Correlation of conventional magnetic resonance imaging features with O6-methylguanine-DNA-methyltransferase gene promoter methylation status and survival outcomes in patients with newly diagnosed glioblastoma: Single-center correlative imaging substudy from a prospective clinical trial

Tejpal Gupta; Anil Tibdewal; Sarthak Mohanty; Torsten Pietsch; Sadhana Kannan; Shashikant Juvekar; Nikhil Merchant; Sridhar Epari; Aliasgar Moiyadi; Prakash Shetty; Goda Jayant Sastri; Rakesh Jalali

doi:10.4103/glioma.glioma_12_18

ORIGINAL ARTICLE

Year : 2018 | Volume : 1 | Issue : 2 | Page : 50-58

Correlation of conventional magnetic resonance imaging features with O6-methylguanine-DNA-methyltransferase gene promoter methylation status and survival outcomes in patients with newly diagnosed glioblastoma: Single-center correlative imaging substudy from a prospective clinical trial

Tejpal Gupta¹, Anil Tibdewal¹, Sarthak Mohanty¹, Torsten Pietsch², Sadhana Kannan³, Shashikant Juvekar⁴, Nikhil Merchant⁴, Sridhar Epari⁵, Aliasgar Moiyadi⁶, Prakash Shetty⁶, Goda Jayant Sastri¹, Rakesh Jalali¹
¹ Department of Radiation Oncology, ACTREC/TMH, Tata Memorial Centre, Mumbai, Maharashtra, India
² Department of Neuro-Pathology, University of Bonn, Bonn, Germany
³ Department of Clinical Research Secretariat, ACTREC/TMH, Tata Memorial Centre, Mumbai, Maharashtra, India
⁴ Department of Radio-Diagnosis, ACTREC/TMH, Tata Memorial Centre, Mumbai, Maharashtra, India
⁵ Department of Pathology, ACTREC/TMH, Tata Memorial Centre, Mumbai, Maharashtra, India
⁶ Division of Neurosurgery, Department of Surgical Oncology, ACTREC/TMH, Tata Memorial Centre, Mumbai, Maharashtra, India

Date of Web Publication

30-Apr-2018

Correspondence Address:
Dr. Tejpal Gupta
Department of Radiation Oncology, Neuro-Oncology Disease Management Group, ACTREC, Tata Memorial Centre, Kharghar, Navi Mumbai - 410 210, Maharashtra
India

Source of Support: None, Conflict of Interest: None

DOI: 10.4103/glioma.glioma_12_18

Abstract

Background: Imaging features may be reflective of inherent disease biology and serve as potentially useful biomarkers in primary brain tumors. This study aimed to correlate conventional magnetic resonance imaging (MRI) features with O6-methylguanine-DNA-methyltransferase (MGMT) promoter methylation status and survival in glioblastoma. Methods: Conventional semantic imaging features were systematically extracted from preoperative MRI of 34 patients with glioblastoma by two reviewers independently and correlated with MGMT methylation status and survival using appropriate statistical tests. Results: MGMT promoter was methylated in 10 (30%) patients, unmethylated in 15 (44%) patients, and invalid or uninterpretable in 9 (26%) patients. Four imaging features, such as border, edema, contact with subventricular zone (SVZ), and necrosis, showed borderline correlation with methylation status. On multivariate logistic regression analysis, the odds of having methylated tumor were significantly reduced for tumors in contact with SVZ and borderline reduced for tumors with sharp borders. With a median follow-up of 18 months (interquartile range, 13–33 months), the median progression-free survival (PFS) and overall survival (OS) were 12.1 months (95% confidence interval [CI]: 9.9–14.3 months) and 17.1 months (95% CI: 12.6–21.5 months), respectively, for the study cohort. Among the semantic imaging features extracted from conventional MRI, only perilesional edema correlated significantly with PFS as well as OS. The hazards of both progression and death were significantly increased for tumors with moderate-to-severe edema on Cox regression analysis. Conclusion: Contact with SVZ and sharp tumor borders shows weak negative correlation with MGMT promoter methylation status in glioblastoma. Among all MRI features investigated in this work, moderate-to-severe edema is the only imaging feature that independently correlates with significantly inferior survival.

Keywords: Correlation, glioblastoma, magnetic resonance imaging, O6-methylguanine-DNA-methyltransferase, survival

How to cite this article:
Gupta T, Tibdewal A, Mohanty S, Pietsch T, Kannan S, Juvekar S, Merchant N, Epari S, Moiyadi A, Shetty P, Sastri GJ, Jalali R. Correlation of conventional magnetic resonance imaging features with O6-methylguanine-DNA-methyltransferase gene promoter methylation status and survival outcomes in patients with newly diagnosed glioblastoma: Single-center correlative imaging substudy from a prospective clinical trial. Glioma 2018;1:50-8

How to cite this URL:
Gupta T, Tibdewal A, Mohanty S, Pietsch T, Kannan S, Juvekar S, Merchant N, Epari S, Moiyadi A, Shetty P, Sastri GJ, Jalali R. Correlation of conventional magnetic resonance imaging features with O6-methylguanine-DNA-methyltransferase gene promoter methylation status and survival outcomes in patients with newly diagnosed glioblastoma: Single-center correlative imaging substudy from a prospective clinical trial. Glioma [serial online] 2018 [cited 2023 Oct 2];1:50-8. Available from: http://www.jglioma.com/text.asp?2018/1/2/50/231497

