Original Article neup_1297 1..9

A study of clinico-pathological parameters andO6 – methylguanine DNA methyltransferase

(MGMT) promoter methylation status inthe prognostication of gliosarcoma

Geetika Singh,1 Supriyo Mallick,2 Vikas Sharma,1 Nikhil Joshi,2 Suvendu Purkait,1 Prerana Jha,1

Mehar Chand Sharma,1 Vaishali Suri,1 Pramod Kumar Julka,2 Ashok Kumar Mahapatra,3

Manmohan Singh,3 Shashank Sharad Kale3 and Chitra Sarkar1

Departments of 1Pathology, 2Radiation Oncology and 3Neurosurgery, All India Institute of Medical Sciences,New Delhi, India

Gliosarcoma is a rare variant of glioblastoma multiforme(GBM) with similar clinical presentation and prognosis buta distinct genetic profile. The clinicopathological featuresof 22 cases of gliosarcoma were analyzed with respect toage, sex, KPS score, operative diagnosis, extent of resectionand histopathological subtype (predominantly sarcoma-tous [PS], predominantly gliomatous [PG] or mixed).Twelve cases were PS, six were PG and four were mixed.The histological subtype did not correlate with the opera-tive diagnosis; however, it did significantly correlate withthe extent of resection (P = 0.014). In 14 cases with avail-able survival data it was found that none of the clinico-pathological parameters significantly correlated withsurvival (P > 0.05). Methyl guanine DNA methyl trans-ferase promoter methylation studies were performed usingmethylation-specific PCR in 16 cases which showed amethylation rate of 31.25% (5/16). The promoter methyla-tion status did not correlate with the histological subtypeand did not significantly affect survival (P > 0.05).Although gliosarcomas continue to be treated in the sameway as GBM, the role of chemotherapy with temozolomideis not clear. This cohort is the largest to date to uniformlyreceive the Stupp’s protocol which is currently “standard ofcare” for GBM. A median overall survival of 18.5 months issubstantially higher than previous studies, suggesting thattemozolomide should be included in gliosarcoma therapy.

Key words: glioblastoma multiforme, glioma, gliosarcoma,MGMT, temozolomide.

INTRODUCTION

Gliosarcoma is a rare variant of glioblastoma multiforme(GBM) defined by its biphasic pattern on pathology.1 Itsclinical presentation, prognosis and therapy are similar toGBM, although differences in the genetic profile2 suggestthat gliosarcomas are a distinct entity.

Various authors have studied prognostic factors inGBM, including O6 – methyl guanine DNA methyl trans-ferase (MGMT) promoter methylation status.3–5 It is awell known epigenetic phenomenon associated with bettertreatment response and prognosis in GBMs,4,6,7 although itsimportance in gliosarcoma has been analyzed in only onestudy to date.8

The significance of the proportion of the glial and mes-enchymal elements in gliosarcoma is another area whichhas been the subject of recent studies. Salvati et al.9 pro-posed two pathological subtypes of gliosarcomas: predomi-nantly gliomatous (PG) and predominantly sarcomatous(PS), with better survival in the PS group probably dueits superficial and well circumscribed “meningioma-like”nature allowing for better resection. Similar observationswere also made by Han et al.10

This study aimed to analyze the clinicopathological fea-tures of a series of uniformly treated gliosarcomas andassess the impact of various prognostic factors, notablypathological subtyping and MGMT promoter methylationstatus on survival.

Correspondence: Chitra Sarkar, MD, Department of Pathology,AIIMS, New Delhi 110029, India. Email: sarkar.chitra@gmail.com

Received 7 November 2011; revised 9 January 2012 and accepted 10January 2012.

Neuropathology 2012; ••, ••–•• doi:10.1111/j.1440-1789.2012.01297.x

© 2012 Japanese Society of Neuropathology

MATERIALS AND METHODS

The neuropathology records of the Department of Pathol-ogy, All India Institute of Medical Sciences, New Delhi,India, from 2003 to 2010 were retrospectively analyzed and22 cases of gliosarcoma were retrieved.

Pathology

Hematoxylin and eosin, reticulin histochemistry andGFAP immunohistochemistry was performed in eachcase. The slides were reviewed by two independentpathologists (CS, GS) and consensus diagnoses reached asper the WHO classification 2007.1 Reticulin and GFAPhelped to delineate the glial and mesenchymal elementswherein the sarcomatous component was reticulin-rich and GFAP-negative compared to the gliomatouscomponent which was reticulin-poor and GFAP-immunopositive. (Fig. 1) The proportion of the sarcoma-tous element was determined by obtaining a percentageon each reticulin-stained section and then summating itand dividing it by the number of slides to obtain theaverage percentage in the whole tumor sampled. Basedon the proportion of the sarcomatous element, tumorswere divided using criteria modified from Salvati et al.9

into PS (>50% sarcomatous), PG (225% sarcomatous)and mixed (25–50% sarcomatous).

