IDH mutant and 1p/19q co-deleted oligodendrogliomas: tumor grade stratification using diffusion-, susceptibility-, and perfusion-weighted MRI

Oligodendrogliomas (ODs) are glial neoplasms originated from oligodendrocytes that primarily affect supratentorial parenchyma [1]. These entities were previously classified into grades II and III according to the 2007 World Health Organization (WHO) classification system of the central nervous system (CNS). The traditional and indispensable approach relies on the histopathologic analysis of proliferative and mitotic activity, nuclear atypia, and cellularity as significant predictors of disease progression [2, 3].

Historically, co-deletion of whole chromosome arms 1p and 19q, namely, 1p/19q co-deletion, has proved to be a diagnostic and prognostic biomarker of ODs [4‐6]. However, progression-free survival (PFS) and overall survival (OS) of ODs between grades are dramatically different after adjustment for genotypes (defined by 1p/19q co-deletion) [7]. Isocitrate dehydrogenase (IDH) mutation is frequently found in WHO grade II–III oligodendrogliomas and astrocytomas [8, 9], and is currently implicated as a prerequisite of tumorigenesis for some types of diffuse gliomas and a precondition of 1p/19q co-deletion [10, 11]. According to the 2016 WHO classification system, the “integrated diagnosis” of ODs requires histological classification, WHO grade, and molecular information (both IDH mutation and 1p/19q co-deletion) [12, 13]. An oligodendroglioma-like tumor lacking diagnostic mutations is given a “not otherwise specified (NOS)” designation, which is strongly discouraged by neuro-oncologists. When dealing with NOS ODs, care should be taken throughout to exclude CNS tumors including astrocytomas, glioblastomas, clear cell ependymoma, and dysembryoplastic neuroepithelial tumor [12]. In addition, nearly all histological “oligoastrocytomas” can be reclassified as either astrocytomas or ODs using genetic testing [12, 14].

Diffusion-weighted imaging (DWI) could reflect the Brownian movement of water molecules and the cytogenetic profile of cerebral tumors through the measurement of apparent diffusion coefficient (ADC) values [15, 16]. As another advanced magnetic resonance imaging (aMRI) approach, susceptibility-weighted imaging (SWI) utilizes intratumoral susceptibility signals (ITSSs) to demonstrate tumor vascularity and blood metabolite sensitively [17, 18]. Regional cerebral blood volume (rCBV) and regional cerebral blood flow (rCBF) derived from dynamic susceptibility contrast perfusion-weighted imaging (DSC-PWI) provide hemodynamic information of microvasculature in glioma grading and genotyping [19‐24]. Non-invasive measurement of rCBV has been confirmed to correlate with tumor vascularity and clinical prognosis of gliomas, which has the potential to provide additional information to conventional MRI (cMRI) [7, 19, 21]. As recently revealed, IDH mutations or 1p/19q co-deletions correlate with aberrant cellularity and angiogenesis by regulating proliferation and vascularization [6, 19, 21, 25]. Therefore, we suspected that the inclusion of mixed genetic profiles in previous studies might lead to a low prediction accuracy of aMRI techniques for grading ODs [7, 26].

During the past two decades, extensive studies have focused on the surrogate MRI characters of tumor grade (defined by histopathology) and genetic subtype (defined by 1p/19q co-deletion or IDH mutation) of CNS tumors [19, 23, 24, 27, 28], whereas these two genic alterations act as diagnostic markers rather than grouping factors for ODs [12]. Thus, it is inappropriate for the neuropathologists or neuroradiologists to “genotype” ODs following the WHO update. To our knowledge, there is no aMRI research in presurgical grading of the redefined tumors. Therefore, the present study aims to investigate whether multiparametric MRI is predictive of tumor grades of the new entity.

Thirty-three cases (17 males and 16 females, age 42.97 ± 9.03 years) were diagnosed as IDH-mutant and 1p/19q co-deleted ODs based on the current WHO criteria. DWI, SWI, and DSC-PWI combined with cMRI were performed on all patients. Five cases of DSC-PWI were of poor perfusion quality and excluded in further analysis.