Introduction

Glioblastoma, the most common and aggressive primary malignant brain tumor in adults, remains essentially incurable despite standard tri-modality therapy consisting of maximal safe resection followed by postoperative focal conformal radiotherapy to the tumor bed with margins with concurrent daily oral temozolomide and 6–12 cycles of monthly adjuvant temozolomide chemotherapy.^[1] The median survival in glioblastoma has reached a plateau (at 15–18 months) with 2-year survival of 29% and 5-year survival rarely exceeding 10%.^[2] The benefit of addition of temozolomide is mediated through methylation of O6-methylguanine-DNA-methyltransferase (MGMT) promoter.^[3] Methylation of the MGMT promoter inhibits the repair of therapeutic DNA damage induced by temozolomide, rendering it more sensitive to alkylating chemotherapy. MGMT promoter is epigenetically silenced (methylated) in approximately 40% of patients with newly diagnosed glioblastoma where it has now consistently been demonstrated to be a strong and independent prognostic factor ^[4],[5] as well as a predictive marker ^[6] for benefit with alkylating agent chemotherapy.

Glioblastomas like other solid cancers have a complex, dynamically evolving, multiscale ecosystem with significant spatial and temporal heterogeneity ^[7] that lends itself naturally to noninvasive characterization using contemporary imaging modalities including molecular functional imaging.^[8],[9] The heterogeneous neuro-imaging and histopathological and molecular biology spectrum of glioblastoma ^[10] provide unique opportunities for robust subclassification, refined prognostication, and precision medicine.^[11],[12] Magnetic resonance imaging (MRI) is currently recommended as the standard of care imaging ^[13] for glioblastoma covering the entire spectrum of management including but not limited to preoperative diagnosis at initial presentation, quantifying the extent of resection postoperatively, assessing response after adjuvant therapies, and as surveillance for the detection of recurrence/progression on follow-up with potential to influence therapeutic decision-making. In addition, MRI also can help in the assessment of specific phenotypic imaging features to serve as noninvasive biomarkers that could offer insights into the underlying molecular biology of brain tumors.^{[13],[14],[15]} For the aforesaid reasons, we sought to assess the potential of presurgical conventional MRI features in predicting MGMT promoter methylation status and survival in patients with newly diagnosed glioblastoma that were treated on a prospective study at an academic neuro-oncology unit.

Materials and Methods

Patients

All patients included in the study had previously provided informed consent after being prescreened at our site for participation in an institutional review board-approved international Phase III multicentric randomized controlled trial (CENTRIC EORTC 26071-22072 study)^[16] testing the addition of cilengitide to standard treatment (radiotherapy/temozolomide) in glioblastoma. Inclusion/exclusion criteria of CENTRIC study have been described in detail previously. Briefly, eligible patients were adults (≥18 years of age), with newly diagnosed, histologically confirmed supratentorial glioblastoma (WHO grade IV), methylated MGMT promoter, and Eastern Cooperative Oncology Group Performance Status (PS) of 0–1. The sponsor of the index CENTRIC study had no role in the data collection, analysis, interpretation, and reporting of this correlative imaging substudy. The flow of patients on the imaging substudy at our center is depicted in [Figure 1].

Figure 1: Patient flow through the correlative imaging study

Click here to view

Extraction of imaging features

Preoperative imaging datasets of all these patients were retrieved from the study archives. Tumors were assessed for location, contrast enhancement, pattern of enhancement, noncontrast-enhancing tumor (nCET), borders, contact with subventricular zone (SVZ), cysts, T2-signal, edema, multifocality, and necrosis based on previous description.^[17],[18] Epicenter of tumor was used to define location as frontal, parietal, temporal, and occipital. Tumors were considered enhancing if there was unequivocal increase in signal intensity on T1-weighted images following intravenous contrast administration. The pattern of enhancement was categorized as peripheral/rim enhancing or solid/nodular enhancement. nCET was defined as tumor area with no obvious enhancement, but intermediate T2-weighted hyperintensity less than cerebrospinal fluid (CSF) and corresponding to a region of T1-weighted hypointensity associated with mass effect and architectural distortion. Tumor borders were categorized as ill defined/diffusely infiltrative or well defined/sharp based on postcontrast T1-weighted images. SVZ, the adult stem-cell niche, was defined as an area 5 mm lateral to the lateral ventricles; direct extension or contiguity of tumor (and not peritumoral edema) with this stem-cell niche was categorized as having contact with SVZ. Cysts were defined as rounded and well defined, often eccentric areas of >1 cm within the tumor showing very high T2-signal and low T1-signal (matching CSF) with thin, regular, smooth, nonenhancing, or regularly enhancing walls. Signal abnormality on T2-weighted images was categorized as homogeneous or heterogeneous based on >80% of tumor (excluding cysts if any) showing similar intensity. Moderate-to-severe edema was defined as signal abnormality on T2/FLAIR images extending beyond 1 cm from the tumor margin; otherwise, edema was scored as mild or none. Multifocality was defined as having more than one area of tumor separated by normal brain signal intensity on T2-weighted images. Necrosis was defined as a region of peripheral and irregular enhancement surrounding areas of high T2-weighted signal intensity. [Table 1] provides a brief description of the conventional imaging features that were extracted from the preoperative dataset for analysis. Some of these features are also depicted in [Figure 2] as representative images. Systematic assessment of the aforesaid semantic imaging features was performed by two qualified reviewers independently (with over 20 years of experience in neuroradiology at an academic comprehensive cancer center) and blinded to the MGMT methylation status; any discrepancy was resolved by consensus through discussion with a third reviewer. No image reprocessing was done by any of the reviewers before assessment. Robust information regarding use of corticosteroids prior to imaging was lacking in majority of patients; however, our prevalent institutional practice is to start dexamethasone only after neuro-imaging, unless clinical/neurological status of the patient mandates urgent medical decompressive therapy.

