Background
NF1 is a dominantly inherited multisystem disorder affecting 1 in 3500 individuals [
1]. It is caused by mutations in
NF1, a large gene located on the long arm of chromosome 17 [
2‐
4]. NF1 is a neurocutaneous disorder characterized by the development of dermal and plexiform neurofibromas and café-au-lait spots [
5]. One of the most serious manifestations of NF1 in children is the development of optic pathway gliomas (OPGs). These tumours affect approximately 20% of all children with NF1 [
6] and can lead to a loss of vision, proptosis, or precocious puberty. Fortunately, however, these tumours remain asymptomatic in the majority of affected children. It is currently not recommended to screen children with NF1 routinely by MRI for optic pathway gliomas, as the vast majority of tumours are indolent, and early detection does not improve visual outcomes [
7].
Factors such as tumour location [
8] and changes in tissue microstructure [
9] have been proposed to predict which OPGs will become symptomatic; however, these findings remain controversial. This is largely because the natural history of OPGs in people with NF1 has not been thoroughly characterized: There are no large studies of adult NF1 patients with OPGs, and we do not even know the prevalence of this tumour in adults with NF1.
In this study, we used routine MRIs to investigate the prevalence and natural history of optic pathway gliomas in children and adults with NF1.
Methods
Patients
All NF1 patients seen in the NF outpatient department of the University Hospital Hamburg-Eppendorf between 2003 and 2015 were offered whole-body and head MRIs as part of a routine tumour monitoring protocol [
10]. Since MRIs were offered to all patients independent of their clinical symptoms, the images are representative of the patient population seen in the clinic. Informed consent was obtained from all subjects, and the ethical committees of the Medical Chamber of Hamburg and the University of British Columbia approved the study.
Magnetic resonance imaging (MRI)
To evaluate the extent of optic gliomas, we defined four locations: intraorbital optic nerves, prechiasmatic optic nerves, chiasm, and optic tracts/optic radiations. A glioma of the optic pathway was diagnosed if there was hyperintensity on T2-weighted images or if they enhanced after contrast injection. A tumour was defined as being multifocal of it was present in two or more unconnected sections of the optic pathway.
Based on the 1775 clinical MRI reports, a list of patients was generated who had been clinically diagnosed with OPG. All head MRIs from these patients were re-evaluated by two neuroradiologists in Canada (M.M. and M.K.S.H.), and the presence of OPGs in each individual MRI study was established by consensus using the criteria described above.
Clinical features extracted from the MRI reports of all 562 patients included: Presence of OPG, presence of non-optic gliomas, presence of unidentified bright objects (UBOs) and presence of plexiform neurofibromas (PNs) on the corresponding whole-body MRI examination. The presence of subcutaneous neurofibromas in patients with OPG was determined from the corresponding whole-body MRI.
The last visual acuity measurements during the study period were used to determine the presence or absence of loss of vision and visual field defects.
Descriptive statistical analysis
Patients were divided into 10-year age groups and counted only once per age group. 95% confidence intervals of OPG prevalence were calculated as ±1.96 standard deviations of a binomial distribution.
Multiple logistic or linear regression was performed to identify factors associated with OPG presence or volume. Natural log transformation was applied to the tumour volumes to achieve normal distribution for linear regression. Predictor variables for logistic regression with presence of OPG as the outcome variable were age at first scan with OPG present and the presence of non-optic gliomas, UBOs, or plexiform neurofibromas. Predictor variables for analysis of OPG volume were age at first scan with OPG present, and the presence of non-optic gliomas, UBOs, plexiform neurofibromas or subcutaneous neurofibromas.
We used the Kaplan-Meier method to calculate the cumulative rates and 95% confidence intervals of OPG progression or regression within 5 years of first MRI diagnosis. The log-rank test was used to assess differences between the rates of progression or regression in patients under 20 years of age and those over 20 years of age. These analyses were performed with IBM SPSS Statistics version 24.
UBO prevalence in patients with or without OPGs was compared using the Mantel-Haenszel test. Each patient was counted once after stratification into a single age group for this analysis: 0–9.99, 10.0–19.99, 20.0–29.99, 30.0–39.99 or ≥ 40 years old at the time of first MRI or first MRI on which an OPG was seen in the study. The calculation was performed using IBM SPSS Statistics version 24.
Results with p ≤ 0.05 were considered to be statistically significant.
Discussion
In this study we report the largest series of head MRIs described to date in unselected NF1 patients. Most previous studies have used convenience samples to estimate the prevalence of OPGs in children with NF1 [
12,
13]. This approach, however, carries the inherent bias of the patients being selected for having clinical symptoms that required them to undergo imaging. In our study, every patient seen in the NF outpatient department was offered MRI, so our series is an unbiased representation of the patients seen in the clinic. Blanchard et al. recently performed a prospective head MRI study of 306 children with NF1 under 6 years of age and found the prevalence of OPG to be 14.7% (95% confidence interval: 11.0% to 19.3%), with 80% of patients being asymptomatic [
13]. Other authors found similar prevalences in children, ranging from 15%–18% [
12,
14,
15]. We found a somewhat higher prevalence (22%) among children with NF1 under 10 years of age: This might be due to different diagnostic criteria (T2 hyperintensity without consideration of thickness or tortuosity of the optic nerve) used in our study compared to the previous studies.
