Comparison of Genetic Profiles and Prognosis of High-Grade Gliomas Using Quantitative and Qualitative MRI Features: A Focus on G3 Gliomas

Abstract

Objective

To evaluate the association of MRI features with the major genomic profiles and prognosis of World Health Organization grade III (G3) gliomas compared with those of glioblastomas (GBMs).

Materials and Methods

We enrolled 76 G3 glioma and 155 GBM patients with pathologically confirmed disease who had pretreatment brain MRI and major genetic information of tumors. Qualitative and quantitative imaging features, including volumetrics and histogram parameters, such as normalized cerebral blood volume (nCBV), cerebral blood flow (nCBF), and apparent diffusion coefficient (nADC) were evaluated. The G3 gliomas were divided into three groups for the analysis: with this isocitrate dehydrogenase (IDH)-mutation, IDH mutation and a chromosome arm 1p/19q-codeleted (IDHmut1p/19qdel), IDH mutation, 1p/19q-nondeleted (IDHmut1p/19qnondel), and IDH wildtype (IDHwt). A prediction model for the genetic profiles of G3 gliomas was developed and validated on a separate cohort. Both the quantitative and qualitative imaging parameters and progression-free survival (PFS) of G3 gliomas were compared and survival analysis was performed. Moreover, the imaging parameters and PFS between IDHwt G3 gliomas and GBMs were compared.

Results

IDHmut G3 gliomas showed a larger volume (p = 0.017), lower nCBF (p = 0.048), and higher nADC (p = 0.007) than IDHwt. Between the IDHmut tumors, IDHmut1p/19qdel G3 gliomas had higher nCBV (p = 0.024) and lower nADC (p = 0.002) than IDHmut1p/19qnondel G3 gliomas. Moreover, IDHmut1p/19qdel tumors had the best prognosis and IDHwt tumors had the worst prognosis among G3 gliomas (p

Conclusion

We found significant differences in MRI features, including volumetrics, CBV, and ADC, in G3 gliomas, according to IDH mutation and 1p/19q codeletion status, which can be utilized for the prediction of genomic profiles and the prognosis of G3 glioma patients. The MRI signatures and prognosis of IDHwt G3 gliomas tend to follow those of IDHwt GBMs.

INTRODUCTION

Diffuse lower grade gliomas (LGGs) are infiltrative brain neoplasms that conventionally include World Health Organization (WHO) grade II (G2) and III (G3) astrocytomas and oligodendrogliomas based on histological characteristics (1). These tumors demonstrate high variability in prognosis, with varied survival that ranges from 1 to 15 years (2). The most recent WHO classification has incorporated the genomic characteristics of these tumors with conventional phenotypic characteristics to produce an integrated diagnosis for diffuse gliomas, as recent studies regarding genomic analysis of gliomas have provided a better understanding of LGGs (1).

The two most well-known genetic mutations incorporated in the newest WHO guidelines are an isocitrate dehydrogenase (IDH)-1 mutation and a chromosome arm 1p/19q (1p/19q) codeletion. The tumors with this IDH mutation (IDHmut) have a better prognosis with a higher sensitivity to chemoradiation therapy than those without this IDH wildtype (IDHwt) (3, 4). Codeletion of 1p/19q is another genetic modification that is associated with a favorable prognosis and treatment response (5, 6) and is commonly found in oligodendroglial tumors.

Magnetic resonance imaging (MRI) is the most reliable imaging modality for the brain and provides not only geographic spatial characteristics, but also functional and molecular information about brain tissue by using various updated techniques, such as perfusion- and diffusion-weighted images. The correlation of genomic information with imaging parameters to identify the characteristics of G2 and G3 gliomas has been investigated in previous studies (7, 8, 9, 10, 11). Even with these interesting findings, the association of genetic profiles with imaging parameters of G3 gliomas still remains somewhat controversial. Even though one previous study has analyzed VASARI features of G3 anaplastic gliomas (12), there is no previous study that includes a large number of patients to fully elucidate both the qualitative and quantitative MR characteristics of G3 gliomas. Therefore, we found the need for a comprehensive investigation of the quantitative and qualitative imaging parameters of G3 gliomas to identify the specific imaging parameters that distinguish G3 glioma subtypes according to their genetic profiles. Moreover, as the IDHwt LGGs are known to behave similarly to IDHwt glioblastomas (GBMs) (7), comparing the imaging findings and prognosis of IDHwt G3 gliomas and GBMs can provide advantages in clinical therapeutic management planning and prognosis prediction for glioma patients.