Table 1: Systematic visual assessment of preoperative conventional magnetic resonance imaging features in the study cohort

Click here to view

Figure 2: Representative axial magnetic resonance imaging sequences from two patients (upper and lower panels, respectively) demonstrating some of the conventional imaging features assessed in the study cohort. Tumor in the upper panel (A-D) shows sharp and well-defined borders (B and C); nodular enhancement (B); large eccentric cyst (B and C); and severe perifocal edema (C and D). In contrast, tumor in the lower panel (E-H) shows ill-defined and infiltrative borders (E and F); peripheral irregular enhancement (F); central necrosis (F); noncontrast-enhancing component (F and G); and moderate perifocal edema (H). Refer Table 1 for a more detailed description of imaging features

Click here to view

Central pathology review

Formalin-fixed paraffin-embedded (FFPE) tumor tissue blocks were shipped to reference laboratory in Europe for centralized pathology review and assessment of MGMT promoter methylation. A dedicated neuropathologist confirmed the diagnosis of glioblastoma on central pathology review. MGMT promoter methylation status of the tumor was also assessed centrally by a licensed laboratory using quantitative methylation-specific-polymerase chain reaction (MS-PCR) as previously described.^[16],[19] Briefly, two sections were deparaffinized from FFPE blocks and tumor DNA was extracted using the phenol/chloroform extraction method and quantified using commercial quantitation kit. Subsequently, sodium bisulfite modification of DNA was done, which selectively deaminates unmethylated cytosine residues resulting in conversion to uracil, whereas 5-methylcytosine residues within the CpG island remain unmodified. Amplification and quantification of the analyte were done using specific primers and primer/detector pairs. The analyte defined in this assay was the MGMT promoter sequence that detects the fully methylated version. β-actin (ACTB) was used as a reference gene using primers that were outside any CpG islands. To compensate for variations in copy number due to differences in sample volume, handling, and DNA isolation, the methylated MGMT copy numbers derived were divided by ACTB copy numbers for that sample to arrive at a ratio value. Patients were classified as MGMT methylated when the ratio of MGMT to ACTB was two folds or higher.

Treatment and follow-up

Only patients with centrally confirmed MGMT promoter methylation and satisfying other eligibility criteria were randomized in a 1:1 ratio to receive either standard treatment (radiotherapy plus temozolomide) alone or combined with cilengitide (standard dose of 2000 mg intravenously twice weekly on days 1 and 4) that was continued for up to 18 months or disease progression or unacceptable toxicity. Standard treatment consisted of focal three-dimensional conformal radiotherapy delivered to the tumor bed with margins on a 6 MV linear accelerator to a dose of 60 Gy in 30 fractions over 6 weeks (2 Gy per fraction, 5 fractions per week) along with concurrent oral temozolomide (75 mg/m ² daily, all 7 days of the week throughout radiotherapy) followed by six cycles of adjuvant oral temozolomide (150–200 mg/m ² for 5 days consecutively, cycled 4 weekly) starting 4 weeks after the end of concurrent chemoradiotherapy. Patients with unmethylated gene promoter or invalid/uninterpretable results were deemed as prescreen failures and treated with standard therapy alone. All patients were periodically followed up clinicoradiologically for assessment of progression-free survival (PFS) and overall survival (OS) as per protocol. Salvage therapy at progression/recurrence was variable and ranged from re-excision, re-challenge with temozolomide, bevacizumab, re-irradiation, and best supportive care.

Statistical analysis

The correlation of conventional imaging features with MGMT promoter methylation status was assessed using binary Fisher's exact test (for all dichotomous variables) except anatomic location (polychotomous) which was correlated using the Chi-square test. No approximation was used and P ≤ 0.05 was considered statistically significant. MRI features that showed some trend toward significance (P < 0.2) were included in a multivariate logistic regression model to calculate the odds ratio of predicting methylation status with 95% confidence intervals (95% CIs). Logistic regression is a statistical method for analyzing a dataset in which there are one or more independent variables (imaging features) that determine an outcome (MGMT methylation status). In this method, the observed values for the dependent outcome and the predicted values are cross-classified at a user-specified cutoff. The outcome was measured as a dichotomous variable (either methylated or unmethylated). Metrics of diagnostic accuracy such as sensitivity and specificity including area under curve (AUC) with 95% CI were also calculated for the predictive accuracy from this multivariate logistic regression model. PFS and OS were calculated from the date of randomization till the event of interest using the product-limit method of Kaplan–Meier. Apart from known prognostic factors such as age, PS, and MGMT promoter methylation status, conventional MRI features were also compared using the log-rank test. Clinicopathological and imaging features showing a trend toward significance (P < 0.2) were further subjected to multivariate Cox regression analysis to compute the hazard ratio of death with 95% CIs. The cutoff date for all time-to-event analyses was April 30, 2016. All statistical analyses were performed using SPSS version 20.0 (SPSS Inc., Chicago, IL, USA) and STATA version 8.0 (StataCorp LP, College Station, TX, USA).

