Under the 2021 WHO classification, Grade 4 gliomas represent a diverse cohort, displaying a spectrum of clinical, radiological, and molecular characteristics. This updated classification has sparked considerable discourse within the research community [12, 13]. In this study, we aimed to delve into the clinicoradiological, molecular, and therapeutic dimensions of WHO Grade 4 gliomas.

In our cohort, mGBM exhibited clinical and neuroradiological profiles largely resembling those of IDH-mutant astrocytomas. Predominantly, these tumors manifested in the frontal and temporal lobes and were often associated with seizures as the initial clinical presentation. The frequency of seizures of patient with mGBM at presentation (55.6%) was slightly lower compared to those with IDH-mutant Grade 2 astrocytomas (70–80%) [29]. Besides, about 38.9% of mGBM patients exhibited mild or patchy contrast enhancement on MRI, consistent with findings reported elsewhere [30]. Moreover, none of non-CE tumor displayed evidence of necrosis (0/14) in mGBM cohort, which corroborated the findings from prior investigations [31]. In the mGBM subgroup, patients exhibited similar ages to those with hGBM, yet tended to present with smaller tumor volumes on MRI. Additionally, a significant proportion of mGBM tumors (over three-quarters) demonstrated cortical involvement, aligning with previous research [32].

Conversely, mGBM displayed molecular features and prognostic outcomes akin to hGBM. In this study, TERT mutations were prevalent in mGBM, consistent with prior literature [10]. Alongside the molecular hallmarks of glioblastomas, the MGMT methylation level of mGBM mirrored that of hGBM. Even upon multivariate analysis, MGMT methylation persisted as a prognostic indicator, corroborating numerous earlier reports [24,25,26]. The relatively low methylation level observed in mGBM may contribute to its unfavorable prognosis. However, in contrast to earlier studies [8], the prognostic relevance of the TERT phenotype diminished in multivariate analysis. Notably, our study revealed no significant disparities in survival outcome between mGBM and hGBM. The median survival of mGBM was 21.2 months, which aligned with the reported range of 20.8–31.9 months in previous research [8,9,10,11]. Moreover, in our series mGBM patients with isolated TERT promoter mutations have a significantly worse OS (19 months) as compared to the results reported by Berzero et al. [12] (88 months). It reinforces the argument for reclassifying the subtype of IDH wild-type astrocytoma as WHO Grade 4. However, our study did not identify a prognostic advantage for histological Grade 2 mGBM compared to Grade 3, contrary to earlier investigations [10, 11, 13].

Our investigation unveiled a notably favorable prognosis for A4 compared to hGBM, marked by a median survival that remained undefined and a median follow-up duration extending to 46.9 months, surpassing previous literature findings [4,5,6, 14,15,16,17, 33]. The high rate of resection and comprehensive radiochemotherapy coverage observed in our study may contribute to this enhanced prognosis. Additionally, distinct features of A4, including younger age, elevated MGMT methylation levels, reduced incidence of TERT mutations, and limited enhancement, likely contribute to its favorable prognosis. Moreover, even upon rigorous multivariate analysis, the prognostic advantage of A4 persisted significantly, which underscored the prognostic implication of IDH mutation.

Furthermore, it appears that many characteristics of low-grade gliomas can also be identified in A4. In terms of radiological characteristics, our investigation underscores that the predominant majority of A4 cases (82%) present CnCE feature, which has been labelled as “low-grade appearance”, consistent with earlier research [31, 34,35,36]. Hence, we included the CnCE in our integrated model to predict A4 phenotype. Similarly, Lasocki et al. [35] found a close relationship between mass-like non-CE tumor morphology in GBM and the IDH mutation phenotype. Additionally, our findings reveal partial enhancement in A4, primarily due to the prevalence of enhancing characteristics in A4 tumors alongside a relatively large proportion of non-CE tumor [34, 37], consistent with prior research findings [31]. From a hispathological perspective, Qiu et al. [33] found that over 50% of A4 histopathologies exhibit components of low-grade gliomas, a proportion significantly higher than that observed in IDH wild-type glioblastoma. In terms of molecular characteristics, previous studies have identified genetic similarities between de novo primary A4 and evolved secondary A4 (initially diagnosed as Grade 2 or Grade 3 astrocytomas [38, 39], suggesting a potential secondary transformation of A4 from low-grade gliomas [7]. All these low-grade gliomas characteristics may contribute to the relatively favorable clinical outcome of A4.

Moreover, our study elucidates that patient with CnCE shows a more favorable prognosis. This protective prognostic attribute even persists after multivariable analysis. In line with previous investigations, primary A4 patients presenting with mixed low-grade gliomas components also demonstrate enhanced clinical outcomes [33]. Both preoperative imaging and histopathological assessments suggest a more favorable prognosis for A4 tumors exhibiting characteristics consistent with low-grade gliomas. This observation may imply an early morphological transition of A4 from a low-grade glioma to a secondary glioblastoma. Although prior studies did not confirm the prognostic advantage of mixed pathology in secondary A4 [33], it is important to acknowledge that these studies primarily assessed survival periods from the confirmation of Grade 4 glioma through a second surgery to the occurrence of events, without considering the disease course of initially diagnosed Grade 2 or Grade 3 astrocytoma.

This study is subject to several limitations. Firstly, its retrospective design introduces inherent biases related to patient selection. Secondly, the relatively low incidence of primary A4 and mGBM limits the depth of consecutive patient enrollment in this single-center study. Thirdly, the predictive models developed for A4 do not encompass Grade 3 gliomas and low-grade gliomas, potentially limiting their clinical applicability to some extent. Lastly, ongoing follow-up until the last patient reaches the endpoint is necessary to validate the results and conclusions drawn from this research.

In conclusion, our study sheds light on the complex landscape of Grade 4 gliomas, particularly A4 and mGBM, under the updated WHO classification. Through comprehensive analysis encompassing clinical, radiological, molecular, and therapeutic aspects, we have uncovered distinct characteristics and prognostic implications associated with these glioma subtypes. Our findings underscore the importance of integrating multimodal approaches, including neuroimaging features and molecular profiling, to accurately diagnose and prognosticate Grade 4 gliomas. Moreover, the development of predictive models for A4 holds promise for enhancing preoperative assessment and guiding personalized treatment strategies.