Thus, the purpose of our study is to investigate the association of the qualitative and quantitative MRI features and major genomic profiles of G3 gliomas and to compare the imaging findings and prognosis between IDHwt gliomas and GBMs.

MATERIALS AND METHODS

Patients

This retrospective study was approved by the Institutional Review Board of Seoul National University Hospital. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. The requirement for informed consent was waived. One hundred thirty-four patients who were initially diagnosed with astrocytomas and oligodendrogliomas between January 2011 and September 2016 at our institution were consecutively included in the study. Using the exclusion criteria, a total of 76 patients were included in our study (47 males, 29 females; mean age, 47.69 ± 14.86 years; age range, 19 to 68 years). A detailed flow diagram of patient selection and classification is shown in Figure 1.

Fig. 1 Flowchart for selection of the study population. G3 = World Health Organization grade III

To compare the imaging features of IDHwt LGGs and GBMs, we utilized the same inclusion and exclusion criteria mentioned above for G3 gliomas, except a requirement of histopathologic diagnosis of GBM was used for the GBM study. A total of 155 IDHwt GBM patients were included in the comparison.

Image Acquisition

For each patient, MR images were acquired by using a 1.5- or 3T MRI scanner with an 8-, 32-, or 64-channel head coil. MR scan parameters are summarized in Supplementary Table 1.

Image Processing

The conventional MR images, apparent diffusion coefficient (ADC) maps, and dynamic susceptibility contrast-enhanced perfusion-weighted imaging were digitally transferred from the picture-archiving and communication system (Infinitt, Infinitt) workstation to a personal computer for further analysis. All imaging analyses were performed by using a dedicated software package (NordicICE, NordicNeuroLab).

Image Analysis

Location, presence of an enhancing portion, necrosis, T2 hypersignal intensity margin of the tumor, and presence of a “T2-fluid attenuated inversion recovery (FLAIR) mismatch sign” were evaluated (Supplementary Material 1). One investigator, (with 3 years in brain MRI experience), who was blinded to the patient information and data, performed all image analyses and drew polygonal regions of interest (ROIs) that contained the T2 high signal intensity lesions in each section of resliced T2-weighted image (T2WI) and FLAIR images, including the cystic and necrotic regions. All of the ROIs were confirmed by an expert, board-certified neuroradiologist, (with 15 years in brain MRI experience).

Detailed information of image analysis is depicted in Supplementary Material 1 and description of the imaging parameters is summarized in Supplementary Table 2.

Molecular/Genetic Analysis

Detailed explanation of molecular genetic analysis is demonstrated in Supplementary Material 2.

Progression-Free Survival

Progression-free survival (PFS) was calculated from the date of initial diagnosis and progression according to the Response Assessment in Neuro-Oncology criteria (13). We only considered the first instance of progression or death, whichever comes first, and all the G3 glioma patients included in the study underwent surgical resection or biopsy, followed by radiation therapy as routine treatment for G3 glioma at our institution.

Validation Data Set

We utilized another set of G3 glioma cases to validate our predictive imaging parameters derived from the multivariable analysis to identify any association between the imaging features of G3 gliomas and the genetic profiles of the tumors. From October 2016 to July 2017, 17 G3 glioma patients with preoperative MRI, histologic diagnoses, and genetic information were available for evaluation. The quantitative imaging parameters were evaluated using the same methodology as mentioned above.

Statistical Analysis

All statistical analyses were performed with MedCalc (version 12.1.0 for Microsoft Windows 2000/XP/Vista/7, MedCalc Software). The results with a p value of less than 0.05 were considered statistically significant.

First, a comparison of the imaging features between the IDHmut group and IDHwt group was performed, and subsequently, a comparison of the imaging features between IDH-mutant, 1p/19q-codeleted (IDHmut1p/19qdel) and IDH-mutant, 1p/19q-nondeleted (IDHmut1p/19qnondel) was performed. Variables were compared between the two groups using unpaired Student's t test, Mann-Whitney U test, and Fisher's exact test. Subsequently, univariable and multivariable regression analyses were used to determine significant associations between the imaging parameters and the molecular information of G3 gliomas.

