Isolated central nervous system relapses in patients with high-risk neuroblastoma. Clinical presentation and prognosis: experience of the Polish Pediatric Solid Tumors Study Group.

Although isolated central nervous system (CNS) relapses are rare, they may become a serious clinical problem in intensively treated patients with high-risk neuroblastoma (NBL). The aim of the study is presentation, assessment of incidence and the clinical course of isolated CNS relapses. Retrospective analysis involved 848 NBL patients treated from 2001 to 2019 at 8 centers of the Polish Pediatric Solid Tumors Study Group (PPSTSG). Group characteristics at diagnosis, treatment and patterns of relapse were analyzed. Observation was completed in December 2020. We analyzed 286 high risk patients, including 16 infants. Isolated CNS relapse, dened as presence of a tumor in brain parenchyma or leptomeningeal involvement, was found in 13 patients (4.5%, 8.4% of all relapses), all of them were stage 4 at diagnosis. Isolated CNS relapses seem to be more common in young patients with stage 4 MYCN amplied NBL, and in this group they may occur early during the rst line therapy. The only or the rst one symptom may be bleeding into CNS, especially in younger children, even without clear relapse picture on imaging, or the relapse may be clinically asymptomatic and found on routine screening. Although incidence of isolated CNS relapses is not statistically signicantly higher in patients after immunotherapy, their occurrence should be carefully monitored, especially in intensively treated infants, with potential disruption of brain-blood barrier.


Background
Neuroblastoma (NBL) is the most common extracranial solid tumor in children and one of the most common causes of death in pediatric oncology 1,2 . The treatment results improved after implementation of new treatment modalities such as high-dose chemotherapy (HDC) followed by autologous stem cell rescue (ASCR), retinoid therapy and immunotherapy, also in the high-risk group 3,4,5,6 . However, concerns are brought that intensive treatment may modify the pattern of relapses.
Central nervous system (CNS) involvement at diagnosis and isolated CNS relapses are rare in children with primary extracranial tumors, including NBL. There are limited data presenting brain metastases in children but clinical reports on different extracranial tumors suggest the frequency of 1.5-4.9% 7,8,9 , being even higher in autopsy studies (6-13%) 10,11 . The estimated 3-year risk of CNS relapses in NBL reaches 8% 12 , with CNS being either isolated or one of involved sites in multiorgan relapse.
CNS can be a "sanctuary site" for cancer cells because the blood-brain barrier may impede penetration of most chemotherapeutic agents 13 . Also, antibodies used in immunotherapy do not penetrate the brain 14 , which may in uence the frequency of isolated CNS relapses in children undergoing immunotherapy.
The aim of the study is presentation, assessment of incidence, and the clinical course of isolated CNS relapses in NBL patients treated at 8 centers of the Polish Pediatric Solid Tumors Study Group (PPSTSG).

Patients And Methods
From 1st January, 2001 to 31st December, 2017, 848 patients with NBL were diagnosed at 8 centers of PPSTSG, including 286 patients with high risk disease. The data were collected till December 2019 and the follow-up was completed in December 2020. The high-risk group included 260 stage 4 patients (including 9 infants with MYCN ampli cation and 251 patients over 1 year of age, irrespective of MYCN status), 23 patients with stage 2 or 3 with MYCN ampli cation (including 4 infants) and 3 infants with 4s MYCN ampli ed tumors. The mean age at diagnosis was 43.4 months (range 0. 2 -186.3). The characteristic of the group is presented in Table 1. Surgery, radiotherapy and 13-cis retinoic acid were used as a standard in all patients. From 2012, children also received immunotherapy. The other protocols used included prolonged chemotherapy mostly according to the Japanese protocol 17 or other individually established chemotherapy protocols, surgery and radiotherapy of the primary tumor site in most cases (dose 21-36.5 Gy). Therapy with 13-cis retinoic acid was introduced depending on clinicians' decisions.
Isolated CNS relapse was de ned as parenchymal (Fig. 1A, B, C) or leptomeningeal involvement, with or without the presence of NBL cells in cerebrospinal uid (Fig. 1D), and/or NBL cells found on histopathological examination of brain hematoma. Patients with CNS metastases originating in the skull bones were excluded from the analysis. The relapse was de ned as isolated if CNS was the only site of relapse in patients in complete remission or the relapse was the only new lesion in patients with stable disease at any other sites. Only the rst relapses were analyzed.
The study was carried out in accordance with the recommendations of the Ethics Committee of Jagiellonian University Medical College, Krakow, Poland. Patients or guardians of patients provided written informed consent for treatment, including data analysis. No additional consent is requested for retrospective analysis. All methods were carried out in accordance with relevant guidelines and regulations.
Statistical analysis was performed with Statistica Software; chi-square test and t-test were used, with signi cance level 0.05.

