Predictive impact of MGMT promoter methylation in glioblastoma of the elderly

Authors


  • Presented at the ASCO Meeting 2011 in Chicago

  • Conflict of interest: MiW has received honoraria for lectures and advisory board participation for MSD

Abstract

O6-methylguanine-DNA-methyltransferase (MGMT) promoter methylation identifies a subpopulation of glioblastoma patients with more favorable prognosis and predicts a benefit from alkylating agent chemotherapy (CT). Little is known about its prevalence and clinical significance in older glioblastoma patients. We studied 233 glioblastoma patients aged 70 years or more (144 males, 89 females, median age: 74 years, range: 70.0–86.6 years), who were prospectively enrolled in the German Glioma Network, for MGMT promoter methylation by methylation-specific PCR (MSP) in all patients and DNA pyrosequencing in 166 patients. MGMT data were correlated with patient outcome. Median progression-free survival (PFS) was 4.8 months (95% CI: 4.3–5.3) and median overall survival (OS) was 7.7 months (95% CI: 6.3–9.0). MGMT promoter methylation was detected by MSP in 134 patients (57.5%). For the whole cohort, PFS was 5.2 versus 4.7 months (p = 0.207) and OS was 8.4 versus 6.4 months (p = 0.031) in patients with versus without MGMT promoter methylation. Patients with MGMT methylated tumors had longer PFS when treated with radiotherapy (RT) plus CT or CT alone compared to patients treated with RT alone. Patients with MGMT unmethylated tumors appeared to derive no survival benefit from CT, regardless of whether given at diagnosis together with RT or as a salvage treatment. Patients treated with RT plus CT or CT alone demonstrated longer OS when pyrosequencing revealed >25% MGMT methylated alleles. Taken together, MGMT promoter methylation may be a useful biomarker to stratify elderly glioblastoma patients for treatment with versus without alkylating agent CT.

Age is the major therapy-independent prognostic factor in gliomas across malignancy grades II–IV. The differential distribution of isocitrate dehydrogenase 1 mutations, which are more common in younger patients, contributes to the negative prognostic effect of age.1 O6-methylguanine-DNA-methyltransferase (MGMT) promoter methylation is a biomarker with both prognostic and predictive aspects, depending on glioma subtype.2 The MGMT status can be determined by the assessment of promoter methylation, mRNA or protein expression or enzymatic activity. Methylation-specific PCR (MSP)-based detection of MGMT promoter methylation has provided the most meaningful results as judged by clinical correlations.2, 3 Yet, the complexity of assessing the MGMT status even by MSP has prevented its broad implementation as a biomarker.2

Predictive power of the MGMT status was delineated in the EORTC NCIC registration trial for temozolomide (TMZ) in newly diagnosed glioblastoma where patients with MGMT promoter methylated tumors derived most benefit when treated with TMZ.4 Moreover, we have recently proposed that specifically patients with MGMT promoter methylated tumors derive benefit from the addition of the integrin antagonist cilengitide to TMZ radiochemotherapy.5 Yet, even in the EORTC NCIC trial, patients with MGMT promoter methylation had longer progression-free survival (PFS) of 5.9 months versus 4.4 months when treated with radiotherapy (RT) alone (p = 0.02).4 A predictive effect of MGMT promoter methylation for PFS after RT alone was also noted in a nontrial glioblastoma cohort.6 Moreover, MGMT promoter methylation was strongly prognostic for PFS in anaplastic glioma patients treated with RT alone in the NOA-04 and EORTC 26951 trials.7, 8 There is little information on the role of MGMT promoter methylation in the large group of older glioblastoma patients who have been excluded from trials in the past and are less often treated with alkylating agent chemotherapy (CT). There is some indication that MGMT promoter methylation is more common in the elderly9 and retains its favorable prognostic power.10, 11

Patients and Methods

Patients

The German Glioma Network (GGN) is a prospective cohort study that enrolled 2,549 newly diagnosed patients with various types of glioma and frozen tissue asservation from October 2004 to October 2010. Here, we identified 233 patients aged 70 years or more with a diagnosis of glioblastoma confirmed by central pathology review and adequate follow-up. Sixty-nine patients were included in previous publications.1, 12 Clinical data were prospectively assembled as outlined before (http://www.gliomnetzwerk.de).12 Extent of resection was assessed by early (<72 hr) postoperative imaging by MRI or by CT when MRI was not feasible or not available. The patients were not enrolled into clinical trials, and treatment decisions were made by the treating physicians, patients and their families, without awareness of the MGMT promoter status. Progression was defined locally according to Macdonald criteria13 and not centrally reviewed. All patients gave written informed consent. All activities of the GGN have been approved by the review boards of the participating institutions.

