In the United States, diffuse low-grade or anaplastic gliomas represent 22.6% of all diagnosed gliomas (1, 2). In France, the standardized incidence rate adjusted on the United States and world population for all grades II–III glioma is 16.0/1 million individuals (3). Since the 2016 WHO classification, molecular features have been included to better classify diffuse gliomas (4). The entity of oligoastrocytoma was deleted compared with the 2007 WHO classification. New entities are diffuse astrocytoma isocitrate dehydrogenase mutated (IDHm) or wild type, oligodendroglioma IDHm, and 1p19q codeleted. Anaplastic gliomas are classified as grade III and non-anaplastic as grade II (4). However, since 2022, the new WHO classification has added more genetic mutations that classify diffuse astrocytoma as grade IV gliomas (homozygous CDKN2A/B deletion, TERT promoter mutation, EGFR gene amplification, and chromosome 7 gain/chromosome 10 loss). These mutations were not systematically used for glioma classification in the WHO 2016 classification (2). Low-grade glioma is characterized by a natural progression toward more aggressive disease (grades III and IV) often characterized by the occurrence of contrast enhancement on MRI (5). In patients with grades II–III glioma, overall survival (OS) is heterogenous reaching almost 20 years in some cases (6–8).

Grades II–III glioma management is not standardized (9). Adjuvant treatment with chemotherapy (temozolomide) or combined chemotherapy (e.g., Procarbazine-CCNU-Vincristine [PCV]) and/or radiotherapy can be proposed upfront to selected patients (10). Similarly, the management of recurrent grades II–III glioma is not standardized. Chemotherapy can be proposed at disease relapse, but only limited data support this strategy (10). Moreover, new treatment modalities are needed due to chemotherapy-induced resistance (11).

The use of anti-angiogenic agents in glioma is based on neovascularization driven mainly by vascular endothelial growth factor (VEGF) signaling and its endothelial receptor VEGFR2 (12). Bevacizumab, a monoclonal antibody against VEGF, has demonstrated clinical benefit in glioblastoma in phase II trials, but not in phase III trials (13–17). Bevacizumab associated with temozolomide showed poor efficacy compared with temozolomide alone in patients with first recurrence of grades II–III glioma without 1p19q codeletion (18). Retrospective studies suggest that bevacizumab-based combination treatments or bevacizumab alone may be a strategy for alkylating agent-refractory grade III glioma, defined as progressive disease after temozolomide (19–22). Few prospective studies (phase II) showed a potential interest of bevacizumab as rescue treatment for recurrent grade III glioma (23, 24). Altogether, these results suggest that bevacizumab may be proposed as rescue treatment in recurrent grades II–III glioma, specifically to heavily pretreated patients (22). However, data on bevacizumab use in recurrent grade II glioma are scarce. Moreover, no established biomarker of the response to bevacizumab is available. Clinical features (performance status and hypertension), tumor profile (proneural subtype, VEGF, VEGFR, and carbonic anhydrase IX expression), tumor mutational status (TP53, IDH1, MGMT, and 1p19q codeletion), and circulating biomarkers (matrix metalloproteinases 2 and 9) have been reported as possible bevacizumab efficacy biomarkers in malignant glioma (25–33).

However, these studies presented limitations and results were conflicting. In routine practice, some patients with recurrent grades II–III glioma who received bevacizumab seem to display long-term benefits, even after its discontinuation. Indeed, a retrospective study of 435 patients with recurrent grades II and III glioma identified 22 patients (eight grade II and 14 grade III) who were alive 3 years after initiation of bevacizumab. However, in this study, recurrence was not clearly defined. Prior treatment history or radiologic criteria were not considered before bevacizumab treatment (e.g., bevacizumab was administered at first recurrence in 10 patients) (34).

Here, we were interested in bevacizumab efficacy in heavily pretreated grades II–III recurrent glioma patients. T1 contrast enhancement was used as a marker for disease malignant transformation. The aim of this retrospective study was, therefore, to describe clinical outcomes after starting treatment with bevacizumab and to determine the patient and tumor characteristics associated with the efficacy of bevacizumab across the whole population. The secondary endpoint was to describe patients who had discontinued bevacizumab without radiographic progression and to examine associated patient characteristics.

This retrospective study evaluated bevacizumab efficacy in heavily pretreated patients (64.2% had received two or more systemic chemotherapy lines, 32.1% had undergone a second surgical intervention, and 23.5% had undergone two or more radiotherapy treatments). All patients had baseline T1 gadolinium enhancement on MRI, highlighting tumor neovascularization or necrosis and the possible tumor upgrading (41). Indeed, progression to higher histological grade was observed in 61.5% of patients after surgical reintervention.

The median PFS and OS since bevacizumab initiation were 4.9 (95% CI 3.7–6.1) and 7.6 months (95% CI 5.5–9.9), respectively, in line with previous studies on oligodendroglioma, high-grade gliomas and also on advanced glioblastomas (14, 23, 24). This suggests that the duration of the response to bevacizumab is about the same magnitude whatever the WHO tumor grade and whatever its position in the patients’ management strategy. Bevacizumab toxicity profile was as previously reported (13–15).

