|Home | About | Journals | Submit | Contact Us | Français|
Recent studies have shown that the clinical outcome of anaplastic oligodendroglial tumors is variable, but also that the histological diagnosis is subject to interobserver variation. We investigated whether the assessment of 1p/19q codeletion, polysomy of chromosome 7, epidermal growth factor receptor (EGFR) gene amplification (EGFRamp), and loss of chromosome 10 or 10q offers additional prognostic information to the histological diagnosis and would allow molecular subtyping. For this study, we used the clinical data and tumor samples of the patients included in multicenter prospective phase III European Organisation for Research and Treatment of Cancer (EORTC) study 26951 on the effects of adjuvant procarbazine, chloroethyl cyclohexylnitrosourea (lomustine), and vincristine chemotherapy in anaplastic oligodendroglial tumors. Fluorescence in situ hybridization was used to assess copy number aberrations of chromosome 1p, 19q, 7, 10, and 10q and EGFR. Three different analyses were performed: on all included patients based on local pathology diagnosis, on the patients with confirmed anaplastic oligodendroglial tumors on central pathology review, and on this latter group but after excluding anaplastic oligoastrocytoma (AOA) with necrosis. As a reference set for glioblastoma multiforme (GBM), patients from the prospective randomized phase III study on GBM (EORTC 26981) were used as a benchmark. In 257 of 368 patients, central pathology review confirmed the presence of an anaplastic oligodendroglial tumor. Tumors with combined 1p and 19q loss (1ploss19qloss) were histopathologically diagnosed as anaplastic oligodendroglioma, were more frequently located in the frontal lobe, and had a better outcome. Anaplastic oligodendroglial tumors with EGFRamp were more frequently AOA, were more often localized outside the frontal lobe, and had a survival similar to that for GBM. Survival of patients with AOA harboring necrosis was in a similar range as for GBM, while patients with AOA with only endothelial proliferation had better overall survival. In univariate analyses, all molecular factors except loss of 10q were of prognostic significance, but on multivariate analysis a histopathological diagnosis of AOA, necrosis, and 1ploss19qloss remained independent prognostic factors. AOA tumors with necrosis are to be considered WHO grade IV tumors (GBM). Of all molecular markers analyzed in this study, especially loss of 1p/19q carried prognostic significance, while the others contributed little prognostic value to classical histology.
Over the last 15 years, oligodendroglial tumors have been recognized as treatment-sensitive tumors with a favorable survival.1,2 Molecular studies have shown that this is of particular concern in the subgroup with an unbalanced translocation of 19p to 1q, der(1;19)(p10;q10), resulting in the 1p/19q codeletion.3–7 The favorable outcome even after radiotherapy (RT) only has recently been further confirmed by two randomized prospective studies on (neo)adjuvant procarbazine, chloroethyl cyclohexylnitrosourea (CCNU; lomustine), and vincristine (PCV) chemotherapy in anaplastic oligodendroglial tumors, which showed a more favorable survival in tumors with combined 1ploss19qloss.8,9 However, not all tumors with an oligodendroglial phenotype have such favorable outcome: up to 20% of patients died within 1 year after diagnosis, an outcome that is more consistent with glioblastoma multiforme (GBM).
European Organisation for Research and Treatment of Cancer (EORTC) study 26951 investigated the benefit of six cycles of adjuvant PCV chemotherapy in anaplastic oligodendroglial tumors.8 We used this study to investigate the correlation between clinical outcome and specific histological and molecular features. We were particularly interested to learn whether necrosis and endothelial proliferation have similar prognostic significance in mixed anaplastic oligoastrocytoma (AOA) and pure anaplastic oligodendroglioma (AOD), and whether the assessment of specific molecular aberrations usually associated with GBM (epidermal growth factor receptor [EGFR] gene amplification, loss of chromosome 10 or of 10q) contributes to histological diagnosis and clinical prognosis. While this research was ongoing, in 2007 a revised WHO classification for glioma was published.10 Because this WHO 2007 classification of brain tumors classifies AOA with necrosis (previously considered grade III) as grade IV GBM, in a further analysis of prognostically important factors AOA with necrosis as diagnosed by the central review pathologist were left out. The first level of this analysis considered only the factors directly related to the analysis of the tissue samples; in the subsequent level of analysis, clinical information was introduced to explore factors with independent prognostic significance.