Results

Between January and December 2010, a total of 34 patients with newly diagnosed supratentorial glioblastoma were consented and prescreened on the CENTRIC study from our institute and their FFPE blocks were shipped to a reference laboratory in Europe for confirmation of diagnosis and MGMT promoter methylation testing. All the 34 patients were confirmed to have glioblastoma or its variants on central pathology review. MGMT promoter was methylated in 10 (30%) patients, unmethylated in 15 (44%) patients, and invalid or uninterpretable in 9 (26%) patients. Correlation of imaging features with MGMT methylation status was restricted to patients with unequivocal results obtained from MS-PCR (n = 25), excluding patients with invalid/uninterpretable results, while correlation with survival outcomes was done for the entire study cohort (n = 34). Of the 10 patients with methylated tumors, only 3 patients were randomized to the investigational arm (cilengitide + standard therapy), while 5 patients were randomized to the control arm (standard therapy alone). Two patients with methylated tumors were deemed screen failures who, along with the remaining 24 patients (either unmethylated tumors or invalid/uninterpretable methylation status), received standard therapy (radiotherapy plus temozolomide) as described earlier. The median age of patients included in the study was 52 years (interquartile range, 40–58 years) and the median Karnofsky status was 90 (range, 80–100).

Although a formal test of agreement between the two readers in assessing conventional imaging features was not performed, it is reassuring to note that any discrepancy necessitating a third review for consensus was necessary <10% of times overall. The results of correlation of conventional imaging features with MGMT methylation status are depicted in [Figure 3]. The Chi-square statistic was not computed for three imaging features, namely enhancement (yes vs. no), T2-signal characteristics (homogeneous vs. heterogeneous), and multifocality (yes vs. no), as all the 34 tumors showed some degree of enhancement as well as T2 heterogeneity and only 1 of the 34 patients had a multifocal tumor. None of the tested conventional imaging features showed any statistically significant correlation with MGMT promoter methylation status. However, a few imaging features such as border, edema, contact with SVZ, and necrosis did show some trend toward borderline significance and were subjected to a multivariate logistic regression analysis [Table 2]. The odds of having a methylated MGMT promoter was significantly reduced for tumors in contact with the SVZ and borderline reduced for tumors with sharp borders. The presence of necrosis and moderate/severe edema increased the odds of having methylated tumors though not statistically significant. The sensitivity (95% CI) of individual imaging feature for predicting MGMT methylation status was acceptably high at 90% (55%–99%), 90% (50%–99%), 80% (44%–97%), and 80% (44%–97%) for moderate/severe edema, presence of necrosis, ill-defined borders, and contact with SVZ, respectively. However, corresponding figures for specificity were rather low at 40% (16%–68%), 40% (16%–84%), 60% (32%–84%), and 60% (32%–84%), respectively, resulting in a modest overall diagnostic accuracy of 60% (39%–79%), 60% (39%–79%), 66% (46%–85%), and 66% (46%–85%), respectively, for each of these individual imaging features. Predictive accuracy improved incrementally by combining imaging features with each other, yielding a sensitivity, specificity, and overall diagnostic accuracy of 80% (44%–97%), 87% (59%–98%), and 84% (64%–95%), respectively, when all the four imaging features were combined. The AUC based on predictive probabilities obtained from the model using all the four features (edema, necrosis, ill-defined borders, and contact with SVZ) was 90.7 (95% CI: 78.6–100).

Figure 3: Correlation of conventional imaging features with O6-methylguanine-DNA-methyltransferase promoter methylation status compared as proportion of patients with and without the particular imaging feature in question using the Fisher's exact test and depicted on a bar diagram. SVZ stands for subventricular zone and nCET represents noncontrast-enhancing tumor

Click here to view

Table 2: Multivariate logistic regression analysis of conventional imaging features predicting O6-methylguanine-DNA-methyltransferase promoter methylation status

Click here to view

A total of 32 of the 34 patients consented to the study had died by the time of this analysis. With a median follow-up of 18 months (interquartile range, 13–33 months), the median PFS and OS were 12.1 months (95% CI: 9.9–14.3 months) and 17.1 months (95% CI: 12.6–21.5 months), respectively, for the entire study cohort. None of the conventional imaging features, except edema, showed any significant correlation with survival outcomes when compared using the Kaplan–Meier log-rank test [Table 3]. Patients with moderate-to-severe edema had significantly inferior survival compared to patients with no or mild edema [Figure 4]. None of the known clinicopathologic prognostic factors (age, PS, and MGMT promoter methylation status) were found significant on univariate analysis [Table 3]. On Cox regression analysis [Table 4], the hazards of progression and death were both significantly increased for tumors with moderate-to-severe edema compared to tumors with no or mild edema. Contact with SVZ was also associated with a nonsignificantly increased risk of progression and borderline significant increased risk of death, while presence of necrosis did not show any significant correlation with survival.