MGMT promoter methylation analyses

DNA was extracted from frozen tumor tissue samples using the Qiagen blood and tissue DNA extraction kit (Qiagen, Hilden, Germany). Each tumor specimen was histologically investigated to assure a tumor cell content of at least 80%. MGMT promoter methylation was assessed by MSP as reported elsewhere in detail.14 Individual tumors showing only very weak PCR products for the methylated MGMT promoter sequence, but strong products for the unmethylated MGMT promoter sequence, were judged “weakly methylated.” In the statistical analyses, these tumors were analyzed within the MGMT promoter methylated tumors. In addition to MSP, we performed quantitative methylation analyses using pyrosequencing of sodium bisulfite-modified DNA in 166 patients, using the PyroMark Q24 MGMT kit (Qiagen).15, 16 The percentage of methylated alleles was calculated as the mean value of the methylation percentage obtained at each of the five investigated CpG dinucleotides.

Statistical analysis

Associations of clinical data and MGMT status were tested by χ2 test. The role of patient age was analyzed by Kruskal–Wallis and Mann–Whitney-U tests. Overall survival (OS) was calculated from the day of surgery until death. PFS was calculated from surgery until progression or death in the absence of a documented progression. Log-rank test was used to compare outcome data. Cox regression models were fitted to assess the independent impact of age (≤75 years versus > 75 years), KPS (≥80 versus < 80), extent of resection (total versus not total), MGMT status (methylated versus unmethylated) and first-line therapy (RT plus alkylating agent CT versus RT alone). Interactions between MGMT promoter status and therapy were evaluated in multivariate models. Data were analyzed using PASW Statistics 18 (Version 18.0.0).

Results

Patient characteristics

Patient characteristics are summarized in Table 1. Higher age and lower KPS were prognostic for PFS and OS, and extent of resection was prognostic for OS (Supporting Information Fig. 1). Median follow-up was 29 months. Progression was documented in 146 patients (62.7%) and death in 204 patients (87.6%). There were no age differences for patients with different extent of resection (p = 0.124) and no difference in age of gross totally resected versus all other patients (p = 0.616). There was an association between therapy and age (p = 0.002): patients without treatment after surgery were 2 years older (median) than patients who received any kind of cytotoxic therapy. Patients aged 75 or more were more likely to receive no further treatment (57% versus 43%) and less likely to receive RT plus CT (31% versus 69%) or CT alone (44% versus 56%; p = 0.011) than patients aged younger than 75 years. MGMT promoter methylation was detected by MSP in 134 of 233 patients (57.5%) and not differentially distributed by age, gender, KPS or extent of resection.

Figure 1.

(a and b) PFS (a) and OS (b) by first-line therapy. (c and d) PFS (c) and OS (d) by the MGMT promoter methylation status as determined by MSP analysis (MGMT meth+: patients with MGMT promoter methylated tumors; MGMT meth−: patients with MGMT promoter unmethylated tumors).

Table 1. Summary of patient characteristics
inline image

Treatment and outcome

Median PFS of all patients was 4.8 months (95% CI: 4.3–5.3), median OS 7.7 months (95% CI: 6.3–9.0). PFS and OS by treatment are depicted in Figs. 1a and 1b. Median PFS, PFS at 6 months, median OS and OS at 1 year by treatment are provided in Table 2. There was no difference in PFS between patients with (5.2 months, 95% CI: 4.3–6.1) versus without (4.7 months, 95% CI: 3.8–5.5) MGMT promoter methylation (p = 0.207; Fig. 1c). OS was longer in patients with MGMT promoter methylation: 8.4 months (95% CI: 6.7–10.1) versus 6.4 months (95% CI: 3.9–8.9; p = 0.031; Fig. 1d). When the 65 patients who received neither RT nor CT were omitted, median PFS was 6.0 months (95% CI: 4.7–7.2) with versus 6.0 months (95% CI: 4.3–7.6) without MGMT promoter methylation (p = 0.376), and median OS was 10.1 months (95% CI: 8.4–11.7) with versus 9.7 months (95% CI: 7.8–11.5) without MGMT promoter methylation (p = 0.065). Among 65 patients treated by biopsy or surgery alone, median OS was 2.3 months (95% CI: 0.8–3.8) with versus 2.0 months (95% CI: 0.6–3.7) without MGMT promoter methylation (p = 0.388).

