The role of irinotecan-bevacizumab as rescue regimen in children with low-grade gliomas: a retrospective nationwide study in 72 patients

At least half of children with low-grade glioma (LGG) treated with first line chemotherapy experience a relapse/progression and may therefore need a second-line chemotherapy. Irinotecan-bevacizumab has been recommended in this setting in France after encouraging results of pilot studies. We performed a retrospective analysis to define the efficacy, toxicity and predictors for response to the combination on a larger cohort. We reviewed the files from children < 19 years of age with progressive or refractory LGG treated between 2009 and 2016 in 7 French centers with this combination. 72 patients (median age 7.8 years [range 1–19]) received a median of 16 courses (range 3–30). The median duration of treatment was 9 months (range 1.4–16.2). 96% of patients experienced at least disease stabilization. The 6-month and 2-year progression-free survivals (PFS) were 91.7% [IC 95% 85.5–98.3] and 38.2% [IC 95% 28.2–51.8] respectively. No progression occurred after treatment in 18 patients with a median follow-up of 35.6 months (range 7.6–75.9 months). Younger patients had a worse PFS (p = 0.005). Prior chemoresistance, NF1 status, duration of treatment, histopathology or radiologic response did not predict response. The most frequent toxicities related to bevacizumab included grades 1–2 proteinuria in 21, epistaxis in 10, fatigue in 12 and hypertension in 8 while gastro-intestinal toxicity was the most frequent side effect related to irinotecan. Bevacizumab-irinotecan has the potential of disease control clinically and radiographically in children with recurrent LGG whatever their previous characteristics; in many cases however these responses are not sustained, especially in younger children.


Introduction
Low-grade gliomas (LGG) are the most common form of childhood brain tumor, accounting for one third of all primary central nervous system tumors in children less than 18 years of age [1]. Prognosis is excellent when complete resection is obtained [2]. However, LGG are frequently not 1 3 amenable to a complete resection because of their location. In patients with incompletely resected tumors, further treatment is required in up to 50% of patients at 2 years [3]. Patients with recurrent LGG usually receive chemotherapy or focal radiotherapy, albeit less frequently due to the risk of long-term toxicity [4].
Despite high response rates, most children relapse within the first 3 years following first-line chemotherapy completion [5][6][7][8][9][10]. In the BBSFOP study, most children with optic pathway glioma that experienced a first relapse subsequently recurred again and up to 25% of the LGG recurring after the first line chemotherapy ultimately died from the disease [11]. New therapeutic strategies have therefore to be developed in order to improve the poor outcome of refractory or recurrent LGG.
Besides conventional chemotherapy, anti-angiogenic drugs have been proposed for recurrent LGG. Furthermore, neoangiogenesis has been associated with more aggressive forms of LGG [12]. Vascular endothelial growth factor (VEGF) is the best characterized pro-angiogenic factor associated with tumor growth [13]. Bevacizumab is a recombinant humanized monoclonal immunoglobulin (IgG1) antibody and has a direct anti-angiogenic effect by binding to and clearing VEGF from the tumor environment; this drug may have a direct role on vascular normalization, improving delivery of cytotoxic drugs. Inhibiting VEGF seems to normalize the vasculature and enable a better blood supply to be transiently established, producing a window where the access and efficacy of chemotherapy will be increased [14]. Moreover, it decreases interstitial pressure and hereby the injury induced by the tumor infiltration [15]. Irinotecan is a campothecin derivate that inhibits the nuclear enzyme, topoisomerase I. It penetrates the blood-brain barrier and has shown single agent anti-tumor efficacy against recurrent glioma [16,17]. It may as well potentiate the action of bevacizumab by inhibiting hypoxia-inducible factor 1 [18].
The strategy of inhibiting angiogenesis by blocking the pathway of the VEGF has proved successful in several adult solid tumors, when combined with systemic cytotoxic treatment, opening new therapeutic perspectives for children as well [19][20][21][22][23][24]. Only few small pediatric series combining these two drugs have been reported so far but with encouraging results and mechanistic studies suggest a crucial role for VEGF receptor-induced angiogenesis in the progression and maintenance of pediatric pilocytic astrocytoma [25][26][27][28]. The initial case series evaluating this treatment approach has demonstrated objective responses and clinical improvements in seven of the ten children treated with recurrent LGG, and no tumor progression while on treatment [25]. These results have been confirmed by the same authors with an expanded group of 35 pediatric patients in a multi-institutional phase II trial conducted by the Pediatric Brain Tumor Consortium [29]. The treatment was thus recommended by the brain tumor committee of the Societe Française des Cancers de l'Enfant (SFCE) as a rescue regimen in children with LGG.
The role of this combination in the management of LGG has to be further defined. We decided therefore to undertake a larger multicentric retrospective population-based study with longer follow-up to define the characteristics associated with sustained response to this therapy.

