the grounds that adequate doses could not be delivered to this anatomic site. Histopathologic types among the other category in Table 1 were extraskeletal chondrosarcoma in 9 patients, alveolar soft part sarcoma in 8 patients, hemangiopericytoma in 8 patients, Ewing/peripheral neuroectodermal tumor in 6 patients, extraskeletal osteosarcoma in 2 patients, low-grade fibromyxoid sarcoma in 2 patients, epithelioid hemangioendothelioma in 2 patients, and malignant rhabdoid tumor in 1 patient. MFH was subclassified as pleomorphic (high grade) in 359 patients and myxoid (intermediate grade) in 101 patients; liposarcoma comprised atypical lipomatous tumor (low grade) in 19 patients, myxoid (intermediate grade) in 103 patients, pleomorphic (high grade) in 38 patients, and dedifferentiated (high grade) in 10 patients. Rhabdomyosarcomas were embryonal in 18 patients, alveolar in 12 patients, and pleomorphic in 23 patients. Leiomyosarcomas were low grade in 10 patients, intermediate grade in 21 patients, and high grade in 46 patients. Fibrosarcomas were low grade in 7 patients, intermediate grade in 21 patients, and high grade in 8 patients. Unclassified sarcomas were low grade in 17 patients, intermediate grade in 28 patients, and high grade in 99 patients. Data on tumor size were available in 1189 patients (97%), with a range 0.5–36.0 cm (mean, 7.5 cm; median, 6.0 cm). Tumors of the lower extremity and trunk (median, 7.0 cm) were significantly larger compared with tumors of the upper extremity and the head and neck (median, 4.0 cm; P < 0.001). Among 77 patients age ≤ 18 years, the most common histologic types were synovial sarcoma in 20 patients (26%), rhabdomyosarcoma in 14 patients (18%), and unclassified sarcoma in 12 patients (16%).
|Characteristic||No. of patients (%)|
|Locally recurrent||194 (16)|
|Superficial trunk||196 (16)|
|Deep trunk||79 (6)|
|Upper extremity||251 (21)|
|Lower extremity||597 (49)|
|T1 (≤5 cm)||484 (41)|
|T2 (>5 cm)||705 (59)|
|Malignant fibrous histiocytoma||460 (38)|
|Synovial sarcoma||133 (11)|
|Neurogenic sarcoma||71 (6)|
|Epithelioid sarcoma||29 (2)|
|Clear cell sarcoma||14 (1)|
|Unclassified sarcoma||144 (12)|
|S XRT||818 (67)|
|XRT S||286 (23)|
|S XRT S||121 (10)|
|Final resection margin|
All 559 patients who presented with macroscopic disease underwent definitive resection at MDACC; 273 patients underwent S followed by XRT, and 286 patients received preoperative XRT. Of 666 patients who presented with no macroscopic disease, 371 patients underwent no further surgery and were given XRT at MDACC; estimated resection margins in these patients were negative in 117 patients (31%), positive in 47 patients (13%), and uncertain in 207 patients (56%). The remaining 295 patients without macroscopic tumor at presentation underwent reexcision at MDACC, with 174 patients undergoing reexcision followed by XRT and 121 patients receiving XRT followed by reexcision (the XRT sandwich). At presentation, the estimated resection margins in these 295 patients were negative in 12 patients (4%), positive in 63 patients (21%), and uncertain in 220 patients (75%). After reexcision, the final margins in the same patients were negative in 257 patients (87%), positive in 35 patients (12%), and uncertain in 3 patients (1%).
XRT consisted of external beam alone in 1167 patients (95%), brachytherapy alone in 34 patients (3%), external beam and brachytherapy boost in 17 patients (1%), and external beam with intraoperative boost in 7 patients (< 1%). External beam doses ranged from 40 Gy to 75 Gy (median, 60 Gy). In 1132 of 1191 patients (95%) who received external beam radiation, the dose per fraction was between 1.8 Gy and 2.2 Gy. Median XRT doses were 50 Gy preoperatively (range, 40–69.5 Gy), 50 Gy for the sandwich prereexcision (range, 46–54 Gy), and 62 Gy postoperatively (range, 40–75 Gy). Patients who received postoperative XRT had significantly higher doses if their resection margins were positive or uncertain (median dose, 64 Gy) than if their margins were negative (median dose, 60 Gy; P < 0.001).