Table 2. Outcome by first-line therapy and MGMT promoter methylation status as determined by MSP analysis1
inline image

Looking at PFS by MGMT status and first-line treatment, patients with MGMT promoter methylated tumors had comparable outcome when treated with RT plus CT or CT alone (p = 0.459), whereas outcome was inferior with RT alone (p < 0.001; Fig. 2a, Supporting Information Fig. 2). In contrast, patients with tumors without MGMT promoter methylation had comparable outcome when treated with RT plus CT (7.2 months) or RT alone (5.2 months; p = 0.336; Fig. 2b, Supporting Information Fig. 2).

Figure 2.

(a and b) PFS and (c and d) OS by treatment in patients with MGMT promoter methylated (a and c) or MGMT promoter unmethylated (b and d) glioblastomas as determined by MSP analysis (MGMT meth+: patients with MGMT promoter methylated tumors; MGMT meth−: patients with MGMT promoter unmethylated tumors).

Looking at OS by MGMT status and first-line treatment, outcome in patients with MGMT promoter methylated tumors was best with combined modality treatment and poorest without further treatment (Fig. 2c, Supporting Information Fig. 2). Interestingly, unlike PFS, patients treated with initial RT alone or initial CT alone had comparable OS. This may be explained by the fact that five of the 31 patients with MGMT promoter methylated tumors treated initially with RT alone had TMZ alone (n = 4) or in combination with nitrosourea (n = 1) at recurrence (Supporting Information Table). In contrast, the outcome was poor for the two patients with MGMT unmethylated tumors who were treated with CT alone (data not shown), whereas similar survival was seen in patients treated with RT alone (8.8 months) or RT plus CT (10.4 months; p = 0.283).

Quantitative MGMT promoter methylation data

Quantitative data on the percentage of MGMT promoter methylated alleles were obtained by pyrosequencing in 166 patients. Using the previously defined cutoff of <8% methylated alleles for MGMT unmethylated tumors,16 there was a high degree of concordance between MSP and pyrosequencing, with 74 of 75 tumors lacking MGMT methylation on MSP showing <8% methylated alleles on pyrosequencing, and 79 of 91 tumors methylated on MSP showing ≥8% methylated alleles (p < 0.001). Both MSP and pyrosequencing data were available for 85 patients treated with CT either upfront or at recurrence. On MSP, 54 patients (63.5%) had methylated tumors, whereas 31 patients (36.5%) had unmethylated tumors. On pyrosequencing, 38 tumors displayed <8% methylated alleles (44.7%) and 47 tumors ≥8% methylated alleles (55.3%). All tumors with <8% methylated alleles lacked MGMT promoter methylation on MSP. Figure 3a shows OS of these 85 patients stratified by MGMT status determined either by MSP or by pyrosequencing (<8% cutoff).

Figure 3.

(ac) OS in the subgroup of 85 patients treated with CT initially or at recurrence, for whom the MGMT promoter methylation status was assessed by both MSP analysis and pyrosequencing. (a) OS stratified according to the MGMT promoter methylation status determined by MSP (MSPmeth+ vs. MSPmeth−) or by pyrosequencing (PSQ) using a cutoff value of 8% methylated alleles. (b) OS according to the MGMT promoter methylation status determined by pyrosequencing using a cutoff value of 25%. (c) OS based on MSP and pyrosequencing results stratified into the following three patient groups: (i) MSPmeth+ and >25% methylated alleles on pyrosequencing, (ii) MSPmeth+ and ≤25% methylated alleles on pyrosequencing and (iii) MSPmeth− and ≤25% methylated alleles on pyrosequencing. (df) OS based on MSP and pyrosequencing results stratified into the three molecular groups defined in (c), but comparing RT alone versus RT plus CT as the first-line treatment in each molecularly defined group.

To assess a prognostic effect of quantitative MGMT promoter methylation data, we divided the 85 tumors into four subgroups: <8%, 8–25%, 26–50% and >50% methylated alleles. Kaplan–Meier plots for OS revealed no difference between the <8% and 8–25% subgroups or the 26–50% and >50% subgroups (data not shown), indicating prognostically different patient groups if separated by a cutoff of 25% (Fig. 3b). Interestingly, the tumors of only 31 of the 54 (57.4%) patients with MGMT promoter methylation by MSP also had >25% methylated alleles upon pyrosequencing. In fact, most of the prognostically favorable effect of MGMT promoter methylation determined by MSP is due to the strongly methylated subgroup of tumors carrying >25% methylated alleles on pyrosequencing (Fig. 3c).