Then, the G3 gliomas were divided into three groups according to the different genetic profiles for the comparison of imaging features using ANOVA: IDHmut1p/19qdel, IDHmut1p/19qnondel, and IDHwt. Additionally, using the Kaplan-Meier method, we compared the PFS according to the different genetic profiles of the tumors. Survival curves were compared using a log-rank test and multivariable analysis was performed using the Cox proportional hazards model to find correlation between imaging features and PFS.

In addition, comparison of imaging features was performed using unpaired Student's t test, Mann-Whitney U test, and Fisher's exact test, and comparison of survival between IDHwt G3 gliomas and IDHwt GBMs was performed using Kaplan-Meier method.

To evaluate the reproducibility of the quantitative analysis and interobserver agreement, we utilized the intraclass correlation coefficient (ICC). Detailed description of statistical analysis for interobserver agreement is depicted in Supplementary Material 3.

RESULTS

Patient Clinical Parameters and Histological and Molecular Characteristics of the Tumor

Table 1 summarizes the patient demographics, pathological diagnosis, and the genetic characteristics of the tumors included in the study.

Table 1 Patient Demographics and Genetic Profiles of the Tumors of 76 Patients

Qualitative Analysis

IDHmut tumors presented with higher proportions of tumors with necrosis than IDHwt tumors (p = 0.008). Among the IDHmut tumors, the IDHmut1p/19del group presented with a higher proportion of enhancement (p = 0.004) (Supplementary Table 3). The T2-FLAIR mismatch sign was present in 28 of the 76 (36.8%) G3 gliomas included in the study. The accuracy, sensitivity, and specificity of the T2-FLAIR mismatch sign for the prediction of IDHmut1p/19qnondel G3 gliomas were 81.6%, 64.3%, and 91.7%, respectively. There was no statistically significant difference in lobar involvement or the proportion of well-defined margins of tumors according to histologic classification or genetic features of the tumor.

Quantitative Analysis

The IDHmut tumors showed a larger tumor volume in the T2WI (p = 0.017) than the IDHwt tumors. Additionally, the IDHmut tumors had a lower cerebral blood flow (CBF)_mean and a higher ADC_mean than the IDHwt tumors (p = 0.048 and 0.007, respectively) (Table 2). Among the IDHmut gliomas, the IDHmut1p/19qdel group had a higher cerebral blood volume (CBV)_mean (p = 0.024) and CBV_85 (p = 0.039). In addition, the IDHmut1p/19qdel group demonstrated a lower ADC_mean, ADC_5, ADC_10, and ADC_15 than the IDHmut1p/19qnondel group (all p values <0.050) (Table 3).

Table 2 Quantitative MRI Features according to IDH Mutation

Table 3 Quantitative MRI Features according to 1p/19q Codeletion in IDH-Mutant Gliomas

The univariable logistic regression analysis revealed significant associations between IDH mutation and T2 volume and between IDH mutation and ADC_mean value. Moreover, among the IDHmut tumors, 1p/19q codeletion demonstrated associations with CBV_mean values and ADC_mean, ADC_5, ADC_10, and ADC_15 values (Table 4). The multivariable regression analysis using the results from the univariable regression analysis revealed associations between IDH mutation and ADC_mean as well as 1p/19q codeletion and CBV_mean (Table 4). From the multivariable analysis, the prediction models for the IDH mutation and 1p/19q codeletion using the quantitative imaging parameters were acquired. The diagnostic performance of the prediction model showed accuracy, sensitivity and specificity values of 66.7%, 66.7%, and 72.7%, respectively, for the IDH mutation and of 66.7%, 100.0%, and 50.0%, respectively, for 1p19q codeletion using the validation dataset.

