Introduction

Optic pathway gliomas (OPGs) account for approximately 5% of all tumor cerebri in children [1], and in approximately 30% of cases, they are associated with neurofibromatosis type 1 (NF1) [2–3]. Although a large majority of these gliomas are pilocytic astrocytomas [3], their evolution is unpredictable, with a 5-year overall survival (OS) rate of more than 90% [4–5] but a lower progression-free survival rate of 40% [6]. Because of their characteristics, they seem to constitute a specific entity among low-grade gliomas (LGGs): indeed, they expose affected patients to neurological, visual and endocrinal risks, and unlike other LGGs in which surgery is often the only course of treatment (for example, posterior fossa), their location does not allow total resection. In addition, recent biological studies suggest different molecular alterations between posterior fossa pilocytic astrocytomas and pilocytic astrocytomas in visual pathways [7].

Currently, the optimal treatment of OPGs remains extremely controversial. The possibility of stabilization, even with a spontaneous regression of lesions, does not seem to justify a systematic treatment of all patients with optic glioma [8]. When the tumor is limited to one optic nerve with loss of vision on that side, surgery is a possible therapeutic strategy [9]. However, the benefit is very questionable when the tumor involves the optic chiasm: partial surgery is possible but with the risk of inducing vision loss and/or hypothalamic disorders. Thus, for a long time, radiotherapy has been the standard treatment, with a dose of 45 Grays or more allowing stabilization of visual disorders and tumor growth [10]. However, this can concomitantly lead to second tumors and significant sequelae, particularly in terms of cerebral vascular radiation [11], dysfunction of hypothalamic-pituitary function [12] and/or slowing of cognitive development [6]. These effects are even more marked as treatment is administered in young children. For these reasons, and to maintain maximum future functions for these children, the French Society of Pediatric Oncology (SFOP) developed, from 1990, a BB-SFOP chemotherapy regimen (Baby Brain-French Society of Pediatric Oncology) that is specifically designed for children under 5 years of age [13]. This treatment included a combination of 6 antimitotics that are administered sequentially for 16 months and include carboplatin, procarbazine, etoposide, cisplatin, vincristine and cyclophosphamide. This chemotherapy was mainly intended to delay or even to avoid irradiation of the tumor: indeed, in cases of secondary progression, second-line chemotherapy or radiation therapy could be proposed depending on the age of the child. Beginning in 1995, this therapeutic strategy was extended to children over 5 years of age, including those carrying NF1, given the deleterious effects of radiotherapy in this type of disorder. Concurrently, other pediatric oncology groups [14–17] have developed a similar therapeutic approach that favors chemotherapy as a first-line treatment for such patients.

At present, these regimens, including first-line chemotherapy, have become the standard for the treatment of OPGs. However, although this therapeutic strategy has been applied for more than 20 years, most studies published in the literature do not consider the outcomes of children beyond 5 or 10 years of age [5,18–19]. These findings are even more surprising given that this tumor is well known for its slow evolution. For this reason, we chose to initiate a large historical cohort study of children treated with BB-SFOP as first-line chemotherapy, which represents a relatively homogeneous population to allow reliable comparisons between patients. These children were subjected to a very long-term follow-up to better define their outcomes. In the present study, we focused on mortality in these patients, with the aim of clarifying overall survival and answering three questions: is the OS of these patients as good as is generally described? What are the main causes of death and the main factors involved in the survival rate?

Patients and Methods

Results

Patient characteristics

Between June 1990 and December 2004, a total of 445 children in France were treated with BB-SFOP chemotherapy. There were various indications (this chemotherapy can be applied in young children for brain tumors other than OPGs), and only 182 patients aged less than 16 years were identified as having been treated for OPG, using BB-SFOP chemotherapy as the first-line treatment. Two patients could not be included in the analysis because too much diagnosis data were missing from their records. Finally, our series included 180 children with a median follow-up of 13.6 years (range: 6.1–23.6). The main characteristics of these patients at the time of diagnosis are summarized in Table 1. Among the 180 patients, 79 had a follow-up ≥ 15 years, and 49 had a follow-up ≥ 18 years. Six patients were lost to follow-up (median follow-up: 11.8 years, range: 6.1–14.6).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of patients at diagnosis.

https://doi.org/10.1371/journal.pone.0127676.t001

Fifteen patients with diencephalic syndrome (DS) were less than 1 year old, 12 were between 1 and 2 years old, and 5 were more than 2 years old. Fifteen patients presented with the spread of disease at diagnosis, including 6 patients with DS. However, spinal axis MRI was not systematically performed, and some metastases may not have been diagnosed.

