Clinical characteristics and outcomes for a modern series of primary gliosarcoma patients




Primary gliosarcoma (PGS) is a rare central nervous system tumor with limited experience reported in the literature. In the current study, the authors present a modern series of confirmed PGS cases treated in the era of magnetic resonance imaging (MRI), after the accepted glioblastoma management of resection, radiation, and temozolomide.


Using a retrospective review, patients with confirmed PGS were identified (1996-2008). Cases were determined to be PGS by central pathology review using the 2007 World Health Organization criteria. Extensive chart review was performed to gather clinical and pathologic data on these cases.


All but 1 patient had undergone a preoperative MRI, with 1 patient receiving a computed tomography scan due to a cardiac pacemaker. A total of 10 patients received radiotherapy with concurrent and adjuvant temozolomide chemotherapy, and 8 patients received radiotherapy alone or in combination with other chemotherapeutic agents. In 2 patients, the history of adjuvant treatment could not be confirmed. The overall median survival was 13.9 months (range, 2.2-22.9 months). Patients with gliosarcomas resembling meningioma were found to have a significantly prolonged median survival compared with patients harboring gliosarcoma resembling glioblastoma multiforme (16 months vs 9.6 months; P = .011). However, no difference in survival was noted between patients who received concurrent radiotherapy and temozolomide compared with those who did not (10.4 months vs 13.9 months; P = .946).


The results of the current study support previous hypotheses that there are 2 distinct types of PGS. The type mimicking the appearance of a meningioma appears to carry a significantly more favorable prognosis, most likely due to an increased chance at achieving macroscopic total resection. Cancer 2010. © 2010 American Cancer Society.

Primary gliosarcoma (PGS) is a central nervous system neoplasm that contains distinct elements of both glioma and sarcoma. It was originally described by Stroebe in 1895, and gained wide acceptance after detailed histologic descriptions by Feigin et al.1, 2 In 2000, PGS was classified by the World Health Organization (WHO) grading scheme as a variant of glioblastoma multiforme (GBM).3 The current accepted definition in the 2007 WHO classification of PGS is a well-circumscribed lesion with clearly identifiable gliomatous and metaplastic mesenchymal components.4 The tumor contains a portion that satisfies the histologic criteria for GBM, and a mesenchymal component that may display a variety of morphologies with origins from fibroblastic, cartilaginous, osseous, smooth muscle, striated muscle, or adipose cell lineage.3

PGS comprises only 1.8% to 2.8% of all cases of GBM.5 Clinically similar to GBMs, they afflict adults in their fifth to seventh decades of life, and men are more commonly affected. However, a growing body of clinical and pathologic evidence supports gliosarcoma as a distinct entity from glioblastoma; recent epidemiologic analysis by Kozak et al has suggested the prognosis of gliosarcoma to be poorer than that of glioblastoma,6 and several studies have discovered a unique genetic profile of gliosarcomas when compared with glioblastomas.7-9 The current body of literature reporting on these rare tumors is limited to case reports and small series. Reported experience with response to treatment modalities is also limited; therefore, PGS patients continue to be managed in a manner similar to patients with GBM.10

In this report, we describe the clinical and radiologic presentation, pathologic diagnosis, and treatment outcomes for a consecutive series of patients with PGS. This study focuses on a series of PGS cases imaged with magnetic resonance imaging (MRI) and treated with the latest GBM regimen using temozolomide, with differential characterization of potential subtypes of PGS.


Patients with PGS were initially identified through the database of our Department of Pathology with dates of diagnosis from 1996 through 2008. The clinical history of the patients was gathered retrospectively by chart review. The cases were considered primary, de novo gliosarcoma and were included in our analysis if: 1) the patient's history did not contain any previous diagnosis of an intracranial neoplasm; and 2) the pathologic diagnosis from surgery confirmed gliosarcoma, using the criteria described earlier. Through the chart review, detailed clinical, radiologic, and pathologic characteristics of the cases were obtained. Histopathologic samples of all cases were reviewed centrally. The histologic diagnosis was established using the 2007 WHO criteria, specifically by determining: 1) the presence of dual morphologies in the tumor on hematoxylin and eosin (H&E) staining (1 of glial morphology and another of spindle morphology); 2) whether the area of glial morphology stained positive using antibodies against glial markers (glial fibrillary acidic protein [GFAP], Olig-2); and 3) whether the area appearing sarcomatous on H&E was negative for glial markers, yet positive for mesenchymal markers (reticulin).

