Fax: (713) 563-4637
Computed tomography scan-driven selection of treatment for retroperitoneal liposarcoma histologic subtypes
Article first published online: 20 JAN 2009
Copyright © 2009 American Cancer Society
Volume 115, Issue 5, pages 1081–1090, 1 March 2009
How to Cite
Lahat, G., Madewell, J. E., Anaya, D. A., Qiao, W., Tuvin, D., Benjamin, R. S., Lev, D. C. and Pollock, R. E. (2009), Computed tomography scan-driven selection of treatment for retroperitoneal liposarcoma histologic subtypes. Cancer, 115: 1081–1090. doi: 10.1002/cncr.24045
- Issue published online: 18 FEB 2009
- Article first published online: 20 JAN 2009
- Manuscript Accepted: 17 JUN 2008
- Manuscript Revised: 13 JUN 2008
- Manuscript Received: 7 MAY 2008
- retroperitoneal liposarcoma;
- atypical lipomatous tumor;
- well differentiation;
Well differentiated (WD) and dedifferentiated (DD) retroperitoneal liposarcoma (RPLS) have distinct biologic behaviors. Consequently, the therapeutic approaches for these tumors differ and mandate an accurate preoperative diagnosis. The authors of this report evaluated whether computed tomography (CT) can be used to differentiate between WD and DD RPLS.
Imaging studies (CT, magnetic resonance imaging, and positron emission tomography-CT) from 78 patients with RPLS who underwent surgery at the University of Texas M. D. Anderson Cancer Center (UTMDACC) between 2001 and 2007 were reviewed by a senior bone and soft tissue sarcoma radiologist who was blinded to the final histopathologic diagnosis. A focal nodular/water density area within an RPLS was interpreted as a marker suggestive of DD. Correlations between imaging diagnosis, histology, and clinical outcome were analyzed.
The study radiologist identified 60 RPLS as DD and 17 RPLS as WD. A radiologic diagnosis of a WD was correlated with preoperative biopsy and postoperative histology in all patients (100%). Focal nodular/water density was a very sensitive marker of DD (97.8%); however, it had relatively low specificity (51.5%). Sixteen WD RPLS (48.5%) contained focal nodular/water density areas, leading to their misdiagnosis as DD; half of those tumors had hypercellular WD. Of 78 preoperative biopsies, 22 (28.2%) were performed at UTMDACC under CT guidance. Preoperative histologic diagnoses obtained from 12 biopsies derived from focal nodular/water density areas were confirmed as unchanged on final pathology; whereas, in 50% of biopsies that were not taken from a suspicious area, DD histology was misdiagnosed as WD.
When CT features are suggestive of WD, no further diagnostic tests are needed for tumor characterization. Moreover, CT can accurately identify most DD, thereby rendering their under-treatment unlikely; however, a CT-guided biopsy is needed to differentiate between DD and WD RPLS that contain focal nodular/water density areas, thereby avoiding their over treatment. Cancer 2009. © 2009 American Cancer Society.
Soft tissue sarcoma (STS) is a rare disease that constitutes less than 1% of all human cancers.1 Of all adult STS histologic subtypes, liposarcoma (LPS) is the most common, accounting for approximately 15% of all STS2; retroperitoneal LPS (RPLS) is the most frequent histologic subtype of sarcoma in this anatomic locus3 (40%). LPS are a heterogeneous group of tumors composed of several histologic subtypes,4 including pleomorphic, myxoid, well differentiated (WD) (also known as atypical lipomatous tumor [ALT]), and dedifferentiated (DD). WD and DD are the most common RPLS subtypes, accounting for 90% of LPS in the retroperitoneum.5, 6
DD was described originally by Evans as an LPS that consists of a combination of atypical lipomatous tumor components and cellular, nonlipogenic, sarcomatous areas that have significant mitotic activity.7 It is unclear whether WD and DD originate from 2 different cellular clones or whether there is a process of progressive evolution from WD to DD; however, it is unequivocally clear that DD is a high-grade tumor and, as such, it has a much worse prognosis than WD because of its propensity to metastasize.3, 8-10 Taking into account the markedly different biologic behaviors of WD versus DD, our therapeutic approach at the University of Texas M. D. Anderson Cancer Center (UTMDACC) is different and reflects these distinctive clinical behaviors. For example, patients who have DD RPLS are treated more frequently with neoadjuvant chemotherapy than patients who have WD RPLS in light of the markedly greater incidence of metastatic activity in the former tumors. Recently, we also demonstrated that patients who have WD tumors frequently benefit from less aggressive surgical approaches that avoid unnecessarily extensive resection, in contrast to our surgical management of tumors with other retroperitoneal histologies.11
In light of these different treatment strategies for WD and DD, an accurate preoperative diagnosis is critical to successful multidisciplinary therapy for these patients. The diagnostic approach for RPLS currently implemented at UTMDACC is based on preoperative computed tomography (CT) scanning and core-needle biopsy: This approach mandates tissue diagnosis in all patients to plan appropriate therapy.
