New Perspectives for Staging and Prognosis in Soft Tissue Sarcoma
- First Online:
- Cite this article as:
- Lahat, G., Tuvin, D., Wei, C. et al. Ann Surg Oncol (2008) 15: 2739. doi:10.1245/s10434-008-9970-6
- 303 Views
Data suggest that the current American Joint Committee on Cancer (AJCC) soft tissue sarcoma (STS) staging criteria merit further evaluation. We sought to identify and validate factors as enhanced descriptors of STS clinical behavior.
Prospectively accrued data were analyzed for 1,091 AJCC stage I to III primary STS patients who had complete macroscopic resection at our institution from 1996 to 2007. Study factors were examined by univariable and multivariable analyses to identify independent prognostic factors for disease related mortality and overall survival (OS).
In contrast to the current AJCC STS staging system, which stratifies size into T1 (≤5 cm) and T2 (>5 cm) groups, we demonstrated three distinct cohorts (P < 0.0001): T1 (≤5 cm; 5-year OS 85%), T2 (5 to 15 cm; OS 68%), and T3 (>15 cm; OS 52%). A two-category system of histologic grade was demonstrably as informative as the current four histologic grade AJCC system. A multivariable Cox proportional hazard model identified tumor size (5 to 15 cm vs. ≤5 cm, P = 0.03; or >15 cm vs. ≤5 cm; P < 0.0001), nonextremity primary site (P = 0.0016), disease of high histologic grade (P = 0.001), specific histology (P = 0.001), and margin positivity (P < 0.0001) as statistically significant adverse independent prognostic factors. Recurrence during follow-up was the most significant risk factor for STS-specific mortality (P < 0.0001).
Tumor size and grade in the AJCC STS staging system need revision; moreover, primary site, histologic subtype, margin status, and recurrence offer additional relevant prognostic insight. Incorporation of these factors may enhance the AJCC staging system, thereby further facilitating individualized therapeutic strategies for STS patients.
KeywordsSoft tissue sarcomaStagingPrognostic factorsRecurrenceSurvival
Soft tissue sarcoma (STS) is a rare form of malignancy, with an annual U.S. incidence of approximately 9,500 new cases.1,2 This extremely diverse cluster of malignancies encompasses >50 distinct histological subtypes.3 STS is further distinguished from other cancers by the capacity to originate in virtually any anatomic locus in the human body.
STS patients have a 5-year survival rate of only 50%, reflecting the disease’s relative propensity for distant metastasis,4 which all too frequently precludes effective therapeutic intervention.5 These high distant metastasis and disease-related mortality rates require a robust staging system that can provide precise prognostic information through accurate grouping of patients.
Numerous studies have identified factors prognostic for STS local recurrence, distant metastasis, and tumor-related mortality6–27; these investigations provide the foundation of the present American Joint Committee on Cancer (AJCC) STS staging system as well as the basis of an STS nomogram recently developed by Kattan et al.23 for use in adult postoperative patients. The current AJCC STS staging system provides useful prognostic information and is also helpful in guiding selection of patients for adjuvant therapy and/or stratification for inclusion in clinical trials.
Despite its general usefulness, the AJCC STS staging system has several important shortcomings that suggest the need for revision. For example, there are several relevant prognostic parameters not currently included in the present AJCC STS staging system that merit attention, such as age, site of primary disease, and margin status. In addition, contemporary data suggest that the currently used T, N, M, and G criteria are not optimal and should be considered for further evaluation.16,23,26 By means of one of the largest available STS prospective databases, we evaluated the prognostic impact of various current and potential staging markers in an attempt to improve the AJCC STS staging system.
Patients and Methods
From January 1996 to July 2007, a total of 6,702 consecutive adult STS patients sought care at the University of Texas M. D. Anderson Cancer Center (UTMDACC). Of these, 3,717 were definitively treated at the UTMDACC; patients with local recurrence (n = 474), synchronous primary and metastatic disease (n = 435), or metastatic-only disease (n = 350) were excluded. Of 1,442 patients with primary STS who were surgically treated at UTMDACC, 1,091 underwent complete gross total resection of a specific STS histology lesion; these were included in the study population. Desmoids and dermatofibrosarcoma protuberans histologies were excluded, as were inconclusive pathologies of mesenchymal tumors of uncertain malignant potential.
All patients included in the study population remained in active clinical or telephone follow-up through the UTMDACC Sarcoma Center.