Table 3: Univariate analysis for correlation of imaging and clinicopathologic features with survival outcomes in newly diagnosed glioblastoma using the Kaplan–Meier log-rank test

Click here to view

Figure 4: Kaplan–Meier curves showing significantly worse progression-free survival (A) and overall survival (B) in patients with moderate-to-severe edema (solid line) versus no or mild edema (dotted line)

Click here to view

Table 4: Multivariate Cox regression analysis correlating imaging features with survival outcomes

Click here to view

Discussion

We attempted to correlate semantic imaging features extracted from conventional MRI with MGMT promoter methylation status and survival outcomes in patients with newly diagnosed glioblastoma. In our study, conventional imaging features, namely, border, edema, contact with SVZ, and necrosis, showed some trend toward correlation with MGMT promoter methylation status. However, their accuracy for predicting methylation status individually was clearly suboptimal. Although it could be improved significantly by combining all the four features, its use in the clinic cannot be recommended presently. Among the imaging features tested for correlation with survival, moderate-to-severe edema and contact with SVZ predicted for shorter survival in our study.

The use of phenotypic imaging features to identify genotype of primary brain tumors and its correlation with outcomes has been an active area of research in contemporary neuroradiology.^[15] Eoli et al.^[20] were among the first group that reported an association between ring enhancement on MRI and unmethylated MGMT promoter in glioblastoma. This association was further confirmed by Drabycz et al.^[21] in a cohort of 59 glioblastomas, wherein ring enhancement was seen in 93% of unmethylated tumors compared to 61% of methylated tumors (P = 0.006). MGMT methylation status was not significantly associated with any other visually assessed MRI features such as tumor margins, presence of cysts, and T2-signal intensity. In addition, the authors performed texture analysis of images using novel quantitative space–frequency transformation algorithms that did not show any significant association with methylation. However, combining ring enhancement with texture analysis improved the predictive accuracy of MGMT promoter methylation. In a cohort of 24 patients with high-grade gliomas, Moon et al.^[22] assessed imaging features such as attenuation on computed tomography (CT), apparent diffusion coefficient (ADC), fractional anisotropy (FA), and relative cerebral blood volume (rCBV) in addition to morphologic features as biomarkers for MGMT methylation. Maximum CT attenuation was significantly lower, while ADC and FA ratios were significantly higher in methylated tumors. The rCBV ratio did not differ between the two groups. Among conventional morphologic imaging features, only ill-defined margins were seen more frequently in methylated tumors than unmethylated ones. Significant positive correlation between higher ADC values and methylation of the MGMT promoter has subsequently been demonstrated by two other groups.^[23],[24] However, Ahn et al.^[25] in a retrospective study on 43 glioblastoma patients reported no significant differences in ADC and FA with regard to MGMT methylation status. Instead, one of the permeability parameters, transfer constant (K_trans) on dynamic contrast-enhanced MRI, was found to be significantly higher in methylated tumors. Notably, none of the conventional imaging features correlated with methylation status. More recently, the utility of positron-emission tomography (PET) imaging using tracers such as 11-C-methionine (MET) and 18F-flouro-deoxyglucose (FDG) has also been explored to predict MGMT methylation with conflicting results. While one study suggested that higher uptake on MET-PET may be suggestive of methylation, the other study did not find any such correlation.^[26] Instead, the authors reported a positive correlation between FDG-uptake and MGMT methylation.^[27]