Patients/eligibility criteria
This study involved children < 19 years of age with progressive or refractory LGG treated between 2009 and 2016 in centers of the Societe Française des Cancers de l'Enfant (SFCE) with the combination of bevacizumab and irinotecan. The study was approved by the IRB of Gustave Roussy and by the Comité Consultatif sur le Traitement de l'Information en matière de Recherche (CCTIRS no 14.817). Accordingly, the parents/guardians gave their written permission for retrospective anonymized data collection and analysis.
Eligible histological diagnosis included all LGG. Histologic confirmation was not required for patients with optic/ hypothalamic tumors, especially in case of neurofibromatosis type 1 (NF1). Histologic confirmation was made by local pathologist review and in some cases central review was performed to confirm the diagnosis in difficult cases. Whenever possible, the status of the BRAF gene (re-arrangement or mutation) was also determined by immunohistochemistry, sequencing or digital droplet PCR (ddPCR) for mutations and by FISH or ddPCR for duplications. The date of diagnosis was the date of the biopsy, or the date of the first magnetic resonance imaging (MRI) scan otherwise. In order to be eligible for the study patients had to have received at least one prior line of non-surgical therapy. Refractory tumors were also eligible. At the time of treatment, all patients had evidence of clinical and/or radiographically progression and had radiographically evaluable disease.

Treatment
All patients were treated with bevacizumab and irinotecan, using the regimen previously published [29,30]. Bevacizumab was given at 10 mg/kg and irinotecan at 125 mg/m 2 every 2 weeks initially. Patients were treated with the drug combination at the discretion of the physician as best available treatment. Patients reports were collected for review of clinical and radiological files. The clinical responses and ophthalmological evaluations were noted from the consultation reports of referring oncologists and ophthalmologists. Patients characteristics, including age at start of bevacizumab, gender, histopathology, molecular biology of the tumor, number of prior treatment lines, duration of treatment, number of cycles administered, previous chemosensitivity, time to best clinical response, time to progression, were collected. The chemosensitivity to prior chemotherapy was defined as the presence of a partial response to at least one previous line of treatment. Same data were also collected in patients rechallenged with bevacizumab. Side effects were collected and their assessment was based on the Common Terminology Criteria for adverse events v 4.0.

Evaluation of responses/assessment of efficacy
All patients underwent an MRI, with and without contrast, prior to treatment initiation and at 2-3 month intervals, thereafter. Response assessed by the treating institution based upon RAPNO criteria were collected [31]. Best response anytime were considered. Modifications of the contrast enhancement only were quoted but not considered as true responses. Cystic lesions were not included in the measurement of tumor response. Clinical responses were also noted separately.

Statistical analysis
The primary endpoint to evaluate prognostic factors was progression-free survival (PFS), defined as the interval from the first day of chemotherapy to the date of the first event (progression, relapse, second malignancy, or death) or to the date of the last follow-up visit. The secondary endpoints were overall survival (OS) and response rate. Survival curves were estimated using the Kaplan-Meier method. Discrete variables were compared using the Fisher exact test. Normally distributed continuous variables were compared using Student t test, and non-normally distributed data were compared using the Wilcoxon rank-sum test. All analyses were performed with R version 3.3.1 (The R Core Team. R: 2016).