Doxorubicin-based chemotherapy was given to 405 patients (33%). Chemotherapy was recommended as a rule for patients with large (T2), high-grade tumors, and the proportion of patients receiving this treatment increased significantly over the years encompassed by this review from 20% in the 1970s, to 36% in the 1980s, and to 41% in the 1990s. According to tumor grade, chemotherapy was used in 5 of 57 patients with low-grade tumors (9%), in 57 of 291 patients with intermediate-grade tumors (20%), and in 343 of 877 patients with high-grade tumors (39%; P < 0.001). Likewise, the use of chemotherapy increased as tumor size increased from 82 of 484 patients with T1 lesions (17%) to 318 of 705 patients with T2 lesions (45%; P < 0.001). The interplay between tumor grade, tumor size, and use of chemotherapy is illustrated by the following data: Among patients with intermediate-grade tumors, chemotherapy was administered to 0 of 20 patients with tumors measuring ≤ 2 cm (0%), to 7 of 83 patients with tumors measuring > 2 cm and ≤ 5 cm (8%), to 25 of 100 patients with tumors measuring > 5 cm and ≤ 10 cm (25%), and to 25 of 81 patients with tumors measuring > 10 cm (31%). Among patients with high-grade tumors, chemotherapy was administered to 11 of 74 patients with tumors measuring ≤ 2 cm (15%), to 64 of 286 patients with tumors measuring > 2 cm and ≤ 5 cm (22%), to 171 of 332 patients with tumors measuring > 5 cm ≤ 10 cm, 51%), and to 93 of 162 patients with tumors measuring > 10 cm (57%). Chemotherapy was administered concurrently with XRT to 99 patients (8%).
Disease Outcome and Patient Survival
At the time of analysis, disease had recurred in 504 patients (41%), and 547 patients (45%) had died, with 372 deaths (30%) due to STS (except for 1 patient who died of brain radionecrosis; however, this death was included as a disease specific death). The actuarial overall, uncorrected, and disease specific survival curves are shown in Figure 1. The sites of recurrence were local in 204 patients (17%), lymph node in 38 patients (3%), and metastatic in 371 patients (30%): One hundred five patients experienced recurrence at more than one site. Local control, lymph node control, metastatic control, and freedom from recurrence (disease free survival) are illustrated in Figure 2. All first recurrences were evident by 15 years, with the majority evident within the first 5 years (Fig. 3). Because this study encompassed a long period, the potential influence of treatment era on outcome was evaluated for patients who were treated in the 1960s, 1970s, 1980s, and 1990s. For all endpoints (local, metastatic, and survival), there were no significant differences according to the time of treatment (data not shown).
Prognostic Factors for Local Control
The results of a univariate analysis of factors that potentially affected local control are presented in Table 2. In that analysis, seven factors were correlated significantly with the likelihood of local control, and the same seven factors were independently significant in a multivariate proportional hazards regression. In order of decreasing significance (favorable feature first), the factors were: final resection margin (negative vs. uncertain/positive), tumor location (extremity/superficial trunk vs. head and neck/deep trunk), presentation (primary vs. locally recurrent), patient age (≤ 64 years vs. > 64 years), histopathology (all others vs. MFH/neurogenic/epithelioid), tumor size (≤ 10 cm vs. > 10 cm), and tumor grade (low and intermediate vs. high). The relative risks, 95%CIs and P values for these factors in the regression model were as follows: The factor resection margin had a relative risk of 2.5 (95%CI, 1.9–3.3; P < 0.001), tumor location had a relative risk of 2.6 (95%CI, 1.8–3.6; P < 0.001), presentation had a relative risk of 2.2 (95%CI, 1.6–3.0; P < 0.001), age had a relative risk of 1.8 (95%CI, 1.3–2.5; P < 0.001), histopathology had a relative risk of 1.7 (95%CI, 1.2–2.3; P = 0.001), tumor size had a relative risk of 1.7 (95%CI, 1.2–2.4; P = 0.002), and tumor grade had a relative risk of 1.5 (95%CI, 1.1–2.2; P = 0.013). The effects of margin status, tumor site, presentation, and age on actuarial local control rates are