The prevalence of OPG in older children and adults with NF1 is lower than in young children but has rarely been evaluated. One retrospective study by Créange et al. found the prevalence in 138 individuals with NF1 over 18 years of age to be 5.8% [
16]. In concordance with this finding, we found the prevalence in adults over 19.9 years of age to be 4.9% (95% confidence interval: 3.3% to 7.2%).
The decline in prevalence of OPG from childhood to adulthood might be explained in several different ways. Firstly, it is important to note that most OPGs are asymptomatic and are never confirmed by biopsy in people with NF1 [
17,
18]. Optic nerve tortuosity and optic nerve sheath thickening are frequent in children with NF1 who do not have OPG [
19], and it is not known if T2 hyperintensity or MRI enhancement of the optic nerves can occur in the absence of other evidence of neoplasia in this setting. Thus, it is possible that some OPGs diagnosed by MRI in children with NF1 are not true neoplasms.
Alternatively, there might be increased mortality in individuals with optic tumours, so that children with OPG are less likely to survive into adulthood. However, a study by Guillamo et al. showed that having an OPG is
not a risk factor for premature death of NF1 patients [
20].
It has been shown that having an OPG predisposes to the development of non-optic gliomas [
21], which are associated with increased morbidity [
20]. We found a strong correlation between presence of OPG and presence of non-optic gliomas: however, this cannot account for the strong decline of OPG prevalence with increasing age seen in our study, as no patients dies from non-optic glioma.
Lastly, tumours might regress spontaneously, as has been described many times for OPG in case reports of children with NF1 (see Additional file
3: Table S3 and [
22‐
24]). Regression seems to occur mostly in tumours involving the optic chiasm, but may also sometimes occur in other sections of the optic pathway. In our study, we identified no instances of complete regression of an OPG but 4 cases of partial regression (see Additional file
1: Table S1). All of the patients in this study who showed tumour regression were under 20 years of age when this occurred. All three of the patients whose MRIs included contrast showed avid enhancement before regression and mild to no enhancement after regression had stopped.
Among the 17 symptomatic OPG patients, 7 were treated. Three of these 7 patients (Patients 5, 13 and 17) received chemotherapy soon after symptom onset, and all showed vision improvement after treatment. Of the remaining 4 patients, the age at symptom onset is unknown for 2 patients, 1 patient received surgery and radiation followed by a decline in vision, and 1 patient received chemotherapy 6 years after symptom onset, followed by stable vision. This may indicate a benefit to starting treatment early after symptom onset; however, our sample is too small to show any definitive benefit. Further research is required to investigate whether early treatment of symptomatic OPG is beneficial in NF1.
Sex has been suggested as a determinant of which NF1 OPGs become symptomatic. Females were reported to receive MRI for visual symptoms significantly more often than males and were 3 times more likely to undergo treatment for visual decline in one study [
25] but not in another [
13]. We did not see a difference in tumour location, tumour frequency symptom status or frequency of treatment initiation between males and females.
We observed a strong association between the presence of OPG and the presence of UBOs after stratification by age (Table
2); this association was also seen when the analysis was restricted to asymptomatic OPGs. Regression analysis showed a similar association between the presence of OPG and UBOs after adjustment for the effect of age (Additional file
2: Table S2).
There are several parallels between UBOs and asymptomatic OPGs: their glial origin, benign nature, usual spontaneous involution, frequent development in early childhood and very infrequent development later in life, and decreased prevalence with increasing age. UBOs are thought to be areas of immature myelin or intramyelinic edema [
11,
26,
27] and
not neoplasms. All studies investigating the histology of OPGs in NF1 patients only include symptomatic patients, as biopsy or surgical removal is not performed in asymptomatic patients. It has generally been assumed that the pathology (and pathogenesis) of symptomatic and asymptomatic OPG is the same in patients with NF1, but there is no direct evidence supporting this assumption. If many asymptomatic OPGs are actually areas of immature myelin instead of true neoplasms, it will change our understanding of NF1 pathology.
Recently, the idea that pediatric gliomas in general are neurodevelopmental disorders has gained traction. Pediatric gliomas vary from their adult counterparts in location (posterior fossa and optic pathway in children, supratentorial compartment in adults), their usual type (low-grade pilocytic astrocytoma in children, high-grade glioblastoma in adults) and their potential for malignant transformation (low in children, high in adults) [
28]. OPGs in NF1 patients are often diagnosed in very young children, and few, if any, cases arise in adults. This stands in contrast to non-optic gliomas in NF1, which
do often arise in adults [
29,
30].
Our study has several limitations. First of all, our study population might not be representative of the NF1 population as a whole. We cannot rule out referral bias or the possibility that symptomatic patients are more likely to consent to participate in an MRI study than asymptomatic patients. Also, patients with more severe phenotypes are likely to receive more frequent clinical follow-up and repeat imaging than patient with less severe manifestations. Another important factor is the diagnosis of OPG based on imaging. There are no generally-accepted diagnostic guidelines for OPGs in NF1 patients, and diagnosis is based on clinical judgment.