Patient characteristics
During the study period, 84 children with recurrent LGG were treated with this bevacizumab-based regimen but only 72 met the inclusion criteria. The 12 patients excluded had been treated with bevacizumab-irinotecan as first line of treatment. Patients characteristics are listed in Table 1. Median age at diagnosis was 3.2 years (range 2.5 months to 13 years) and median age at initial treatment was 7.8 years (range 1 to 19 years). Forty-seven patients (65%) had an optic pathway glioma. Of these patients, 13 (18%) suffered from a multifocal disease, of which 8 had NF1. Seven patients (9.7%) presented with a leptomeningeal dissemination. Fifteen patients (20.8%) had NF1. Histopathology was obtained in 59 patients (82%) with a diagnosis of pilocytic astrocytoma for 37 patients (Table 1). BRAF molecular status explorations found 28 KIAA1549:BRAF fusion (47.5%), 6 BRAFV600E mutation (10.2%), 12 BRAF wild type (20.3%) and 22% of undetermined status. Prior to this treatment, all patients had failed at least one prior chemotherapy and/or radiotherapy. The median number of prior lines of non-surgical treatment was 2 (range 0 to 6 lines). Forty-one tumors (57%) had received prior surgical resection. Various chemotherapeutic regimens have been used prior to bevacizumab-irinotecan including carboplatin and vincristine for 63 patients (87.5%), combination of carboplatin/procarbazine, cisplatin/etoposide and vincristine/cyclophosphamide in the BBSFOP study for 19 patients (26.4%) and vinblastine for 35 patients (48.6%). The combination of BBSFOP chemotherapy and vinblastine was performed in 13 patients (18%) [11]. In addition, the combination of cyclophosphamide and cisplatin was used in 4 patients, thioguanine /procarbazine/CCNU/vincristine combination in 2 and temozolomide in 5.

Responses and progression-free survival
The median duration of treatment was 9 months (range 1.4-16.2 months) with a median number of doses of bevacizumab received of 16 (range 3-30 doses). The median follow-up from the first dose of irinotecan and bevacizumab was 43.6 months (range 7.6-90 months). Objective responses defined as a radiological partial response, were noted in only 15 patients (21%). Minor responses were  (Fig. 2) The median PFS of the 54 patients who progressed was 16 months. Only 12 patients (16.6%) were retreated for isolated radiological progression. In these patients, visual or neurological stability or improvement persisted despite tumor progression. No progression occurred in 18 patients with a median follow-up of 35.6 months (range 7.6-75.9 months). The median PFS for patients with diffuse leptomeningeal disease at diagnosis was 17.6 months (7.9-36.9). Fourteen patients had brainstem involvement. The median PFS of these patients was 13 months (2.3-54.1) Three of the four patients who progressed on treatment had brainstem involvement.

Prognostic factors for response to therapy
The study of risk factors as a function of radiological response (PR + MR vs. SD + PD) showed that age at first treatment initiation with bevacizumab and irinotecan between 5 and 10 years was statistically associated with radiological response (OR 5.78 [1.3-37.5] p = 0.03). Treatment duration longer than 9 months and the number of irinotecan-bevacizumab doses received (> 15 doses) were also associated with an improved radiological response (respectively, OR 4.6 [1.4-16.4] p = 0.005 and OR 6.2 [1.8-26.0] p = 0.002). Patients who progressed on treatment were removed from the analysis. Previous chemosensitivity, NF1 status or the status of the BRAF gene were not associated with the radiological response.
Age was associated with PFS; patients younger than 5 had a worse PFS than those older than 10 years, while those aged between 5 and 10 years old had an intermediate PFS, p = 0.005 (Fig. 3). The analysis of the PFS according to the treatment duration demonstrated that most patients with PFS > 18 months were treated more than 9 months (67%). The median PFS of patients treated more than 9 months was higher than those treated for a shorter time: 19.6 months (range 11-50 months) vs. 15.8 months (range 4.6-75.8 months). This difference, however, was not statistically significant, (p = 0.145). Time off treatment and before progression, i.e. PFS after stopping therapy, was the same in the two groups: 9.1 months in patients treated < 9 months and 9.3 months in patients treated > 9 months. Progression during therapy and patients treated recently and still on therapy were excluded from this analysis. Outcomes of these analyses are presented in Table 2 and Supplementary Data Figs. 4, 5, 6, 7 and 8.