Patients were eligible for EORTC study 26951 if they had been diagnosed by the local pathologist with AOD or AOA with at least 25% oligodendroglial elements according to the 1994 edition of the WHO classification of brain tumors,11 had at least three of five anaplastic characteristics (high cellularity, mitoses, nuclear abnormalities, endothelial proliferation, and necrosis), were between 16 and 70 years of age, had an Eastern Cooperative Oncology Group performance status (PS) of 0 to 2, and had not undergone prior chemotherapy or RT to the skull. The clinical details of these studies have been published elsewhere.12 Since no statistically significant differences in overall survival were observed between the patients assigned to RT and those assigned to RT followed by six cycles of adjuvant PCV chemotherapy, the patients in both arms were studied together.
After inclusion and randomization, central pathology review took place (J.M.K.). Patients were then regrouped in three data sets: (1) all patients as diagnosed by the local pathologist using the local diagnosis of both histology and anaplastic features (“local diagnosis”), (2) centrally confirmed AOD and AOA according to the WHO 1994 classification (“WHO 1994”), and (3) all centrally confirmed AOD and only the centrally confirmed AOA without necrosis (“WHO 2007”). In groups 2 and 3, the central review diagnosis of anaplastic features was used. For comparison, a group of patients with GBM obtained from EORTC 26981 was used.13
Probes to 1p36 (D1S32), centromere 1 (pUC1.77), 19p (equivalent amounts of bacterial artificial chromosome [BAC] RPCI 11-959O6, 11-957I1, and 11-153P24), 19q (BAC 426G3), PTEN (PAC 190P6), CEP10 (CEP10; D10Z1), EGFR (BAC RPCI 11-148p17, a kind gift of Dr. A. Perry), centromere 7 (CEP7; P7t1), and CEP12 (CEP12; Pα12H8) were labeled with biotin-16-dUTP (pUC1.77; Roche Diagnostics, Mannheim, Germany), digoxigenin-16-dUTP (D1S32, 959O6, 957I1, 153P24, 190P6, 148p17; Roche Diagnostics), Spectrum Green (P7t1; Vysis Inc., Downers Grove, IL, USA), Spectrum Orange (D10Z1; Vysis Inc.), or Cy5 (Amersham Biosciences, Piscataway, NJ, USA). Tumor sections were deparaffinized, dehydrated, and microwave treated in citrate buffer (pH 6.0) and then digested in 0.4% pepsin solution (Sigma-Aldrich, St. Louis, MO, USA) in 0.9% NaCl (pH 1.5–2.0), as previously described.14 Subsequently, slides were dehydrated, and probe solutions were applied. Tumor sections and probes were codenaturated on a slide moat preheated to 80°C for 5 min, and then cooled on ice and incubated at 37°C for 48 h in a moistened chamber. After incubation, slides were washed in 1.5 M urea saline–sodium citrate (SSC) at 45°C for 30 min and rinsed in 2× SSC. Probes were detected using antirhodamine-conjugated digoxigenin (426G3, 148p17; Roche Diagnostics, Mannheim, Germany), antidigoxigenin-conjugated with fluorescein isothiocyanate (D1S32, 959O6, 957I1, 153P24, 190P6; Roche Diagnostics), or Cy3-conjugated avidin (pUC1.77; Brunschwig Chemie, Amsterdam, the Netherlands) antibodies at a concentration of 4 and 15 μg/ml diluted in phosphate-buffered saline. Nuclei were counterstained with diamidinophenylindole in antifade solution (Vector Laboratories, Burlingame, CA, USA).