We also considered the risk reduction regarding OS afforded by the addition of CT to RT as the first-line treatment versus RT alone in the three populations: it was 0.63 (26 versus 26 patients, 95% CI: 0.33–1.18) for patients with unmethylated tumors, 0.25 (18 versus 8 patients, 95% CI: 0.09–0.67) for patients with methylated tumors and ≤25% methylated alleles and 0.28 (26 versus 16 patients, 95% CI: 0.12–0.65) for patients with methylated tumors and >25% methylated alleles (Figs. 3d–3f). Thus, any degree of MGMT promoter methylation may be sufficient to predict benefit from CT in the first-line treatment.

Multivariate analyses

Multivariate Cox models were built to assess the relative risk (RR) for age, KPS, extent of resection, MGMT status and first-line treatment on PFS and OS. To determine which treatment provides most benefit, multivariate models were restricted to patients treated after surgery by RT or RT plus CT. Compared to RT alone, treatment with RT plus CT showed the strongest effect on PFS as well as on OS (Fig. 4). Patients younger than 75 years had a decreased risk for progression and death.

Figure 4.

PFS (left) and OS (right) for patients with RT or RT plus CT in first-line therapy. Cox model for age ≤ 75 vs. >75 (ref.), KPS ≥ 80 vs. < 80 (ref.), resection total vs. no total (ref.), MGMT promoter methylated (MGMT meth+) vs. MGMT promoter not methylated (MGMT meth−) (ref.), first-line therapy RT plus CT vs. RT (ref.); lower panel: RT plus CT vs. RT (ref.) in patients with MGMT meth+ tumors, RT plus CT vs. RT (ref.) in patients with MGMT meth− tumors of Cox model with interaction (first-line therapy × MGMT). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

On the basis of the differences in PFS and OS by MGMT status (Fig. 2), we performed an additional analysis with interaction term for treatment and MGMT status. The effect of treatment was different for patients with methylated versus unmethylated tumors for PFS (p = 0.033) and OS (p = 0.018). Compared to patients with MGMT unmethylated tumors treated by RT alone, the risk for death decreased with RT plus CT with a RR of 0.81 (95% CI: 0.45–1.45). For patients with MGMT methylated tumors, this RR was 0.31 (95% CI: 0.18–0.55). For PFS this interaction provided similar results (Fig. 4).

Discussion

Soon more than half of all glioblastoma patients will be “elderly.” Our current concepts on the role of MGMT promoter methylation stem from data obtained in younger clinical trial populations and cannot simply be extrapolated to the elderly. Moreover, the elderly tolerate cancer therapy less well. Therefore, a biomarker that allows to differentially allocate treatment would be a major step ahead specifically in the elderly. MGMT promoter methylation has been advanced as the first predictive biomarker in neuro-oncology because the benefit derived from TMZ in newly diagnosed glioblastoma was largely restricted to patients with this molecular feature.4 A strong prognostic impact of this biomarker had previously been determined in malignant glioma patients treated with RT plus nitrosourea-based CT.17 Yet, the MGMT status has not been incorporated into routine diagnostics. Multiple methods to determine the MGMT status have been advocated, but consensus regarding the optimal technique has not been established. Although MSP analysis is most widely used, there is no standardization, and novel, complex techniques like pyrosequencing are considered superior to MSP by some.2

Our study was performed to define the prognostic and predictive value of MGMT promoter methylation in elderly glioblastoma patients. The major strengths of this study include the large sample size, the prospective data collection, the standardized use and comparison of two techniques to assess the MGMT promoter methylation status and the opportunity to dissect prognostic and predictive aspects of MGMT promoter methylation as a biomarker. We confirm that MGMT promoter methylation in glioblastomas may be more prevalent in the elderly.10 Interestingly, MGMT promoter methylation is absent in the cerebral cortex of young individuals but increases in older individuals starting at the age of 50 years.18 However, it remains to be investigated whether and how an aging-associated increase in MGMT promoter methylation relates to the higher prevalence of this aberration in glioblastomas of older patients, as detected in our series, and possibly higher toxicity of TMZ in this age group.19 In fact, it cannot be ruled out that contamination from MGMT promoter methylated normal cerebral tissue contributes to the apparent increase in MGMT promoter methylation in the elderly population studied here. However, this issue is difficult to address experimentally because of the limited availability of normal, i.e., completely tumor-free cerebral tissue samples in resection specimens of glioblastomas that could be separately evaluated for MGMT promoter methylation.