The prognostic impact of specific imaging features on survival in glioblastoma was systematically assessed by Pope et al.^[17] in 110 patients with glioblastoma. Among the 15 conventional imaging features studied, only nCET predicted better survival. Edema, satellite lesions, and multifocality were all associated with poorer survival on univariate analysis. In a subsequent analysis,^[18] the same group retrospectively demonstrated the relationship between imaging features (nCET and edema); underlying molecular abnormalities (IDH1 mutation and MGMT promoter methylation); and survival in a cohort of 202 patients with glioblastoma. A higher proportion of nCET, frontal location, larger tumor size, presence of cysts, and satellite lesions all correlated with IDH1 mutation. However, none of the imaging features correlated with MGMT promoter methylation status. Both IDH1 mutation and MGMT methylation were associated with longer survival as expected. Edema stratified survival in methylated tumors but not in unmethylated ones. Patients with MGMT methylation and little/no edema had particularly long survival. In another study by Li et al.,^[28] degree of contrast enhancement (<5% vs. more), perilesional edema (absent vs. present), and intensity on T2-weighted images (hyper- to iso-intense vs. iso-to-hypo-intense) correlated with survival in a retrospective cohort of thirty glioblastoma patients pointing their utility as prognostic imaging biomarkers. More recently, Wu et al.^[29] demonstrated independent adverse prognostic impact of major edema and necrosis (seen on preoperative imaging) on survival in 87 patients with primary glioblastoma. The most comprehensive analysis of imaging predictors of molecular profile and survival is derived from The Cancer Genome Atlas ^{More Details} (TCGA) glioblastoma dataset.^[30] Preoperative images of 75 patients of glioblastoma with known genetic data in the TCGA portal were assessed for size, location, and tumor morphology using a 30-item standardized feature set named Visually AcceSAble Rembrandt Imaging (VASARI) that was developed to normalize grading of visual/subjective MRI features of malignant glioma.^[15] Associations between survival, tumor size, and morphology were determined using multivariate Cox regression models; associations between imaging features and genomics were studied using the Fisher's exact test. Contrast-enhanced tumor volume (P = 0.025) and longest axis length of tumor (P = 0.009) were strongly associated with poor survival even after adjusting for Karnofsky PS. Glioblastomas of proneural subtype had significantly lower levels of contrast enhancement (P = 0.02) compared to others; while mesenchymal subtype glioblastomas showed lower levels of nonenhancing tumor (P = 0.01). The authors concluded that comprehensive assessment based on standardized set of MRI features is predictive of OS in glioblastoma and that this can be combined with genetic alterations and gene expression profiles to provide deeper insights into the underlying molecular biology of glioblastoma. Other groups have also subsequently confirmed that the addition of imaging features including hemodynamic imaging biomarkers ^[31] and imaging invasive phenotypes ^[32] to genetic markers strengthens the glioblastoma survival prediction within TCGA dataset.^[33] More recently, Jamjoom et al.^[34] in a cohort of 46 glioblastoma patients demonstrated that MRI diffusion metrics indicative of high focal cellularity and steeper transition from high cellular tumor edge to low cellular edema defined a more aggressive subtype of glioblastoma with poorer prognosis. A previous systematic review of seven studies examining the relationship between preoperative peritumoral edema and survival in newly diagnosed glioblastoma was inconclusive; the author concluded that the available evidence was not sufficient to definitely support or refute a putative relationship due to contradictory results and significant heterogeneity between studies.^[35] However, subsequent reports have demonstrated the utility of peritumoral edema as an independent prognostic factor, something that is also corroborated by the present study.^[28],[29] The borderline prognostic impact of contact with SVZ on survival in our study is likely due to the small numbers. Aggressive phenotype of glioblastomas in close contact with SVZ has been previously demonstrated and is widely accepted;^[36] such tumors are more likely to be multicentric at initial diagnosis, have noncontiguous distant brain failure at recurrence, commonly display stem cell-associated gene signatures, and are invariably associated with poorer survival.^[37],[38]

Caveats and limitations

The number of patients included in our analysis was rather low (n = 34), precluding robust conclusions. Furthermore, over a quarter of them had invalid/uninterpretable MGMT methylation status, further reducing the sample size and strength of inference. Preoperative imaging on these patients was done at various radiology centers on different machines. Hence, lack of uniformity and standardization in image acquisition and processing protocols with a consequent effect on imaging features cannot be completely ruled out. It was largely a qualitative analysis and not quantitative analysis. Analysis of MRI was limited to semantic features (by visual assessment), excluding advanced MRI techniques such as spectroscopy, perfusion, and diffusion imaging that may have a higher predictive value. Subjectivity in such visual assessment can introduce variable degree of bias in interpretation that can be significantly reduced using a standardized feature set such as VASARI which was lacking in our study. Finally, we did not attempt to use automated texture-based features that are nowadays being increasingly applied for such correlative studies.^[15],[39] Newer MRI techniques such as spectroscopy and diffusion and perfusion imaging and novel methodologies such as automated texture analysis and PET may have potential to further improve radiogenomic correlation.

Radiomics workflow in the contemporary era has evolved tremendously from manual methods to autosegmentation and from visual interpretation of imaging features to automated high-throughput feature extraction. Despite improvements in predictive modeling of imaging datasets, several logistic, computational, and clinical challenges remain.^[15] It is imperative to develop robust statistical models correlating imaging features with outcomes for the radiomics approach to be useful in clinical practice. More studies with larger number of patients correlating phenotypic imaging features with genotypes and disease outcomes are needed to provide deeper insights into the biology of glioblastoma.

Systematic visual assessment of conventional imaging features though subjective is feasible in the clinic. Contact with SVZ and sharp tumor borders show weak negative correlation with MGMT promoter methylation status. However, their accuracy in predicting methylation status singly is somewhat suboptimal at present which improves when used in combination with edema and necrosis. Perilesional edema emerged as the only independent prognostic imaging feature among all MRI features investigated in this work, with moderate-to-severe edema being associated with significantly inferior survival outcomes.

Financial support and sponsorship

MGMT promoter methylation status was assessed for all patients at a central laboratory in Europe through participation in the CENTRIC study sponsored by Merck. The funders had no role in the data collection, analysis, interpretation, and reporting of this correlative imaging substudy.

Conflicts of interest

There are no conflicts of interest.