Results
Among 286 children with diagnosed HR-NBL, relapse was diagnosed in 161 (56.3%), including 103 (50.2%) relapses in 205 children treated with SIOPEN protocol and 59 (72.8%) relapses in 81 children treated with conventional chemotherapy. The difference between these 2 groups was statistically signi cant (p=0.0001). From 2001 to 2019, isolated CNS relapses were con rmed in 13 patients (8.4% of all relapses). Incidence of isolated CNS relapses was 10.7% in the rst group (11 out of 103 relapses) and 3.4% in the second group (2 out of 59 relapses). The difference between groups was not statistically signi cant (p=0.1).
Taking into consideration the employment of immunotherapy, 28 (30.8%) relapses occurred in 91 patients treated with immunotherapy and in 118 (63.1%) of 195 patients treated without immunotherapy in the rst line treatment. The difference between groups was statistically signi cant (p<0.0001). Six (6.6%) isolated CNS relapse was diagnosed in 91 patients in whom the immunotherapy was employed in the rst line therapy and 7 (3.6%) in 195 patients without immunotherapy in the rst line treatment. The difference between groups was not statistically signi cant (p=0.26). ). In the group treated with SIOPEN protocols, the CNS relapse occurred before the time of scheduled immunotherapy in four patients, during immunotherapy in three patients and after immunotherapy (3-6 months) in three patients.
Clinical data on patients with isolated CNS relapse are presented in Tables 2 and 3.  Pathologic veri cation of CNS lesions at relapse was obtained in 11 patients, including four with NBL cells found in hematomas, three with NBL cells presented in cerebrospinal uid and ve with histopathological evaluation of tissue received after tumor removal (one of them had also previous positive CSF evaluation).

Characteristics of CNS relapse
The time from the diagnosis to relapse was 5-26 months (mean 12.5 months, median 12.0 months).
The time to relapse was much shorter in young children with intracranial bleeding being the rst presentation of relapse or accidental relapse diagnosis (5-12 months, mean 7.2) than in children with symptomatic relapse (8-26 months, mean 17.4). The difference is statistically signi cant (p=0.0002). In 5 (38.5%) out of 13 patients, the relapse occurred during intensive treatment, including one case that occurred just after induction, two cases before ASCR, one case just before the therapy of minimal residual disease with anti-GD2 antibodies and one case during this treatment.
Eight relapses were diagnosed in 16 infants with MYCN ampli cation, seven in stage 4 and one in stage 3 patients. Four of them (50%), all in stage 4 infants, were isolated CNS relapses. In three infants with isolated CNS relapse, ASCR was not carried out because of disease progression that occurred during intensive induction chemotherapy or just before planned HDC.

