Extent of MGMT promoter methylation modifies the effect of temozolomide on overall survival in patients with glioblastoma: a regional cohort study

Abstract Background MGMT methylation in glioblastoma predicts response to temozolomide but dichotomizing methylation status may mask the true prognostic value of quantitative MGMT methylation. This study evaluated whether extent of MGMT methylation interacts with the effect of temozolomide on overall survival. Methods We included consecutive glioblastoma patients aged ≥16 years diagnosed (April 2012–May 2020) at a neuro-oncology center. All patients had quantitative MGMT methylation measured using pyrosequencing. Those with MGMT methylated tumors were stratified into high and low methylation groups based on a cut-off using Youden index on 2-year survival. Our accelerated failure time survival models included extent of MGMT methylation, age, postoperative Karnofsky performance score, extent of resection, temozolomide regimen, and radiotherapy. Results There were 414 patients. Optimal cut-off point using Youden index was 25.9% MGMT methylation. The number of patients in the unmethylated, low and high methylation groups was 223 (53.9%), 81 (19.6%), and 110 (26.6%), respectively. In the adjusted model, high (hazard ratio [HR] 0.60, 95% confidence intervals [CI] 0.46–0.79, P = 0.005) and low (HR 0.67, 95% CI 0.50–0.89, P < 0.001) methylation groups had better survival compared to unmethylated group. There was no evidence for interaction between MGMT methylation and completed temozolomide regimen (interaction term for low methylation P = 0.097; high methylation P = 0.071). This suggests no strong effect of MGMT status on survival in patients completing temozolomide regimen. In patients not completing the temozolomide regimen, higher MGMT methylation predicted better survival (interaction terms P < 0.001). Conclusions Quantitative MGMT methylation may provide additional prognostic value. This is important when assessing clinical and research therapies.


Extent of MGMT promoter methylation modifies the effect of temozolomide on overall survival in patients
with glioblastoma: a regional cohort study O-6 Methylguanine-DNA methyltransferase (MGMT) gene promoter methylation is an important biomarker in glioblastoma management. It is associated with better survival and is proposed to enhance efficacy of the alkylating agent temozolomide. [1][2][3][4] Radiotherapy, with concurrent and adjuvant temozolomide, is the standard care "Stupp" regimen for younger and fitter patients following glioblastoma diagnosis.
Measurements of methylation at the 5′-cytosinephosphate-guanine-3′ (CpG) sites along the MGMT promoter determine the methylation status. Quantitative assays generate the percentage of methylated CpG sites, which is dichotomized against assay-specific thresholds. 5 This dichotomy may mask additional prognostic values associated with extent of MGMT methylation.
Some studies have reported a positive association between extent of MGMT methylation and survival, [6][7][8][9] but their smaller cohorts precluded analyses adjusting for confounders. They were also underpowered for examining effect modification of MGMT methylation on temozolomide, as predicted by the postulated mechanism of molecular interaction. This study assessed the association between extent of MGMT methylation and overall survival and evaluated the interaction between MGMT methylation and temozolomide.

Study Design and Setting
This retrospective cohort study included consecutive patients reviewed by a regional neuro-oncology multidisciplinary team in Scotland, UK, between 1 April 2012 and 31 May 2020. Data collection was approved by Southeast Scotland Research Ethics Committee (reference 17/ SS/0019).

Participants
We included patients aged ≥16 years with histologically confirmed glioblastoma following surgery (biopsy or debulking). All patients had a first diagnosis of glioblastoma without prior low-grade glioma. Eligible patients were identified from weekly neuro-oncology multidisciplinary meetings. Data were collected from electronic patient records. Exclusion criteria included patients who: had incomplete treatment data, were lost to follow-up (>1 year since last review/treatment without documented follow-up and no evidence of death), had unavailable pre-operative neuroimaging, and enrolled in a therapeutic clinical trial.