The objective of the current study was to evaluate this diagnostic approach and to validate the accuracy of CT scanning in differentiating WD from DD RPLS. We also sought to propose an algorithm for the diagnosis and treatment of RPLS.
MATERIALS AND METHODS
Eighty-eight patients with RPLS who underwent surgery at UTMDACC between July 2001 and July 2007 were identified using an institutional STS database. RPLS histologies were confirmed by a UTMDACC STS pathologist. Seventy-eight patients who had pretreatment digital CT scans available had a final postoperative pathologic diagnosis of WD or DD; these individuals were included in this study, and their medical records formed the basis of this retrospective review. All patients underwent preoperative biopsy (22 biopsies were conducted at the UTMDACC, and 56 biopsies were conducted at other locations); the results from all biopsies were recorded in our study database.
All CT scans were obtained originally with a multichannel, multidetector CT scanner using a 5-mm or 2.5-mm slice thickness (several earlier CT scans had 7-mm slice thickness). Examinations were performed with intravenous contrast injection of 125 to 150 mL of contrast media (Omnipaque 350, Nycomed Amersham; later, Optiray 350, Mallinckrodt) at a rate of 3 mL per second. An abdominal postcontrast CT was obtained with an approximate delay of 60 seconds. Patients received oral barium sulfate suspension (900 mL) for bowel contrast and rectal barium sulfate suspension as needed. For the purpose of this study, a board-certified radiologist with specialized interest and expertise in soft tissue tumors (J.E.M.) who was blinded to the final histopathologic diagnosis reviewed the preoperative CT scans. An initial review of representative cases was performed and vetted by other sarcoma specialists. Axial and reformatted images were reviewed on a PACS workstation (iSite; Stentor Inc., Brisbane, Calif) with soft tissue window/level (500 window width/55 window level).
CT features that previously were reported as possible discriminators between low-grade tumors versus high-grade tumors were evaluated.12-14 The study radiologist recorded the largest axial tumor dimension. Tumors were subcategorized by the study radiologist according their fat content using a scale that ranged from 1 to 5, in which 1 represented <5% fat, 2 represented 6% to 25% fat, 3 represented 25% to 50% fat, 4 represented 50% to 75% fat, and 5 represented tumors that were >75% fat. The tumor margin radiologic appearance was characterized by the study radiologist as either smooth or irregular. A tumor was evaluated as ‘infiltrating’ if the study radiologist determined that a clear line of demarcation between an organ and the tumor was absent. Major vessels were considered to be involved when either encasement and/or infiltration were observed by the study radiologist. Tumors were categorized as ‘heterogeneous’ when nonadipose elements were detectable on the CT scan. Specific radiologic tumor descriptors were defined as follows: focal nodular/water density was defined as a nodular area within the tumor with a muscle density (Fig. 1a), ground-glass opacities were more dense than fat but less than water or muscle, and hypervascularity was used to describe areas of higher density compared with muscle (Fig. 1b). Hypodense ‘cystic’ areas within the tumor that were less dense than water or muscle, were not fat, and did not enhance were considered necrotic areas. Calcifications were diagnosed using a noncontrast scan; septations or ‘streakiness’ were considered as present when thin septa of uniform thickness were detectable within the tumor.
Twenty-two CT-guided biopsies that were taken at UTMDACC were analyzed. Core biopsies were taken using 14-gauge, side-cut needles. A pathologist reviewed and confirmed the adequacy of all tissue samples that were included in the analysis.