A tumor was considered to be a primary STS if it was previously untreated and if there was no evidence of metastasis at the time the patient sought care. Evaluation methods for these clinical determinations included various radiographic (e.g., computed tomographic scan, magnetic resonance imaging, positron emission tomography, and ultrasound) and clinical examinations. Some patients received systemic chemotherapy and/or radiation as per the recommendations of the multidisciplinary UTMDACC Soft Tissue Sarcoma group. Surgical, chemotherapeutic, and radiation treatments were ultimately applied on the basis of an evaluation of relevant and clinical prognostic factors. Therefore, outcomes did not necessarily reflect the effectiveness of a specific treatment. In addition, there were no relevant control groups; consequently, treatment-related factors were not included in the multivariable analysis.
The median follow-up time among the survivors in this STS study cohort was 53.3 months. Data on the following factors were correlated with the rates of local recurrence, distant recurrence, disease-specific survival, and overall survival (OS). Patient factors included patient age and sex; tumor factors included tumor depth and anatomic site (extremity vs. nonextremity). Pathologic factors included the following: tumor size, histopathologic subtype, grade, and status of microscopic surgical margins. Only a few patients had nodal involvement, so therefore, we did not include nodal status in this data analysis.
All tumors were categorized as extremity (upper/lower), intra-abdominal (abdominal/pelvic), visceral, thoracic, superficial trunk, or head and neck. Tumor size was defined as the maximum diameter of the tumor at pathologic analysis; tumors were classified as <5, 5 –10, 10–15, or >15 cm. Tumor grade was assessed as either low grade or high grade; this determination was based on histopathologic diagnosis. The following histologies were considered as low grade unless specifically designated otherwise by the pathology report: atypical lipomatous tumor/well-differentiated liposarcoma (WD), hemangiopericytoma/solitary fibrous tumor, low-grade fibromyxoid sarcoma, and inflammatory myofibroblastic sarcoma. Other histologies specifically designated as low grade in the pathology report were also included in this category. All other STS were considered to be high grade unless specifically designated in the pathology report as intermediate grade. No formalized grading system (French Federation of Cancer Centers Sarcoma Group [FNCLCC], National Institutes of Health [NIH], or other) was used either prospectively or retrospectively. The exception to this was gastrointestinal stromal tumor, where the NIH consensus risk assessments were listed in the reports; these were used as a surrogate for histologic grade.28 Unclear or confusing grade designations were reviewed by a soft tissue pathologist (A.J.L.) for final assignment.
The anatomic depth of tumors was evaluated relative to the investing fascia of the extremity or trunk, with all head and neck tumors being considered as deep; all tumors were characterized as either superficial or deep. A microscopically positive surgical margin was defined as tumor present at an inked surgical margin on permanent histologic examination.
Thirty different histologic subtypes were identified in the study cohort. However, the relative proportion of most histologic subtypes was too small for these to be independently included in a statistical analysis. Consequently, to facilitate a statistically meaningful analysis, STS histologies were clustered into four large groups, as follows: (1) undifferentiated pleomorphic sarcoma (UPS), (2) dedifferentiated liposarcoma (DD), (3) WD, and (4) others.
Tumors originally designated as malignant fibrous histiocytoma (MFH) were of the storiform, pleomorphic, inflammatory, and myxoid histologic subtypes. High-grade sarcomas not having such definite features were categorized by our pathology group as unclassified sarcomas. Both histological subtypes were included in the UPS group. Only a few MFH cases in this series were classified as myxofibrosarcoma; these tumors were also included in the UPS group.
Estimates of OS curves were established according to Kaplan–Meier product-limit method29 and were calculated from the date of diagnosis of the primary disease to the event date or last follow-up if censored. Missing data elements of unknown status were not included in the analysis.
Estimate of cumulative incidence of sarcoma-specific mortality was calculated according to the method by Gooley et al.30 with a publicly available SAS macro created by Erik Bergstralh of the Mayo Clinic.31 The figures were generated by S-plus software (Insightful, Seattle, WA). Five-year cumulative incidence rate of these events was estimated by the subdistribution analysis method for competing risks by Gray32 with the Cmprsk R package.33
For sarcoma disease-specific survival, all deaths due to sarcoma were considered to be events, whereas deaths from other causes were deemed as competing risks. Patients who were alive at last follow-up were censored. Time-to-event analyses were performed to assess whether there were associations between independent variables and sarcoma-specific mortality. Each independent variable was first separately examined in a univariable Cox proportional hazard model with competing risk by means of the stacked data set technique.31 All univariable Cox models were fitted with all possible data points. Only the independent variables that had P values of ≤ 0.10 in the univariable Cox model analyses were examined in multivariable Cox models.