illustrated in Figures 4 and 5. The selection of locally adverse histopathologies was based on univariate results. For patients with MFH, the 5-year and 15-year actuarial local control rates were 78% and 75%, respectively; the corresponding results for patients with neurogenic sarcoma were 76% and 71%, and the results for patients with epithelioid sarcoma were 65% and 65%. These results were did not differ significantly from one another. The local control rates for patients with leiomyosarcoma at 5 years and 15 years were 90% and 90%, respectively; the corresponding rates for patients with synovial sarcoma were 88% and 84%; for patients with liposarcoma, the corresponding rates were 85% and 78%; for patients with fibrosarcoma, the corresponding rates were 83% and 79%; for patients with rhabdomyosarcoma, the corresponding rates were 80% and 80%; and, for patients with unclassified sarcoma, the corresponding rates were 83% and 83%. These differences were not significantly different from one another but were significantly superior compared with the results from the first three histotypes. The selection of 64 years as an age cut-off point was suggested by recursive partitioning. Actuarial local control rates at 5 years and 15 years by age decade were as follows: age ≤ 10 years (n = 14 patients), 92% of patients at both 5 years and 15 years; age > 10 years and ≤ 20 years (n = 92 patients), 81% and 79% of patients at 5 years and at 15 years, respectively; age > 20 years and ≤ 30 years (n = 169 patients), 88% and 85% of patients at 5 years and 15 years, respectively; age > 30 years and ≤ 40 years (n = 182 patients), 83% and 78% of patients at 5 years and 15 years, respectively; age > 40 years and ≤ 50 years (n = 222 patients), 86% and 84% of patients at 5 years and 15 years, respectively; age > 50 years and ≤ 60 years (n = 226 patients), 85% and 81% of patients at 5 years and 15 years, respectively; age > 60 years and ≤ 70 years (n = 200 patients), 76% and 70% of patients at 5 years and 15 years, respectively; and age > 70 years (n = 120 patients), 71% and 67% of patients at 5 years and 15 years, respectively. The last 2 decades were associated with significantly poorer local control rates compared with all other decades. With respect to tumor size, the following were 5-year and 10-year actuarial control rates using 2-cm, 5-cm, and 10-cm cut-off points: ≤ 2 cm, 84% and 83% at 5 years and 10 years, respectively; > 2 cm and ≤ 5 cm, 84% and 82% at 5 years and 10 years, respectively; > 5 cm and ≤ 10 cm, 83% and 79% at 5 years and 10 years, respectively; and > 10 cm, 77% and 73% at 5 years and 10 years, respectively. There were no differences in local control rates according to location of tumor in the extremities. Patients who had lesions of the proximal upper extremity (shoulder and upper arm) had 5-year and 15-year control rates of 81% and 80%, respectively; and patients who had lesions of the distal upper extremity (elbow and distal) had 5-year and 15-year control rates of 80% and 77%, respectively (P = 0.772). Patients who had lesions of the proximal lower extremity (groin and thigh) had 5-year and 15-year control rates of 88% and 85%, respectively; and patients who had lesions of the distal lower extremity (knee and distal) had 5-year and 15-year control rates of 84% and 79%, respectively (P = 0.084). The 5-year and 15-year local control rates for 99 patients who received concomitant XRT and chemotherapy (82% and 82%, respectively) were no different from the control rates for patients who did not receive concomitant chemotherapy (83% and 79%, respectively; P = 0.993). For the 601 patients who had disease that fell into a relatively favorable group of extremity or superficial trunk tumors with no prior recurrence and with negative resection margins, the 5-year and 15-year local control rates were 91% (95%CI, 88–93%) and 89% (95%CI, 86–92%), respectively. At the unfavorable end of this spectrum, there were 56 patients who had tumors of the head and neck or the deep trunk with no prior recurrence but with positive or uncertain resection margins. The 5-year and 15-year local control rates for this group were 55% (95%CI, 40–70%) and 51% (95%CI, 35–67%), respectively.