Table 4 Univariable and Multivariable Regression Analyses of Imaging Parameters and Genomic Profiles

From the comparison of quantitative MRI features between 3 groups of genetic profiles of G3 gliomas, IDHmut1p/19qnondel group revealed the largest tumor volume, lowest CBV_mean, CBV_85, and CBF_mean, as well as the highest ADC_mean, ADC_5, ADC_10, and ADC_15 values (Supplementary Table 4). The IDHmut1p/19qnondel group demonstrated the highest tumor signal intensity ratio between T2WI and FLAIR images, representing the presence of T2-FLAIR mismatch sign from the quantitative analysis (1.46 ± 0.25, 1.53 ± 0.29, and 1.29 ± 0.17 in IDHmut1p/19qdel, IDHmut1p/19qnondel, and IDHwt group G3 gliomas, respectively; p <0.001).

The representative cases of G3 gliomas according to their genomic profiles are depicted in Figure 2.

Fig. 2 Representative cases of G3 gliomas according to their genetic profiles. A representative case of IDHmut1p/19qdel G3 glioma demonstrated higher CBV and lower ADC values than IDHmut1p/19qnondel tumors. Note that the IDHmut1p/19qnondel group showed a T2-FLAIR mismatch sign, which was defined as a high signal intensity of the tumor in T2WI and relatively low signal intensity on FLAIR, except for the peripheral rim. IDHwt G3 gliomas showed lower ADC values than IDHmut G3 gliomas. ADC = apparent diffusion coefficient, CBV = cerebral blood volume, FLAIR = fluid attenuated inversion recovery, IDH = isocitrate dehydrogenase, IDHmut1p/19qdel = IDH mutation and a chromosome arm 1p/19q-codeleted, IDHmut1p/19qnondel = IDH mutation, 1p/19q-nondeleted, IDHwt = IDH wildtype, T2WI = T2-weighted image

Survival Analysis

The median PFS of the patients included in the study was 38.93 months (95% confidence interval [95CI]: 30.42–43.88 months). The IDHmut1p/19qdel group had the longest median PFS (45.50 months, 95CI: 40.20–50.80). The IDHmut1p/19qnondel group showed a longer median PFS (38.93 months, 95CI: 30.37–49.63) than the IDHwt group (11.90 months, 95CI: 7.70–36.83). The survival curves between the three groups showed statistical significance (p <0.001) (Fig. 3).

Fig. 3 Kaplan-Meier survival graphs according to genetic profiles. Note that patients with IDHmut1p/19qdel G3 gliomas demonstrate the longest median PFS, followed by patients with IDHmut1p/19qnondel G3 gliomas. Lastly, the IDHwt G3 gliomas had the shortest median PFS. The survival curves between the three groups showed a significant difference (p <0.001). PFS = progression-free survival

We found statistically significant associations between PFS and the genetic profile of the tumors, such as O6-alkylguanine DNA alkyltransferase (MGMT) promoter methylation status (p = 0.020), CBV_mean (p = 0.010), and CBF_mean (p = 0.029) (Table 5).

Table 5 Multivariable Cox Hazards Regression Analysis of MRI Features with Progression-Free Survival

Comparison between IDHwt GBM and Grade III Gliomas

From the Kaplan-Meier survival analysis of 155 IDHwt GBMs and 29 IDHwt G3 gliomas, the median PFS did not show any significant difference (median PFS: 9.83 vs. 11.90 months, respectively; p = 0.107). From the comparison of quantitative imaging features of IDHwt GBMs and IDHwt G3 gliomas, the CBV_mean (1.74 ± 0.86 vs. 1.75 ± 0.65, p = 0.948) and ADC_mean values (1.48 ± 0.50 vs. 1.36 ± 0.49, p = 0.256) did not reveal any significant differences. However, IDHwt GBMs showed higher tumor volume on T2WI, higher values of CBV_95, CBV_90, and CBV_85, as well as lower ADC_5, ADC_10, and ADC_15 values than IDHwt G3 gliomas (all p values <0.050).

Interobserver Agreement

Interobserver agreement between the three radiologists revealed an ICC of 0.97 for the CBV_mean and 0.98 for the ADC_mean values, indicating almost perfect agreement between the readers. From Pearson's chi-square test, there was no significant difference in the evaluation of the presence of a T2-FLAIR mismatch sign between the three readers from the interobserver analysis.