Clinical data

In all patients, age, sex, location of the tumor and type ofsurgery (total or subtotal resection) were obtained. Wher-ever possible, operative diagnoses and follow-up data wasretrieved. All the patients received “standard of care”therapy comprising of maximally safe surgical resectionfollowed by uniform adjuvant therapy based on Stupp’sprotocol.11 This consisted of adjuvant radiation to a doseof 60 Gy in 30 fractions, five fractions a week with con-

current temozolomide 75 mg/m2. This was followed 4weeks later by adjuvant temozolomide 150 mg/m2 for 5days (cycle 1) and if well tolerated, 200 mg/m2 for 5 daysevery 28 days for a total of six cycles. Patients not pro-gressing at the end of this standard treatment wereoffered further temozolomide at the same dose and fre-quency till tolerance was reached or progression wasnoted.

Tissue procurement and DNA preparation

In 19 of the 22 cases blocks available had sufficient mate-rial for DNA extraction. Corresponding blocks of slideswith adequate amounts of tumor were then serial-sectioned, with approximately 10 serial sections of 10microns thickness each. The tumor area was marked bycomparing with the corresponding HE section and thenwas scraped off, taking care to avoid excess surroundingparaffin wax. The tumor scrapings were stored in sterilemarked vials at 4°C. DNA isolation was done usingEX-WAX DNA extraction kit (M/s Milllipore, Billerica,MA, US) according to the manufacturer’s protocol. DNAisolation from each case was done in duplicate. Briefly, foreach single isolation, the tissue was placed in a 1.5 mLtube and digested using 150 mL digestion solution, 50 mLprotein-digesting enzyme and was incubated at 52°C for6 h. The 100 mL extraction solution was added and tissuewas centrifuged at 135 000 ¥ g for 10 min. The middlesupernatant layer was collected in a new collecting tubeand 150 mL precipitation solution was added to it fol-lowed by 1 mL ethanol addition, 8 mL glycogen (10 mg/mL) addition and was kept overnight at -20°C. Afterthis the mixture was centrifuged at 135 000 ¥ g /4°C for35 min. Next the supernatant was discarded and the pelletwas dried at 60°C/10 min. The DNA pellet was dissolvedin 40 mL of re-suspension solution and was stored at-20°C till further use.

Fig. 1 A case of predominantly sarcoma-tous (PS) gliosarcoma composed of fasciclesof spindle cells (A, HE, 10¥) which is richin reticulin (B, reticulin, 10¥) and GFAP-negative (C, GFAP, 10¥) except for a smallisland in the upper right corner of the figure.This tumor had approximately 90% sarcoma-tous component. In contrast, another case ofPS gliosarcoma with 75% sarcomatous com-ponent shows an “intermixed” pattern atplaces (D, HE, 10¥) with an intimate ad-mixture of the reticulin-rich sarcomatouscomponent (E, reticulin, 10¥) and the GFAP-reactive glioblastoma component (F, GFAP,10¥).

A B C

D E F

2 G Singh et al.

© 2012 Japanese Society of Neuropathology

Methylation-specific PCR for MGMT promotermethylation status

DNA methylation pattern of the MGMT gene promoterwas determined by methylation-specific (MS)-PCR. Thisprocedure involved chemical modification of unmethy-lated cytosine to uracil, followed by a nested two-stagePCR. For each case this was done in duplicate for compari-son and validation of the results.

Genomic DNA (~500 ng) from each sample wasmodified by sodium bisuphite treatment (EZ Gold DNAmethylation kit; M/s. Zymo Research, Orange, CA, USA).Enzymatically methylated DNA was used as positivemethylation control whereas normal lymphocytic DNAserved as a non-methylation control. Methylation-specificprimers were used for PCR. First stage primer recognizesthe bisulphite-modified template flanking the MGMT genebut does not discriminate between methylated and non-methylated alleles. Primer sequences of the first PCR were5′GGATATGTTGGGATAGTT 3′ (forward primer) and5′ CCAAAAACCCCAAACCC 3′ (reverse primer). PCRamplification procedure for stage 1 was as follows: aninitial denaturation step of 5 min at 94°C; followed by 40cycles of 30 s at 94°C, 30 s at 52°C and 30 s at 72°C and afinal elongation step of 7 min at 72°C in DNA EnginePeltier Thermal Cycler (M/s. Bio-Rad, Hercules, CA, USA)using recombinant Taq DNA polymerase (M/s. FermentasLife Sciences, Glen Burnie, MD, USA). A 25 mL volumewas used in all PCR reactions. The stage 1 PCR productwas diluted 20-fold and 2 mL of this dilution was subjectedto a stage 2 PCR. Methylation- and non-methylation-specific primers were used separately for each test. Primersequences for the non-methylated reaction were 5′-TTTGTGTTTTGATGTTTGTAGGTTTTTGT-3′ (forwardprimer) and 5′-AACTCCACACTCTTCCAAAAACAAAACA-3′ (reverse primer) and for the methylatedreaction were 5′-TTTCGACGTTCGTAGGTTTTCGC-3′(forward primer) and 5′-GCACTCTTCCGAAAACGAAACG-3′ (reverse primer).