References

1.	Stupp R, Mason WP, van den Bent MJ, Weller M, Fisher B, Taphoorn MJ, et al. Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med 2005;352:987-96. [PUBMED]
2.	Stupp R, Hegi ME, Mason WP, van den Bent MJ, Taphoorn MJ, Janzer RC, et al. Effects of radiotherapy with concomitant and adjuvant temozolomide versus radiotherapy alone on survival in glioblastoma in a randomised phase III study: 5-year analysis of the EORTC-NCIC trial. Lancet Oncol 2009;10:459-66. [PUBMED]
3.	Hegi ME, Diserens AC, Gorlia T, Hamou MF, de Tribolet N, Weller M, et al. MGMT gene silencing and benefit from temozolomide in glioblastoma. N Engl J Med 2005;352:997-1003. [PUBMED]
4.	Chen Y, Hu F, Zhou Y, Chen W, Shao H, Zhang Y, et al. MGMT promoter methylation and glioblastoma prognosis: A systematic review and meta-analysis. Arch Med Res 2013;44:281-90.
5.	Zhang K, Wang XQ, Zhou B, Zhang L. The prognostic value of MGMT promoter methylation in glioblastoma multiforme: A meta-analysis. Fam Cancer 2013;12:449-58. [PUBMED]
6.	Olson RA, Brastianos PK, Palma DA. Prognostic and predictive value of epigenetic silencing of MGMT in patients with high grade gliomas: A systematic review and meta-analysis. J Neurooncol 2011;105:325-35. [PUBMED]
7.	Gillies RJ, Anderson AR, Gatenby RA, Morse DL. The biology underlying molecular imaging in oncology: From genome to anatome and back again. Clin Radiol 2010;65:517-21. [PUBMED]
8.	Mahajan A, Goh V, Basu S, Vaish R, Weeks AJ, Thakur MH, et al. Bench to bedside molecular functional imaging in translational cancer medicine: To image or to imagine? Clin Radiol 2015;70:1060-82. [PUBMED]
9.	von Neubeck C, Seidlitz A, Kitzler HH, Beuthien-Baumann B, Krause M. Glioblastoma multiforme: Emerging treatments and stratification markers beyond new drugs. Br J Radiol 2015;88:20150354.
10.	Aubry M, de Tayrac M, Etcheverry A, Clavreul A, Saikali S, Menei P, et al. From the core to beyond the margin: A genomic picture of glioblastoma intratumor heterogeneity. Oncotarget 2015;6:12094-109.
11.	Prados MD, Byron SA, Tran NL, Phillips JJ, Molinaro AM, Ligon KL, et al. Toward precision medicine in glioblastoma: The promise and the challenges. Neuro Oncol 2015;17:1051-63. [PUBMED]
12.	Ree AH, Redalen KR. Personalized radiotherapy: Concepts, biomarkers and trial design. Br J Radiol 2015;88:20150009.
13.	Bohman LE, Swanson KR, Moore JL, Rockne R, Mandigo C, Hankinson T, et al. Magnetic resonance imaging characteristics of glioblastoma multiforme: Implications for understanding glioma ontogeny. Neurosurgery 2010;67:1319-27. [PUBMED]
14.	Ellingson BM. Radiogenomics and imaging phenotypes in glioblastoma: Novel observations and correlation with molecular characteristics. Curr Neurol Neurosci Rep 2015;15:506. [PUBMED]
15.	Narang S, Lehrer M, Yang D, Lee J, Rao A. Radiomics in glioblastoma: Current status, challenges, and potential opportunities. Transls Cancer Res 2016;5:383-97.
16.	Stupp R, Hegi ME, Gorlia T, Erridge SC, Perry J, Hong YK, et al. Cilengitide combined with standard treatment for patients with newly diagnosed glioblastoma with methylated MGMT promoter (CENTRIC EORTC 26071-22072 study): A multicentre, randomised, open-label, phase 3 trial. Lancet Oncol 2014;15:1100-8. [PUBMED]
17.	Pope WB, Sayre J, Perlina A, Villablanca JP, Mischel PS, Cloughesy TF, et al. MR imaging correlates of survival in patients with high-grade gliomas. AJNR Am J Neuroradiol 2005;26:2466-74.
18.	Carrillo JA, Lai A, Nghiemphu PL, Kim HJ, Phillips HS, Kharbanda S, et al. Relationship between tumor enhancement, edema, IDH1 mutational status, MGMT promoter methylation, and survival in glioblastoma. AJNR Am J Neuroradiol 2012;33:1349-55. [PUBMED]
19.	Vlassenbroeck I, Califice S, Diserens AC, Migliavacca E, Straub J, Di Stefano I, et al. Validation of real-time methylation-specific PCR to determine O6-methylguanine-DNA methyltransferase gene promoter methylation in glioma. J Mol Diagn 2008;10:332-7. [PUBMED]
20.	Eoli M, Menghi F, Bruzzone MG, De Simone T, Valletta L, Pollo B, et al. Methylation of O6-methylguanine DNA methyltransferase and loss of heterozygosity on 19q and/or 17p are overlapping features of secondary glioblastomas with prolonged survival. Clin Cancer Res 2007;13:2606-13. [PUBMED]