Locus-specific fluorescence in situ hybridization (FISH) probes were enumerated in 60 nonoverlapping nuclei per hybridization utilizing a Leica DM-RXA fluorescence microscope (Leica, Wetzlar, Germany). Images were captured using a COHU 4910 series monochrome CCD camera (Cohu Inc., San Diego, CA, USA) attached to the fluorescence microscope equipped with a PL Fluotar ×100, numerical aperture 1.30–0.60 objective, I3 and N2.1 filters (Leica), and Leica QFISH software (Leica Imaging Systems, Cambridge, UK). Ratios were calculated for 1p versus CEP1, 19q versus 19p, or 10q versus CEP10 (10qloss) by dividing the number of signals of the marker by the number of signals of the reference; a ratio of less than 0.80 was considered allelic loss. If a borderline ratio was obtained (0.75–0.90), spots in 200 nuclei were counted. For ratios of EGFR versus CEP7, CEP7 versus CEP12, or CEP10 versus CEP12, different cutoff levels were used. A ratio of EGFR/CEP7 >2 was considered EGFR amplification (EGFRamp), a ratio of CEP7/CEP12 > 1.1 displayed polysomy of chromosome 7 (7poly), and a ratio < 0.85 for CEP10/CEP12 was indicative of monosomy 10 (10loss). These cutoff values were determined in a set of nontumoral controls that displayed <10% nuclei with more than two signals for the investigated marker/nucleus.
The prognostic significance of the tissue variables (diagnosis, anaplastic features, and molecular features) were first analyzed without taking any nontissue (clinical) factors into account. For the multivariate analysis, the following major prognostic clinical variables were used: WHO PS (0, 1, 2), age (<50, ≥50), type of surgery (biopsy or resection), and type of adjuvant treatment (none or PCV). For the histological factors, the diagnosis (AOD or AOA) and the five anaplastic features were used, diagnosed either by the local pathologist for the local diagnosis or by the central review pathologist for centrally confirmed tumors. For the molecular factors, 1ploss, 19qloss, 1ploss19qloss, EGFRamp, 7poly, 10loss, and 10qloss were used.
Association between factors was assessed by the Spearman correlation coefficient (SCC); Fisher’s exact test was used for inference. Survival analyses in the three populations were performed with the log-rank test and the Cox regression analysis stratified by the treatment, with and without backward selection. Internal validation was performed by bootstrap resampling technique (5% confidence) to assess the generalizability of the models. Factors with a probability of inclusion (PI) in regression models of less than 60% based on 1,000 bootstrap samples were considered not confirmed as independent prognostic factors. Patients with missing values in at least one factor were removed from the analyses. No formal adjustment for multiple testing was performed; nevertheless, a conservative significance level of 1% was considered for all comparisons.
In this study, 368 patients were randomized; 265 had been diagnosed by the local pathologist with an AOD and 100 with an AOA (three missing). At central pathology review, in 257 patients the diagnosis of an anaplastic oligodendroglial tumor was confirmed (175 AOD, 82 AOA). Other frequent diagnoses at central review were low-grade tumors (39 patients) and high-grade astrocytic tumors (anaplastic astrocytoma or GBM, 39 patients); for 22 patients no material was received for review. Table 1 specifies the pathology findings, presence of necrosis, and endothelial proliferation in each of the three data sets.
Table 2 shows the distribution of molecular characteristics and Table 3 the correlations between the molecular characteristics. 1ploss and 19qloss were highly correlated with each other (SCC = 0.51). EGFRamp was correlated with 7poly (SCC = 0.40) and with 10loss (SCC = 0.48). EGFRamp and 1ploss19qloss were poorly anticorrelated (SCC = −0.24): of the 227 patients with both measures, 59 patients had 1ploss19qloss, 50 had EGFRamp, 3 had both, and 121 had neither (p = 0.0001).
Of the 82 AOA tumors, 55 (67%) showed necrosis. Endothelial proliferation was present in almost all AOA tumors (80 of 82) and in all 55 AOA tumors with necrosis. Table 4 shows the survival of patients with AOD and AOA with or without necrosis, and the reference group of GBM patients treated with RT only, and Fig. 1 shows the survival curves of these patients in the RT arms of both studies. After correction for extent of resection, PS, and age, in the patients treated with RT alone survival for AOA with necrosis was in a similar range as the survival for GBM (hazard ratio [HR] = 1.53; 95% confidence interval [CI], 1.02–2.31; p = 0.042). Survival in the 25 patients with AOA showing endothelial proliferation but no necrosis was better than in patients with GBM (p = 0.007). The outcome of AOD patients without tumor necrosis was better than for those without tumor necrosis, but in the latter category survival was still much more favorable than for patients with GBM or with AOA without necrosis (Fig. 1).