Patients were divided into 2 groups based on macroscopic descriptions of the tumor at the time of surgery by the surgeon, and the survival of each group was compared. Second analysis compared the survival of patients who received the currently recommended therapy for GBM of radiotherapy with concurrent and adjuvant temozolomide.

Findings on preoperative imaging were reviewed through the documentation of the radiologist's report and review of the films by the senior author (A.T.P.). The length of survival was calculated from the date of diagnosis to the date of death as acquired through the Social Security Death Index (SSDI) and chart review. The log-rank test (univariate analysis) was used for statistical comparison of survival of groups of patients. The Pearson chi-square test was used when comparing radiographic characteristics. The Student t test for paired data was used when appropriate. The P value was considered statistically significant at the 5% level (P = .05). Multivariate analysis was not attempted due to the small number of patients. The study protocol was reviewed and approved by the Committee for Human Research (CHR#: H41995-32867-01).


Clinical Characteristics

From 1996 through 2008, 20 patients with histologically confirmed PGS received treatment at the study center. The mean age of the cohort in the current study was 60.8 years (range, 42‒90 years) and there were 12 men and 8 women. The patients presented with signs and symptoms consistent with an expanding intracranial mass, including headache, seizure, aphasia, hemiparesis, hemianopsia, and visual hallucination. The location of the PGS was frontal in 9 patients (45%), temporal in 6 patients (25%), parietal in 2 patients (10%), parieto-occipital in 2 patients (10%), and cerebellar in 1 patient (5%). The current series did not include any patients with findings of gliosarcoma metastases.

Radiographic Characteristics

Preoperative MRI was performed in 19 patients, and computed tomography (CT) was performed in 1 patient due to the presence of a cardiac pacemaker. The MRI results described a heterogeneously enhancing mass with rim enhancement in 5 patients, a heterogeneously enhancing mass without rim enhancement in 10 patients, a homogeneously enhancing mass with associated rim enhancement in 1 patient, and a homogeneously enhancing mass without rim enhancement in 1 patient. A thick rim of enhancement alone was described in 2 patients. Prominent surrounding edema detected by T2 prolongation was found in 10 patients. A hyperintense mass with surrounding edema was observed on the single CT performed. Representative radiographic images are shown in Figure 1.

Figure 1.

Preoperative magnetic resonance imaging of gliosarcoma is shown. (A) A gadolinium-enhanced, T1-weighted axial image demonstrated a heterogeneously rim-enhancing right parieto-occipital lesion with associated edema evident on (B) T2-weighted and (C) fluid-attenuated inversion recovery (FLAIR) sequences.

Surgical Findings

Surgical resection was performed for 19 patients, and a biopsy was performed in 1 patient with a cerebellar gliosarcoma (Fig. 2). A gross (macroscopic) total resection (GTR) was achieved in 13 patients, and subtotal resection (STR) of the tumor was performed in 5 patients. One patient underwent the initial surgical resection at an outside institution, and the reports were not available.

Figure 2.

Histology of gliosarcoma is shown. (A) Hematoxylin and eosin (H&E) stain of the astrocytic component of the tumor (× 400) is shown. (B) Strong diffuse glial fibrillary acidic protein (GFAP) staining is evident in the astrocytic portion of the gliosarcoma (× 400). (C) H&E stain is shown (× 100). A portion of the tumor demonstrated sarcomatous, spindle morphology. (D) The sarcomatous component was negative for GFAP (× 400) and (E) demonstrated focal, weak CD34-positive staining (× 400). (F) Individual cell investment of basement membrane is highlighted by Laidlaw reticulin impregnation in the sarcomatous component of the tumor (× 400; scale bar = 100 microns).