The associations between consolidated histology and each CT feature were evaluated using a generalized Fisher exact test to determine P values for categorical variables and the Wilcoxon test to determine P values for continuous variables.15 Sensitivity and specificity were calculated for a comparison of consolidated histology with each feature. Multivariate logistic regression was used to identify the combination of CT features that was most predictive of a DD diagnosis using stepwise selection. Kaplan-Meier product-limit survival probability estimates of progression-free survival (PFS) and overall survival (OS) were calculated,16 and log-rank tests17 were performed to compare PFS and OS between simple noncellular WD and the hypercellular WD variant.
Description of Study Population
Seventy-eight patients with WD and DD RPLS who underwent resection at UTMDACC between July 2001 and July 2007 were included; their preoperative CT scans formed the basis of this study evaluation. Of the 78 patients who were included in the study, there were 52 men and 26 women with a median age of 61.5 years (range, 28-86 years). All patients had a preoperative CT scan, 9 patients (11.5%) also had a magnetic resonance imaging (MRI) study, and 4 patients (5.1%) also had a positron emission tomography (PET) scan. Sixty-seven tumors were retroperitoneal (85.9%), 6 tumors were retroperitoneal and mesenteric (7.7%), and 5 tumors were mesenteric only (6.4%).
Histologic Assessment Based on CT Interpretation
Blinded to the histology of each tumor, an STS radiology specialist (J.E.M.) analyzed the preoperative CT scans from all 78 patients; a focal nodular/water density area within a lipomatous mass was interpreted as a marker suggestive of DD (Fig. 1a). The study radiologist identified 60 such cases (76.9%), which he diagnosed as DD; 17 tumors (21.8%) that lacked a focal nodular/water density area were diagnosed as WD, and 1 tumor (1.3%) was inconclusive. CT images of the latter tumor demonstrated a suspicious area that appeared as a dense ground-glass nodule, a feature that may be seen in WD; however, the density in this specific instance was greater, and the study radiologist stated that the area should be biopsied. Other CT features also were analyzed, scored, and correlated with the postoperative histologic diagnosis.
Correlation Between Radiologic Findings and Histologic Subtype
We evaluated whether various CT scan features that previously were reported as distinctive of the RPLS subtype12-14 were correlated with postoperative histologic subtype determinations; all features that we used in this evaluation are presented in Table 1. Eleven patients with DD tumors (24.4%) and 3 patients with WD tumors (9.1%) had irregular margins (P = .13). According to the impression of the radiologist, 34 of 45 DD tumors (75.6%) and 17 of 33 WD tumors (51.5%) infiltrated adjacent organs (P = .03); the kidney was deemed infiltrated in approximately 33% of these tumors. Focal/nodular water density was observed in 44 of 45 DD tumors (97.8%) and in 16 of 33 WD tumors (48.4%; P <.0001); whereas a ground-glass radiologic appearance was more common in WD tumors (15.1%) compared with DD tumors (4.4%; P = .13). Hypervascularity was observed in 43 DD tumors (95.6%) compared with only 12 WD tumors (36.4%; P <.0001). Cystic/necrotic lesions were observed in 39 DD tumors (86.7%) compared with 17 WD tumors (51.5%; P <.0001). Calcifications were observed in 13 DD tumors (28.9%) compared with 5 WD tumors (15.1%; P = .18). Vascular involvement, heterogeneity, septations, fat content, and size did not differ between WD tumors and DD tumors (Table 1). In this series, the radiologic diagnosis of a RPLS as DD versus WD was based on the presence or absence of focal nodular/water density areas, respectively.
|No. of Patients (%)|
|Variable||DD, n=45||WD, n=33||P*|
|Median tumor size cm [range]||17.2 [4.5-45]||18.4 [6.3-35.5]||.71|
|Average fat content, 1-5†||3.1||3.6|
|Irregular margins||11 (24.4)||3 (9.1)||.13|
|Infiltrated organs (imaging)||34 (75.6)||17 (51.5)||.03|
|Vascular involvement||15 (33.3)||11 (33.3)||.1|
|Heterogeneity||45 (100)||30 (90.1)||.07|
|Focal nodular/water density||44 (97.8)||16 (48.4)||<.0001|
|Ground-glass nodule||2 (4.4)||5 (15.1)||.13|
|Hypervascularity||43 (95.6)||12 (36.4)||<.0001|
|Cystic/necrotic area||39 (86.7)||17 (51.5)||<.0001|
|Septations||41 (91.1)||31 (93.9)||1.00|
|Calcifications||13 (28.9)||5 (15.1)||.18|
Table 2 depicts the correlation between the radiologic diagnosis suggested by the study radiologist and the postoperative histologic diagnosis. Forty-four of 45 patients (97.8%) who had a postoperative histologic diagnosis of DD were identified as such radiologically by the study radiologist based on the presence of this feature. This implies that the sensitivity of a focal nodular/water density area as a marker of DD is high (97.8%); however, because 16 histologically proven WD tumors were diagnosed as DD based on this criterion, its specificity is quite low (51.5%). The positive predictive value (PPV) of a focal nodular/water density area for the prediction of DD histology is 73.3% (Table 3). Remarkably, the negative predictive value (NPV) of this imaging criterion was 100% (Table 3); all 17 tumors that were assessed radiologically as WD based on the lack of a focal nodular/water density area ultimately were demonstrated pathologically to be WD tumors.