Within the whole cohort of 233 patients, MGMT promoter methylation was prognostic for OS but not for PFS (Figs. 1c and 1d), although the significance for OS was lost when patients never treated with RT or CT were omitted from the analysis. However, there was an intriguing difference in PFS in response to the various treatment regimens—RT alone, alkylating agent CT alone or combined modality treatment—by MGMT status. Patients with MGMT promoter methylated tumors had better outcome when treated with RT plus CT or CT alone, whereas patients with unmethylated tumors had better outcome when treated with RT or RT plus CT (Fig. 2, Table 2, Supporting Information Fig. 2). Whether the small increase in PFS and OS with combined modality treatment over RT alone in patients with unmethylated tumors is relevant is difficult to determine because this was not a randomized study and selection bias may have shifted the use of combined modality treatment to patients with favorable prognostic factors.20, 21

There is an ongoing controversy on the optimal approach to determine the MGMT status. Therefore, in addition to MSP, we investigated tumors from 166 patients for MGMT promoter methylation by pyrosequencing of sodium bisulfite-modified DNA. This technique provides quantitative data on the percentage of methylated alleles.15, 22, 23 Using a previously defined threshold of <8% versus ≥8% methylated alleles,16 we found a good concordance between MSP and pyrosequencing. Only 13 of 166 tumors (8%) showed discrepant results, including 12 tumors judged methylated by MSP but displaying <8% methylated alleles on pyrosequencing. Because MSP and pyrosequencing assess different CpG sites within the MGMT promoter, regionally heterogeneous methylation patterns most likely explain these discrepancies.2 The extent of MGMT promoter methylation may have an impact on prognosis of glioblastoma patients treated with RT and TMZ, with tumors carrying >35% methylated alleles demonstrating the best outcome.23 Here, patients with strongly methylated tumors (>25% MGMT methylated alleles) showed a significantly better outcome than patients with tumors with ≤25% methylated alleles, too, if treated with alkylating agent CT. Importantly, only 57.4% of the tumors methylated on MSP had >25% MGMT methylated alleles on pyrosequencing, and this group of tumors was responsible for most of the prognostically favorable effect. OS of CT-treated patients with tumors that were methylated on MSP, but demonstrated ≤25% methylated alleles on pyrosequencing, did not differ from that of patients with MGMT unmethylated tumors on MSP (Fig. 3c). Accordingly, the cutoff of 25% may be prognostically relevant among patients uniformly treated with alkylating agent CT. Yet, the OS benefit derived from the addition of CT to RT as first-line treatment was comparable in patients with high- versus low-degree methylation (Figs. 3d–3f), supporting the hypothesis that even weak or partial MGMT promoter methylation may be of clinical relevance because methylation affects the subpopulation of tumor cells with tumor-initiating capacity.24 However, these findings are based on relatively small numbers, and studies on larger cohorts are needed to define the clinically relevant extent and pattern of MGMT promoter methylation in glioblastoma. For now, our data indicate that for clinical routine use, conventional MSP analysis is not inferior to pyrosequencing, which is less readily available on a community level.

Age, KPS and extent of resection were prognostic on univariate analysis when all patients were considered (Supporting Information Fig. 1). However, when only patients treated with RT or RT plus CT were considered, multivariate analysis identified age and MGMT promoter methylation as the two most important prognostic factors, whereas neither KPS nor extent of resection was significant (Fig. 4). KPS also failed to become significant when the MGMT status was omitted from the multivariate analysis. Thus, MGMT status and type of treatment override the prognostic power of KPS and extent of resection in this selected patient population considered eligible for surgery.

Our study represents an important step toward the definition of standards of care for elderly glioblastoma patients. The value of RT has been confirmed in a small randomized trial comparing best supportive care versus RT alone.25 Several smaller series have explored the safety and efficacy of TMZ alone in the elderly, with encouraging results.26,27 Two phase III trials comparing RT alone with TMZ alone arrived at different conclusions: the Nordic trial reported no difference,28 whereas NOA-08 trial failed to demonstrate noninferiority of dose-dense TMZ alone versus RT.29 These trials may only be adequately interpreted when data on outcome by MGMT promoter methylation status become available. Until then, MGMT promoter methylation may be a useful biomarker to stratify older glioblastoma patients for treatment with versus without alkylating agents.

Acknowledgements

The German Glioma Network (GGN) is supported by the German Cancer Aid (Deutsche Krebshilfe). The authors acknowledge the support of the GGN teams in each of the participating Clinical Centers and the Center for Biometry, Documentation and Bioinformatics, University of Leipzig.

Ancillary