Postsurgical Treatment

There were 18 patients who received adjuvant radiotherapy, and 3 of these patients also received gamma knife radiosurgery. All patients who received radiotherapy received fractionated focal external beam radiotherapy. In 2 patients, the radiotherapy history could not be confirmed. There were 10 patients who were diagnosed more recently (2003 and later) and received adjuvant chemotherapy with temozolomide, a protocol described by Stupp et al for patients with glioblastoma.11 One patient received chemotherapy with carmustine, and another patient received the regimen of procarbazine, CCNU (lomustine), and vincristine (PCV). A total of 6 patients were initially treated with radiotherapy alone after surgical resection. At the time of disease progression, 4 patients received carmustine, 1 received CCNU, 3 received temozolomide, 3 received tamoxifen, 2 received carboplatin, 2 received procarbazine, and 1 received irinotecan. There were 3 patients who went on to receive bevacizumab at the time of disease recurrence after temozolomide therapy. There were 11 patients who underwent repeat surgery at the time of disease progression.


The precise length of survival from the time of diagnosis of PGS was determined for 16 patients, and 2 patients were confirmed to be alive at the time of last follow‒up (Fig. 3). The date of death was unavailable through the SSDI and chart review for 2 patients. The mean survival from the time of PGS diagnosis was 12.4 months, and the median survival was 13.9 months (range, 2.2-22.9 months). The mean time to disease progression after initial resection was 4.71 months (median, 3 months; range, 2-11 months). The clinical characteristics of the patients are summarized in Table 1.

Figure 3.

The overall survival of all patients (n = 20) is shown. The median survival was 13.9 months.

Table 1. Clinical Characteristics of Patients
CaseAge, Years/SexLocation of GSType of SurgerySurgical FindingRadiotherapyChemotherapyTime to Disease Progression, MonthsSurvival, Months
  1. GS indicates gliosarcoma; P, parietal; GTR, gross (macroscopic) total resection; Menin, resembling a meningioma; GKRS, gamma knife radiosurgery; BCNU, carmustine; T, temporal; STR, subtotal resection; G, resembling a glioblastoma multiforme; NA, the data were not obtainable; F, frontal; —, no treatment; PCV, procarbazine, CCNU (lomustine), and vincristine; Bx, biopsy; O, occipital; OSH, outside hospital; CPT-11, irinotecan.

153/ManRight PGTRMeninExternal beam; GKRS on recurrenceBCNU; procarbazine on recurrence316
390/WomanRight FGTRMeninExternal beam35.6
452/WomanLeft FSTRGExternal beamPCV; temozolomide, tamoxifen on recurrence313.8
575/ManLeft PGTRGExternal beam; GKRS on recurrenceNA15.2
664/ManRight TGTRMeninNANANA17.1
744/WomanRight FGTRMeninExternal beam; GKRS on recurrenceTemozolomide, BCNU, tamoxifen on recurrence722.9
857/ManLeft FGTRMeninExternal beamTemozolomide, BCNU on recurrence513.9
968/ManRight cerebellumBx onlyGExternal beamConcurrent and adjuvant temozolomide; CCNU, thalidomide on recurrence39.6
1065/WomanRight FGTRMeninExternal beamConcurrent and adjuvant temozolomide; BCNU on recurrence310.4
1159/WomanLeft FGTRMeninExternal beamConcurrent and adjuvant temozolomide; BCNU, thalidomide, carboplatin on recurrence820.1
1262/ManRight FGTRGExternal beamTemozolomide on recurrence39.6
1371/ManRight TSTRGExternal beam22.2
1462/ManRight TSTRGExternal beamConcurrent and adjuvant temozolomide22.5
1569/WomanRight FSTRGExternal beamConcurrent and adjuvant temozolomide54.7
1651/WomanRight P, OGTRMeninExternal beamConcurrent and adjuvant temozolomide915.9
1749/ManRight P, OGTRGExternal beamConcurrent and adjuvant temozolomide; bevacizumab on recurrence3NA
1849/ManRight TResection at OSHMeninExternal beamConcurrent and adjuvant temozolomide; carboplatin, CPT-11, bevacizumab on recurrence1118.4
1963/ManRight FGTRGExternal beamConcurrent and adjuvant temozolomide; bevacizumab on recurrence5+11 (alive)
2070/WomanLeft TGTRGExternal beamConcurrent and adjuvant temozolomide; bevacizumab on recurrence5+8.5 (alive)
Mean60.75/60% Men40% F, 25% T68% GTR45% Menin  4.7112.4