|Radiologic Diagnosis||WD Histology||DD Histology|
|Variable||Sensitivity, %||Specificity, %||PPV, %||NPV, %|
|Infiltrated organs (imaging)||75.6||48.5||66.7||59.3|
|Focal nodular/water density||97.8||39.4||73.3||100|
Taken together, these data suggest that CT scanning per se is very sensitive for the diagnosis of DD; however, its specificity is relatively low. Moreover, because the NPV of this test was 100%, it seems that the diagnosis of WD can be based on CT scanning alone.
Hypercellular WD Variant Commonly Is Misinterpreted on CT Scans as DD
Hypercellularity is a pathologic finding that recently was described by Evans as occurring against the backdrop of classic, simple WD18; moreover, the outcome of patients with these tumors is similar to that of patients with simple WD tumors. Compounding this issue, the hypercellular WD variant frequently is misdiagnosed pathologically as DD, with obvious prognostic implications. A high incidence of the hypercellular WD variant was discovered in the group of WD tumors that was radiologically identified incorrectly as DD tumors; 16 patients who had a postoperative pathologic diagnosis of WD had their tumors identified as DD by radiologic criteria; it was demonstrated pathologically that 8 of these 16 tumors (50%) were of the hypercellular WD variant. Conversely, all 17 patients who preoperatively had their tumors identified radiologically as WD had their tumors identified pathologically as simple WD (P < .0001).
The hypercellular WD variant tends to appear as a high-grade tumor on CT scans (Fig. 2); and although, histologically, it also resembles DD, its clinical behavior is consistent with that of WD RPLS. In light of this dichotomy, we evaluated the outcome of patients with hypercellular WD versus simple WD. After a median follow-up of 38 months, recurrence-free survival and overall survival did not differ statistically (P = .09). These results support the previous report by Evans,18 who demonstrated that the long-term outcome of patients who had hypercellular WD and simple WD were equivalent. The differing histologic and radiologic appearance of DD compared with hypercellular WD does not appear to be a point of clinical or therapeutic divergence, in that both simple WD and hypercellular WD seemingly confer similar rates of RPLS survivorship; therefore, by implication, the therapy for these 2 entities should be comparable.
Re-evaluation of Potential CT Scan Markers for RPLS Histologic Subtypes
Margin irregularity, infiltration into adjacent organs, calcification, necrosis (manifesting as intratumoral cysts), and hypervascularity all are radiologic imaging markers of malignant behavior.14 Table 3 depicts the sensitivity and specificity of each parameter. The sensitivity of focal nodular/water density and hypervascularity as markers of DD was quite high (97.8% and 95.6%, respectively); however, the specificity of both markers for DD was relatively low (51.5% and 60%, respectively). Cysts suggestive of necrosis were less sensitive (86.7%) and also had a relatively low specificity (48.5%) as correlates of a DD diagnosis. All other potential markers were significantly less sensitive (Table 3).
We re-evaluated the 60 tumors that contained focal nodular/water density areas, including 44 (73.3%) ‘true’ DD tumors and 16 (26.7%) ‘false’ DD tumors (ie, radiologic DD that turned to be pathologic WD), trying to discriminate further between both groups using radiologic criteria. Multiple nodules or larger areas of focal nodular/water density did not correlate with DD. We used multivariate logistic regression with stepwise selection to identify the best combination of CT features for the diagnosis of DD; the combination of hypervascularity and focal nodular/water density was the only 1 that was recognized by the final model. Using this combination, we observed that all 6 hypovascular tumors that contained an area of focal nodular/water density were WD tumors; therefore, using hypovascularity as an additional criterion of WD, it was possible to increase the number of WD tumors that were diagnosed as such by the radiologist from 17 (51.5%) to 23 (69.7%).