Variables and Data Sources
Clinical characteristics included age at radiological diagnosis, sex, postoperative Karnofsky Performance Status (KPS), extent of resection, temozolomide administration, and MGMT methylation. Primary outcome was overall survival from date of radiological diagnosis. Secondary outcome was 2-year survival. Censoring date for survival data was 16 January 2021.
Experienced board-certified neuropathologists made the histopathological diagnoses of glioblastoma. Isocitrate dehydrogenase 1 (IDH1) mutation testing using immunohistochemistry or sequencing was routine since 2017 and 2020, respectively. Routine magnetic resonance imaging (MRI) is performed within 72 h postoperatively and extent of resection is reported by specialist neuroradiologists. A core was taken from a region of the tumor sample with maximal tumor cellularity for quantifying MGMT promoter methylation. Pyrosequencing using Qiagen MGMT assay generated the mean percentage methylation of the CpG sites (76-79) as the extent of MGMT methylation. The threshold for MGMT methylation was 6.4%, above which indicated the presence of MGMT methylation. We used the nonparametric method detailed in the Clinical and Laboratory Standards Institute guideline to generate the clinical threshold. Our institution offers patients radiotherapy with concurrent and adjuvant temozolomide as standard care. 10 We recorded the number of weeks of concomitant and cycles of adjuvant temozolomide received. Temozolomide use was categorized into none, not completed (less than 6 weeks of concurrent or six adjuvant cycles) and standard "Stupp" regimen

Importance of Study
O-6 Methylguanine-DNA methyltransferase (MGMT) gene promoter methylation is associated with better survival in patients with glioblastoma. It is postulated that this increases efficacy of the alkylating agent temozolomide.
Studies have yet to demonstrate this effect modification in clinical cohorts and to examine the biological gradient to substantiate this proposed mechanism. This study included 414 consecutive patients with histologically diagnosed glioblastoma in Southeast Scotland between April 2012 and May 2020 where quantitative MGMT promoter methylation measurement was routine. Our findings did not demonstrate an interaction between MGMT methylation and completed temozolomide regimen, which suggests no strong effect of MGMT status on survival in patients completing temozolomide regimen. In patients not completing the temozolomide regimen, higher MGMT methylation predicted better survival. Quantitative MGMT methylation may provide additional prognostic information in clinical and therapeutic studies.
(based on completed cycles). Whilst not specifically analyzed in this study, patients who had previously received temozolomide within 12 months received procarbazine, lomustine, and vincristine, or single agent lomustine on disease recurrence. Radiotherapy referred to the primary treatment received after glioblastoma diagnosis and was categorized into 60 Gy, <60 Gy (40 Gy in 15 fractions or 30 Gy in six fractions over 2 weeks), and none.

Handling of Variables
We reported categorical variables as frequencies with percentages and continuous variables as median values with quartiles and ranges. Postoperative KPS was categorized into <70 and ≥70 because of its relevance in informing treatment options. We classified the extent of resection into biopsy only, subtotal resection (<90% debulking), and gross total resection (≥90% debulking) of the contrast enhancing tumor, based on the residual volume identified in the standard care postoperative MRI. When considering the extent of MGMT methylation as a continuous variable, we applied a logarithmic transformation because of the positively skewed distribution. We performed stratification of MGMT methylation to assess whether there was a secondary threshold of prognostic value among patients with a methylated (>6.4%) glioblastoma. We constructed a receiver-operating characteristic curve of logarithmically transformed MGMT methylation with 2-year survival as the dependent variable. We determined the optimal cutoff by maximizing the Youden Index calculated parametrically. The resulting stratified MGMT methylation groups included unmethylated, low methylation, and high methylation groups. Clinical MGMT status refers to the dichotomous MGMT status based on the clinical threshold (6.4%); logarithmically transformed MGMT level is the raw MGMT percentage methylation after natural logarithmic transformation.

Bias and Missing Data
In the literature, pre-operative KPS is usually reported as a prognostic marker. However, oncologists use postoperative KPS as a marker of functional status to inform treatment selection. The inclusion of postoperative KPS made our analyses more robust in reflecting clinical decision making but not consistent with some studies. We did not have missing data to report because of our inclusion and exclusion criteria.