The PCR protocol for stage 2 was as follows: an initialdenaturation step of 5 min at 94°C, followed by 35 cycles of15 s at 94°C, 15 s at 62°C and 15 s at 72°C and final elon-gation step of 7 min at 72°C. MS-PCR was performed toamplify 83 base pair fragments of the methylated MGMTgene promoter and 91 base pair fragments of the non-methylated product.Amplified products were separated on4% agarose gel, ethidium bromide-stained and visualizedunder UV illumination.

Statistics

Statistical analysis was performed for correlation of extentof resection, operative diagnosis, histopathology andMGMT methylation status with survival. Variables were

analyzed by the log rank test, chi-square test or indepen-dent t-test as applicable and Kaplan-Meier survivalcurves were generated. A P-value of <0.05 was consideredsignificant.

RESULTS

Over 8 years, 22 cases of gliosarcoma were retrieved(Table 1), which comprised 2.64% of 832 GBMs.

Clinical features

Age and sex

Mean age at presentation was 37.5 years (range 7–65years). Most cases fell in the 21 to 40 years bracket, that is,i.e. the third to fourth decades. Three cases of pediatricgliosarcoma (7 years, 10 years and 12 years of age) werealso included. A male predominance was noted with 16males and six females (M : F = 2.6:1.0).

Site

Nearly half of the tumors were located in the temporallobe (47.6%), followed by the frontal (23.8%) and parietallobes (14.28%). One tumor each was noted in occipital,intraventricular and cerebellar locations.

Operative impression

Operative impression (based on the surgeon’s operativenotes) was divided into “meningioma-like” versus “GBM-like” tumors. Of the 21 cases with available details, 15 were“GBM-like”, that is, soft, necrotic and ill-defined and sixwere “meningioma-like”, that is, firm, well-circumscribedand peripherally located.

Extent of resection

In 19 cases extent of resection was known, of which 10underwent gross total resection (GTR) and nine under-went subtotal resection (STR). GTR was defined asresidual tumor 210% and STR was defined as residualtumor 250%. Of the six “meningioma-like” tumors, four(66%) underwent GTR and two (33%) underwent STRcompared to the 13 “GBM-like” tumors in which nearly50% each underwent STR and GTR (6 cases GTR and 7cases STR).

Treatment

All the patients received “standard of care” therapy com-prising of maximally safe surgical resection followed byuniform adjuvant therapy based on Stupp’s protocol.11 Inevaluation prior to the start of adjuvant therapy, a Karnof-sky performance status (KPS) score was assigned which

Prognostication of gliosarcoma 3

© 2012 Japanese Society of Neuropathology

was 70 and above for all patients. Chemoradiotherapy waswell-tolerated in all cases.

Follow-up

Eight patients were lost to follow-up. The rest were underregular follow-up which included at least two contrast-enhanced MRIs at the third and sixth months of adjuvantchemotherapy. Median follow-up was 13.15 months (range1.5–46 months).

Endpoints were defined as expired due to progressivedisease (E) or no evidence of disease (NED). NED wasbased on absence of symptoms and absence of tumor onradiology or both. Seven patients expired due to progres-sive disease and their follow-up ranged from 1.5 months to24.8 months (mean 12.8 months). Seven patients had noevidence of disease on follow-up at the time of this study,with follow-up ranging from 6.3 months to 46 months(mean 21.2 months).

Pathology and genetics

Histopathology (Fig. 1 a–f)

All the tumors showed biphasic pathology. The sarcoma-tous element was fibroblastic in 100% of cases and com-posed of fascicles of malignant oval to spindle cells withmoderate eosinophilic cytoplasm. No heterologous ele-ments were noted.The glial component fulfilled the criteria

for GBM. Twelve cases (54.5%) fell into the PS and six(27.3%) fell into the PG category. Four cases (18.2%) hada sarcomatous component between 26% and 50% andwere placed in the mixed category. Unlike Salvati et al.,9

complete concordance was not noted between the opera-tive impression and the pathological subtype. In fact, sevenof the 15 tumors which were “GBM-like” according to thesurgeon turned out to be PS on histology, five were PG andthree were mixed. Of the six “meningioma-like” tumorsfour were PS, one was PG and one was mixed. Thus nosignificant correlation was noted between the histologyand operative diagnosis (P = 0.681).

However, a highly significant association was notedbetween the extent of resection and pathological subtype(P = 0.014). All the PG tumors underwent STR comparedto only 2/8 PS tumors; 75% (6/8) of the PS tumors under-went GTR. In the mixed tumors, 50% underwent GTR and50% underwent STR.Thus PS tumors were associated withbetter degree of resection, probably owing to their firmnessand circumscription compared to PG tumors. This associa-tion was even stronger if the cut-off for PS was reduced to325% (P = 0.011).