Table 1 shows the summarization of the main clinical and cMRI features of molecularly defined (IDH-mutant and 1p/19q co-deleted) ODs between groups. All kappa coefficients of interobserver measurements were greater than 0.75 (an excellent agreement). Out of 33 patients, 11/18 (61.1%) LGOs and 13/15 (86.7%) HGOs situated in the frontal lobe with insignificant difference between grades. Heterogeneous appearance (including hemorrhage and cystic degeneration) and infiltrating borders were non-specifically observed in ODs regardless of grades. It is noteworthy that LGOs revealed a decreased level of edema compared to anaplastic counterparts (P = 0.005). Additionally, HGOs were related to the nodular or ring-like enhancement patterns on CE-T1WI (P = 0.002).

Tables 2 and 3 give the (semi-)quantitative measurements of aMRI for LGOs and HGOs. Relevant MRI images are depicted in Figs. 1 and 2. nADC values from DWI were significantly higher in LGOs than in HGOs (P < 0.001). ITSS levels of HGOs were substantially greater than that of LGOs with seven (46.7%) HGOs which exerted an agglomerated or nodular appearance of ITSSs (P = 0.006). Both nCBV and nCBF values in HGOs were significantly higher than the values of LGOs (P < 0.001 and P = 0.005).

The results of ROC analysis of the parameters are illustrated in Table 4. Optimal cutoff values of quantitative nADC from DWI and quantitative nCBV from DSC-PWI in tumor grading were determined as 1.11 and 2.90, respectively. The sensitivity/specificity of nADC and nCBV were 100.00/77.78 and 92.31/80.00%, respectively. With a threshold value of ≥1 for semi-quantitative ITSSs from SWI, the sensitivity and specificity in the diagnosis of HGOs were 60.00 and 83.33%, respectively.

Logistic regression and ROC analyses were conducted for cMRI (combining edema and enhancement) and aMRI (combining DWI, SWI, and DSC-PWI by two or three parameters). The combination of SWI and DSC-PWI resulted in the highest sensitivity and specificity of 100.00 and 93.33%, respectively. Nevertheless, combined DWI, SWI, and DSC-PWI techniques yielded similar diagnostic performance.

Significant concordance between the histopathological grading and aMRI-based grading was found in our reclassified cases. This concordance may be of more clinical value as we followed the updated WHO classification standards which have already redefined many infiltrating gliomas. Our study has revealed non-invasive MRI techniques including DWI, SWI, and DSC-PWI which may improve the diagnostic accuracy in grading redefined ODs.

Although molecular and genetic factors have been introduced recently, a large body of evidence has suggested that glioma grading based on histopathological evaluation is still important for determining therapeutic options and predicting clinical outcomes [2, 5, 12, 31, 32]. Shaw et al. suggest that OS (approximately 4 years of HGOs and 10 years of LGOs) and treatment planning are different between grades [31‐33]. Patients with LGOs can benefit from subtotal resection and early radiotherapy with small fraction doses, or remain stable without any treatment [5, 34, 35]. In contrast, several randomized controlled trials of HGOs show that adjuvant chemotherapy increases PFS while postoperative radiotherapy produces minimum benefit [33, 36]. Besides, preoperative MRI-based grading and guidance may affect the surgical strategy to maximize safe resection [31, 37].

Frontal location, characteristic calcification, poorly defined border, and internal heterogeneity are highly suggestive of molecularly defined ODs as unveiled in the latest studies [27, 38‐41], which may serve as reliable and robust biomarkers in the differential diagnosis prior to tumor grading. The results of this study display that molecularly defined ODs have a predilection for the frontal lobes that is consistent with previous findings [27]. Peritumoral edemas in HGO patients are frequently encountered in our study, mainly attributed to the high frequency of cellular/vasogenic pathogenesis in high-grade tumors [42]. Up to date, Johnson et al. have suggested that the tumor borders of molecularly defined ODs are fairly infiltrative [41, 43]. While peritumoral edema was determined based on its non-enhancement on CE-T1WI and higher signal on FLAIR in contrast to clearly enhanced tumor solid component, radiologically, definite differentiation of tumor infiltration from peritumoral edema is still a technical challenge in the field. Although nodular or ring-like enhancement in association with blood-brain barrier breakdown and angiogenesis [37, 42] is frequently observed in HGOs, it should be further noted that such enhancement patterns of HGOs are more homogenous than that of glioblastoma (WHO grade IV, most are IDH wild types), whereas it is quite difficult to stratify tumor grades by utilizing cMRI only.