Clinical Characteristics by Differential Surgical Findings

For patients undergoing attempted resection, at surgery the tumor macroscopically appeared firm and distinct from normal tissue, with sharp, well-demarcated borders in 9 patients. Often, the surgeon described a tumor that was easily removed and appeared to be similar to a meningioma. The other 11 tumors more closely resembled GBM, and were often described as hypervascular, necrotic, infiltrative, and difficult to distinguish from underlying normal tissue. In the patients with meningioma-appearing tumors, there was a greater proportion of tumors that displayed significant rim enhancement on MRI (5 of 10 tumors vs 3 of 10 tumors, respectively); however, these results were not found to be statistically significant (chi-square test, P = .650). In the group of patients with tumors with meningioma-like features, there were a greater number of tumors that demonstrated homogeneous enhancement compared with the GBM-appearing group (2 of 10 tumors vs 0 of 10 tumors, respectively); however, these results were also not found to be statistically significant (chi-square test, P = .474).

The survival of the 2 groups of patients with these types of tumors differed significantly (Fig. 4). The median survival of the patients with meningioma-like PGS was 16.0 months, and the median survival of patients with tumors resembling GBM was 9.5 months (P = .011). The median time to disease progression was also longer for patients with PGS resembling meningioma compared with patients with PGS that resembled GBM (median of 6 months and 3 months, respectively; P = .025). A greater proportion of meningioma-resembling tumors was removed with macroscopic total resection (8 of 8 tumors compared with 5 of 10 for GBM‒like tumors). The age and proportion of patients who received radiotherapy and chemotherapy was similar for both groups (Table 2).

Figure 4.

Comparison of the survival of patients with tumors macroscopically resembling a meningioma (n = 9) and the survival of patients with tumors resembling a glioblastoma (n = 11) is shown. The median survival times were 16.0 months and 9.6 months, respectively. Statistical significance was determined using the log-rank test: P = .011.

Table 2. Comparison of 2 Types of Gliosarcoma
Type Based on Surgical FindingNo.Mean Age, YearsProportion Who Underwent GTRProportion Who Received Stupp ProtocolMedian Survival, MonthsMean Survival, MonthsStatistical Significance of SurvivalMedian Time to Disease Progression, MonthsMean Time to Disease Progression, MonthsStatistical Significance of Time to Disease Progression
  • GTR indicates gross (macroscopic) total resection; GBM, glioblastoma multiforme.

  • a

    One patient underwent surgical resection at an outside institution and information regarding the degree of resection was unavailable.

  • b

    One patient received a biopsy only.

Resembling meningioma959.18/8a4/916.015.6P=.0116.06.1P=.025
Resembling GBM1162.15/10b6/

Response to Radiotherapy With Concurrent Temozolomide

There were 10 patients in the current series who were treated with the currently accepted therapy for GBM, which includes surgical resection followed by radiotherapy and adjuvant temozolomide. Compared with the group of 10 patients who did not receive this protocol, there was no significant difference noted with regard to the groups' mean or median survival (Fig. 5). The group that received radiotherapy and temozolomide had a median survival of 10.4 months compared with a median survival of 13.9 months noted in the other group (P = .946). The time to disease progression between these 2 groups did not differ significantly (median of 5 months and 3 months, respectively; P = .13). Other clinical characteristics and prognostic factors, including the patients' demographics and proportion of macroscopic total resections achieved, were similar between the 2 groups (Table 3).