21.	Drabycz S, Roldán G, de Robles P, Adler D, McIntyre JB, Magliocco AM, et al. An analysis of image texture, tumor location, and MGMT promoter methylation in glioblastoma using magnetic resonance imaging. Neuroimage 2010;49:1398-405.
22.	Moon WJ, Choi JW, Roh HG, Lim SD, Koh YC. Imaging parameters of high grade gliomas in relation to the MGMT promoter methylation status: The CT, diffusion tensor imaging, and perfusion MR imaging. Neuroradiology 2012;54:555-63. [PUBMED]
23.	Romano A, Calabria LF, Tavanti F, Minniti G, Rossi-Espagnet MC, Coppola V, et al. Apparent diffusion coefficient obtained by magnetic resonance imaging as a prognostic marker in glioblastomas: Correlation with MGMT promoter methylation status. Eur Radiol 2013;23:513-20. [PUBMED]
24.	Sunwoo L, Choi SH, Park CK, Kim JW, Yi KS, Lee WJ, et al. Correlation of apparent diffusion coefficient values measured by diffusion MRI and MGMT promoter methylation semiquantitatively analyzed with MS-MLPA in patients with glioblastoma multiforme. J Magn Reson Imaging 2013;37:351-8. [PUBMED]
25.	Ahn SS, Shin NY, Chang JH, Kim SH, Kim EH, Kim DW, et al. Prediction of methylguanine methyltransferase promoter methylation in glioblastoma using dynamic contrast-enhanced magnetic resonance and diffusion tensor imaging. J Neurosurg 2014;121:367-73. [PUBMED]
26.	Okita Y, Nonaka M, Shofuda T, Kanematsu D, Yoshioka E, Kodama Y, et al. (11)C-methionine uptake correlates with MGMT promoter methylation in nonenhancing gliomas. Clin Neurol Neurosurg 2014;125:212-6. [PUBMED]
27.	Choi H, Bang JI, Cheon GJ, Kim YH, Park CK, Park SH, et al.¹⁸ F-fluorodeoxyglucose and ¹¹C-methionine positron emission tomography in relation to methyl-guanine methyltransferase promoter methylation in high-grade gliomas. Nucl Med Commun 2015;36:211-8. [PUBMED]
28.	Li WB, Tang K, Chen Q, Li S, Qiu XG, Li SW, et al. MRI manifestions correlate with survival of glioblastoma multiforme patients. Cancer Biol Med 2012;9:120-3. [PUBMED]
29.	Wu CX, Lin GS, Lin ZX, Zhang JD, Liu SY, Zhou CF, et al. Peritumoral edema shown by MRI predicts poor clinical outcome in glioblastoma. World J Surg Oncol 2015;13:97.
30.	Gutman DA, Cooper LA, Hwang SN, Holder CA, Gao J, Aurora TD, et al. MR imaging predictors of molecular profile and survival: Multi-institutional study of the TCGA glioblastoma data set. Radiology 2013;267:560-9. [PUBMED]
31.	Jain R, Poisson L, Narang J, Gutman D, Scarpace L, Hwang SN, et al. Genomic mapping and survival prediction in glioblastoma: Molecular subclassification strengthened by hemodynamic imaging biomarkers. Radiology 2013;267:212-20. [PUBMED]
32.	Colen RR, Vangel M, Wang J, Gutman DA, Hwang SN, Wintermark M, et al. Imaging genomic mapping of an invasive MRI phenotype predicts patient outcome and metabolic dysfunction: A TCGA glioma phenotype research group project. BMC Med Genomics 2014;7:30. [PUBMED]
33.	Nicolasjilwan M, Hu Y, Yan C, Meerzaman D, Holder CA, Gutman D, et al. Addition of MR imaging features and genetic biomarkers strengthens glioblastoma survival prediction in TCGA patients. J Neuroradiol 2015;42:212-21. [PUBMED]
34.	Jamjoom AA, Rodriguez D, Rajeb AT, Manita MA, Shah KA, Auer DP, et al. Magnetic resonance diffusion metrics indexing high focal cellularity and sharp transition at the tumour boundary predict poor outcome in glioblastoma multiforme. Clin Radiol 2015;70:1400-7.
35.	Liu SY, Mei WZ, Lin Z ×. Pre-operative peritumoral edema and survival rate in glioblastoma multiforme. Onkologie 2013;36:679-84.
36.	Lim DA, Cha S, Mayo MC, Chen MH, Keles E, VandenBerg S, et al. Relationship of glioblastoma multiforme to neural stem cell regions predicts invasive and multifocal tumor phenotype. Neuro Oncol 2007;9:424-9. [PUBMED]
37.	Adeberg S, Bostel T, König L, Welzel T, Debus J, Combs SE, et al. A comparison of long-term survivors and short-term survivors with glioblastoma, subventricular zone involvement: A predictive factor for survival? Radiat Oncol 2014;9:95.
38.	Adeberg S, König L, Bostel T, Harrabi S, Welzel T, Debus J, et al. Glioblastoma recurrence patterns after radiation therapy with regard to the subventricular zone. Int J Radiat Oncol Biol Phys 2014;90:886-93.
39.	Alobaidli S, McQuaid S, South C, Prakash V, Evans P, Nisbet A, et al. The role of texture analysis in imaging as an outcome predictor and potential tool in radiotherapy treatment planning. Br J Radiol 2014;87:20140369.