Table 5 shows the molecular findings in the centrally confirmed AOA without necrosis, AOD, and AOA with necrosis. Combined 1ploss19qloss was more frequent in AOD or AOA without necrosis; AOA with necrosis more often had EGFRamp, 7poly, and 10loss. Despite the observed differences in frequencies, none of the items clearly separated both subgroups. Frontal tumors and previous resection for a low-grade tumor were more frequently observed in tumors with combined 1ploss19qloss (p = 0.0021 and p = 0.0087) and less frequent in tumors with EGFRamp (p = 0.0004 and p = 0.0045). No other clinical factor, including age, was related to any of the molecular factors (data not shown).
Table 6 shows the univariate analysis of all histological and molecular factors in all patients. All molecular factors except for loss of 10q were correlated with outcome (p < 0.01). Outcome for AOA with EGFRamp was similar to that for AOD with EGFRamp (p = 0.354). Except for combined 1ploss19qloss, in all three data sets, multivariate analysis using tissue (molecular and histological) factors showed none of the other molecular factors to be of prognostic significance. The presence of necrosis was of significance, when assessed by the local pathologist or by the central reviewing pathologist. Furthermore, the central review histopathological diagnosis (AOD, AOA with or without necrosis) was of significance.
Selected were 1ploss, combined 1ploss19qloss, necrosis, 7poly, and 10loss. With bootstrap resampling, not confirmed were 7poly (PI = 52%) and 10loss (PI = 49%). 1ploss was borderline not confirmed (PI = 58%).
Selected were 1ploss, combined 1ploss19qloss, AOA, and necrosis. With bootstrap resampling, 1ploss was not confirmed (PI = 52%).
Selected were 1ploss and combined 1ploss19qloss; both were confirmed with bootstrap resampling.
Selected were PS, age, 1ploss, 1ploss19qloss, necrosis, 7poly, and 10loss (Table 7). With bootstrap resampling, 1ploss (PI = 56%) and 10loss (PI = 38%) were not confirmed. The PI of 7poly was of borderline significance (59.7%); the other factors were confirmed.
Selected were PS, age, 1ploss19qloss, AOD diagnosis, and necrosis (Table 7). All factors were confirmed by bootstrap resampling.
Selected were PS, age, 1ploss, 1ploss19qloss, and 10loss (Table 7). With bootstrap resampling, PS and 1ploss19qloss were confirmed. The other factors had PIs < 60%.
In the present study, losses of 1p (41%) and/or 19q (33%) were the most common genomic alterations. In addition, no less than 22% of cases displayed EGFRamp, which was inversely related to the 1p/19q codeletion. Loss of 10q and/or copy number aberrations of chromosomes 7 (gain) and 10 (loss) were also observed in a substantial number of anaplastic oligodendroglial tumors, predominantly in tumors with EGFRamp. The molecular, histological, and clinical properties identified two subgroups of tumors with distinct prognostic characteristics: (1) oligodendroglial tumors with 1ploss19qloss, mainly located in the frontal lobe, with a predominant AOD histology and a favorable prognosis, and (2) tumors with EGFRamp, often with copy number alterations of chromosomes 7 and/or 10, located outside the frontal region, with a mixed oligoastrocytoma phenotype and with a less favorable prognosis. The subgroup with EGFRamp and loss of chromosome 10 resembles the previously described “GBM with oligodendroglial phenotype.”15–21 The same genetic lesions (and often also with EGFRvIII mutations) have been described in the small-cell GBM variant, which is characterized by monomorphous, deceptively bland nuclei and is often misdiagnosed as AOD.22,23 Together with PTEN (phosphatase and tensin homolog) mutations and deletions, EGFRamp, 7 poly, and 10loss are the most common genotypic alterations in GBM.24 As expected, after correction for prognostic variables, comparison of this subset of patients with tumors with EGFRamp to a group of GBM patients from EORTC study 26981 shows no statistically significant difference in survival.25 Clearly, despite the intent of the EORTC study 26951 to include chemosensitive oligodendroglioma, this analysis shows that a large number of less sensitive GBM with some oligodendroglial features was entered into this clinical study.