Methylation status of the MGMT promoter

Methylation studies were performed in 16 cases with suf-ficient DNA (Fig. 2). Methylated promoter was detectedin five cases (31.25%), and non-methylated in 11 cases

Table 1 Clinicopathological features and methyl guanine DNA methyl transferase (MGMT) promoter methylation status of cases ofgliosarcoma

SNO. Age Sex KPS Surgery Operativediagnosis

Site Histology % S MGMT Status Follow-up(months)

1 10 M 90 GTR GBM L Frontal PS 90 UM NED 462 37 M 80 STR GBM R Temporal PG 25 UM E 8.53 65 F 90 STR GBM R Temporal PS 60 UM E 1.54 7 F 100 GTR Meningioma R Lat Ventricle Mixed 40 UM NED 19.55 34 M 70 STR GBM R Temporal Mixed 40 – E 24.86 50 M 80 GTR Meningioma L Temporal PS 50 UM E 18.57 35 M 90 STR GBM L Temporal PG 25 UM E 15.48 62 M 100 STR Meningioma R Temporal PS 55 UM NED 10.99 35 M 90 GTR GBM† L Temporal PS 75 UM E 7.8

10 50 M 90 GTR Meningioma R Temporal PS 50 M NED 6.311 35 M 90 GTR GBM R Parietal PS 75 UM NED 25.612 18 F 70 GTR Meningioma L Parietal PS 90 M NED 2313 36 M 80 STR GBM R Frontal Mixed 30 – NED 714 43 M 70 – GBM R Fronto-parietal PS 90 – E 715 32 M – STR GBM R Temporal PG 25 M – –16 58 F – STR Meningioma R Frontal PG 25 – – –17 40 M – GTR GBM L Parietal Mixed 30 – – –18 26 M – – GBM Cerebellum PG 5 UM – –19 12 F – STR GBM – PG 5 – – –20 62 M – – – R Occipetal PS 50 M – –21 30 F – GTR GBM L Frontal PS 65 UM – –22 49 M – GTR GBM‡ L Temp PS 85 M – –

†Previous AA 5 years ago, ‡Previous GBM 1 year ago. %S, percentage sarcomatous area; –, not available; E, expired; F, female; GBM,glioblastoma multiforme; GTR, gross total resection; KPS, Karnofsky performance score; L, left; M, male; M, methylated; NED, no evidence ofdisease; PG, predominantly gliomatous; PS, predominantly sarcomatous; R, right; STR, subtotal resection; UM, non-methylated.

4 G Singh et al.

© 2012 Japanese Society of Neuropathology

(68.75%). In both the PS and PG groups, cases were pre-dominantly non-methylated; however, among the methy-lated tumors, 4/5 (80%) were PS in histology. However, nosignificant correlation was noted between the promotermethylation status and histological subtype (P = 0.719).

Survival data and correlations

Survival data was available in 14 cases. The median overallsurvival was 18.5 months (2 years survival rate of 48.6%).

Age and sex did not correlate with survival (P > 0.05).

Operative impression

Sixty-six percent of patients with “GBM-like” tumorsexpired with a median survival of 15.4 months versus only20% of “meningioma-like” tumors (P = 0.22) (Fig. 3a)

Extent of resection

Seventy-one percent of cases with GTR were without evi-dence of disease compared to 33% of cases with STR.However, again no significant correlation was foundbetween extent of resection and survival (P = 0.094)(Fig. 3b).

Histopathology subtype

PG tumors had a lower median overall survival of 8.5months compared to PS tumors (18.5 months) (P = 0.372)(Fig. 3c). Changing the arbitrary cut-offs of percentage sar-comatous area proposed by Salvati et al. to: (i) <50% sar-comatous (PG) and 350% sarcomatous (PS); or (ii) <25%sarcomatous (PG) and 325% sarcomatous (PS), also didnot affect survival (P = 0.823 and P = 0.184, respectively).

MGMT promoter methylation status

Two cases with methylated promoter were alive withoutevidence of disease, whereas in cases of non-methylatedpromoter nearly equal numbers of patients had expired(n = 5) or had no evidence of disease (n = 4). No significantcorrelation of MGMT promoter methylation status withsurvival was noted (P = 0.333) (Fig. 3d).

The differences in clinical and pathological featuresbetween the patients who survived and those that expiredare shown in Table 2. As only two pediatric patients hadsurvival data, a valid statistical analysis and comparisonwith adult patients was not possible. After excluding thepediatric cases, there was still no significant survival advan-tage based on operative impression, extent of resection,pathological subtype and MGMT promoter methylationstatus in the adult cases (P > 0.05).