|Factor||5-yr control (%)||15-yr control (%)||P valuea|
|H/N and deep trunk||68||64||<0.001|
|Extremity and superf trunk||85||82|
|Low and intermediate||88||85||0.003|
|Prior local recurrence|
|S XRT S||83||80||0.687|
Prognostic Factors for Metastatic Recurrence
Table 3 summarizes the univariate analysis of factors that potentially affected metastatic recurrence. Although tumor location, treatment sequence, the use of adjuvant chemotherapy, tumor size, tumor grade, and histopathology each correlated significantly with metastatic recurrence in univariate analysis, the effect of the first three of these variables on metastatic outcome was entirely explicable by their correlation with tumor size, grade, and histopathology. In multivariate analysis, intermediate versus low tumor grade had a relative risk of 8.9 (95%CI, 1.2–64.1; P < 0.001), and high versus low tumor grade had a relative risk of 22.5 (95%CI, 3.2–160.1; P < 0.001); tumor size (> 5 cm vs. ≤ 5 cm) had a relative risk of 2.9 (95%CI, 2.3–3.8; P < 0.001); and histopathology (unfavorable vs. favorable) had a relative risk of 1.4 (95%CI, 1.1–1.7; P = 0.006) that was correlated independently with metastatic recurrence. The effects of tumor grade and size are illustrated in Figure 6. For intermediate-grade tumors, there was no evidence that a finer subdivision of size was correlated better with metastatic recurrence compared with the traditional 5-cm cut-off point: The 5-year and 15-year freedom from metastasis rates for tumors measuring ≤ 2 cm, > 2 cm and ≤ 5 cm, > 5 cm and ≤ 10 cm, and > 10 cm were 95% and 95%, 92% and 88%, 79% and 77%, and 80% and 74%, respectively, and only the differences between tumors measuring ≤ 5 cm and tumors measuring > 5 cm were significant (P = 0.003). For high-grade tumors, however, there was an increasing metastatic gradient from small to large, and the 5-year and 15-year freedom from metastasis rates for tumors measuring ≤ 2 cm, > 2 cm and ≤ 5 cm, > 5 cm and ≤ 10 cm, and > 10 cm were 94% and 88%, 75% and 72%, 57% and 51%, and 42% and 38%, respectively; and each of the successive differences was significant (P < 0.01). The selection of four unfavorable, highly metastatic histotypes—leiomyosarcoma, synovial sarcoma, neurogenic sarcoma, rhabdomyosarcoma, and epithelioid sarcoma—was based on univariate analysis. For patients with leiomyosarcoma, the 5-year and 15-year freedom from metastasis rates were 64% and 57%, respectively; for patients with synovial sarcoma, the respective rates were 62% and 53%; for patients with neurogenic sarcoma, the respective rates were 63% and 59%; for patients with rhabdomyosarcoma, the respective rates were 59% and 55%; and, for patients with epithelioid sarcoma, the respective rates were 64% and 64%. These results did not differ significantly among one another. Among patients with less metastatic histotypes, the 5-year and 15-year freedom from metastasis rates were as follows: MFH, 71% and 69%, respectively; liposarcoma, 81% and 78%, respectively; fibrosarcoma, 80% and 77%, respectively; and unclassified sarcoma, 72% and 63%, respectively. In this group, liposarcoma was significantly less metastatic, but this difference disappeared when analysis was confined to patients with intermediate-grade or high-grade tumors only.
|Factor||5-yr metastatic control (%)||15-yr metastatic control (%)||P valuea|
|H/N and deep trunk||66||61||0.285|
|Extremity and superf trunk||71||67|
|Intermediate grade and tumor size|
|High grade and tumor size|
|Prior local recurrence|
|S XRT S||74||69||0.058|
Prognostic Factors for Lymph Node Recurrence
All lymph node recurrences occurred within 5 years; and, of the factors listed in Tables 1 and 2, only tumor grade was correlated with lymph node recurrence. The 5-year rates for freedom from lymph node metastasis were 99% and 96% for patients who had low-grade and intermediate-grade tumors compared with patients who had high-grade tumors, respectively (P = 0.003). However, the actuarial incidence of lymph node recurrence was very low (≤ 3%) among patients with all histologic types except for rhabdomyosarcoma (17% at 5 years), epithelioid sarcoma (18% at 5 years), and clear cell sarcoma (15% at 5 years). These 3 histotypes (lymphogenous histotypes) accounted for 96 patients with an overall 5-year freedom from lymph node metastasis rate of 82% (95%CI, 73–89%), whereas the remaining 1129 patients had a freedom from lymph node recurrence rate of 98% (95%CI, 97–99%; P < 0.001). In multivariate analysis, only histotype was independently significant, with a relative risk of 9.3 (95%CI, 4.8–18.0; P < 0.001).