Of 1,091 patients, 788 patients were alive at last follow-up. A total of 303 patients (27.8%) died during the follow-up period; 210 patients died of sarcoma disease, 49 died of other causes, and 44 death events were of unknown cause. It was not considered appropriate to simply exclude the deaths designated as “unknown cause” because it could inappropriately reduce the estimate of the hazard for deaths due to sarcoma disease. To impute the missing value of the causes of death, a logistic regression model was fitted to the subset of patients who had died at last follow-up, with the cause of death (sarcoma disease vs. other causes) being the response variables, and other clinical variables such as histology, type of recurrence, tumor size, primary site, margin status, and age serving as predictor variables. A predicted probability was calculated on the basis of the logistic regression model for each of the unknown cause of death cases. For each case, a random number was generated from the standard uniform distribution. If the randomly generated number was less than the predicted probability of death due to disease, the patient was assigned to have died from disease. The patient was assigned to have died from other causes if the randomly generated value was greater than the predicted probability. This process was repeated 50 times, and each time, a univariable and a multivariable Cox model with competing risks31 was then fitted to the imputed data. The above describes a multiple imputation algorithm; multivariable inference based on Wald tests was then performed incorporating these combined results from this algorithm by using the multiple imputation method described in the SAS Mianalyze procedure.34 The causes of the death of 13 patients who died of unknown causes could not be imputed because of missing values in the predicting variables.
Clinicopathologic characteristics in 1,091 primary STS
Age ≥ 60 years
Head and neck
Undifferentiated pleomorphic sarcoma
Malignant fibrous histiocytomaa
Gastrointestinal stromal tumor
Gross total resection
All patients included in this cohort had complete macroscopic resection (n = 1091); surgery was the only treatment in 520 patients (47.7%). Three hundred ten patients (28.4%) received chemotherapy, and 473 (43.3%) received radiation.
Disease-specific survival according to univariable Cox proportional hazard models
5–15 versus <5 cm
>15 versus <5 cm
High grade versus low grade
Positive versus negative
Nonextremity versus extremity
Deep versus superficial
DD versus WD
UPS versus WD
Others versus WD
Continuous in 10-year increments
Male versus female
Distant versus no recurrence
Local versus no recurrence
The AJCC STS staging system stratifies grade into four categories. By using our data, it was possible to demonstrate that disease-specific survival for G1 and G2 tumors was equivalent (Fig. 1B); therefore, all tumors were classified as being either low grade or high grade. High grade correlated with an increased 5-year STS-related mortality and with decreased 5-year OS (P < 0.0001), 31% and 62% compared with 7% and 87.4% for patients with low-grade tumors. We examined how the interplay between tumor size and grade affects STS-related mortality. There were no differences between the three size groups for low-grade tumors; however, the 5-year cumulative incidence rates of sarcoma-related mortality for high-grade tumors <5 cm versus 5 to 15 cm versus >15 cm were 13%, 32%, and 59.5%, respectively (Fig. 1C).
An analysis of postresection margin status was also conducted. Margins were positive in 247 cases (22.6%), negative in 750 (68.7%), and not designated in 94 (8.6%). The time-dependent increase of sarcoma-related deaths changed statistically significantly as per margin status (Fig. 3B). The 5-year STS-related mortality and 5-year OS (P < 0.0001) rates for patients with disease-negative margins versus disease-positive margins were 18% and 74% compared with 31% and 63%, respectively. The impact of the interplay between margin positivity and grade on STS-related mortality was examined. Margin status did not affect STS-related mortality for low-grade tumors; however, the 5-year cumulative incidence rates of sarcoma-related mortality for high-grade tumors with postresection disease-negative margins versus disease-positive margins (P < 0.0001) were 24%, and 46%, respectively (Fig. 3C).