In terms of diagnosis, 109 of 180 patients (60.5%) underwent surgery (biopsies and/or partial resections) initially or during the course of treatment. Among the 71 patients without histological diagnosis, 46 were NF1 patients, and 25 were non-NF1 patients: for these patients, the clinical data did not specify the reasons for not performing surgical biopsy and/or partial resection. When a histological diagnosis was available, it confirmed LGG in 104 patients and showed grade II or III astrocytoma in 5 patients; however, unfortunately, a pathology review could not be performed. For the other children, MRI imaging was considered sufficient to establish a diagnosis of OPG.

Overall Survival

Computed from the date of diagnosis, the 5- and 10-year OS rate were 95% [95% CI: 90.6–97.3] and 91.6% [95% CI: 86.5–94.8], respectively, for the whole population. Among the 79 patients who could have a follow-up of ≥ 15 years, the OS rate at 15 years was 80.7% [95% CI: 72.7–86.8], and among the 49 patients who could have a follow-up of ≥ 18 years, the OS rate at 18 years was 75.5 [95% CI: 65.6–83] (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Overall Survival for the whole population with optic pathway glioma.

https://doi.org/10.1371/journal.pone.0127676.g001

We analyzed different clinical criteria that might influence the results: sex, age at diagnosis, NF1 associated with OPG, presence of DS or intracranial hypertension at diagnosis. The results are summarized in Fig 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Overall survival according to age at diagnosis (A), NF1 status (B), presence of DS (C) or intracranial hypertension (IH) (D) at diagnosis.

https://doi.org/10.1371/journal.pone.0127676.g002

According to the Cox univariate analysis and the Kaplan-Meier survival analysis (Table 2), males had a better survival than females, although this difference was not statistically significant (HR: 0.4 [95% CI: 0.2–1], p = 0.06). We observed a better prognosis in children with NF1 associated with OPG (HR: 0.8 [95% CI: 0.6–0.9], p = 0.035). Patients with intracranial hypertension at diagnosis also had a worse prognosis (HR: 4 [95% CI: 1.2–13.3], p = 0.023). The risk of death was increased in children less than 1 year of age (HR: 2.6 [95% CI: 1.3–5.4]), p = 0.006) and in those with DS at diagnosis (HR: 2.8 [95% CI: 1.3–6], p = 0.005). We also analyzed the potential influence of the applied treatments (initial partial surgery, chemotherapy and/or radiotherapy administered after initial BB-SFOP treatment). There was no statistically significant difference in the survival rates and the risk of death according to whether the patients were or were not eligible for initial partial resection (HR: 1.2 [95% CI: 0.5–2.7], p = 0.65), or had or had not received radiotherapy (HR: 0.9 [95% CI: 0.4–1.9]) (p = 0.88) or chemotherapy (HR: 1.04 [95% CI: 0.9–1.2]) (p = 0.56) after BB-SFOP. Survival rates according to treatments are shown in Table 2 and S1 Fig.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Overall Survival distributions and Hazard Ratios depending on the clinical data at diagnosis and treatments (p values according to univariate Cox regression analysis).

https://doi.org/10.1371/journal.pone.0127676.t002

In the Cox multivariate analysis (Table 3), an age at diagnosis of less than 1 year (HR: 3.1 [95% CI: 1.5–6.2], p = 0.002) and the presence of initial intracranial hypertension (HR: 5.2 [95% CI: 1.5–17.6], p = 0.007) were associated with a worse OS. Neither NF1 status nor the presence of DS at diagnosis appeared to be discriminant prognostic factors. When introducing an interaction term between the group of patients with DS and the variable sex, the association between the DS group and OS varied by gender. Boys without DS had a better prognosis compared with girls with or without DS and boys with DS (HR: 0.3 [95%CI: 0.1–0.8], p = 0.007).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Multivariate Cox regression analysis of factors associated with Overall Survival in patients with OPG.

https://doi.org/10.1371/journal.pone.0127676.t003

Using these factors associated with a poor prognosis (age, intracranial hypertension, sex and DS), we were able to describe 2 groups of patients: B (absence or presence of one of these poor prognosis factors); C (presence of 2 or more of these factors). There was a statistically significant difference in the 5, 10, 15 and 18-year OS between group B and C (HR: 4.5 [2.1–9.2], p<10⁻³). The OS rate at 18 years was 87.8% in the first group [95%CI: 78.4–93.3] compared to 50.4% in the second group [95%CI: 31.4–66.6]. These results are summarized in Fig 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Overall survival according to the number of risk factors present at diagnosis: whole population (A), 0 or 1 risk factor (B), 2 or more risk factors (C).

https://doi.org/10.1371/journal.pone.0127676.g003

Discussion

To our knowledge, this is one of the largest series ever published describing children who were diagnosed with an OPG and initially treated uniformly using the same chemotherapeutic regimen with a long-term follow-up.