Prognostic Factors for Disease Free Survival
In univariate analysis, the factors that were significant for local recurrence (Table 2) and metastatic recurrence (Table 3) also were significant for disease free survival (freedom from recurrence; data not shown). On multivariate regression analysis, seven factors were correlated independently with disease free survival: In order of decreasing significance, these were (favorable feature first): tumor grade (low vs. intermediate vs. high), tumor size (≤ 5 cm vs. > 5 cm), resection margin (negative vs. positive/uncertain), presentation (no prior local recurrence vs. prior local recurrence), tumor site (extremity/superficial trunk vs. head and neck/deep trunk), patient age (≤ 64 years vs. > 64 years), and histotype (all others vs. lymphogenous). The relative risks, 95%CIs and P values for these factors in the regression model were: Intermediate versus low tumor grade had a relative risk of 2.0 (95%CI, 1.1–4.1; and high versus low tumor grade had a relative risk of 4.5 (95%CI, 2.2–9.0; P < 0.001); tumor size had a relative risk of 2.2 (95%CI, 1.8–2.7; P < 0.001); resection margin had a relative risk of 1.6 (95%CI, 1.3–1.9; P < 0.001); presentation had a relative risk of 1.6 (95%CI, 1.3–2.0; P < 0.001); tumor site had a relative risk of 1.5 (95%CI, 1.2–1.9; P = 0.001); patient age had a relative risk of 1.4 (95%CI, 1.1–1.7; P = 0.003); and histotype had a relative risk of 1.4 (95%CI, 1.0–1.9; P = 0.048).
Prognostic Factors for Disease Specific Survival
Three hundred seventy-two patients died of sarcoma (including 1 fatal therapeutic complication). Among these 372 patients, 326 patients developed metastatic recurrences (88%), but 46 patients died of local disease alone (n = 37 patients), lymph node disease alone (n = 4 patients), local and lymph node disease (n = 4 patients), or complications (n = 1 patient). Thirty-one of 46 patients who died without metastases had primary tumors that arose in the head and neck or the deep trunk. The results of a univariate analysis of factors that potentially were correlated with disease specific survival are summarized in Table 4. Similar to the development of metastatic recurrence, several factors that were related significantly to disease specific survival exerted this effect by virtue of their correlation with other significant factors. Multivariate analysis revealed that six factors were correlated independently with disease specific survival: in order of decreasing significance (favorable feature first), these were tumor grade (low vs. intermediate vs. high), tumor size (≤ 5 cm vs. > 5 cm), tumor site (extremity and superficial trunk vs. head and neck and deep trunk), histopathology (others vs. lymphogenous), patient age (≤ 64 years vs. > 64 years), and resection margin (negative vs. positive/uncertain). The relative risks, 95%CIs, and P values for these factors in the regression model were as follows: For the factor tumor grade, intermediate versus low tumor grade had a relative risk of 4.6 (95%CI, 1.1–19.0), and high versus low tumor grade had a relative risk of 12.1 (95%CI, 3.0–48.8; P < 0.001); tumor size had a relative risk of 2.9 (95%CI, 2.3–3.7; P < 0.001); tumor site had a relative risk of 1.9 (95%CI, 1.4–2.4; P < 0.001); histotype had a relative risk of 1.6 (95%CI, 1.1–2.2; P = 0.013); patient age had a relative risk of 1.4 (95%CI, 1.1–1.8; P = 0.018); and resection margin had a relative risk of 1.3 (95%CI, 1.0–1.6; P = 0.025).The actuarial 5-year and 15-year disease specific survival rates for the major histopathologic types were as follows: MFH, 73% and 68%; liposarcoma, 80% and 72%; unclassified sarcoma, 74% and 63%, respectively; synovial sarcoma, 76% and 55%, respectively; leiomyosarcoma, 70% and 53%, respectively; neurogenic sarcoma, 64% and 61%, respectively; rhabdomyosarcoma, 52% and 46%, respectively; fibrosarcoma, 77% and 74%, respectively; epithelioid sarcoma, 74% and 64%, respectively; and clear cell sarcoma, 60% and 48%, respectively. Although patients with tumors of the head and neck appeared to fare better compared with patients who had tumors arising deep in the trunk (Table 4), this was not borne out by multivariate analysis, because trunk tumors were significantly larger compared with tumors that arose above the clavicles. The influence of tumor size and grade are illustrated in Figure 7. These same factors, except for age, also were significant for overall uncorrected survival (data not shown).