In our study, IDHm, 1p19q codeletion (or ATRX protein expression), percentage of Ki67-positive cells and histological grade were associated with survival of patients with recurrent grades II–III glioma from diagnosis, but not from bevacizumab initiation. Oligodendroglioma IDHm/1p19q codeletion, grade II gliomas, and low Ki67 are known good prognostic factors (2, 36, 42). However, in our sample, they were not predictive of bevacizumab efficacy. This may be related to disease progression toward glioma transformation, as indicated by T1 contrast enhancement and upgrading in reoperated patients. These patients’ tumors may have the same evolution as grade IV gliomas; however, 1p19q codeletion had no impact on OS after bevacizumab. Moreover, our study failed to demonstrate a relationship between the efficacy of bevacizumab and IDHm. This could be related to missing data and lack of statistical power. Only intracranial hypertension at diagnosis was associated with PFS since bevacizumab initiation. This could be explained by bevacizumab antiangiogenic activity related to neovascularization that induces cerebral edema and intracranial hypertension. Our multivariate analysis did not find any association between patient characteristics and OS. This suggests that patient response to bevacizumab and survival since starting bevacizumab are not correlated with the initial histological tumor grade or mutational status.

In addition, a subgroup of patients discontinued bevacizumab without radiological progression. Oligodendroglioma and age ≥ 38 years at diagnosis were associated with this subgroup and might be a biomarker for discontinuing bevacizumab without radiological progression. It is known that survival is longer in patients with oligodendrogliomas than with astrocytomas (2). Conversely, age ≥ 38 years at diagnosis was an unexpected association as older age is usually negatively associated with survival (2, 13). We may hypothesize that these patients, who had been on heavy treatment before starting bevacizumab, were more frail and were more prone to discontinuing bevacizumab due to clinical degradation, bevacizumab toxicity, or presence of comorbidities that were incompatible with continuing bevacizumab. Indeed, the main reasons for discontinuing bevacizumab in this subgroup were patient/physician decision and severe toxicity. Among these 24 patients, 10 were alive at 12 months. Most of these long-term responders did not receive any other treatment after bevacizumab. Moreover, two patients were still alive 8 years after initiating bevacizumab as a rescue treatment. A previous study also reported long-term survivors among patients on bevacizumab for recurrent grades II–III gliomas (34). However, to the best of our knowledge, this is the first report of maintained bevacizumab response after treatment discontinuation without radiological progression in patients with recurrent grades II–III glioma. This long-term effect might be explained by bevacizumab’s capacity to activate the anti-tumor immune response by promoting the infiltration of anti-tumor lymphocytes and decreasing the pro-tumor lymphocyte infiltrate (43–46). In agreement, in patients with glioblastoma on bevacizumab, an increase in inflammatory markers in peripheral blood has been associated with better survival (47). Different mechanisms may be involved, such as promotion of diapedesis, maturation of dendritic cells, a decrease in myeloid suppressor cell infiltration, or the regulation of glutamate transport (45, 46). Alternatively, bevacizumab’s inhibitory effects on the VEGFR pathway (angiogenesis) and on glioma stem cells might be implicated. Pathways leading to bevacizumab resistance (e.g., cMET upregulation, induction of other neo-angiogenic pathways) and their regulation also should be investigated (48, 49).

Our study has certain limitations. First, the retrospective study design limits the extrapolation of our results. Second, some data were missing (particularly IHC data), and tumor methylation profiles were not investigated. Moreover, IDH gene sequencing was only available for a few patients’ and was not reported. Third, the 2016 WHO classification has been replaced by a new WHO classification (50). However, the distinction between grades II and III glioma has not been fundamentally changed. The only exceptions are also the presence of CDKN2A/B homozygous deletion for IDHm astrocytoma, and of TERT promoter mutation, EGFR gene amplification, and chromosome 7 gain/chromosome 10 loss for IDH wild-type diffuse astrocytomas, which classify them as grade IV, which were missing in our study. However, the IHC analysis at inclusion gave similar IDHm and 1p19q codeletion rates as previously reported (51). Fourth, our patient cohort, with its distinct molecular pattern and bevacizumab administration, was not homogeneous. Half patients had undergone chemotherapy combined with bevacizumab and bevacizumab was administered at 10 or 15 mg/kg every 2 or 3 weeks. However, that both administration schemas are not different in terms of treatment efficacy (52). Finally, the RANO high-grade glioma criteria were used, based on the assumption that the study sample included mainly patients with transformed diffuse gliomas. In the vast majority of the cases, the increasing of enhancement was followed by spread of the infiltration of the tumor. In the other cases, FLAIR hyper-signal was considered as stable. However, radiological progression was assessed by MRI every 3 months. Considering PFS on radiological evaluation was a limitation. Indeed, among patients who discontinued bevacizumab without radiological progression, application of the RANO criteria would probably have enabled them to be classified as progressive solely on the criterion of clinical degradation (39).

Bevacizumab alone or in combination can be proposed as a rescue treatment for heavily treated patients with recurrent grades II–III glioma and T1 contrast enhancement on brain MRI. PFS and OS were the same as reported for glioblastomas. In our study, the response to bevacizumab was not associated with the histological grade or tumor biology at diagnosis. Few patients who responded to bevacizumab experienced prolonged survival after discontinuing bevacizumab, without any subsequent treatment. We now need to identify the clinical, biological, and/or radiological factors that predict the long-term response to bevacizumab in these patients.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by SA, MJ, AD, and MF. The first draft of the manuscript was written by SA. and all authors commented on previous versions of the manuscript. All authors contributed to the article and approved the submitted version.