DISCUSSION

The 2007 WHO classification of tumors of the CNS definesgliosarcoma as a glioblastoma variant characterized by abiphasic tissue pattern with alternating areas displayingglial and mesenchymal differentiation.1

Gliosarcoma has a poor prognosis with a median sur-vival between 6.25 and 11.5 months on therapy (Table 3).No statistically significant difference in survival was foundbetween GS and GBM,13,14,17,18,20 although these studiespredate the current treatment protocol for GBM. Similarto GBM, post-operative radiotherapy (RT) has beenfound to benefit cases of gliosarcoma.14,18 However, therole of chemotherapy is still unclear. In more recentstudies on gliosarcoma, temozolomide has been used. Inthe study by Han et al.10 of 20 cases, 10 received temozo-lomide in addition to RT (median overall survival [OS]10.4 months) and 10 cases received RT with or without

Fig. 2 Methylation status of the methyl guanine DNA methyltransferase (MGMT) promoter in gliosarcoma specimens by themethylation-specific PCR (MS-PCR) assay. DNA from normalperipheral blood lymphocytes (PBL) was used as a control for thenonmethylated MGMT promoter (UM), a known methylatedcase of GBM (SW48) from a previous study (7) served as a posi-tive control for the methylated MGMT promoter (M), and waterwas used as a negative control for the PCR. A 100-bp markerladder (L) was loaded to estimate molecular size. (M, PCRproduct amplified by methylated-specific primers; UM, PCRproduct amplified by non-methylated-specific primers; L, ladder;SW48, methylated control DNA; PBL, non-methylated controlDNA). Gels marked 1 to 19 represent the following cases in order:20, 17, 7, 8, 6, 2, 3, 21, 15, 22, 12, 4, 10, 9, 18, 11, 19 and 16 (seeTable 1). Cases 5, 13 and 14 did not have sufficient DNA and havenot been shown. For example, Case 10 is methylated with thepresence of a band in the M lane along with a faint band in theUM lane and Case 3 is non-methylated with no band in the Mlane. In controls used, PBL has a band only in the UM lane, SW48has a band only in the M lane and water has no bands.

Prognostication of gliosarcoma 5

© 2012 Japanese Society of Neuropathology

other chemotherapy (median OS 13.9 months) with nosignificant difference in survival (P = 0.946). Salvati et al.9

studied 11 cases, four of which received RT and temozo-lomide with a median survival of 17.4 months and fivereceived RT alone with a median survival of 15.7 months.They concluded that chemotherapy seems to have anadded benefit. The present study of 22 cases is the largestcohort to date to have uniformly received the Stupp’sprotocol.11 Survival data was available in 14 cases with amedian survival of 18.5 months, which is higher thanprevious reports on gliosarcoma with any chemoradio-

therapy regimen. It is important to note that this cohorthad a mean age of 37.5 years (mean age in the Westernpopulation is reported between 50 and 60 years, Table 3).In addition, in all cases with survival data the KPS scoreswere above 70. Younger age at presentation and highKPS scores may also contribute to better survival,although statistically age did not affect survival in thisstudy (P > 0.05). Salvati et al.9 also found that age andKPS had no influence on survival. In this scenario, it issafe to suggest that temozolomide should be included inthe standard of care for gliosarcomas.

Fig. 3 Comparison of survival of patientswith different clinicopathological param-eters. (A) Between tumors with an operativediagnosis of GBM (n = 15) and meningioma(n = 6) no significant difference in mediansurvival was found (log rank test P = 0.22).(B) Between tumors with gross total resec-tion (n = 10) and subtotal resection (n = 9) nosignificant difference in median survival wasfound (log rank test P = 0.094). (C) Betweentumors with histological subtype of predo-minantly sarcomatous (PS, n = 12) andpredominantly gliomatous (PG, n = 6) andmixed (n = 4) no significant difference inmedian survival was found (log rankP = 0.372). (D) Between tumors with methy-lated methyl guanine DNA methyl trans-ferase (MGMT) promoters (n = 5) andnon-methylated MGMT promoters (n = 11)no significant difference in median survivalwas found (log rank P = 0.333).

1.0

0.8

0.6

0.4

0.2

0.0

Cum

Sur

viva

l

1.0

0.8

0.6

0.4

0.2

0.0

Cum

Sur

viva

l

1.0

0.8

0.6

0.4

0.2

0.0

Cum

Sur

viva

l

1.0

0.8

0.6

0.4

0.2

0.0

Cum

Sur

viva

l

0 20 40months

0 20 40months

0 20 40months

0 20 40months

GBM

Mixed

Gross total resection

Meningioma

Overall survival by operative diagnosis

Overall survival by histological subtype Overall survival by MGMT promoter methylation status

Overall survival by extent of resection

Subtotal resection

PS

PG

Methylated

Unmethylated

A B

C D

Table 2 Comparison of clinicopathological parameters and methyl guanine DNA methyl transferase (MGMT) promoter methylationstatus in patients who expired or had no evidence of disease

