Liposarcoma (LS) is the single most common soft tissue sarcoma and accounts for at least 20% of all sarcomas in adults.¹ Classification of LS into 4 types, based on morphologic features and cytogenetic aberrations, is now widely accepted.² These 4 types are (1) well-differentiated; (2) dedifferentiated; (3) myxoid/round cell; and (4) pleomorphic. The extent of differentiation, as reflected by histologic grade, remains the most important determinant of clinical course and of ultimate prognosis for patients with LS after resection. The anatomic distribution of LS appears to be closely related to histologic type: myxoid/round cell and pleomorphic LS have a predilection for the extremities, whereas well-differentiated/dedifferentiated LS occurs predominantly in the retroperitoneum. Well-differentiated LS often recur locally but have minimal metastatic potential, with a 5-year survival probability of 90%. Deaths result from local effects on critical adjacent organs, usually in the retroperitoneum.^3-6 Dedifferentiated lesions have similar local effects. Although they metastasize systemically (eg, to lungs), they have a significantly lower metastatic rate (10-15%) than would be predicted given their high-grade morphology, and a 5-year survival probability of 75%.⁷ The myxoid/round cell LS represents a morphologic continuum, and histologic grading is based on the extent of round cell component. Recent work from our group at Memorial Sloan-Kettering Cancer Center (MSKCC) found that patients with low-grade myxoid LS, defined as <5% round cell areas, have a 5-year survival of 90%, whereas those patients with high-grade myxoid/round cell LS (>5% round cells) has a 5-year survival probability of 60%.⁸ Pleomorphic LS, the least common subtype, are high-grade, very aggressive tumors with a high metastatic potential and a 5-year survival probability of 30-50%.⁹

The treatment of retroperitoneal LS continues to be a challenge for the surgeon, given their large size and often central location. The ability to completely resect a retroperitoneal sarcoma remains the most important predictor of local recurrence and overall survival.^4,5,10 Most retroperitoneal LS are large (>10 cm) at the time of diagnosis and because of the high degree of adipocyte differentiation, are difficult to distinguish from normal retroperitoneal fat. Thus, it is often difficult to obtain a margin of normal tissue around the tumor. The large size and extent of these tumors make treatment with tumoricidal doses of adjuvant radiation difficult and are often associated with substantial morbidity. For the low-grade, well-differentiated LS, adjuvant chemotherapy has little to offer, given the very low mitotic rate of these lesions. The use of adriamycin and ifosfamide chemotherapy for the higher grade subtypes, such as dedifferentiated, myxoid/round cell, or pleomorphic LS results in partial responses in up to 50% of patients¹¹ without a demonstrable improvement in overall survival. Furthermore, complete response rates to chemotherapy have been seen in less than 10% of patients. Thus, complete surgical resection continues to be the most important component of therapy. The present study was undertaken to determine the influence of histologic subtype and extent of surgical resection on patterns of recurrence and overall survival in a large series of patients with primary retroperitoneal LS.

All patients treated for soft-tissue sarcoma at Memorial Sloan-Kettering Cancer Center between July 1, 1982, and December 31, 2001, were entered in a prospectively maintained database. For the present study, all patients with primary retroperitoneal LS and malignant fibrous histiocytoma were considered for histologic review. In the present study primary retroperitoneal LS was classified into 4 histologic types based on recent World Health Organization criteria as (1) well-differentiated, (2) dedifferentiated, (3) myxoid/round cell, and (4) pleomorphic. After histologic reclassification, there were 177 patients with confirmed primary retroperitoneal LS that had surgical resection at MSKCC and had complete clinicopathologic data for correlation to clinical outcome. None of the patients had distant metastasis at the time of presentation. Standard evaluation of patients included a history and physical examination, complete blood count, and serum chemistry analyses. Computed tomography scans of the chest, abdomen, and pelvis were also obtained. Patient, tumor, treatment, and survival data were prospectively acquired and entered into our sarcoma database. Approval was obtained from the MSKCC IRB for this study.

Patient demographics included age at diagnosis and sex. Tumor burden was determined by the sum of the maximum diameter of the primary tumors as reported at the time of initial surgical resection. The type of surgery performed (total or subtotal resection) as well as adjacent organs removed was ascertained from the operating surgeons assessment as documented on the operative report. Specimen margins were analyzed for evidence of microscopic disease. Fourteen (8%) of the patients received postoperative external beam radiation therapy. None of the patients received chemotherapy. Patients were observed in our soft tissue sarcoma program at 4-month intervals during the first 2 years and at 6-month intervals thereafter. Information obtained during follow-up included status of disease (alive without disease, alive with recurrent disease, dead of other causes without evidence of recurrent disease, dead of other causes with disease or dead as a result of sarcoma treatment or with sarcoma) and local and distant disease recurrence.