|Factor||5-yr disease-specific survival (%)||15-yr disease-specific survival (%)||P valuea|
|H/N and deep trunk||58||47||<0.001|
|Extremity and superf trunk||76||67|
|Intermediate grade and size|
|High grade and size|
|Prior local recurrence|
|S XRT S||81||65||0.090|
STS tends to be an aggressive and rapidly progressive disease. Figure 3 shows that recurrences, if they are to occur, tend to be early, mostly within the first 5 years. Although the phenomenon of late recurrence (> 5 years) is real,16, 29 it should not be overemphasized, because its magnitude is small—in this study, 36 patients (7% of whom developed recurrent disease) experienced their first recurrence beyond 5 years, and only 6 patients had their first recurrence beyond 10 years. Conversely, disease specific death may be delayed: In our study, the disease specific survival rate fell from 73% at 5 years to 63% at 15 years, and a downward trend in the actuarial curve was evident even up to 25 years (Fig. 1). Two patients died of their sarcoma more than 20 years after initial treatment. The most common failure pattern was metastatic recurrence, followed by local recurrence (Fig. 2). Lymph node recurrence was uncommon, with an actuarial incidence of only 3%. Overall, by 15 years, nearly 50% of our patients (45%) had experienced at least one episode of disease recurrence.
Because the main goal of this analysis was to elucidate potentially significant disease and patient factors, no attempt was made to comprehensively analyze treatment effects. However, as it turned out, there were no significant differences in outcome by any of the overall treatment parameters. The potential influence of factors such as reexcision of tumor beds,29 the utility of preoperative XRT versus postoperative XRT,30 and the significance of XRT dose31–33 will be reported separately. It should be reiterated that this series included relatively few patients with STS arising in the deep trunk, particularly in the retroperitoneum, because it was not general policy to offer adjuvant XRT to patients with such disease.
Our analysis revealed that different disease endpoints were influenced significantly by different factors, as reported previously by other investigators.5, 18, 19 Local recurrence was related to microscopic resection margin, tumor location, locally recurrent presentation, patient age, histopathologic subtype, tumor size, and tumor grade. The status of the resection margin is recognized generally as an important determinant of local control.2, 5, 10, 11, 16, 17, 34 In our study, a significant proportion of patients (19%) had uncertain margins due to the practice of offering XRT instead of reexcision when the latter may compromise function. Unless the pathologic report actually evaluates and stipulates that the resection margin as negative, the patient must be regarded as being at increased risk for local recurrence. The issue of reresection for positive or uncertain margins is addressed in a separate report.35
The second most significant determinant of local recurrence was tumor location. Patients who had lesions of the head and neck and the deep trunk had significantly higher recurrence rates compared with patients who had lesions arising in the superficial trunk or extremity. This also has been observed by others.7–9 Our data strongly suggest that patients with STS of the extremities and superficial trunk can be placed into one group when reporting local control rates. Overall local control rates for STS of the extremity and superficial trunk combined were 85% and 81% at 5 years and 15 years, respectively, and were significantly superior to those for STS of the head and neck and the deep trunk (68% and 64%, respectively) (Table 2). The precise reason for higher recurrence rates among patients with STS of the head and neck and the deep trunk is not entirely clear. It is conceded generally that STS in these locations cannot be excised as adequately as STS that arises in other sites; and, in our series, the incidence of positive or uncertain resection margins in these sites was 43% (77 of 181 patients), compared with 33% (341 of 1044 patients) with tumors at all other sites (P = 0.010). However, this cannot be the entire explanation, because disease site was significant in multivariate analysis in which, ostensibly, adjustment was made for margin status. However, the assessment of resection margins at these sites generally is problematic due to large tumor size and piecemeal removal—even microscopically negative margins may not be truly negative. Another factor that contributes to recurrence may be the inability to adequately irradiate regions where critical dose-limiting structures abound.
The third most significant determinant of local recurrence was prior recurrence. This also has been reported as significant in other analyses.5, 18, 19 It must be understood that this group of locally recurrent tumors is highly selected: local recurrence in the absence of distant metastatic recurrence. It seems most plausible that an STS that has recurred once is demonstrating its particular pathobiologic propensity. Indeed, there appeared to be a trend toward higher local recurrence rates with greater numbers of previous recurrences (Table 2). Mechanistically, it is reasonable to suppose that these tumors are associated with more residual clonogens and that higher XRT doses may be beneficial. An analysis of this treatment parameter will be reported separately.