Expired No evidence of disease P-value

Age 1 mean 42.7 1 11.38 years 31.14 1 20.61 years 0.218Age range 35–65 years 7–62 yearsSex 6 M, 1 F 5 M, 2 FGTR 2 5 P = 0.286STR 4 2GBM-like 6 3 P = 0.266Meningioma like 1 4PG 2 0 P = 0.295PS 4 5Mixed 1 2Methylated MGMT – 2 P = 0.455Unmethylated MGMT 5 4Follow up (mean) 12.8 months 21.2 monthsFollow up (range) 1.5 months–24.8 months 6.3 months–46 months

GBM, glioblastoma multiforme; GTR, gross total resection; PG, predominantly gliomatous; PS, predominantly sarcomatous; STR, subtotalresection.

6 G Singh et al.

© 2012 Japanese Society of Neuropathology

Pathological subtyping

The first indication of two subtypes of gliosarcoma in theliterature was by Maiuri et al.21 in 1990 who described CTscan findings in five cases with greater survival rates in caseswhich resembled meningioma. Cervoni and Celli22 in theirstudy of six gliosarcomas found that three which resembledmeningiomas (based on radiological and operative find-ings) had a median survival of 14 months while the otherthree which resembled glioblastoma only had a mediansurvival of 7 months. However, Parekh et al.15 did not findany correlation between radiological and operative findingsin 15 cases of gliosarcoma which were surgically resected.Han et al.10 in their study of 20 cases relied solely on opera-tive findings to classify their cases into PS and PG with signi-ficant survival differences noted (16 months vs. 9.5 months,respectively, P = 0.011). Radiological and operative diag-noses did not correlate significantly in this study, although alarger proportion of patients of PS type on surgery hadwell-circumscribed, intensely enhancing meningioma-liketumors on radiology and underwent gross total resection.The first study to look at the histology in addition to theoperative findings was by Salvati et al.9 Incidentally, in alltheir cases the histology corresponded to the surgical andradiological findings, that is, PG tumors resembled GBMand PS tumors resembled meningioma. They found a sig-nificantly better median survival in the PS group versus thePG group (71 1 6 weeks vs. 63 1 6 weeks, P = 0.0417).

In the present study, histology was used as the solecriteria to determine if the percentage of sarcomatous areaaffected biological behavior of the tumor. A third categoryof “mixed” tumors with sarcomatous component between26% and 50% was created. Unlike Salvati et al.,9 no signifi-cant difference in survival was found between the PS,mixed or PG groups. In addition, the operative impressionswere not concordant with the histological or the radiologi-cal findings, highlighting its subjectivity. As this was a ret-

rospective study, the surgeon’s operative notes were reliedupon and therefore subjective impressions like “firmerthan usual GBM” could not be assessed and only clear-cutmeningioma-like tumors were classified in a separate cat-egory. A prospective study with prior education of thesurgeons and a more accurate assessment of firmness andcircumscription may help better classify the two subgroups.PS histology correlated with better resection, which isexpected as the tumor would be comparatively firmand well-circumscribed; however independently neitherhistology nor extent of resection correlated with survival(P > 0.05).

It appears from prior studies that there is a subset ofgliosarcomas which has a significantly better survival;however, there appears to be no consensus on criteriafor definition, that is, surgical, radiological, histological ora combination of these. All the criteria used in moststudies, including the current study, have not alloweda clear demarcation between the groups. Radiologicalassessment has not correlated with operative diagnosis;operative diagnoses are subjective and histology is subjectto sampling errors. Better survival has been attributed to amore complete resection of the well-circumscribed, firm,meningioma-like PS group of gliosarcomas; however, dueto small numbers, a multivariate analysis has not been pos-sible to determine whether the PS group is a prognosticfactor independent of extent of resection.10,22 Unlike GBM,there are no well-controlled studies comparing survivalfollowing GTR or STR in gliosarcomas.

In the current study no survival advantage was found intumors with PS histology, operative impression of menin-gioma or GTR.

MGMT promoter methylation status

Methyl guanine DNA methyl transferase is a DNA repairprotein which mediates the repair of O6-guanine adducts

Table 3 Gliosarcoma case series with comparison of therapies employed and survival outcome. Series using temozolomide are in italics

Study, year Ref no. n Mean age Sx RT CT Agents Median OS (months)

Morantz et al., 1976 12 24 54 All 18 9 Mithromycin, ametophterin 4 (all); 7 (Sx + RT)Meis et al., 1991 13 26 50%

>60NA All 17 Nitrosurea, Dacarbazine,

Misonidazole8.3

Perry et al., 1995 14 32 66 31 12 6 Nitrosurea 6.25 (all); 11.5 (Sx + RT)Parekh et al., 1995 15 17 52 All 16 3 NA 9Sarkar et al., 1997 16 29 42 All All Nil None <18Galanis et al., 1998 17 18 61.5† 16 All 18 Nitrosurea 8.75Lutterbach et al., 2001 18 12 56† All All Nil None 11.5Salvati et al., 2005 9 11 52† All 9 4 Temozolomide 16.2 (all); 15.75 (Sx + RT);