OPGs are known to have a slow but unpredictable course, and although rapid progression can be observed in some patients, most authors indicate a very good 5-year OS for this tumor, varying between 92% and 98% [5,18–19]. After 5 years of follow-up, our study revealed comparable results to these series. However, because we extended the follow-up beyond 5 years, we observed that after this period, survival significantly decreased over time without ever stabilizing. After 15 years, the OS rate was 80.7%, and after 18 years, it was only 75.5%. A number of patients died as a result of tumor progression more than 10 years after diagnosis, which is probably due to the very slow evolution of this type of glioma. Therefore, based on our data, we think that survival, when evaluated 5 or 10 years after diagnosis, may not be representative of the true survival experience of children with OPG because of the characteristics of this tumor. Consequently, there is a need to study survival over a longer period.

The role of NF1-associated OPG remains controversial. While some authors did not find a significant difference between progression-free survival (PFS) rates in children with or without NF1 [15–22], Fouladi et al. [23], in 2003, discovered a 6-year OS rate of 96% in NF1 patients compared with 80% in non-NF1 patients, but this series included patients with active treatment immediately after diagnosis and patients without initial therapeutic intervention. More recently, Gnekow et al, in a series of 415 supratentorial midline LGG patients [24], did not find any differences in OS in NF1 compared with non-NF1 patients, but these results included all localizations of LGG. In our series, we did not detect statistically significant differences between NF1 and non-NF1 patients in the Cox multivariate analysis when considering NF1 as predictive factor in the survival rate. Despite a trend toward a better prognosis in NF1 patients during the first years of follow-up, this difference completely disappeared over time with the same OS rate between the 2 groups at the 15-year follow-up. A majority of deaths in NF1 patients occurred after the 10-year follow-up and following tumor progression, which suggested that these children could die later than the non-NF1 patients. They did not die of other phakomatosis complications but of an OPG, and the administered treatments, especially radiotherapy [25–26], did not seem to have an effect on this long-term evolution. However, these results could be influenced by the observation that the percentage of irradiated children, in our series as in others, was generally lower in NF1 than in non-NF1 patients. When considering our results and those in the literature, the slower progression might have been easier to initially control in these NF1 patients but was likely to worsen over time. This finding again appears to highlight the need for the sufficient follow-up of these tumors to accurately assess the outcomes of the children.

Age is another prognostic factor noted by many authors: in numerous series, the analysis of pediatric patients with an OPG has demonstrated that an age of less than 1 year is an independent negative prognostic factor [5, 19, 23–24, 27–28]. These results are very similar to ours. The difference observed in OS was statistically significant at the first 5-year follow-up and did not change thereafter. In other words, with the knowledge that the deaths were mostly due to tumor progression, the risk of treatment failure appeared to be higher for children less than 1 year old than for older children and did not improve over time.

Although it is a rare disorder, the presence of a DS at diagnosis is a third prognostic factor noted in the literature, especially when it is associated with leptomeningeal spread [13, 24, 29]. In our series, the number of children with a spread of disease at diagnosis was too small to draw any conclusions concerning the association between DS and disseminated disease. Recently, Kilday et al. [30] reported a series of 9 children with DS who were treated with up-front chemotherapy followed, in cases of progression, by several lines of chemotherapy and/or surgery without radiotherapy: these patients demonstrated a 5-year OS of 100%, but the conclusions appeared to be difficult to determine because the number of patients was very small and the follow-up was short. Our study results did not confirm these findings, and moreover, our conclusions might seem surprising given that DS and female appeared to be a poor prognostic factor. This result can be compared with the recent study reported by Fisher et al. [20], who found that gender may be a disease modifier in the development of clinically significant OPG in children with NF1. However, to our knowledge, no study specifically investigating very long-term OS in these children with DS has ever been published. The number of children with an OPG and DS in our cohort is one of the largest reported to date, and the distribution between males and females is similar. Because these children also had a very long follow-up (median: 13.6 years, range: 9.6–23.5), we believe that our results merit interest. Nevertheless, we do not have an explanation for the present findings, and we did not find any studies in the literature comparing boys and girls with an OPG and DS: a prospective study of these patients is likely needed to better understand the evolution of these two groups of children.