Figure 5.

Comparison of the survival of patients with primary gliosarcoma who underwent surgical resection and radiotherapy with concurrent and adjuvant temozolomide (n = 10) and those who did not (n = 10) is shown. The median survival times were 10.4 months and 13.9 months, respectively. Statistical significance was determined using the log-rank test: P = .946.

Table 3. Comparison of Patients Receiving Different Modalities of Treatment
TreatmentNo.Mean Age, YearsProportion Who Underwent GTRProportion Who Have Meningioma- Resembling TumorsMedian Survival, MonthsMean Survival, MonthsStatistical Significance of SurvivalMedian Time to Disease Progression, MonthsMean Time to Disease Progression, MonthsStatistical Significance of Time to Disease Progression
  • GTR indicates gross (macroscopic) total resection.

  • a

    One patient underwent surgical resection at an outside institution and information regarding the degree of resection was unavailable, and another patient received a biopsy only.

Radiotherapy plus concurrent and adjuvant temozolomide1060.56/8a4/1010.411.6P=.9465.05.4P=.13


To our knowledge to date, the published literature contains only 8 reports that include ≥10 cases of gliosarcoma.5, 10, 12-17 However, the majority of these studies did not specify a diagnosis of PGS, and included a mixture of both de novo and secondary gliosarcomas. The report by Perry et al in 1995 suggested a more favorable prognosis in patients with secondary gliosarcoma, underscoring the importance of keeping the primary type and gliosarcoma diagnosed at disease recurrence of prior malignant glioma distinct.15 Their report included 25 patients with PGS and 7 patients with secondary gliosarcoma.15 The report by Meis et al, published in 1991, described 26 patients with PGS.12 Lutterbach et al also reported 11 cases of PGS in their series.5 Because to the best of our knowledge experience with the management of PGS is limited to case series, these patients continue to be managed in the same manner as GBM patients. Advances in the treatment for GBM have also led to changes in the treatments for gliosarcoma. To the best of our knowledge, all the large published case series, including the latest by Salvati et al published in 2005, predate the currently accepted treatment regimen using surgical resection followed by radiotherapy with adjuvant temozolomide for patients with GBM.16

The mean age in the sixth decade and the higher proportion of males noted in the patients in the current study are features consistent with previous reports describing the demographic similarity of patients with PGS and GBM.5, 15 Several studies have found that PGS is most commonly localized to the temporal lobe,5, 10, 14, 16 whereas other reports indicate a highest incidence in the frontal lobe.12, 15 In the current series, the frontal lobe was the most frequent location, accounting for 45% of cases. In contrast to GBM, it is likely that PGS does have a slight temporal lobe predilection, but it may not be strong enough to be noted in series with relatively small patient numbers such as ours. The radiographic characteristics of PGS are similar to those of GBM, with prominent associated peritumoral edema and mass effect.

The development of metastasis from PGS is well established through numerous case reports, and the rate of metastasis found in literature is 11%.1, 18-30 Given the relative rarity of PGS, these reports support the clinical experience that PGS may have a greater potential for metastasis than GBM. However, in our series of 20 consecutive patients, there were no documented cases of metastases. Determining the true rate of PGS metastases will require a larger case series than those currently published in literature, and postmortem autopsy to detect asymptomatic PGS metastasis.

Even in the early descriptions by Feigin et al, surgeons noted that the macroscopic appearance of PGS was often firm and well circumscribed, clearly distinct from the appearance of GBM, whereas at other times, the tumor appeared indistinguishable from GBM.13 Subsequently, several authors likened this firm, well-demarcated appearance to that of a meningioma.16, 31, 32 These reports correlated the meningioma-like macroscopic appearance to a homogeneously contrast-enhancing mass with sharply defined borders observed on CT. In the current series, there was a larger proportion of tumors that demonstrated homogeneous enhancement patterns and significant rim enhancement on MRI noted in patients whose tumors were found to appear to be similar to a meningioma at surgery. However, this correlation was not statistically significant, most likely due to the underpowered nature of our analysis.