17.4 (Sx + RT + CT)Han et al., 2010 10 20 52 All All 12 Temozolomide(10), BCNU (1)

and PCV (1)13.9 (Sx + RT 1 other CT);

10.4 (Sx + RT + TZ)Biswas et al., 2011 19 17 50† All 15 2 Temozolomide 8.27Current study, 2011 14 37.5 All All All Temozolomide 18.5

†Median age. CT, chemotherapy; OS, overall survival; RT, radiotherapy; Sx, surgery.

Prognostication of gliosarcoma 7

© 2012 Japanese Society of Neuropathology

which are synthesized during therapy with alkylatingagents. Several studies have shown that silencing of thegene by promoter methylation leads to improved survivalin patients with GBM receiving alkylating agents.6 In addi-tion, it has been found to be an independent favorableprognostic factor.23 Methylation rates in GBM range from16.8%24 to 89%.25 In a study at our institution, MGMTmethylation was noted in 57% of GBMs and did not cor-relate with any other known genetic alteration.7 In anotherstudy of GBMs with long-term survival, MGMT promotermethylation was noted in 5/6 tumors (83.33%).26

Studies on MGMT promoter methylation status ingliosarcomas are lacking. In the only other study Kanget al.8 performed MS-PCR and immunohistochemistry forMGMT promoter methylation on 12 cases and found thatsix (50%) were methylated by MS-PCR and 58.3% showedimmunoreactivity. Cases with methylated promoter hadsignificantly longer overall survival (15 months) comparedto those with non-methylated promoter (11.3 months)(P = 0.045). In our study we found lower methylation rates(5/16, 31.25%) and no correlation with survival (P > 0.05).Sasai et al.27 have demonstrated MGMT expression in non-neoplastic cells such as lymphocytes, vascular endothelialcells and macrophages/microglia and have recommendedimmunohistochemical evaluation of MGMT for correctinterpretation in tumor cells. Although immunohistochem-istry was not performed in this study, care was taken toexclude surrounding non-neoplastic tissue during tissueprocurement. That the MGMT promoter status of intratu-moral lymphocytes and endothelial cells altered the resultsis a possibility that cannot be excluded in this study.

To conclude, this study on a cohort of gliosarcomasshowed a median survival substantially higher thanprevious studies, suggesting that temozolomide should beincluded in the “standard of care” for this tumor. Patho-logical subtyping correlated with the extent of resection(P < 0.05). None of the clinicopathological factors studied,that is, extent of resection, operative diagnosis or patho-logical subtyping, correlated with survival. Pathologicalsubtyping is hindered by sampling error, does not seemto accurately prognosticate the tumor and is not recom-mended in the opinion of the authors. Operative impres-sions also suffer from subjectivity. MGMT promotermethylation was found in a smaller number of cases com-pared to the study by Kang et al.8 (31.25% vs. 50%) and didnot correlate with survival.

ACKNOWLEDGMENTS

The authors thank Mr Manoharan from the Delhi Cancerregistry, Dr B.R.A Institute Rotary Cancer Hospital(IRCH), All India Institiute of Medical Sciences (AIIMS),New Delhi for his help with the statistics

REFERENCES

1. Kleihues P, Burger PC, Aldape KD et al. Glioblas-toma. In: Louis DN, Ohgaki H, Wiestler OD, CaveneeWK, eds. WHO Classification of Tumors of the CentralNervous System, 4th edn. Lyon: International agencyfor Reseach on Cancer, 2007; 48–49.

2. Reis RM, Könü-Lebleblicioglu D, Lopes JM, KleihuesP, Ohgaki H. Genetic profile of gliosarcomas. Am JPathol 2000; 156: 425–432.

3. Krex D, Klink B, Hartmann C et al. Long-term survivalwith glioblastoma multiforme. Brain 2007; 130: 2596–2606.

4. Martinez R, Schackert G, Yaya-Tur R, Rojas-Marcos I,Herman JG, Esteller M. Frequent hypermethylation ofthe DNA repair gene MGMT in long-term survivorsof glioblastoma multiforme. J Neurooncol 2006; 83:91–93.

5. Shaw E, Seiferheld W, Scott C et al. Reexamining theradiation therapy oncology group (RTOG) recursivepartitioning analysis (RPA) for glioblastoma multi-forme (GBM) patients. Int J Radiat Oncol Biol Phys2003; 57: S135–S136.

6. Suri V, Jha P, Sharma MC, Sarkar C. O6-methylguanine DNA methyltransferase gene pro-moter methylation in high-grade gliomas: a review ofcurrent status. Neurol India 2011; 59: 229–235.