Disease-specific survival according to multivariable Cox proportional hazard models
5–15 cm versus <5 cm
>15 cm versus <5 cm
High grade versus low grade
Positive versus negative
Nonextremity versus extremity
DD versus WD
UPS versus WD
Others versus WD
Distant versus no recurrence
Local versus no recurrence
Tumor, node, metastasis systems of staging have been applied to most malignant tumors in an effort to group together patients with similar clinical and/or pathologic characteristics. Such clustering, clinicians hope, enables accurate estimation of prognosis, which is critical in selecting appropriate therapies and for comparing clinical experiences across time or institutions. A useful staging system must be practical and applicable to the diverse needs of all relevant medical practitioners and their patients. Moreover, staging criteria must accurately reflect disease biology, as demonstrated by consistent outcomes for comparable patients treated equivalently at different institutions. These criteria must be evidence based and incorporate statistically significant prognostic factors, as may be identified by Cox multivariable regression analyses.35
The AJCC STS staging system dates to 1977.36 Subsequent alternations have been made, such as incorporating tumor depth as a prognostic factor in the 5th edition of the AJCC manual37 or simplifying the staging system by eliminating several substages in the 6th edition of the AJCC manual.38 These changes have been facilitated by the development of large prospective relational clinical databases useful in multivariable analyses of disease outcome. The current AJCC STS staging system includes the T, N, M, G, and “a” versus “b” criteria. Several additional prognostic markers were incorporated into a postoperative adult extremity STS nomogram developed at the Memorial Sloan-Kettering Cancer Center (MSKCC) and subsequently validated using other large databases.17,18,21,24 These additional factors include age, histologic subtype, and primary site in addition to size, grade, and depth. Other considerations may be of prognostic value, as demonstrated above. The data presented in this report are based on an analysis of prognostic factors for different oncologic end points using prospectively collected data. This is in contrast to most previous analyses, which have generally been based on retrospectively retrieved data sets that have included 400 or fewer patients.
The Royal Marsden group analyzed 316 previously untreated STS patients who received therapy at their institution; multivariable regression analysis demonstrated that further stratification by tumor size was additionally informative of prognosis.16 Our data support this observation; patients diagnosed with STS >15 cm are at increased risk for distant recurrence and disease-specific mortality compared with patients with smaller tumors. Interestingly, when we analyzed the interplay between size, grade, and STS-specific mortality, we found that size >15 cm is a statistically significant risk factor for high-grade tumors only, whereas it does not affect the survival of patients with low-grade STS.
On the basis of their results, the Royal Marsden group proposed to stratify size into four groups: <5, 5–10, 10–15, and >15 cm. We did not find statistically significant differences in disease-specific survival for patients with STS size 5–10 and 10–15 cm; therefore, we suggest that size be further stratified into the following three categories: ≤5, 5–15, and >15 cm.
STS grade has always been considered an important prognostic factor; grade is stratified into four categories in the present AJCC STS staging system and into three categories according to the FNCLCC or NIH systems.12,39 However, our data demonstrate that two-grade stratification may be useful, and such dichotomous grade stratification has been already implemented in the STS nomogram developed by MSKCC.23,40 The College of American Pathologists40 has recently recommended that the FNCLCC system be incorporated as standard in the AJCC STS staging system.
Primary tumor location has been previously reported as an important prognostic marker in STS, and it is also incorporated into the MSKCC nomogram23; further, head and neck as well as retroperitoneal location greatly increase STS-specific mortality.21,23,41 In the series reported here, nonextremity location (mainly retroperitoneal) was a statistically significant independent risk factor for STS-specific mortality. Our data confirm the above previous observations, suggesting the possible inclusion of tumor site as a criterion in the AJCC STS staging system.
Historically, the specific histopathologic subtype has been considered to be of secondary importance because of the general impression that individual histologic subtypes of comparable histologic grade behave similarly.42,43 However, several reports have established the independent adverse prognostic significance of specific histologic subtypes15,23,41,44,45; for example, the prognosis of patients with malignant peripheral nerve sheath tumor differs greatly from that of patients with high-grade fibrosarcoma.23 Our data further support the observation that different high-grade sarcomas possess different biological behaviors, consistent with the known heterogeneity of STS. Our univariable analysis demonstrates that different histological subtypes correlate with distinct disease specific mortality rates; the outcome of DD patients was the worst. This may be attributed to their retroperitoneal location; however, in our multivariable analysis, histology was still an independent prognostic factor. We have shown that DD patients have a distinct outcome compared with UPS or other high-grade sarcomas. These data suggest that possible inclusion of histologic subtype in the AJCC STS staging system should also be considered.