DISCUSSION

From this study, we found a significant correlation between qualitative and quantitative MRI features and major genetic and histologic features in G3 gliomas. We found that IDHmut G3 gliomas tend to have more necrosis, a larger tumor volume, lower CBF, and higher ADC values than IDHwt tumors. Among the IDHmut G3 gliomas, the IDHmut1p/19qdel group tended to have more enhancement, higher CBV, and lower ADC values than the IDHmut1p/19qnondel group. The PFS was the longest in the IDHmut1p/19qdel group, followed by the IDHmut1p/19qnondel and IDHwt groups. Additionally, our study discovered that the CBV and CBF values have a significant association with the prognosis of G3 gliomas, regardless of the genetic profiles of the tumors. We found significant associations between the quantitative and qualitative imaging features of G3 gliomas according to the presence of an IDH mutation and 1p/19q codeletion, which were included in the most recent WHO guidelines for glioma classification. With the findings of this study, we were able to create a prediction model. This model can be used in molecular and prognosis prediction of G3 gliomas, providing a better guidance for the patient management of G3 glioma patients.

In LGGs, higher tumor volumes and ADC values for IDHmut tumors than for IDHwt tumors have been reported previously (8, 9, 14, 15). These observations potentially represent the slowly growing nature and lower cellularity of IDHmut G3 gliomas compared to those of IDHwt gliomas. Furthermore, regression analysis revealed significant associations between IDH mutation and ADC values, suggesting the potential role of this particular imaging feature in prediction of IDH mutation status in G3 gliomas. However, the G3 gliomas included in our study did not show a significant difference in the proportions of tumors with contrast-enhancing portions according to the presence of the IDH mutation, inconsistent with findings by Leu et al. (8). This discrepancy might be due to the differences in the number of patients included in the study, since a previous report included only 34 cases of G3 gliomas compared to the 76 cases included in this study. Instead, we found that IDHmut G3 gliomas included in this study showed a high proportion of tumors with necrosis, which is usually present in aggressive tumors. We believe that this observation is due to the increased heterogeneity of IDHmut gliomas as compared to IDHwt (16), which may also result in individual variations from slow growing gliomas with low cellularity to focal aggressive foci within tumors. In addition, previous research has found significantly higher proportions of necrosis in GBMs due to intravascular thrombosis caused by hypoxia-mediated coagulation system activation, resulting in endothelial cell proliferation and necrosis of IDHmut tumors (17, 18). Our result are in line with these studies done on GBM, suggesting the similar pathophysiological changes in G3 gliomas.

Among the IDHmut G3 gliomas, a higher proportion of tumors with contrast-enhancing portions, higher CBV and lower ADC values were observed in the IDHmut1p/19qdel group, consistent with a previous report (8) which enhanced the role of these imaging markers in prediction of the genetic profiles of G3 gliomas. Additionally, the regression analysis showed a significant association with the CBV value and the presence of a 1p/19q codeletion in IDHmut G3 gliomas. Since oligodendrogliomas are comprised of only IDHmut1p/19qdel gliomas according to the WHO guidelines (1), our results depicted similar findings to those described in previous studies that reported higher CBV values in oligodendrogliomas compared with astrocytomas (19, 20). Also, we found that T2-FLAIR mismatch sign was able to distinguish IDHmut1p/19qnondel G3 gliomas among IDHmut tumors, similar to the result of previous study that included LGGs (11).

From the three group comparison of the MRI features of the G3 gliomas according to their genomic profiles, we have found that IDHmut1p/19qnondel group has a tendency to have higher tumor volume, lower CBV, and higher ADC values compared with IDHmut1p/19qdel and IDHwt groups, similar to that of a previous report (8). However, these findings can only be useful in distinguishing IDHmut1p/19qdel group from IDHmut1p/19qnondel group and IDHmut1p/19qnondel group from IDHwt G3 gliomas. Furthermore, they did not correlate with the prognosis of G3 gliomas. We believe that our stratified approach (prediction of IDH mutation status first, and then to predict 1p/19q codeletion of G3 gliomas) is a more clinically applicable and plausible way to predict the genetic profiles and prognosis of G3 gliomas using MRI features.