Power Calculation
The hazard ratio (HR) of death associated with MGMT methylation was 0.49 (95% confidence interval 0.41-0.59) in a systematic review and meta-analysis of clinical trial participants. 11 We expected attenuation of this effect size outside the clinical trial setting. To estimate the power of our cohort, we set the hypothesized HR as 0.7, α at 5% and proportion of participants who died during follow-up as 95%. The calculated power was 94% with our sample size. We anticipated the power of our analyses for detecting difference in MGMT methylation subgroups and for assessing interaction was lower.

Statistical Methods
We did not report P-values in descriptive tables to avoid multiple testing. We used the nonparametric Kaplan-Meier estimator to describe the survival function of MGMT methylation groups. We considered age, postoperative KPS, extent of resection, temozolomide use, and radiotherapy as potential confounders based on the literature. 12 The primary survival analysis of the association between MGMT methylation and overall survival was performed using the parametric accelerated failure time (AFT) model with Weibull distribution because of suspicion of hazard nonproportionality. Time ratios were converted to HRs for easier interpretation. We further performed these analyses in patients who had methylated (>6.4%) glioblastoma to assess the relevance in this subgroup, and in patients who received 60 Gy of radiotherapy to assess confounding effect of radiotherapy. We also evaluated the effect of MGMT methylation in older patients aged 65 years or above.
For 2-year mortality, we used multiple logistic regression adjusting for the same confounders other than radiotherapy and postoperative KPS because of complete data separation. Collinearity between variables in the multiple logistic regression was checked by examining the variance inflation factors.
To assess the effect modification between MGMT methylation and temozolomide use on survival, we introduced an interaction term between the two variables in the AFT model. Interaction was considered using the Wald test of the interaction term, and by comparing the interaction model with the main effects model using likelihood ratio test. We compared temozolomide cycles in patients receiving not-completed temozolomide regimen by MGMT methylation groups using Kruskal-Wallis test. We performed all data analyses in R (v4.1.0) 13 using packages "cutpointr" (v1.1.0), 14 "survival" (v3.2.7), 15 and "survminer" (v0.4.8). 16

Participant Characteristics
Of the 439 patients diagnosed with glioblastoma who met inclusion criteria, we excluded 25 patients due to: enrolment in a therapeutic clinical trial (N = 7), lost to follow-up (N = 15) and no pre-operative MRI scan (N = 3). The final study cohort included 414 patients with total follow-up of 526 person-years (

MGMT Promoter Methylation
The median extent of MGMT methylation was 5.0% (IQR 3.0-27.8%). Among the 191 patients with methylated glioblastoma, the median MGMT methylation was 30.0% (IQR 16-42.8%). Figure 1 shows the distributions of MGMT methylation and its logarithmically transformed values. Based on the Youden Index, the optimal cut-off point of MGMT methylation for 2-year survival among patients with a methylated glioblastoma was 25.9% (3.25 in the logarithmic scale) with sensitivity and specificity of 67.8% and 46.6%, respectively. The resulting three groups comprised of unmethylated (N = 223), low methylation (N = 81), and high methylation (N = 110) patients. Characteristics of these groups are presented in Table 1. Higher proportions of patients in the low (23.5%) and high (28.2%) methylation groups received complete concomitant and adjuvant temozolomide compared to the unmethylated group (7.2%). Of the 15 patients with IDH1 mutation, 1 patient had unmethylated, 9 had low methylation, and 5 had high methylation glioblastoma.