Clinical manifestation of isolated central nervous system relapse
In intensively treated patients, especially the youngest ones, isolated CNS involvement may be asymptomatic or rapid symptoms of intracranial bleeding may occur, without previous symptoms of CNS involvement.
Intracranial bleeding was the rst clinical symptom in 4 (30.2%) out of 13 patients, including three infants, all with MYCN ampli cation, with acute symptoms of increased intracranial pressure and early invagination. Two of them were diagnosed before planned ASCT and 2 either just before or during immunotherapy. Imaging (CT and/or MRI) done in three children revealed no evident tumor mass, only intracranial bleeding and in 1 patient there was a suspicion of bleeding into the tumor. Assessment of meninges on imaging was di cult as no contrast was given. The presence of disseminated neuroblastoma cells in hematoma together with in ltration of meninges with NBL cells were revealed on histopathological examinations in all patients. One patient had a surgery at the time of symptoms occurrence, and NBL in ltration was found on post mortem examination in three patients, as rapid progression of invagination and brain death occurred. In all these patients the symptoms progressed very rapidly. Two of them were hospitalized at the time of episode, the other one, with no symptoms, was seen in an outpatient clinic on the day of episode. Only one out of four had a mild headache a few days before episode which was interpreted to be caused by teething. All but one patient was younger than 18 months at the time of relapse, so the potential symptoms were di cult to assess. No CNS lesions were visible at MIBG imaging or/and on CT/MRI done at any time before the relapse.
In two out of four patients, there was an episode of mild head trauma on the day of bleeding, such as falling down while walking (13 and 15 months old at the time of relapse). In both cases the accident happened in the presence of parents, with no loss of consciousness or any other symptoms directly after the accident.
In 2/13 patients (15.4%) there were no clinical symptoms of CNS involvement and the relapse was diagnosed on routine examinations after induction chemotherapy (stage 4 infant with MYCN ampli cation, leptomeningeal involvement (Fig. 1A, B)) and after 2 cycles of antiGD2 (stage 4 infant with MYCN ampli cation at diagnosis, only mild headache on a day of planned CT). The rst patient, in spite of given chemotherapy, had bleeding to tumor and died in the course of this episode three weeks after CNS relapse. In the second one, the complete remission was obtained after surgery and chemotherapy, but the second isolated CNS relapse occurred on treatment at the time of the last evaluation.
The other 7 (53.8%) children presented with neurological symptoms of increased intracranial pressure (nausea, vomiting, headache) or focal symptoms, mainly disturbances of cranial nerves (strabismus, paresis of facial nerve). Symptoms lasted for a few days before the diagnosis and were the reason of performing CNS imaging (Fig. 1C). All of them were over 18 months old at the time of diagnosis and the time to progression was longer (8-26 months, mean 17.4 months from the rst diagnosis).
Only 2 (15.4%) out of 13 patients are disease free, one of them is alive without the disease >5 years from the relapse diagnosis and the second one 2.5 years from the relapse diagnosis. The rst patient received prolonged chemotherapy as the rst line treatment, without HDC, and CNS relapse was treated with chemotherapy according to brain tumor protocol and radiotherapy without a surgery as the tumor regressed after treatment. The second one received HRNBL-1/SIOPEN protocol with immunotherapy as the rst line treatment, and CNS relapse was treated with surgery, chemotherapy, RTX and intraventricular compartmental radioimmunotherapy.