TABLE 1. Clinicopathologic and Treatment Characteristics in 177 Patients With Primary Retroperitoneal Liposarcoma

FIGURE 1. DSS for 177 patients with primary retroperitoneal LS.

TABLE 2. Univariate Analysis of Clinicopathologic Variables for Disease-specific Survival in 177 Patients With Primary Retroperitoneal Liposarcoma

FIGURE 2. DSS by LS histologic subtype.

FIGURE 3. DSS by margin status.

FIGURE 4. DSS by age.

Variables significant at the 0.25 level, such as histologic type, margin of resection, nephrectomy, contiguous organ resection (excluding nephrectomy), and age (as a continuous variable), were entered into a multivariate Cox proportional hazards model to identify independent predictors of DSS (Table 3). Given the small number of patients with myxoid and round cell tumors, only patients with well-differentiated and dedifferentiated tumors were included. Because all well-differentiated tumors were low grade and all dedifferentiated tumors were high grade, grade was not included in the multivariate models. The following factors were significantly associated with DSS by Cox regression analysis: dedifferentiated histology versus well differentiated (HR 6, P < 0.0001), gross positive margins versus negative margins (HR 4, P < 0.0001), contiguous organ resection (excluding nephrectomy) versus no contiguous organ resection (HR 2, P = 0.05), and age as a continuous variable (HR 1.03, P = 0.03).

TABLE 3. Predictors of Disease-specific Survival for Retroperitoneal Liposarcoma From Multivariate Analysis (n = 177)

TABLE 4. Univariate Analysis of Time to Local Recurrence (Patients With Positive Gross Margins are Excluded, n = 144)

FIGURE 5. Probability free of local recurrence by histologic subtype.

Histologic type and contiguous organ resection were entered into a multivariate Cox proportional hazards model to identify independent predictors of local recurrence (Table 5). Both factors were significantly associated with local recurrence by Cox regression analysis: dedifferentiated histology versus well-differentiated (Hazard ratio [HR] 3.6, P < 0.0001) and contiguous organ resection versus no contiguous organ resection (HR 1.7, P = 0.04).

TABLE 5. Predictors of Local Recurrence for Retroperitoneal Liposarcoma From Multivariate Analysis (n = 144)

TABLE 6. Univariate Analysis: Time to Distant Recurrence

FIGURE 6. Probability free of distant recurrence by histologic subtype.

Histologic type and contiguous organ resection were entered into a multivariate Cox proportional hazards model to identify independent predictors of distant recurrence (Table 7). Both factors were significantly associated with distant recurrence by Cox regression analysis: dedifferentiated histology versus well-differentiated (HR 15, P < 0.0001) and contiguous organ resection versus no contiguous organ resection (HR 3, P = 0.02).

TABLE 7. Predictors of Distant Recurrence for Retroperitoneal Liposarcoma From Multivariate Analysis (n = 177)

LS is the most common mesenchymal tumor of the retroperitoneal space but continues to pose a challenge with regard to diagnosis, prediction of clinical behavior, and treatment of disease recurrence within the abdominal/retroperitoneal space. In the present series, the majority of patients (93%) had tumors of the well-differentiated/dedifferentiated subtypes. Dedifferentiated LS may be particularly difficult to recognize because they exhibit a variable histologic picture, but most frequently they resemble unclassified malignant fibrous histiocytoma (MFH)-like pleomorphic sarcoma or intermediate- to high-grade myxofibrosarcoma.^16,17 Careful and extensive sampling is therefore mandatory, particularly in large retroperitoneal lesions as the well-differentiated component may be overlooked. Primary myxoid/round cell LS are relatively uncommon and account for only 7% of patients with primary retroperitoneal LS. Pleomorphic LS are very rare in the retroperitoneal location and not a single case was found in the present study.