Extent of MGMT Methylation and Overall Survival
Considering the extent of MGMT methylation as a logarithmically transformed continuous variable, higher MGMT methylation was associated with better survival in both unadjusted (HR 0.73, 95% CI 0.67-0.79, P < 0.001) and adjusted AFT models (HR 0.85, 95%CI 0.78-0.92, P < 0.001) ( Table 2). When restricting the analyses to 191 patients in the methylated group, without stratifying into high and low extent of MGMT methylation, the logarithmically transformed MGMT methylation was not associated with overall survival (HR 0.89, 95% CI 0.71-1.12, P = 0.314) (Supplementary Table 2). Using the methylation subgroups informed by the Youden Index, both low methylation (HR 0.67, 95% CI 0.50-0.89, P = 0.005) and high methylation (HR 0.60, 95% CI 0.46-0.79, P < 0.001) groups were associated with better survival compared with the unmethylated group (

Extent of MGMT Methylation in Older Patients
There were 168 patients aged ≥65 years at the time of glioblastoma diagnosis (Supplementary Table 5

Effect Modification of Extent of MGMT Promoter Methylation by Temozolomide Regimen
Comparing models with and without interaction, likelihood ratio tests showed preference for the models with interaction terms between the extent of MGMT promoter methylation and temozolomide regimen (P < 0.001). In

Neuro-Oncology Advances
the AFT model with additional interaction terms, logarithmically transformed MGMT methylation showed interaction with not-completed (interaction term P < 0.001) and completed (P = 0.028) temozolomide regimen (Table  3). Interaction between MGMT methylation and notcompleted temozolomide regimen on overall survival was demonstrated when considering the three methylation groups (interaction terms P < 0.001; Table 3). The HRs associated with not-completed temozolomide regimen were 0.79 (95% CI 0.58-1.09, P = 0.153) in unmethylated group, 0.29 (95% CI 0.18-0.45, P < 0.001) in low methylation group and 0.27 (95% CI 0.18-0.40, P < 0.001) in high methylation group. There was no evidence for interaction between MGMT methylation and standard temozolomide regimen All models adjusted for age at diagnosis, post-operative Karnofsky Performance Score, extent of resection and radiotherapy. Hazard ratios were parameterized from the models. Model 1 used the logarithmically transformed MGMT methylation values; Model 2 used the 3 methylation groups determined by clinical threshold and the Youden Index. § Standard temozolomide was in accordance with the Stupp protocol, not-completed regimen referred to less than the total dose of the standard regimen. Model 3 was the same as Model 2 but included patients who received 60 Gy radiotherapy only. HR = hazard ratio; CI = confidence interval; MGMT = O-6 methylguanine-DNA methyltransferase; TMZ = temozolomide; Ref = reference.
(interaction term for low methylation group P = 0.097; high methylation group P = 0.071). The HRs associated with standard temozolomide regimen in all MGMT methylation groups were similar (Table 3). Supplementary Tables 7 and  8 show the adjusted HRs for other variables in the models. Sensitivity analysis of patients who received 60 Gy of radiotherapy suggested a gradient across unmethylated (HR 0.77, 95% CI 0.50-1.19, P = 0.235), low methylation (HR 0.36, 95% CI 0.19-0.68, P = 0.002) and high methylation (0.21, 95% CI 0.12-0.35, P < 0.001) groups in those receiving notcompleted temozolomide (Table 3), though the smaller cohort precluded robust testing of an interaction. We examined whether the interaction with notcompleted temozolomide regimen was due to differences in the number of temozolomide cycles received in each methylation group. Comparison between the groups did not demonstrate evidence of differences (P = 0.092) (Supplementary Figure 1).