Cervoni and Salvati correlated the prevalence of the sarcomatous component found on histology (>50%) with the firm, well-demarcated appearance of PGS. The authors did not state the specific methods of their histologic analysis; however, drawing conclusions regarding the prevalence of 1 component over another in a large tumor by histologic analysis suffers heavily from the challenges of sampling error. Hence, this type of analysis was not pursued in the current study. En bloc resections of all tumors are likely necessary for adequate analysis of the relative prevalence of the 2 components of gliosarcoma.

Maiuri et al first noted a trend toward longer survival in patients with PGS that appeared similar to a meningioma on imaging and at surgery.32 Cervoni and Celli also compared the survival of patients with PGS that resembled meningioma on imaging and at surgery with that of those whose PGS did not.31 The patients with tumors appearing similar to a meningioma on imaging and at surgery had a longer survival of 14 months compared with 7 months in the patients whose tumors did not resemble meningioma (no P value or statistical method provided).31 In a series of 11 patients, Salvati et al also proposed the existence of 2 distinct types of PGS based on appearance, and these authors were able to demonstrate a statistically significant difference in survival between the 2 groups.16

In the current series, there was a significantly longer survival noted in the group of PGS patients with tumors resembling meningiomas. Patients in both groups underwent similar treatments. In the group of patients with PGS that resembled meningioma, however, macroscopic total resection of the tumor was achieved in a greater proportion of patients. This difference was also noted in the series by Salvati et al, and likely represents the relative ease of resection of this type of PGS compared with the type resembling GBM. Therefore, the results of the current study support the existence of 2 distinct types of PGS with a survival difference that may be related to the extent of resection performed.

Half of the patients in the current study received the currently accepted protocol using radiotherapy with concurrent and adjuvant temozolomide for the treatment of GBM. The landmark report by Stupp et al in 2005 found a 2‒month increase in survival in patients with GBM who were treated with radiotherapy and adjuvant temozolomide compared with patients who received radiotherapy alone.11 In the current series, there was no difference in the length of survival noted between patients with gliosarcoma who received radiotherapy with temozolomide compared with those who did not. However, the current study lacks statistical power for adequate analysis of the potential therapeutic benefit of radiotherapy and temozolomide, and the small size of our study significantly limits recommending a standard of therapy for these rare tumors.

The limitations of the current study include the heterogeneity in baseline prognostic factors, and treatment strategies used for patients, largely due to the retrospective nature of our analysis. This heterogeneity and the lack of robust statistical power of the study may confound our analysis of prognostic factors. Ideally, a multivariate proportional hazards model would have been used. Specifically, multivariate analysis would determine whether tumors of the meningioma-like subtype are a prognostic factor independent of macroscopic total resection, and the impact of achieving macroscopic total resection and having a meningioma-like subtype on survival would be compared to determine which factor is of greater significance. However, the small sample size of the current series prevented the appropriate building of this technique. Large prospective trials remain difficult due to the rarity of these tumors. Despite these difficulties, further studies, inclusive of large numbers of patients, are necessary to adequately study the role of this treatment protocol while accounting for age, extent of surgical resection, baseline performance status, and other relevant prognostic factors.


In the current study, we describe a modern series of 20 consecutive patients diagnosed with PGS. This study supports previous findings that there are 2 distinct types of PGS: 1 that is similar to a GBM and 1 that macroscopically mimics a meningioma. The latter type's similarity to meningioma appears to allow for a greater extent of resection and likely facilitates a more favorable prognosis. As molecular attributes that distinguish PGS from GBM are elucidated, there may be a continued need to consider PGS as a unique entity with potential novel therapeutic targets.


Dr. Han received funding from the Howard Hughes Medical Institute Research Training Fellowship.

Dr. Yang received a National Research Service Award.

Dr. Parsa is a Reza and Georgianna Khatib Endowed Chair in Skull Base Tumor Surgery.