The overall DSS survival of 73% at 3 years and complete resection rate of 81% in this study of primary retroperitoneal LS was similar to the results of other published series of retroperitoneal sarcoma.^10,18-20 Patients with retroperitoneal LS who were treated with incomplete resection had a 3-year DSS of only 43% and thus emphasize the importance of achieving a complete resection even if sacrifice of a contiguous organ is required. In the present study the 2 most important predictors of DSS on multivariate analysis were histologic subtype (dedifferentiated histology) and the presence of gross positive margins. Whether age was examined as a continuous variable or categorical variable based on the median, older age was associated with a significant reduction in DSS (P = 0.02) on univariate analysis. Of interest is that for patients greater than 63 years of age 47% presented with dedifferentiated histology compared with a 32% incidence of dedifferentiation at presentation in the younger age group. In the multivariate analysis when age was analyzed as a categorical variable using the median age of 63 as a cutoff it was not found to be an independent prognostic factor for DSS on multivariate analysis (HR 1.2, P = 0.43). This may well be due to the association of age with dedifferentiated histology. However, when age is analyzed as a continuous variable despite adjusting for dedifferentiated histology and gross positive margins age remained a significant prognostic factor on multivariate analysis. Contiguous organ resection was associated with a 2-fold (P = 0.05) increased risk of disease-specific death compared with patients not undergoing contiguous organ resection, having adjusted for other prognostic factors. This suggests that involvement of adjacent organs other than kidney portends more aggressive tumor biology for patients with retroperitoneal LS. Tumor burden and nephrectomy had no influence on DSS, suggesting that as long as complete resection can be achieved, these variables do not influence survival in retroperitoneal LS. When we just looked at patients who had a complete resection tumor burden was still not prognostic for DSS. With a median tumor burden of 26 cm, it is evident that the majority of patients with retroperitoneal LS present with tumors >10 cm in size largely because of their deep location. This may suggest that once a retroperitoneal sarcoma exceeds a size of 10 cm that they all behave in a similar high-risk fashion.

A previous study of nephrectomy for retroperitoneal sarcoma (all histologic types) demonstrated that renal capsular invasion was present in 15% of patients, renal parenchymal invasion in 9% and renal vein invasion in 3% of patients with the majority of patients 73% having no evidence of direct kidney invasion.²¹ Although in the present LS series we did not examine the precise incidence of renal invasion, the general impression was that LS (particularly the well-differentiated subtype) pushed against adjacent kidney without direct invasion. In such cases our general approach was to strip off the outer capsule of the kidney together with the adjacent LS so that a clean but close margin could be achieved without sacrificing kidney parenchymal function. In cases of LS with circumferential kidney hilar involvement nephrectomy was generally required to achieve a complete resection. The present study demonstrates that kidney preservation with capsular resection was not associated with any reduction in DSS. Thus, nephrectomy should only be performed if required to accomplish a complete gross resection.

Multivariate analysis showed that histologic subtype and contiguous organ resection were independent prognostic factors for local recurrence. Dedifferentiated LS was associated with a 4-fold increased risk of local recurrence compared with well-differentiated histology. In fact, after 3 years of follow-up over 80% of the patients with primary dedifferentiated LS will have recurred locally. Previous studies of retroperitoneal sarcoma have suggested that LS histologic type is associated with a 3-fold higher risk of local recurrence compared with other histologic types that was independent of sarcoma grade.¹⁸ The high rates of local recurrence observed with LS may relate to the often multifocal involvement of disease throughout the retroperitoneal space as well as the difficulty in detecting LS from adjacent normal fat. Of the 99 patients with well-differentiated LS, 39 patients developed at least 1 local recurrence at the time of last follow-up. Of these first-time local recurrences who underwent resection, 83% remained well-differentiated and 17% recurred as high-grade dedifferentiated LS. Of the patients with well-differentiated first local recurrences who then developed a second recurrence, 44% recurred as dedifferentiated LS and 56% remained well differentiated. Thus, the fraction of well-differentiated tumors that progress and dedifferentiate seems to increase with each subsequent recurrence.

Involvement of LS of an adjacent organ other than kidney that requires resection was associated 2-fold higher risk of local recurrence and 3-fold higher risk of distant recurrence compared with those who did not require contiguous organ resection. This suggests the need for contiguous organ resection, excluding kidney relates to the local invasiveness and perhaps multifocal nature of a given LS. In addition, these data suggest that tumors with more extensive local involvement requiring adjacent organ resection are biologically more aggressive and thus are more likely to recur locally as well as distantly.

