Current treatment strategies after relapse yield unsatisfactory outcomes. Re-resection and focal re-irradiation have proven effective [8, 34], which is also reflected in our cohort, where the best response in the second PFS was achieved by re-resection in every case and was followed by radiotherapy in 7 cases, with 3 of them receiving radiotherapy for the first time. However, both resection and radiation carry risks of severe morbidity and should be carefully considered. While re-irradiation could be performed with acceptable short-term toxicity [35], its feasibility could be limited by short intervals between prior radiation courses or proximity to critical structures in the radiation field. Modifying focal re-irradiation to cerebrospinal irradiation (CSI) in focally relapsed patients has demonstrated greater efficacy [36], but is limited by worse long-term toxicity of CSI [37] and was not used in this cohort.

The role of chemotherapy is currently evaluated in the SIOP-EP II trial [18], where it serves as a frontline therapy in different combinations of radiotherapy and residual disease. Early results of the recent ACNS 0831 trial [27] randomizing patients with and without adjuvant chemotherapy have shown possible effects in some ependymoma patients, but further analyses are needed and proceeding. Historically, the effectiveness of frontline chemotherapy in ependymoma was considered uncertain, even with high-dose chemotherapy regimens [10], in many phase II studies indicating prolonged remission in individual cases but no significant global improvement in OS. In relapsed treatment, adjuvant chemotherapy failed to improve OS in the large cohort of ependymoma patients from German HIT-REZ studies [38]. However, chemotherapy remains an option for relapsed patients who can tolerate it and for whom local control cannot be achieved with radiation or surgery [39]. It may also serve to delay radiotherapy or to reduce tumor size before the second-look surgery [40]. In this cohort, we used chemotherapy mainly in frontline therapy except for patient 3 where we followed the ACNS0831 protocol [27] after resection without significant effect.

Molecularly driven targeted therapies have become a promising option for R/R solid tumors. However, this strategy is limited in ependymomas, as these tumors are generally known to harbor very few alterations and lack genomic findings that could serve as predictive markers [15]. Nonetheless, the role of protein kinase signaling has been implied in ependymoma tumorigenesis [41], prompting several ongoing studies evaluating TKIs targeting RET, ALK, ROS, IGFR, PDGFR, FGFR, or EGFR, as well as inhibitors targeting PI3K, mTOR, KIT, and CDK4 [25]. In our cohort, no targetable variants were found on the DNA level. However, phosphoproteomic analyses enabled the detection of increased phosphorylation of EGFR, PDGFRβ, or Ins-R, in three patients, one of whom received a corresponding TKI without a recognizable benefit. Despite prior evidence supporting receptor tyrosine kinase involvement in ependymoma, phase II studies, as well as our own clinical experience have not proven the efficacy for specifically erlotinib and sunitinib [23, 24].

Other potential therapeutic strategy involves targeting the tumor’s interaction with its microenvironment, such as PD-L1 inhibition (discussed later) or anti-angiogenic therapy with anti-VEGF antibody bevacizumab. While early data on bevacizumab suggested promise [40], subsequent studies in ependymoma didn’t confirm its efficacy [22, 42]. This might be explained by findings demonstrating that the ependymoma vasculature has less angiogenic activity and is generally more mature compared to aggressive high-grade gliomas [25]. In this cohort, bevacizumab was used only in combination with other antiangiogenic agents and intrathecal chemotherapy in MEMMAT protocol which is discussed further.

Immunotherapy is increasingly considered for R/R pediatric brain tumor patients, with current reports largely consisting of pilot studies or phase I trials utilizing immune checkpoint inhibitors (ICIs) focused primarily on establishing safety and dose tolerance while monitoring survival endpoints [19, 43]. Similarly, in our cohort, we considered anti-PD-1 blockade if there was a biomarker-based rationale [44] for its administration. We assume a positive impact of nivolumab in a maintenance setting in two ZFTA + patients, in whom their second CR was achieved through resection and radiotherapy. Notably, the subsequent addition of nivolumab led to a prolongation of their second PFS. PD-L1 expression and the possibility of the anti-PD-1 blockade in the ZFTA + subtype have already been described by some studies [19, 20] and might suggest a viable future direction. On the other hand, nivolumab was unsuccessful in PFA patient 3 despite high PD-L1 positivity. It correlates with findings that the immunosuppressive phenotype of PFA ependymoma is rather associated with wound healing and tissue remodeling [25] than PD-L1 interaction. Furthermore, nivolumab was used during tumor progression in this case, not as a maintenance in CR, suggesting the importance of the proper position of the ICIs in the treatment schedule. Apart from ICIs, DC vaccines have also been utilized in some patients. Data regarding their effect on pediatric ependymomas are limited, but reports of mixed R/R patient cohorts showed promising preliminary results [45, 46]. We have observed significantly increased PFS3/PFS2 ratios in PFA patients 1 and 3, who used the vaccine in CR. Patients 2 and 8 started vaccination with residual disease and progressed without significant prolongation of the interval. The future direction of immunotherapy for pediatric brain tumors represents CAR-T cell therapy. However, only the potential targets of ependymoma cells for chimeric receptors, have been recently discovered [47], and many further studies are needed.