Discussion
This retrospective cohort study of surgical patients diagnosed with glioblastoma in Scotland (UK) showed that MGMT promoter methylation was associated with better survival, but this effect was lost in patients aged ≥65 years old. MGMT promoter methylation modified the effects of temozolomide on survival for not-completed, but not standard temozolomide regimen, suggesting the interaction effect waned with increasing temozolomide up to the standard dose. This has implications for our interpretation of observational studies and clinical trials where MGMT status is only dichotomized and completion of standard temozolomide regimen is not considered. We need to clarify the presumptions about molecular mechanisms between MGMT methylation and temozolomide and whether the optimal number of temozolomide cycles vary between patients.
Our findings are consistent with previous studies that reported temozolomide is more beneficial in patients with MGMT methylated glioblastoma than in those with MGMT unmethylated glioblastoma. [17][18][19][20] There is evidence supporting the biological mechanism of this observation. The MGMT protein removes alkyl groups from the O-6 guanine position, repairing the DNA damage caused by alkylating chemotherapies, such as temozolomide. 21 Downregulating MGMT expression by methylation of the MGMT promoter sequence therefore increases temozolomide sensitivity. 21,22 Previous studies have reported an association between higher extent of MGMT promoter methylation and better overall survival. [6][7][8]23 We observed longer survival in the higher methylation group compared with the low methylation group. Our larger cohort also demonstrated an interaction between the effect of multimodal therapy and MGMT methylation on overall survival. This contrasted a study using clinical trial data that reported no evidence of interaction, though the authors recognized the lack of power to test for effect modification. 17 In our study, MGMT methylation modified the effect of temozolomide when the total concomitant and/or adjuvant dosage was less than the standard therapy. The absence of evidence for an interaction with standard temozolomide regimen may reflect the ceiling of benefit at this dose. This may have implications for our assumptions about the mechanism of interaction between temozolomide and MGMT methylation.
Two analyses have investigated the value of continuing temozolomide beyond adjuvant cycles after standard radiotherapy in glioblastoma. 24,25 Both studies used a binary MGMT methylation status and reported no evidence of benefit from extended temozolomide in patients with methylated glioblastoma compared those with unmethylated glioblastoma (HR 0.89, 95%CI 0.63-1.26 24 and HR 1.45, 95%CI 0.89-2.33). 25 We observed similar benefit from standard temozolomide regimen across MGMT methylation groups (Table 3, Model 3). However, there may be a gradient effect (lower HRs with higher methylation) across methylation groups in those receiving notcompleted temozolomide regimen. If this represents a biological interaction between MGMT methylation and temozolomide and our findings suggested a ceiling effect with standard temozolomide, the mix of patients with high and low methylation might have masked the benefit of extended temozolomide in previous studies. This warrants further interrogation. Temozolomide administration in MGMT unmethylated patients remains controversial. Secondary analyses of the EORTC-NCIC trial data suggested minimal benefit for addition of temozolomide to radiotherapy in MGMT unmethylated patients. 17 That study reported median survival of 11.8 months with radiotherapy only and 12.7 months with multimodal therapy. Withholding temozolomide based on MGMT methylation is debated because of limitations in methylation assays and lack of alternative therapies. 26 We have shown that median survival for patients with unmethylated glioblastoma completing multimodal therapy was 26.1 months. The median survival for these patients was 17.1 months in another retrospective study. 7 These observations may reflect careful patient selection for multimodal therapy in those with unmethylated tumors, although our analysis adjusted for factors affecting decision making such as age at diagnosis, functional status, and extent of resection. Standard multimodal therapy is still valuable for selected patients with MGMT unmethylated glioblastoma.

Strengths and Limitations
Our study measured quantitative MGMT methylation in consecutive glioblastoma patients over 8 years, which allowed better power to examine the interaction between MGMT methylation and temozolomide. Because measuring MGMT methylation was routine, there was no missing data that would introduce bias. Residual confounding from other molecular markers and clinical variables could affect our results. However, we adjusted for the most relevant variables associated with survival and showed there was no differential distribution of IDH1 mutation in the high methylation group. Selecting patients based on their clinical characteristics could have inflated benefits from multimodal therapy, particularly in the unmethylated group. Due to smaller cohort size and Neuro-Oncology Advances different distributions in treatment received in this study, future work should clarify the role of MGMT promoter methylation in older patients.

Conclusion
Quantitative extent of MGMT promoter methylation has prognostic value in those receiving 60 Gy radiotherapy. Patients receiving completed standard temozolomide regimen have comparable benefit regardless of MGMT methylation level. Assessing the extent of MGMT methylation in observational studies and clinical trials is necessary to permit accurate interpretation.

Supplementary material
Supplementary material is available at Neuro-Oncology Advances online.