The most important prognostic factors for distant recurrence include histologic subtype and contiguous organ resection. Multivariate analysis showed that dedifferentiated histology had a 15-fold increased risk of distant recurrence compared with well-differentiated histology. Only 3 of 99 patients with well-differentiated histology recurred distantly and all of these were of the sclerosing subtype. In contrast, 14 of 65 dedifferentiated LS recurred distantly for a probability free of distant recurrence at 3 years of 70% (55-86%). This risk is about 15% higher than generally quoted in the literature^3,12 and may be a result of the fact that our current analysis is on a homogeneous group of patients with primary disease followed in a prospective database with computed scans at defined intervals. The median tumor size of these primary retroperitoneal dedifferentiated LS was 21 cm (5-96 cm). In comparison, the probability free of distant metastasis for all patients with greater than 5-cm, high-grade MFH tumors of the extremity seen at MSKCC from 7/1/82-6/30/02 (n = 231) at 3 years is 59%. Thus, despite the large median size of retroperitoneal dedifferentiated LS the risk of distant metastasis appears to be about 10% less at 3 years than the typical high large grade extremity MFH. Longer follow-up on this homogeneous population of dedifferentiated retroperitoneal LS will be needed to determine if with long-term >5-year follow-up the patients continue to develop distant metastasis and in the end approach the same distant recurrence rates as seen for the typical large, high-grade nonlipogenic extremity tumor.

In this study of patients with retroperitoneal LS, histologic subtype, incomplete resection, contiguous organ resection (excluding nephrectomy) and increasing age are strongly associated with death from tumor. Tumor burden and nephrectomy were not associated with disease specific survival. Surgical treatment of retroperitoneal LS should consist of an aggressive approach to achieve a complete surgical resection. En bloc resection of adjacent organs should be performed if necessary to achieve a complete resection. However, kidney parenchymal sparing renal capsular resections can be performed without any measurable influence on DSS as long as complete resection is achieved. Despite an aggressive surgical approach over 80% of patients with dedifferentiated histology will recur locally and 30% will metastasize to distant sites within 3 years of diagnosis. Combination of surgery with new systemic therapies, such as PPARγ ligands, and locoregional therapies, such as preoperative intensity modulated radiation therapy, are needed to improve outcomes in patients with retroperitoneal LS. The prognostic variables and outcome data generated from the present study should serve as the “benchmark” for future trials in retroperitoneal LS as no patients received chemotherapy and only 8% of patients were treated with adjuvant radiotherapy. In addition, this work emphasizes the importance of stratification by histology and histologic subtype in the design of future prospective randomized trials in retroperitoneal sarcoma.

Dr. Raphael E. Pollock (Houston, Texas): I appreciate having had the opportunity to review the manuscript prior to this meeting and appreciate the privilege of the floor at this time. A couple of thoughts:

First of all, it is clear on the basis of this work that the surgical services at Sloan-Kettering can continue to successfully engraft organ transplants from Boston to New York; congratulations to both Dr. Singer and Dr. Brennan.

This is a very important study. We are considering a very rare disease treated in a standard fashion at 1 institution that is intimately familiar with the disease and its natural history. The patients reported on here have all been followed prospectively after having been treated in a standard fashion. And I think it does show the value of prospective data collection, because it suggests that in some circumstances histologic subtype may be a or even the critical driver in terms of multiparametric patient outcomes.

Dr. Singer has demonstrated that differentiated pathologies have a very different behavior than do de-differentiated: the bottom line is a 4-fold increase in local recurrence in de-differentiated patients; as a matter of fact, after 3 years, 80% of the de-differentiated patients will have local occurrence. So my questions really focus on how we might use this information rather than the basis of its derivation per se. I have 3 brief interrelated questions:

First: Given the differing biologies under consideration, how does 1 justify adjacent non-nephrectomy organ resection in de-differentiated patients if they are doomed to local recurrence anyway? Or, on the other hand, if well-differentiated patients behave in such an indolent fashion, is there any role for subtotal debulking prospective palliation strategies in these individuals?

The second question is: Given the biologic impact of de-differentiated versus well-differentiated histology, as a routine practice do you biopsy areas of suspected de-differentiation either pre or intraoperatively, and as a practical matter how do you make use of the knowledge of de-differentiated status? For example, does this drive you to neoadjuvant chemo- or radiotherapy approaches? Does it govern the extent of resection? Or does it lead you to recruit or think about recruiting intraoperative radiotherapy or other such intraoperative strategies?

And the third question, which really ties directly to this, is: Do you use intraoperative frozen section as a guide to the extent of resection, particularly in non-nephrectomy contiguous organ resection in de-differentiated patient?

This is a critically important study because it has the potential to change our approach to a very devastating oncology problem, and I appreciate the opportunity to pose these questions this morning. Thank you.