A few reports have been conducted to investigate intracranial hypertension in patients with an OPG [31–32], but none of them considered this symptomatology at diagnosis as a possible predictive factor of the patient outcomes. Thus, in our series, it was surprising to note a slightly significant difference in the OS rate with a worse prognosis in patients with these symptoms. Intracranial hypertension did not correlate with other factors. Half of the patients with intracranial hypertension died before 6 years of follow-up. The tumor characteristics (location, speed of development, leptomeningeal dissemination) could explain why but we were not able to demonstrate it. This question should likely be investigated in future studies.

There are several limitations of the present study that should be acknowledged. First, this is a historical cohort study in which only OS was analyzed without reference to event-free survival (EFS) or progression-free survival (PFS). However, the main purpose of this study was to assess the mortality of children with OPG who were treated with up-front chemotherapy. In addition, analyses of EFS and PFS would require a consensus regarding the definition of the clinical and radiological criteria used. At present, there are still too many variations of these criteria and of the tumor measurements, and discrepancies in response categorizations among readers have been well demonstrated in several studies [33–34]. Under these conditions, comparisons between different published series lack reliability. The second limitation is that this series included only children with initial symptoms that justified treatment. By definition, it excluded children with no or few symptomatic OPGs, in whom the condition might have remained stable over time, and thus, only monitoring was needed. Therefore, our findings relate only to this type of patient with initially progressive disease and cannot be considered as a prognostic evaluation of all OPGs. A third limitation of this study is the number of patients with a histological diagnosis. Most studies concerning optic pathway glioma have reported a histological diagnosis in a limited number of patients, and the diagnosis was usually established based on MRI. For example, in the study reported by Mishra et al. [5], 445 patients were identified through the SEER database as being diagnosed with an OPG, but histology was available in only 131 patients (29.5%). Thus, because there was no central review, we cannot definitively exclude that some of the tumors had been misclassified or that some patients were carrying tumors other than LGG. A fourth limitation was that the cause of death was noted by the clinician in the patient’s record, and in the present study, it was impossible to verify these data long after death. Therefore, we cannot exclude the possibility that, for some patients, the cause of death differed from that identified in the file. A fifth limitation was the analyzed prognostic factors. We limited our study to clinical criteria because a radiological review (in progress) would not be able to answer all of our questions, mainly because of differences in the techniques used, depending on the centers and the dates on which the MRIs were performed. In addition, the number of deceased patients in our series was not large enough to enable the detection of other risk factors. We can no longer take into account recent advances in the tumor biology of LGG [35–41]. Nevertheless, there is no doubt that, in the future, advances in pathology and molecular biology will greatly improve the understanding of LGG and, consequently, of OPGs. This progress will facilitate the design of new treatments that are better targeted and adapted to different types of tumors so that significantly improved results in terms of survival are realized in the future.

Nevertheless, the results of this study allowed us to draw three conclusions. First, the prognosis of OPG in children, in terms of OS, is probably not as favorable as previously described when patients have a follow-up that is long enough to account for later deaths (more than 10 years after diagnosis). If we want to properly assess the evolution of these children, it seems necessary not to limit the follow-up to 5 or 10 years but to extend it for a very long period of 15 years or more. Second, tumor progression appears to be the main cause of death, especially at time points longer than 6 years after diagnosis. Deaths related to vasculopathy or second tumors are much less frequent and do not appear to be systematically linked to NF1 status and/or radiotherapy. Third, taking into account the risk factors that we have identified, children with OPG appear to be classified into 2 groups, each with different issues. In the first group, which had a very good 18-year OS, the key question appears to concern long-term disease sequelae and/or treatment (especially visual outcomes and hypothalamic syndrome). In contrast, in the second group with poor long-term survival, the main problem appears to be the effectiveness of the treatments. Future advances in therapeutics resulting from a better understanding of tumor biology should focus on patients with a worse prognosis.