CT-Guided Biopsy for RPLS
Therapy was determined by an image-guided core-needle biopsy for all patients who were included in this study. We reviewed 78 preoperative biopsies; of them, 56 biopsies were taken outside UTMDACC, and only 22 CT scan-guided biopsies were performed at UTMDACC. All 17 RPLS that were identified radiologically as WD by the study radiologist also were diagnosed as WD on the preoperative biopsy final postoperative pathology, suggesting that preoperative biopsy in such tumors may not be needed. Of the 60 RPLS that were identified radiologically as DD based on the presence of a focal nodular/water density area, biopsy showed DD or other high-grade sarcoma in 34 tumors (56.7%) and WD in 26 tumors (43.3%); in 13 of these 26 WD preoperative biopsies (50%), the diagnosis was incorrect, as determined by comparing the biopsy both with the radiologic impression of the study radiologist and with the final postoperative pathology report.
Next, we reviewed the CT-guided biopsy images that were performed at UTMDACC. Twelve of 22 guided biopsies (54.5%) that were performed at UTMDACC were taken from suspicious areas within the tumor, such as focal nodular/water density or hypervascular foci (Fig. 3); of these, all 6 biopsies that were taken from tumors identified as DD in the postoperative pathology report were interpreted as DD or unspecified high-grade sarcoma in the preoperative CT scan-guided biopsy report. Conversely, 9 CT scan-guided biopsies that were taken from a nonsuspicious RPLS area (fatty or ground-glass opacities) were interpreted as WD; whereas 1 such biopsy had inflammatory changes only, and 5 biopsies (50%) were diagnosed as DD on final postoperative pathology analysis. Such preoperative biopsies, mostly taken outside UTMDACC in this series, may explain the high rate (50%) of incorrect WD preoperative diagnoses.
Taken together, these data suggest that a preoperative, CT scan-guided biopsy is highly sensitive and specific for DD when it is taken from an area of focal nodular/water density or from a hypervascular focus. Such CT scan-guided biopsy of suspicious areas in RPLS should be performed to differentiate between DD and WD tumors when the diagnosis is uncertain. In contrast, CT scan-guided biopsy of nonsuspicious RPLS areas may result in an under-identification of DD preoperatively compared with final pathology diagnosis.
Retroperitoneal WD and DD are 2 distinct histologic subtypes of LPS that are treated differently at UTMDACC; patients who have DD tumors frequently are treated with systemic chemotherapy followed by an aggressive surgical approach, whereas patients who have WD tumors usually are treated using somewhat less aggressive surgery alone, avoiding resection of contiguous organs when possible. We recently confirmed that a less aggressive surgical approach does not impair the long-term outcome of patients with WD RPLS.11 Tailoring the appropriate treatment for each patient with RPLS must be based on precise diagnostic assessments.
Several reports have correlated CT imaging findings with specific extremity LPS histologic subtypes19-21; WD in the extremities has been described as a tumor with CT attenuation equivalent to that of fat. These tumors can have thickened, irregular septa and minor nodular components with CT scan attenuation approximating that of skeletal muscle. It has been suggested that the greater the fat component within an LPS, the lower the likely tumor grade.19, 21, 22 Murphey et al reported that 75% of WD tumors are composed of fat-like tissues; focal nodular nonlipomatous regions that measure >1 cm to 3 cm are suspicious for DD.14
In the current series, we evaluated numerous parameters that potentially may reflect tumor aggressiveness for their utility as possible radiologic markers of DD. We observed that focal nodular/water density and hypervascularity were highly sensitive but had relatively low specificity. By using the presence of focal nodular/water density areas as a marker of dedifferentiation, nearly all DD tumors were identified (97.8% in our series); moreover, the lack of focal nodular/water density areas excludes the diagnosis of DD.
The results presented here led us to propose a diagnostic algorithm for treatment selection in patients with RPLS (Fig. 4). When RPLS is identified by CT scan, evidence of a focal nodular/water density area should be ascertained by the radiologist; in the absence of such a feature, the tumor is WD, and a biopsy is unnecessary. In the presence of a focal nodular/water density area, a biopsy is needed to differentiate between DD and WD; according to the data presented in this report, such a biopsy always should be taken from the suspicious area; otherwise, accuracy is significantly compromised. A preoperative biopsy of a simple or hypercellular WD variant should be followed by a less aggressive surgical approach,11 and neoadjuvant chemotherapy should be considered for patients who have DD tumors.