Toxicity
The common grade 1-3 toxicities related to bevacizumab and irinotecan are listed in Table 3 (Supplementary data). The most common toxicities related to bevacizumab were grade 1-3 proteinuria, grades 1-2 epistaxis and hypertension in 22, 10 and 8 patients, respectively. The most common adverse events in initial treatment courses related to irinotecan were gastrointestinal toxicity. Gastrointestinal toxicity resolved when irinotecan was discontinued (n = 6). Treatment was discontinued in 15 patients for the previously mentioned toxicity: proteinuria (7), gastrointestinal toxicity (6) and fatigue (2). Proteinuria resolved over a period of 2 weeks to 9 months following cessation of bevacizumab. Treatment was never discontinued because of hypertension. One patient came off treatment because of a grade 2 thrombosis of a deep vein after 5 months. The thrombosis did recur when the patient was retreated by bevacizumab and irinotecan for a clinical progression. No child in the study developed intracerebral hemorrhage. There was no toxic death.

Bevacizumab re-treatment
Nineteen patients (26%) were re-treated with bevacizumab. For 3 patients, the treatment had been interrupted because of toxicity, for 6 others because of a radiological stability and for the 10 remaining patients systematically after a treatment duration of 1 year. Thirteen patients had no other therapy in the interval between the two bevacizumabbased treatments. Bevacizumab monotherapy was used for 3 patients, associated to irinotecan for 8 and to vinblastine for 2 patients. Despite previous exposure to bevacizumab, all patients had clinical and/or radiographic improvement or stability with re-treatment. The median duration of the second bevacizumab treatment regimen was 8 months (range 2.6-12.1 months). Median PFS after secondary therapy with bevacizumab was 12 months (range 4-36 months). Five patients were re-treated a third time with bevacizumab associated for 2 of them with vinblastine and clinical improvement and radiological stability were again noted. One patient was treated a third and a fourth time with bevacizumab for subsequent clinical or radiographic progression. Within few months, his vision again improved in conjunction with radiographic improvement. The PFS after each bevacizumabbased therapy were respectively 12 and 20 months. Bevacizumab related toxicities reported were hypertension in two, the two times at the first bevacizumab-based regimen line but not during re-treatment, and proteinuria in eight, four of them during the two lines of treatment, three only during the first and one during the re-treatment.