Interestingly, an antiangiogenic and intrathecal MEMMAT regimen, which has already shown a prolongation of long-term survival in recurrent medulloblastoma patients [48] led PFA patient 7 into PR. Although this effect was not reflected in the PFS ratio it is the only instance in which we were able to demonstrate curative potential instead of using it in a maintenance setting after achieving CR by other modalities. In patients 2 and 8, we did not see the benefit of modified MEMMAT despite the presence of bevacizumab, possibly due to the inability to use its intrathecal component, which represents an integral part of this regimen.

Empiric therapeutic options included heterogeneous repurposed agents with potential activity in preclinical models or other oncological diagnoses. The most extensively studied compound is valproic acid, evaluated in stratum III of the SIOP-EP-II trial [18], known for its ability to inhibit histone deacetylase activity [30]. Another potential epigenetic agent, 5-AZA, inhibits DNA methylation and has been tested in several pediatric cancer models, including brain tumors [29]. Its CSF penetration via oral administration remains uncertain; however, an intraventricular delivery was tested by Sandberg et al. [49] and demonstrated safety. Other repurposed modalities lacked ependymoma-specific rationale but were considered based on their CSF penetration and preclinical activity in other CNS tumors. Namely, mebendazole was chosen for its good CSF penetration and preclinical activity in MBL models [31] with possible antiangiogenic effect. Rifabutin (ansamycin) was selected due to its putative targeting of HSP-90 and NF-κB pathways [32], which showed increased activity in transcriptomic data in our cohort. Since these agents were primarily administered as a supportive treatment in combination with other treatment strategies, independent efficacy remains difficult to assess. The only clear effect in the context of a residual tumor was seen in patient 5, as supported by prolonged PFS4/PFS3 and PFS5/PFS4 ratios. However, in patients 1 and 2, the sole administration of empiric agents didn’t yield any benefit. Therefore, such agents might potentially serve as a supportive backbone to other approaches aiming at maximal multimodal disease control, rather than as stand-alone therapies.

Compared to other studies reporting a 5-year SAR of around 40% [8, 50], we achieved a 66% rate. However, this observation might be influenced by the small size of our cohort. From an individual perspective, we attained a prolongation of PFS in selected cases, suggesting a potential benefit of the respective chosen strategies. Based on our experience, surgery, and radiotherapy remain the cornerstone of relapse treatment, but their use should be carefully weighed against the risk of potential severe morbidity. Early molecular testing of tumor samples and their microenvironment should be integrated into initial management to guide subsequent therapeutic decisions. In our cohort, immunotherapy was the most promising approach, particularly as maintenance therapy in CR. ICIs demonstrated efficacy in PD-L1 positive ZFTA + tumors, while in cases with negative microenvironment biomarkers, DC vaccines may serve as an alternative. When immunotherapy is unavailable or significant residuum is present, the MEMMAT protocol might be considered, though it serves only as palliative care aimed at slowing the tumor growth and comes with the risk of neurocognitive impairment and requires frequent outpatient visits. Other modalities, like repurposed valproic acid or 5-AZA, while showing potential, require further evaluation to confirm their anti-cancer effect and determine the optimal scheduling, as we observed discordant responses among our patients. However, it could be used as a support to other modalities or an alternative in case of the unavailability of previous approaches. TKIs and other molecularly targeted agents remain marginal in relapse management of pediatric ependymoma due to the lack of underlying actionable alterations.

In conclusion, identifying a rational therapeutic target based on the current understanding of the molecular background of ependymoma is still unsatisfactory. More studies are needed to elucidate signaling pathways crucial for ependymoma development that could be explored therapeutically. The application of additional strategies presented within this report appears feasible with a favorable safety profile, but their efficacy warrants further investigation.