Although nADC values were not significantly different between WHO grades in morphologically defining ODs (based on WHO 2007 criteria) [26], molecularly defined ODs could be stratified by grades applying nADC as shown in our research. Diffusion MRI enables studies in tumorous physical properties microscopically: increased proliferation elevates cellularity and restricts water diffusion [15, 44]. Mutations of the IDH gene family produces oncometabolite 2-hydroxyglutarate, leading to a slower growth of tumor cells compared to the wild types [9]. Conversely, a poor correlation between 1p/19q co-deletion and cell proliferation was confirmed in an early study [27]. Consequently, it is reasonable to speculate that so-called IDH wild-type ODs of various proportions in both grades may play as confounding factors and result in a complexity of heterogeneity in preceding studies.

SWI appears to be a promising imaging method in grading gliomas based on the different frequencies and appearances of ITSSs [18]. The dot-like hypointense signal is regarded as microhemorrhage and the linear signal as an intralesional vessel, both of which are common in malignant CNS tumors [30]. Similarly, the results of our study indicate a significant difference in ITSSs between grades, suggesting that ITSSs can be a valuable biomarker for a precise diagnosis of HGO. However, medium-to-high levels of ITSSs in LGOs were occasionally observed, which may be attributed to the moderately increased neovascularity in both LGOs and HGOs as a consequence of 1p/19q co-deletion [21, 40].

Our study introduced multiple diagnostic parameters derived from DSC-PWI to identify high-grade IDH-mutant and 1p/19q co-deleted ODs. We believe that nCBV and nCBF values can be used as surrogates of angiogenic features of molecularly defined ODs. A significant positive correlation was found between tumor grades and nCBV values. However, most of the previous researchers [7, 24, 25, 28, 45] indicated that nCBV was either inefficient or overlapping in grading ODs when using the gold standard of 2007 WHO criteria. With improved understanding of the pathogenic mechanism, scientists suggest that IDH mutation can reduce the activation of hypoxia-inducible factor-1 alpha and inhibit downstream signaling pathways of angiogenesis [19, 46]. Similarly, some studies indicated that gliomas with 1p/19q co-deletion were likely to have mildly elevated rCBV due to increased microvascular proliferation [21, 25]. Indeterminate proportions of “genotypes” may substantially increase the overlap of perfusion parameters between LGOs and HGOs suggested in previous studies [7, 25].

The combination of ITSSs and nCBV remarkably improved the grading accuracy, and may serve as an optimal method neuroradiologically. However, the combination of all aMRI methods provided no improvement in differentiating anaplastic tumors from LGOs. Although nADC is an ideal parameter for grading molecularly defined ODs, additional DWI procedures may not be necessary when both SWI and DSC-PWI are available.

There are some limitations in our study such as the retrospective nature and small sample. Although the case number of IDH-mutant and 1p/19q co-deleted OD is relatively limited due to its low prevalence and the complexity of multiple cytogenetic characteristics required for diagnosis, further study is needed in multicenter large-scale MRI-based radiomics approaches. Moreover, a possible contrast extravasation in DSC-PWI may generate an imprecise estimation of glioma nCBV. Both preload of gadolinium and leakage correction may be necessary to minimize the leakage contamination. In addition, intratumoral calcific, cystic, necrotic, and hemorrhagic areas may potentially interfere with our manual measurements despite the efforts made to minimize the interference. Finally, a long-term follow-up is desired for analysis of patients’ prognosis.

The study demonstrates that IDH-mutant and 1p/19q co-deleted ODs can be stratified by grades using cMRI and aMRI techniques including DWI, SWI, and DSC-PWI with high specificity and sensitivity. HGOs are prone to prominent edema and enhancement. Furthermore, adding ITSSs to nCBV in clinical practice seems to be a promising option for grading molecularly defined ODs.