An additional prognostic factor evaluated was the resection margin status achieved; this factor is relevant to the subset of patients who were treated by surgery alone and may not be applicable in a universal staging system. However, several studies suggest that margin positivity is a marker of adverse prognosis.17,18,46 An analysis of this question was provided by the MSKCC group in 2002,18 when it was demonstrated that a positive microscopic margin was associated with a 1.6-fold increase in sarcoma-related death. Margin status was also identified as having prognostic importance in a 2003 series from UTMDACC.21 In this latter study of 1,225 stage III patients treated at UTMDACC, 5- and 15-year local control rates for STS resected with negative versus positive margins were 88% versus 64% and 86% versus 56%, respectively. Our current data further support these observations; in the multivariable analysis that did not include recurrence as a variable margin positivity was associated with a 5.9-fold increase in STS-related death (P < 0.0001). As with STS size, margin positivity did not affect the survival of patients with low-grade tumors; however, it was a critical prognostic factor for high-grade STS patients. These data suggest the possible value of including margin status in a revised STS staging system.
Most analyses identify previous local recurrence as a major risk factor for subsequent local recurrence. A 2003 report from the University of California, Los Angeles, examined their experience with 753 patients with intermediate- and high-grade extremity STS who received all their treatment at that institution.22 Development of local recurrence was the most important factor associated with subsequent local recurrence and impaired STS-specific survival; an individual STS local recurrence patient was threefold more likely to die from STS than one who remained free of recurrence. It is not clear whether local recurrence is causally related to disease-specific mortality; local recurrence could possibly be a marker for a more aggressive biology, which ultimately results in disease dissemination and patient death. The previously cited UTMDACC experience with 1,225 patients with localized STS identified a similar prognostic effect of local recurrence,21 demonstrating that recurrence was the greatest single risk factor of a shorter disease-specific survival. Furthermore, distant recurrence increased the rate of disease-related mortality twofold compared with local recurrence. We have shown that recurrence during follow-up is the most important risk factor for adverse prognosis; patients with local recurrence had a marked increase in STS-specific mortality, and this risk was doubled when patients experienced recurrence with distant metastases. Taken together, these results suggest the possible value of incorporating recurrence during follow-up into STS staging systems. However, recurrence, which is mainly related to tumor biology, may also be associated with inadequate treatment. Such distinction may be important and should be further evaluated.
Our cohort included only a few patients with lymph node metastasis, so nodal status per se was not analyzed. However, the question of whether N1 disease is indicative of stage III or stage IV status needs to be addressed. Several reports have considered the prognostic impact of lymph node status in patients whose STS was of a histologic subtype with a known predilection for lymphatic dissemination. In a 2004 analysis, the Princess Margaret group of the University of Toronto identified 39 of 1,066 STS surgical patients who developed lymph node metastases, including epitheloid sarcoma, rhabdomyosarcoma, clear cell sarcoma, and angiosarcoma histologies.25 Patients with isolated lymph node metastasis had an estimated 4-year OS rate of 71%, which was equivalent to the survival of AJCC stage III patients. In contrast, those patients who had synchronous systemic and lymph node involvement had far worse outcomes that were comparable to those of stage IV patients. In 2003, the Royal Marsden group reported similar results.26 These contemporary series suggest that isolated lymph node metastases are most likely to be associated with an AJCC stage III rather than stage IV survival pattern.
Specific molecular prognostic markers for staging may be particularly useful in this era of fresh insight into the molecular biology of cancer. The detection of such markers could potentially be based on high-throughput assays. These techniques can be used for parallel in-situ detection of DNA, RNA, and protein targets for a large set of specimens examined at the same time. Results could possibly be integrated with measurements of disease progression, treatment response, and survival data, leading to even more prognostically relevant sophisticated and dynamic STS staging systems.47
In summary, the AJCC STS staging system has undergone many changes since 1977. However, contemporary large-scale data analyses suggest that the current system could be improved by incorporating additional validated prognostic factors. Our analysis of prospectively collected data from a single institution has identified and confirmed several specific adverse prognostic factors for local recurrence, distant recurrence, and tumor-related mortality. These data suggest that the criteria underlying size and grade should be revised in the future AJCC STS staging system, and N1 patients should be categorized as stage III rather than stage IV. In addition, age, primary site, histologic subtype, final pathology margin status, and recurrence are also relevant and may enhance estimates of prognosis if incorporated into a future STS staging system. Finally, identification of molecular markers may further improve the stratification of STS patients into clinical/prognostic staging categories; research is currently being conducted to achieve these goals.
Partially sponsored by a grant from the AJCC (to R.E.P.) on behalf of the AJCC Soft Tissue Sarcoma Committee (2006–2008).