Discussion
In this expanded cohort, the largest available in the literature for this disease, we confirm that the combination of bevacizumab and irinotecan produces at least a disease stabilization in over 95% of patients who had previously failed standard chemotherapy and/or radiotherapy. In terms of response rate and disease control rate, with the caveats of the retrospective nature of the study, this regimen compares favorably with existing treatments and warrants further consideration. Bevacizumab-based therapy demonstrated partial radiological responses in only 15 of 72 (21%) of these refractory or recurrent cases, a lower response rate than in the study from Hwang (5 out of the 14 patients) [27]. Conversely, the Pediatric Brain Tumor Consortium phase II reported only 2 partial radiological responses in the 35 patients studied, i.e. 5.7% [29]. Only four tumors progressed under irinotecan / bevacizumab treatment. This result is similar to what has been observed in a previous study [29]. However, radiological responses did not necessarily predict sustained disease stability. In our study, PFS at 6 and 24 months were 91.7% [IC 95% 85.5-98.3] and 38.2% [IC 95% 28.2-51.8] respectively. In the PBTC phase II study including 35 LGG patients the 6-month and 2-year progressionfree survivals were similar, with 85.4% (SE + 5.96%) and 47.8% (SE + 9.27%), respectively [29].
This combination also leads to rapid clinical benefits. Almost all patients (96%) described an improvement or a stabilization of their clinical state. The relatively short time to maximal response and noted clinical improvement highlights the advantage of utilizing this regimen in situations where there is a functional threat or a progressive neurologic/visual deterioration.
Some patients were re-treated with bevacizumab and despite previous exposure, all patients had clinical and/ or radiographic improvement or stability. There did not appear to be a decrease in efficacy with repeated courses of bevacizumab, implying that the tumors did not gain resistance to this approach. This suggests that the tumor vasculature might still be responsive to bevacizumab on repeated exposure. Moreover, toxicity of the second or third chemotherapy with bevacizumab was not increased.
The evaluation of potential prognostic factors demonstrated that the age at initiation of treatment affected the response and improvement of PFS. Patients younger than 5 years had a lower PFS than older patients. This observation confirms the data from the literature obtained with standard chemotherapies. Stokland et al. demonstrated on a cohort of 639 children with LGGs that age is a strong risk factor for disease progression [11,32].
In our study, chemoresistance or anterior chemosensitivity did not impact PFS after treatment with irinotecanbevacizumab. De Haas et al. studied relapses after a first line of chemotherapy in 68 patients treated for an optic pathways glioma. This study showed that patients with an objective radiological response to first-line chemotherapy had a greater chance of being good responders to secondline chemotherapy [11]. The use of bevacizumab has similar efficacy in patients previously chemosensitive or not. Consequently, bevacizumab-irinotecan appear to be a good rescue regimen for patients who did not respond to the previous chemotherapy(ies).
NF1 status did not appeared to predict a more sustained disease stability after this regimen, unlike what is usually observed with standard chemotherapy [33,34]. These patients did not respond longer than non-NF1 patients despite differences in tumor vascularization on MRI [35,36].
The study of PFS according to the pathological diagnosis seems to show a poorer PFS in patients with gangliogliomas compared to patients with pilocytic astrocytoma. This difference was not statistically significant. This trend seems to be confirmed after the molecular biology study of BRAF status with poorer PFS in patients with tumors carrying the BRAF V600E mutation, but again without statistical significance. In the literature, patients with BRAF V600E PLGG exhibited poor outcomes after chemotherapy and radiation therapies and constitutes a distinct population with poor prognosis when treated with current adjuvant therapy [11]. In the absence of significant differences, this therapeutic combination could therefore be considered as a rescue regimen, irrespective of the nature of the tumor.
The pattern of toxicities in this study is similar to that observed in adult and pediatric patients treated with bevacizumab [37,38]. In our study, hypertension, fatigue, and proteinuria were relatively common and usually manageable. Nevertheless our understanding of long-term side effects in patients exposed to bevacizumab is evolving and both shortand long-term toxicity warrants further studies [27].
The results of this study indicate that bevacizumab-irinotecan is a useful therapeutic modality for children with recurrent LGG who have failed standard chemotherapy regimens. Disease control is obtained in most patients, irrespective of their clinical characteristics. Sustained responses are mostly observed in older children. Duration of therapy does not affect the PFS after stopping the therapy. With the bevacizumab-irinotecan combination, interval extension between two doses could prevent therapy discontinuation to allow for prolonged therapy and further optimized clinical efficacy and safety. However, the tolerance of irinotecan has often been poor, justifying the discontinuation of the combination. It was previously described in a small number of cases that monotherapy with bevacizumab has its own benefits, as removal of irinotecan from the treatment regimen decreases combined toxicity without significant effect on good outcomes [27]. Combination of bevacizumab with other agents have been used empirically but not reported precisely. Gorsi et al. studied bevacizumab as a single agent in the treatment of recurrent or refractory LGG, and as in our study, even though almost all patients experienced radiological progression of the disease after the discontinuation of bevacizumab, clinical responses were meaningful and sustained in most of cases [39]. Their study showed a favorable response of LGG to single agent bevacizumab. Regarding the optimal duration of treatment, it is not defined and need to be studied further; however, it seems plausible to consider longer courses of therapy would be possible with careful monitoring of proteinuria.
Author contributions JG, ATE, PV and CdM designed and conceptualized the report, acquired and interpreted patient data, drafted the manuscript for intellectual content, reviewed and revised the manuscript. AC, CP, MPR, KB, SP, NB, DFB, EdC, FB, PL, FF, NA, AIB, TB, CD and DVC acquired and interpreted patient data, reviewed and revised the manuscript. All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations
Conflict of interest No funds, grants, or other support was received. The authors have no relevant financial or non-financial interests to disclose. Dr Grill has received grant support from Hoffmann La Roche for a study with bevacizumab in pediatric high-grade gliomas.
Ethical approval Ethical approval was waived by the local Ethics Committee of Gustave Roussy (CCTIRS no 14.817) in view of the retrospective nature of the study and all the procedures being performed were part of the routine care.
Informed consent Informed consent was obtained from all individual participants included in the study.