The implications of these findings are that, if CT imaging alone is used as a predictor of histologic subtype, then DD tumors will not be misdiagnosed as WD and, thus, patients with these tumors will not be under-treated; moreover, patients who are diagnosed radiologically with WD tumors have ‘true’ WD tumors and, thus, may be treated as such with certainty. However, based on the current series, using CT imaging criteria alone to arrive at a preoperative histologic subtyping of RPLS results in a high rate of over-diagnosis: This clearly is an unacceptable outcome given the therapeutic implications.
Unlike previous reports in the radiologic literature,14 half of the WD tumors in our series contained focal nodular/water density areas, and all but 1 of these areas measured >3 cm. It is noteworthy that almost half of these were hypercellular WD tumors, which are deemed DD tumors at many institutions; however, it can be demonstrated that the clinical outcome of these patients is similar to the outcome of patients who have the simple WD variant.18
An additional modality (radiologic/histologic) may increase the accuracy of preoperative diagnosis and avoid the over-treatment of WD tumors. MRI is used commonly for RPLS, and previous data suggest that MRI features may be used to differentiate between LPS histologic subtypes.13, 19, 23, 24 Data indicate that a nonlipomatous component within RPLS on MRI correlates with DD histology13; Song et al also reported that certain MRI features may be used to differentiate between RPLS histologic subtypes.23 Neither of those reports demonstrated that MRI was superior to CT scanning; however, evaluating the diagnostic utility of MRI for RPLS was not part of our study design. Moreover, the number of MRIs that we reviewed was too small; therefore, conclusions regarding its efficacy could not be made. There are limited data regarding the role of PET-CT in STS; however, several reports have reported a correlation between uptake and grade25 or clinical outcome.26 Other data demonstrated that PET-CT was better than CT or MRI in depicting lymph node involvement and STS bone manifestations.27 A single report by Suzuki et al suggested that PET-CT may provide accurate assessment of RPLS histologic subtype, and this possibility merits prospective evaluation.28
Tumor histologic subtype and grade can be determined in 95% to 99% of core-needle biopsies if they are interpreted by pathologists who are experienced in examining STS1, 29, 30; however, RPLS can be highly heterogeneous. Consequently (and as a caveat), preoperative CT image-directed core-needle biopsies of several different areas within the same tumor may differ from the final histologic diagnosis, when the entire tumor is available for analysis.
We have demonstrated that a preoperative CT-guided biopsy taken from focal nodular/water density and/or hypervascular areas is accurate; all such biopsies were correlated with postoperative histologic reports and were useful in discriminating between WD and DD; therefore, such a biopsy should be performed routinely in all patients who have RPLS with a focal nodular/water density area. In contrast, a single biopsy taken from a radiologically nonsuspicious area within the tumor frequently is inaccurate, uninformative, and, thus, should be avoided.
In conclusion, RPLS actually is comprised of 2 radiologically, histopathologically, and clinically distinct tumors that, today, usually are treated in the same manner, resulting in either under treatment or over treatment, depending on differentiation status. Appreciating these differences has led to our policy of tailoring RPLS treatment to differentiation status; consequently, an accurate preoperative assessment of this status is critical. We have demonstrated that the diagnosis of WD can be based on CT studies in the absence of a focal nodular/water density alone, and a biopsy is not needed before surgery. In addition, CT scanning can identify almost all DD accurately, thereby rendering their under treatment unlikely. However, a CT-guided biopsy is needed to differentiate between DD and WD tumors that contain focal nodular/water density areas to avoid over treatment. At UTMDACC, such discrimination enables patients who have WD tumors to receive less aggressive surgical approaches while accurately selecting DD patients to receive neoadjuvant chemotherapy followed by surgery.
Conflict of Interest Disclosures
The authors made no disclosures.
- 5Liposarcoma: Soft Tissue Tumors. 3rd ed. St. Louis, Mo: CV Mosby; 2001., .
- 6Pathology and Genetics of Tumors of Soft Tissue and Bone. Lyon, France: IARC Press; 2002., , .
- 15Statistical Methods. 7th ed. Ames, Iowa: Iowa State University Press; 1980., .