Patient age has been reported as a significant determinant of local control in relatively few series.5, 9, 14 One group has consistently reported an age of 50 years as their cut-off point,5, 14 but it is not clear how that age was selected. In our series, recursive partitioning suggested a cut-off point in the middle 60s. Actuarial analysis (Table 2) supports the use of an age of approximately 64 years. Apart from a significantly higher incidence of MFH among patients age > 64 years (62%; 151 of 244 patients) compared with patients age ≤ 64 years (32%; 309 of 981 patients; P < 0.001), there appeared to be no significant tumor differences. The difference in pathologic subtype could not explain the age affect, because age remained significant in multivariate analysis with pathology included. For the present, this age effect on local control remains inexplicable. Three histopathologic subtypes—MFH, malignant peripheral nerve sheath tumor (neurogenic sarcoma), and epithelioid sarcoma—were associated with local recurrence. Other studies have found that malignant peripheral nerve sheath tumors were adverse for local control.5 The influence of tumor subtype likely is mediated by differences in local growth patterns. The influence of tumor size on local control was relatively weak and was evident only when a cut-off size of 10 cm was used. Very small tumors (≤ 2 cm) were as likely to recur as tumors of moderate size (> 5 cm and ≤ 10 cm). Most reports have found that tumor size did not have a significant influence on local control.5, 7, 8, 14, 35, 36 It is likely that the relatively large size of our sample revealed some factors that are significant statistically but relatively unimportant clinically. The same may apply to tumor grade. Although low-grade and intermediate-grade tumors had significantly lower recurrence rates compared with high-grade tumors, the difference was not large. Most reports have failed to find that tumor grade was a significant determinant of local recurrence.5, 7, 14, 35 The idea that low-grade sarcomas are particularly unlikely to recur is erroneous.
Unlike local control, which was affected by many factors, metastatic recurrence was determined by only three independent factors: tumor grade, tumor size, and histopathology. The retrospective nature of this review precluded an analysis of tumor depth, which was not recorded consistently and cannot be reconstructed reliably. From other reports,5, 11, 14 it is certain that tumor depth relative to the investing fascia is a fourth prognostic factor. The significance of tumor grade and size is well-recognized. Grading systems, however, are controversial. The AJCC classification system15 uses a four-tier system but amalgamates it into two tiers for the purposes of stage grouping—there is no allowance for an intermediate grade. At our institution, a three-tier grading system long has been in use,1, 21 and the current analysis supports this grading system. Intermediate-grade STS, such as myxoid MFH and myxoid liposarcoma, are clinically useful designations for tumors that have a higher metastatic propensity than atypical lipomatous tumors or well-differentiated fibrosarcomas but that have distinctly less metastatic potential compared with pleomorphic MFH, pleomorphic liposarcoma, synovial sarcoma, or rhabdomyosarcoma. Tumor size was a significant determinant of metastatic outcome for patients with intermediate-grade and high-grade STS. Although the traditional size cut-off point is 5 cm, very small tumors (≤ 2 cm) had significantly less metastatic potential compared with tumors of intermediate size (> 2 cm and ≤ 5 cm). Even large (> 10 cm) intermediate-grade tumors had a metastasis rate of only 26% at 15 years, whereas large (> 10 cm) high-grade tumors had a 62% 15-year metastasis rate. Data on the influence of histopathologic subtype on metastatic outcome are provided rarely, although it is part of general oncologic knowledge that some STS tumors (rhabdomyosarcoma, Ewing sarcoma, and primitive neuroectodermal tumor) are highly metastatic. In our series, five histologic subtypes emerged as especially metastatic: leiomyosarcoma, synovial sarcoma, rhabdomyosarcoma, malignant peripheral nerve sheath tumor, and epithelioid sarcoma. We had too few patients to evaluate Ewing sarcoma and primitive neuroectodermal tumor. Leiomyosarcoma has been reported as a particularly metastasizing tumor5 as well as synovial sarcoma.7 The significance of histopathologic subtype of STS suggests the need to include this parameter in treatment decision trees.