Dr. Samuel Singer (New York, New York): Thank you, Dr. Pollock. In terms of the issue of how the data presented would change our approach to retroperitoneal sarcoma and how we would use this information in treating patients. Basically, I see that there are 3 potential uses of this information. The first 1 is that there is now a significant problem with consistently grading these tumors even by experienced pathologists after having the entire specimen to review. So the idea would be to use molecular genetics combined with morphology to see if we can more objectively grade these tumors to assess a patient’s risk of local and distant relapse.

Second, by understanding the underlying biology of each histologic subtype we can figure out which patients, for example, with well-differentiated liposarcoma at high risk to dedifferentiate to a dedifferentiated liposarcoma of higher grade; in other words which of these patients would progress to a more aggressive form in the future. Thirdly, this data on histologic subtype can serve as a benchmark for future trial designs in terms of assessing high-risk groups for both local and distant recurrence. This data will permit us to identify how we should stratify patients in these randomized trials so as to optimize their design and accurately balance risk in each arm.

To address the second question of whether our operative approach would change based on histologic subtype. For example, would our operative approach for organ preservation and kidney sparing nephrectomy change based on histologic subtype (whether it was well-differentiated or dedifferentiated). We would strive for a complete resection of all gross disease independent of the histologic subtype. In kidney sparing resections, most often the kidney parenchyma is not involved and you can often excise the kidney capsule with the adjacent tumor (for a clean but close margin) whether it is well-differentiated or dedifferentiated. If the renal hilum is involved with either well-differentiated or dedifferentiated tumor then 1 will need to completely remove the kidney to achieve a complete resection of all liposarcoma. We would follow the same approach for other contiguous organ resections. If there is obvious organ involvement then that needs to be resected regardless of the histologic subtype. Occasionally it is required that 1 do a contiguous organ resection for exposure so that 1 may obtain a complete resection but again this is typically unrelated to the histologic subtype. Finally, often when you take out these very large liposarcomas there is no way of knowing whether there is a section of the tumor, which has dedifferentiated until you have completely removed it.

A third question was whether there is value in doing an intraoperative frozen section analysis. In general frozen section analysis of fatty tissue is not particularly useful. It is very hard to freeze fat without significant artifact and obtain an accurate histologic diagnosis. Accurate histologic analysis on frozen section is extremely difficult for well-differentiated lipoma like and sclerosing liposarcoma given their high fat content. For dedifferentiated high-grade sarcomas with less fat content frozen section is more accurate. We are presently working in our laboratory using magic angle spinning and NMR spectroscopy techniques on fresh tissue specimens to do real time analysis of normal fat versus well-differentiated liposarcoma to see if we can accurately distinguish tumor fat from normal fat. In this analysis, we are looking for biochemical differences between normal fat and liposarcoma. We want to see whether we can in a blinded fashion compare biochemical analysis of the same specimen used for NMR to morphology on permanent histologic sections to see how accurate we can do this in real time. However, presently we do not have the capability yet of doing intraoperative frozen sections accurately on these tumors that contain high levels of fat. We are also working on developing molecular genetic techniques in real time for diagnostic analysis of these liposarcomas.

The fourth question is how we would make practical use of knowing that a patient had dedifferentiated histology preoperatively and whether this would alter our choice of adjuvant therapy and the extent of resection. In my experience dedifferentiated liposarcomas are not particularly responsive to neo-adjuvant chemotherapy and at present we have no protocols for that at Memorial and so if we knew an area was dedifferentiated we would not treat that patient with neo-adjuvant chemotherapy. In the past, it was believed that the systemic risks from these dedifferentiated liposarcomas, even though they were high grade was relatively low and they were mainly prone to local recurrence problems with a distant recurrence rate of only 15%. In the present study we show that the distant recurrence risk is as high as 30% and thus it would not be unreasonable to consider potential systemic therapy in future trials. This has not been our practice to date. In terms of extent of resection again our goal would be complete resection and that would not change whether it was well-differentiated or dedifferentiated. You are only as good as your most limited margin and so extended organ resection to achieve more than a complete resection is usually not indicated unless it is needed to make sure that you have removed all gross tumor. Brachytherapy techniques have been associated with high morbidity in our hands and so we have not used those. We now plan to examine the role of preoperative IMRT radiotherapy at Memorial in a randomized trial in patients with primary retroperitoneal liposarcoma. In this trial, patients would be stratified by histologic subtype, which clearly will be important in analyzing the results and balancing treatment arms in the trial. The primary end point would be local recurrence free survival. The idea being if we could improve local recurrence free survival for both the well-differentiated and dedifferentiated tumors then that would prove overall survival as well in these patients where they primarily die of local control problems in the abdomen and retroperitoneum.