In addition, we found significant differences in the prognosis between the three genetic groups of G3 gliomas. Additionally, our data revealed significant associations between prognosis and not only MGMT promoter methylation status, but also imaging parameters, such as CBV and CBF, which are consistent with previous reports (21, 22). This enhances support for the function of imaging parameters as promising biomarkers for predicting the prognosis of G3 glioma patients. Interestingly, the direction of CBV and CBF in relation to survival was opposite from the results of our study, which was similar to one previous study that have analyzed associations between survival and perfusion/diffusion parameters. Also, this previous report did not find a significant association between ADC and survival in LGGs, which were consistent with the results of this study (22). Unexpectedly, IDH mutation and 1p/19q codeletion status were not significantly associated with survival according to multivariable analysis. We believe that further studies with a large data set is needed to demonstrate the role of multiple variables such as genetic profiles and imaging features in predicting survival in G3 glioma patients.

Furthermore, our data depicted that there was no significant difference in the prognosis of IDHwt G3 gliomas and GBMs, which is in concordance with previous reports (7, 23, 24). Additionally, the imaging parameters of IDHwt G3 gliomas and GBMs were similar to one another, with a wider distribution of values in GBMs due to the heterogeneity of these tumors. Recently, a new classification of LGG, known as the Consortium to Inform Molecular and Practical Approaches to central nervous system Tumor Taxonomy (cIMPACT-NOW), has been introduced and states that only a subset of LGG with specific molecular subtype can be termed as IDHwt, with molecular features of GBM, that follow the prognosis of IDHwt GBMs (25). Therefore, we believe that there is a need for a future study that includes the specific subset of IDHwt LGGs according to the new classification stated by cIMPACT-NOW, which investigates their imaging features compared with IDHwt GBMs.

There are some limitations in this study beyond the nature of its retrospective design. First, the patients in this study underwent different durations of radiation therapy, which might have affected the PFS of the G3 glioma patients included in the study. Second, we analyzed MR images from different MR scanners from various manufacturers with different field strengths. However, we normalized the CBV, CBF, and ADC values to the normal white matter area of the corona radiata on the contralateral side of the tumor to diminish the potential bias of the CBV and ADC measurements. Third, some genetic information was missing for a small number of patients. However, the number of patients with missing genetic information was small. Fourth, we included internal cystic and necrotic lesions of the tumor for analysis to identify features and to develop a model that will represent the entire tumor characteristics rather than parts of the tumor. In effect, our results might contradict those of previous studies. However, we believe that since the majority of the tumors included in the study contained relatively large portions of cystic or necrotic areas, segmenting all parts of the tumor was more feasible for the analysis and future application of the prediction model. Fifth, we only included G3 gliomas for analysis in this study, which is slightly different from studies that included LGGs (G2 and G3 gliomas). So far, our study is the only study with enough number of G3 glioma patients included that evaluated quantitative characteristics of G3 gliomas, and we believe this gives a unique value to this study.

In conclusion, we found significant differences in the MRI features in G3 gliomas according to different molecular subtypes, which can be utilized for the prediction of the genomic profiles and prognosis of G3 glioma patients. Additionally, the MRI signatures and prognosis of IDHwt G3 gliomas tend to follow those of IDHwt GBMs. We believe that our application can be used in precision medicine for the diagnosis and treatment planning for G3 glioma patients.

Supplementary Materials

The Data Supplement is available with this article at https://doi.org/10.3348/kjr.2020.0011.

Supplementary Table 1

MR Scan Parameters

Supplementary Material 1, 2, 3

Supplementary Table 2

Imaging Parameter Descriptions

Supplementary Table 3

Qualitative MRI Features according to Histologic Classification and Genetic Profiles

Supplementary Table 4

Quantitative MRI Features according to Genetic Profiles

Notes

This study was supported by a grant from the Korea Healthcare technology R&D Projects, Ministry for Health, Welfare & Family Affairs (HI16C1111), by the Brain Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT & Future Planning (NRF-2016M3C7A1914002), by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT & Future Planning (NRF-2020R1A2C2008949 and NRF-2020R1A4A1018714), by Creative-Pioneering Researchers Program through Seoul National University (SNU), and by the Institute for Basic Science (IBS-R006-A1).

Conflicts of Interest: The authors have no potential conflicts of interest to disclose.