Lymph node recurrence is uncommon in patients with STS37 and occurred with an incidence rate of 3% in our series. Three histopathologic subtypes—rhabdomyosarcoma, epithelioid sarcoma, and clear cell sarcoma—were more prone to lymph node spread compared with all other subtypes. The propensity for lymph node metastasis by these three histotypes has been well documented, but synovial sarcoma also often is included.37 We found no evidence in this analysis or in a previous study38 that synovial sarcoma spreads to the lymph nodes with any significant frequency. A similar lack of lymph node spread from this subtype was reported in at least one other study.39
Because disease free survival reflects freedom from local, lymph node, and metastatic recurrence, factors that are significant for each of those endpoints were expected to play a role in this endpoint. Moreover, because metastasis was the most common recurrence pattern, factors that are significant for metastatic recurrence were expected to be most significant. This was borne out by our analysis, and seven factors were correlated independently with disease free survival: tumor grade, tumor size, microscopic resection margin, presentation, tumor site, patient age, and histologic subtype. The histopathologic subtype that was correlated best with disease free survival was lymphogenous, doubtless because each of the three subtypes included were high grade as well as metastasizing to lymph nodes.
Finally, we considered disease specific survival. This endpoint reflects the effects of disease recurrence and the outcome of salvage therapy. Because the majority of deaths from STS are associated with metastatic recurrence, factors that determine this recurrence are likely to influence disease specific survival—an effect that is magnified by the relatively poor salvage of patients who have metastatic recurrences compared with patients who have local recurrences. Outcome after first recurrence and salvage therapy in this series will be reported separately; however, it should be noted that a small but significant fraction of patients who died from STS (46 of 372 patients; 12%) did so with only local recurrence. Thus, some of the factors that affect local recurrence are likely to influence disease specific survival as well. In multivariate analysis, the independent factors that affected disease specific survival were tumor grade, tumor size, tumor site, histopathology, patient age, and resection margins. The factors tumor grade, size, and histopathology (lymphogenous vs. other) exerted their effect largely through their influence on metastatic recurrence; tumor site, patient age, and resection margins exerted their effect through predisposing to local recurrence. One other large study found that tumor grade, tumor size, tumor depth, tumor site, tumor subtype, and patient age were significant determinants of disease specific survival in patients with STS.12
In conclusion, STS comprises a heterogeneous group of tumors with diverse outcomes determined by numerous factors. Local recurrence, metastatic recurrence, and lymph node recurrence are influenced by different, largely nonoverlapping variables (as summarized in Table 5). Disease free survival and disease specific survival are influenced by a conglomerate of factors that exert their effects by way of predisposing to various recurrence patterns. Treatment strategies should be designed to take into account the various prognostic factors; although, in a disease as uncommon as STS, it will likely require multiinstitutional studies to evaluate appropriate treatment for different patient subgroups.
|Resection margin positive or uncertain|
|Tumor of head and neck or deep trunk|
|Locally recurrent tumor|
|Patient age >64 yrs|
|MFH, neurogenic, epithelioid sarcoma|
|Tumor >10 cm|
|High grade tumor|
|Large size (>5 cm)|
|Adverse histotype (leiomyosarcoma, synovial sarcoma, neurogenic sarcoma rhabdomyosarcoma, epithelioid sarcoma)|
|Lymph node recurrence|
|Rhabdomyosarcoma, epithelioid sarcoma, clear cell sarcoma|
|Large size (>5 cm)|
|Tumor of the head and neck or deep trunk|
|Rhabdomyosarcoma, epithelioid sarcoma, clear cell sarcoma|
|Patient age >64 yrs|
|Resection margin positive or uncertain|
- 15American Joint Committee on Cancer. AJCC cancer staging manual. 6th ed. New York: Springer, 2002.
- 20Association of Directors of Anatomic and Surgical Pathology. Recommendations for the reporting of soft tissue sarcoma. Hum Pathol. 1999; 30: 3–7.
- 22Soft tissue sarcoma. In: FletcherGH, editor. Textbook of radiotherapy, 3rd ed. Philadelphia: Lea and Febiger, 1980: 922–942..
- 24Applied survival analysis. regression modeling of time to event data. New York: John Wiley & Sons, Inc., 1999., .
- 25Classification and regression trees. Chapman and Hall: Boca Raton, 1998., , , .
- 26Practical statistics for medical research. London: Chapman and Hall, 1991..
- 28U.S. Department of Health and Human Services. Vital statistics of the United States, 1984. Hyattsville: U.S. Public health Service, National Center for Health Statistics, 1988.
- 32Management of extremity soft tissue sarcomas with limb-sparing surgery and postoperative irradiation: do total dose, overall treatment time, and the surgery-radiotherapy interval impact on local control? Int J Radiat Oncol Biol Phys. 1995; 32: 969–976., , , et al.
- 36Preoperative radiotherapy in the treatment of soft tissue sarcomas. Clin Orthop. 2002; 397: 177–189., , , , , .