Supporting Information

Acknowledgments

The members of the Brain Tumor Committee of SFCE (Principal Investigator: Xavier Rialland, email: famille.rialland@gmail.com) are as follows:

Claire Alapetite, Department of Pediatric Oncology, Curie Institute, France;

Nicolas André, Department of Pediatric Oncology, AP-HM, Marseille, France;

Valérie Bernier, Department of Radiotherapy, A. Vautrin Center, Vandoeuvre-les-Nancy, France;

Anne-Isabelle Bertozzi, Department of Pediatric Oncology, University Hospital, Toulouse, France;

Franck Bourdeaut, Department of Pediatric Oncology, Curie Institute, Paris, France;

Christian Carrié, Department of Radiotherapy, Leon Berard Center, Lyon, France;

Pascal Chastagner, Department of Pediatric Haematology-Oncology, University Hospital of Nancy, France;

Cécile Conter, Department of Paediatrics, Institute of Hematology and Pediatric Oncology, Lyon, France;

Emilie De Carli, Department of Pediatric Oncology, University Hospital, Angers, France;

François Doz, Department of Pediatric Oncology, Curie Institute and University Paris Descartes, Sorbonne Paris Cite, France;

Christelle Dufour, Department of Pediatric and Adolescent Oncology, Gustave Roussy Institute, Villejuif, France;

Natacha Entz-Werlé, Department of Pediatric Oncology, University Hospital, Strasbourg, France;

Dominique Figarella Branger, Service d'Anatomie Pathologique et de Neuropathologie, Timone Hospital, APHM, Marseille, France;

Fanny Fouyssac, Department of Pediatric Haematology-Oncology, University Hospital of Nancy, France;

Didier Frappaz, Adult and Pediatric Neurooncology, Léon Bérard Center, Lyon, France;

Jean-Claude Gentet, Department of Pediatric Oncology, Timone Hospital, APHM, Marseille, France;

Jacques Grill, Department of Pediatric and Adolescent Oncology, Gustave Roussy Institute, Villejuif, France;

Karien Guichardet, Department of Pediatric Oncology, University Hospital, Grenoble, France;

Jean Louis Habrand, Radiotherapy Department, University Hospital, Caen, France;

Claire Hoyoux, Pediatric Hematology-Oncology, CHR Citadelle, Liege, Belgium;

Céline Icher, Department of Pediatric Hematology-Oncology, University Hospital, Bordeaux, France;

Justine Kanold, Department of Pediatric Haematology-Oncology, University Hospital, Clermont Ferrand, France;

Véronique Laithier, Department of Pediatric Oncology, University Hospital, Besançon, France;

Pierre Leblond, Department of Pediatric Oncology, Oscar Lambret Center, Lille, France;

Véronique Mosseri, Department of Biostatistics, Curie Institute, Paris, France;

Xavier Muracciole, Radiotherapy Unit, La Timone University Hospital, Marseille, France;

Laetitia Padovani, Radiotherapy Unit, La Timone University Hospital, Marseille, France;

Anne Pagnier, Pediatric Department, University Hospital, Besançon, France;

Gilles Palenzuela, Timone Hospital, Marseille, France;

Yves Perel, Department of Pediatric Hematology-Oncology, University Hospital, Bordeaux, France;

Dominique Plantaz, Department of Pediatric Haematology-Oncology, University Hospital, Grenoble, France;

Xavier Rialland, Department of Pediatric Oncology, University Hospital, Angers, France;

Eric Sariban, Hemato-Oncology Unit and Laboratory of Pediatric Oncology, Hôpital Universitaire des Enfants, Brussels, Belgium;

Pascale Schneider, University Hospital Charles Nicolle, Rouen, France;

Christine Soler. Department of Pediatric Haematology-Oncology, University Hospital, Nice, France.

We are grateful to the investigators of SFCE for providing details regarding the follow-up: Y Perel CHU Bordeaux, C Chappé CHU Rennes, N Sirvent CHU Montpellier, P Schneider CHU Rouen, C Berger CHU St Etienne, P Lutz CHU Strasbourg, P Blouin CHU Tours, C Piguet CHU Limoges, P Lemoine CHU Brest, F Millot CHU Poitiers, O Minckes CHU Caen, F Demeocq CHU Clermont-Ferrand. We are also grateful to F de Vathaire, INSERM U1018, Institut Gustave Roussy, Villejuif, for providing data base BB-SFOP. We are grateful to E Berardi and S Chalal, CHU Angers, for collecting data, to C Alberti, INSERM CIE5, Robert Debré Hospital, Paris for reviewing the study design and to I Pellier, S Proust and P Mercier, CHU Angers, for reviewing the manuscript. The authors thank Joanne Walker for English language advice.