Discussion
There are few data reported in the literature on the incidence of isolated CNS relapses in patients with high-risk NBL. The previously reported rate of incidence of CNS relapses was from 2.3% 18 to 16% 19 and up to 25% in small institution studies 20 , with a mean of 3.8% 12 . However, most of studies did not report only isolated CNS relapses. It is estimated that they account for about 50% of all CNS relapses 12,18 . In the German Childhood Cancer Registry, 85 patients with CNS involvement were identi ed, including 57 with isolated CNS relapse 21 .
Among 127 patients with stage 4 NBL over one year of age diagnosed in Memorial-Sloan-Kettering, eight patients (6%) developed CNS relapses. In this cohort, there was a tendency towards the higher number of CNS relapse among patients treated with immunotherapy without HDC than in patients that received HDC without immunotherapy (7/67 vs 1/60, respectively) 22 . The retrospective analysis performed by the Children's Oncology Group showed the CNS relapses at rst recurrence in 8/434 patients (2%) 18 , and in the French cohort, CNS relapses were diagnosed in 8/127 patients (6%) 12 . In the German study, the incidence of CNS relapses was reported in 49/451 patients (11%), treated with HDC 23 . In the Chinese series, brain metastases occurred in 11/106 patients (10.4%), accounting for 20% of all relapses 24 . The frequency of CNS relapses seems not to change in time (1985-2000 12 , 1990-2010 21 ). However, the in uence of immunotherapy is still unclear. There are data supporting the thesis that by reducing the incidence of systemic relapses, the number of isolated CNS relapses increases 25 , but in the SIOPEN analysis the increased risk of CNS relapse after immunotherapy was not con rmed 26 . In our study although the incidence of isolated CNS relapses was higher in intensively treated patients and after immunotherapy, the difference was not statistically signi cant Studies referred to above, except for Matthay et al 12 , reported on data of studies including children over one year of age at the time of diagnosis. As the relatively high number of infants were observed in our cohort, and in this group isolated CNS relapses were 50% of all relapses, the number of CNS relapses may be underestimated in the whole high-risk group, also including patients under one year of age with MYCN ampli cation. Additionally, in some studies, only patients who received HDC were assessed 23 , which excluded early relapses, that seem to be more common in young intensively treated patients.
In the current study, we presented 13 patients with isolated CNS neuroblastoma relapse. The pattern of relapses seems to change with treatment intensi cation, resulting in early relapses, taking the form of intracranial bleeding, especially in infants.
There are no clear risk factors for isolated CNS relapse. The data in the literature are inconsistent 18,22 .
Identi cation of statistically signi cant risk factors for CNS relapse was presented in two previously published studies. The risk factors for CNS recurrence reported by the previous research were lumbar puncture at diagnosis and LDH in one cohort 22 or age, lumbar puncture and MYCN ampli cation in another cohort 12 . In our group, disseminated disease in infants with MYCN ampli cation was the risk factor for the early CNS relapse (isolated CNS relapse was found in 25% of all MYCN ampli ed infants).
HDC alone, probably, does not prevent the CNS relapse 27 . In the presented data, occurrence of isolated CNS relapses is slightly higher in intensively treated patients, with employment of HDC, but the difference is not statistically signi cant.
Treatment of this kind of relapse is not established. Chemotherapy has been previously reported to be effective in single cases of patients with CNS NBL relapses, especially when there was the possibility of removing tumor 24 . With the employment of new treatment modalities, such as protocol employing intrathecal radiolabelled antibodies 28 , this kind of relapse is potentially curable.
In all reported patients, the parenchymal or leptomeningeal involvement were isolated settings. It supports the theory that CNS metastases are mainly blood-borne 22,29 and they may be related to the genetic alterations of NBL cells on intensive therapy 12 . Speci c genomic lesions, like 18q22.1 gain may predispose to CNS metastases 30 . These lesions are recurrently acquired during metastatic progression.
TERT gene (5p), associated with telomere maintenance and poor prognosis in NBL 31 , is one of the candidate genes associated with CNS involvement 30 . Patients with CNS relapse have also different speci c pattern of microRNA expression, with downregulation of miR-29a as a potential biomarker, with its potential role in CNS progression 32 . NBL cells penetrating to CNS on hematogenous way may keep their proliferative potential and may be a source of metastases 19,29,33 . In some studies, increased incidence of CNS metastases was observed after chemotherapy intensi cation 22,98 . Our results also suggest that natural history of disease may be changed by intensive treatment. The reason for it may be disruption of blood -brain barrier especially in very young children.
The symptoms of relapse are like described in previous reports. However, it's important to underline that symptoms are not always present, especially in young children. In our setting, we found patients who either had no clinical symptoms and relapse was found on routine imaging (two patients) or had severe neurological symptoms of bleeding and invaginations was the rst clinical presentation (four patients).
The intraparenchymal hematoma occurring prior to radiologically detectable CNS metastases are rarely described in the literature 34 .
The disease control in other tissues may be improved in case of carrying out HDC followed by ASCR and immunotherapy. It may decrease the incidence of disseminated relapses and increase the number of isolated relapses, including CNS relapses 25 . The diagnosis and treatment of isolated CNS relapse may be a more di cult challenge than disseminated relapses involving CNS.
Routine CNS examination must be taken into consideration at the scheduled time of disease assessment, even without having reported clear symptoms of CNS involvement. Also, whenever the neurological symptoms occur during therapy, the relapse should be excluded The MIBG examinations performed on routine time points may not reveal metastases. Evaluation of CNS relapses in MIBG scintigraphy may be di cult -the CNS lesions may be interpreted as skull lesions and leptomeningeal in ltration may be not seen. Moreover, it was described in the literature that the MIBG scintigraphy may be negative in con rmed relapses, even in nodular lesions 12  Patients or guardians of patients provided informed consent for treatment, including data analysis. No additional consent is requested for retrospective analysis.
Availability of data and material: The datasets generated and/or analysed during the current study are not publicly available due to General Data Protection Regulations but are available from the corresponding author on reasonable request.
Funding: There were no funding sources for this work.
Con ict of interest: The authors declare that the research was conducted in the absence of any commercial or nancial relationships that could be construed as a potential con ict of interest.

Author Contributions
Conception and design of the study: AW, WB,