The EORTC study on anaplastic oligodendroglial tumors was initiated because studies on recurrent disease showed anaplastic oligodendroglial tumors as opposed to GBM to be sensitive to chemotherapy. Over the past years, it has become clear that the diagnosis of WHO grade III, including classical AOD, is subject to a considerable interobserver disagreement, and the exact delineation of AOD and AOA from GBM with oligodendroglial features is unclear.26,27 During the conduct of the clinical study, the strong relationship between 1ploss19qloss and response to chemotherapy was discovered, limiting the subset of chemotherapy-sensitive tumors to the 1p19q-codeleted tumors. The poor clinical outcome observed in some of the patients made us hypothesize that the use of molecular diagnostics aiming at genetic abnormalities associated with GBM could identify tumors with some oligodendroglial morphology but with an outcome similar to GBM (and to be treated like GBM). Our results show that molecular diagnostics can indeed serve this purpose, although the added benefit is less than anticipated.
Simultaneously, retrospective studies have shown that survival of AOD differs from the survival of AOA, and that the presence of necrosis identifies a subgroup of AOA with unfavorable outcome.28,29 Because of the latter finding, by simply leaving out the word “necrosis” in the section on AOA, the 2007 edition of the WHO classification of brain tumors considers the presence of necrosis no longer consistent with the diagnosis of AOA.30,31 In the present prospective setting, the outcome of AOA with necrosis is in a similar range and clinically more or less equivalent to the outcome of GBM with a risk-adjusted p-value that did not meet preset levels of statistical significance (p = 0.04). The GBM-like nature of “AOA with necrosis” is further corroborated by the presence of EGFRamp in 44% and 10loss in 35% of the tumors, which is similar to the incidence of these genetic aberrations in studies of GBM. On the other hand, AOA patients with only endothelial abnormalities have a somewhat better outcome, justifying the inclusion of this population in the present WHO definition of AOA. Similarly, although for AOD necrosis had a prognostic impact, the outcome was much better than for GBM. In another review series, “grade IV” AOA (the diagnosis of which required necrosis) had a better survival than GBM, but median survival in the AOA grade IV group was 15.6 months, compared to 10.9 months in the GBM group.32 Our findings (see Table 4) are also consistent with that conclusion, and the limited number of AOA with necrosis (55 patients) may have affected the power to reach statistically significant differences with the GBM group. In two previous studies, necrosis did affect outcome of AOA (or nonclassic oligodendroglioma) but not that of AOD (or classical oligodendroglioma).27,33 In the pathology review of Radiation Therapy Oncology Group study 94-02, age, multifocal disease, histology (classical vs. nonclassical), and 1p/19q status were independent prognostic factors but not necrosis. In our analysis, necrosis was of prognostic significance in locally diagnosed tumors and also in the WHO 1994 data set and in centrally confirmed AOD. In multivariate analysis of the entire study population, necrosis remained an independent prognostic factor (data not shown). With changing views on the diagnosis of oligodendroglial tumors, it will be interesting to repeat this analysis based on a repeated pathology review with stricter criteria for oligodendroglioma.
Our findings show that 1ploss19qloss is the most powerful molecular predictor of outcome. Although highly correlated to outcome in univariate analyses, the other molecular factors have little additional value when considered together with all other available information, in particular, the histological diagnosis and the histological features (the presence of necrosis). Both clinically and molecularly, the WHO 1994 “AOA with necrosis” classification indeed equals GBM, and previous studies have shown that in GBM the presence of EGFR amplification has no additional prognostic significance.34 Of note, in the local diagnosis polysomy of chromosome 7 and loss of chromosome 10 were selected in multivariate analysis but not confirmed. This suggests that molecular studies may be more relevant when the pathological diagnosis is less certain (e.g., in a setting where the diagnosis is not made by an experienced neuropathologist or when only small biopsies are available for histopathological diagnosis).