7. Jha P, Suri V, Jain A et al. O6-methylguanine DNAmethyltransferase gene promoter methylation status ingliomas and its correlation with other molecular alter-ations: first Indian report with review of challenges foruse in customized treatment. Neurosurgery 2010; 67:1681–1691.

8. Kang SH, Park KJ, Kim CY et al. O6-methylguanineDNA methyltransferase status determined by pro-moter methylation and immunohistochemistry ingliosarcoma and their clinical implications. J Neuroon-col 2011; 101: 477–486.

9. Salvati M, Caroli E, Raco A, Giangaspero F, Delfini R,Ferrante L. Gliosarcomas: analysis of 11 cases do twosubtypes exist? J Neurooncol 2005; 74: 59–63.

10. Han SJ, Yang I, Ahn BJ et al. Clinical characteristicsand outcomes for a modern series of primary gliosar-coma patients. Cancer 2010; 116: 1358–1366.

11. Stupp R, Hegi ME, Mason WP et al. Effects of radio-therapy with concomitant and adjuvant Temozolomideversus radiotherapy alone on survival in glioblastomain a randomized phase III study: 5-year analysis ofthe EORTC-NCIC trial. Lancet Oncol 2009; 10: 459–456.

12. Morantz RA, Feigin I, Ransohoff J 3rd. Clinical andpathological study of 24 cases of gliosarcoma. J Neu-rosurg 1976; 45: 398–408.

8 G Singh et al.

© 2012 Japanese Society of Neuropathology

13. Meis JM, Martz KL, Nelson JS. Mixed glioblastomamultiforme and sarcoma. A clinicopathologic study of26 radiation therapy oncology group cases. Cancer1991; 67: 2342–2349.

14. Perry JR, Ang LC, Bilbao JM, Muller PJ. Clinicopatho-logic features of primary and postirradiation cerebralgliosarcoma. Cancer 1995; 75: 2910–2918.

15. Parekh HC, O’Donovan DG, Sharma RR, Keogh AJ.Primary cerebral gliosarcoma: report of 17 cases. Br JNeurosurg 1995; 9: 171–178.

16. Sarkar C, Sharma MC, Sudha K, Gaikwad S, Varma A.A clinico-pathological study of 29 cases of gliosarcomawith special reference to two unique variants. Indian JMed Res 1997; 106: 229–235.

17. Galanis E, Buckner JC, Dinapoli RP et al. Clinicaloutcome of gliosarcoma compared with glioblastomamultiforme: North Central Cancer Treatment Groupresults. J Neurosurg 1998; 89: 425–430.

18. Lutterbach J, Guttenberger R, Pagenstecher A. Glio-sarcoma: a clinical study. Radiother Oncol 2001; 61:57–64.

19. Biswas A, Kumar N, Kumar P et al. Primary gliosar-coma – clinical experience from a regional cancercentre in north India. Br J Neurosurg 2011; 25: 723–729.

20. Han SJ, Yang I, Tihan T, Prados MD, Parsa AT.Primary gliosarcoma: key clinical and pathologic dis-tinctions from glioblastoma with implications as aunique oncologic entity. J Neurooncol 2010; 96: 313–320.

21. Maiuri F, Stella L, Benvenuti D, Giamundo A,Pettinato G. Cerebral gliosarcomas: correlation ofcomputed tomographic findings, surgical aspect, patho-logical features, and prognosis. Neurosurgery 1990; 26:261–267.

22. Cervoni L, Celli P. Cerebral gliosarcoma: prognosticfactors. Neurosurg Rev 1996; 19: 93–96.

23. Hegi ME, Diserens AC, Gorlia T et al. MGMT genesilencing and benefit from temozolomide in glioblas-toma. N Engl J Med 2005; 352: 997–1003.

24. Capper D, Mittelbronn M, Meyermann R, Schitten-helm J. Pitfalls in the assessment of MGMT expressionand in its correlation with survival in diffuse astrocy-tomas: proposal of a feasible immunohistochemicalapproach. Acta Neuropathol 2008; 115: 249–259.

25. Pollack IF, Hamilton RL, Sobol RW et al.O6-methylguanine-DNA methyltransferase expres-sion strongly correlates with outcome in childhoodmalignant gliomas: results from the CCG-945 Cohort.J Clin Oncol 2006; 24: 3431–3437.

26. Das P, Puri T, Jha P et al. A clinicopathological andmolecular analysis of glioblastoma multiforme withlong-term survival. J Clin Neurosci 2011; 18: 66–70.

27. Sasai K, Nodagashira M, Nishihara H et al. Carefulexclusion of non-neoplastic brain components isrequired for an appropriate evaluation ofO6-methylguanine-dna methyltransferase status inglioma: relationship between immunohistochemistryand methylation analysis. Am J Surg Pathol 2008; 32:1220–1227.

Prognostication of gliosarcoma 9

© 2012 Japanese Society of Neuropathology