Retrospectively, the inclusion criteria of the EORTC phase III study on adjuvant PCV chemotherapy failed to reach its objective: to enter a subset of chemotherapy-sensitive tumors into the study. The inclusion of many GBM-like tumors is likely to have affected the power of this study, and thus the outcome of the study. Of note, the analysis of the pathology review within the similar North American Intergroup Radiation Therapy Oncology Group Trial 9402 suggests that adjuvant PCV treatment may be beneficial in the so-called classical AOD as diagnosed by microscopy.27 In that study, it was suggested that inclusion based on central review is pivotal to keep the study population homogeneous. It is of note, however, that in that study the interobserver agreement was also moderate, and it remains unclear how a study population based on inclusion by central review reflects the patients locally diagnosed (and treated) in everyday clinical practice with such a tumor.
There are a number of potential shortcomings of the present study. First, FISH has a limited sensitivity and specificity. In particular, GBM may have partial deletions of 1p, which is picked up by FISH for 1p36.6.35–37 This may explain why we observed 1p/19q codeletions together with EGFRamp in some patients. Also, FISH for 19q can be troublesome due to the often weak fluorescence signals that are obtained. Second, limitations in available material may have caused sample bias. More advanced techniques using better conserved tissue samples are likely to yield better results. In a comparable project including 60 patients also treated within EORTC study 26951 and from whom frozen tumor samples were available, BAC array-based comparative genomic hybridization (aCHG) was performed (Ibdaih et al., unpublished observations). This analysis revealed four genomic subgroups with prognostic information (combined 1p/19q loss, EGFRamp, loss of chromosome 21, and neither of these). Multivariate analysis with all relevant prognostic factors identified age (p = 0.0002) and genomic profile (p < 0.0001) as independent prognostic factors. The interobserver variation between the aCHG data from that study and the FISH data used in the present study was considered good for both EGFRamp (κ = 0,796) and 1p/19q codeletion (κ = 0,612). Still, a few samples were classified differently by these techniques. The two studies suggest that, in addition to pathological features (notably necrosis), molecular data are of interest for anaplastic oligodendroglial tumors. Both studies confirmed the 1p/19q codeletion as a strong biomarker in anaplastic oligodendroglial tumors and identified additional biomarkers requiring further investigations as candidates in the non-1p/19q-codeleted anaplastic oligodendroglial tumors.
Methylation status of the MGMT (O6-methylguanine– DNA methyltransferase) promoter gene was no part of this analysis; with the limited amount of tissue available (usually slides only) and stored for many years, this was not yet possible at the time of this study, although it is currently being studied in a subset of patients. This is likely to be an additional prognostic or predictive factor, especially in patients treated with chemotherapy (although the recent German NOA4 study suggests it may also be of prognostic value in patients managed with RT only).38,39
In conclusion, this study shows that, at the molecular level, EORTC study 26951 of anaplastic oligodendroglial tumors included a heterogeneous group of tumors, with almost 25% of tumors more resembling GBM, which should not have entered the study. Particularly, combined 1ploss19qloss contained additional prognostic significance, while most of the additional prognostic information of the other investigated molecular characteristics was already covered by the histopathology. Our study confirms that AOA with necrosis should indeed be considered GBM (WHO grade IV). The clinical outcome of patients with EGFR amplification was similar to that of a control group with GBM.
M.C.M.K and T.G. contributed equally to this work.
This article was written on behalf of the European Organisation for Research and Treatment of Cancer (EORTC) Brain Tumor Group and the Medical Research Council Clinical Trials Group. The authors and coauthors acknowledge the financial support to this work provided by the EORTC Translational Research Fund grant TRF 01/02, by AstraZeneca EORTC Translational Research grant AZ/01/02, and by Dutch Cancer Society grant DDHK 2005-3416. This article was supported by grants 2U10CA11488-25 through 2U10CA11488-35 from the National Cancer Institute (Bethesda, MD, USA) and by a donation from the Dutch Cancer Society from the Netherlands through the EORTC Charitable Trust. Its content is solely the responsibility of the authors and does not necessarily reflect the official views of the National Cancer Institute.