Patients

A cross-sectional convenience sample with prospective data collection identified candidate subjects between 6 and 21 years of age during their routine clinical visits to the neuro-ophthalmology or neuro-oncology clinics at Children’s Hospital of Philadelphia from July 2009 through January 2010. Children between 6 and 17 years of age required parental or guardian informed consent and, when appropriate, child assent before study enrollment. Participants 18 to 21 years of age provided their own consent. Children in the NF1-associated optic pathway gliomas group were required to have both a diagnosis of NF1 based on established National Institutes of Health criteria and magnetic resonance imaging (MRI) results of the brain demonstrating the presence of optic pathway gliomas. Children in the sporadic optic pathway glioma group required documented MRI findings characteristic of low-grade optic pathway gliomas or diagnostic biopsy results. For purposes of descriptive comparison and a secondary analysis, children with a diagnosis of NF1 but without optic pathway gliomas also were enrolled. Patients were excluded if they had a history of ophthalmologic or neurologic disease, other than optic pathway gliomas, that could have affected their VA or their optic nerve function (eg, amblyopia, cataracts, glaucoma, retinopathy of prematurity, or elevated intracranial pressure requiring ventriculoperitoneal shunting). Healthy control subjects with no significant history of ophthalmologic or neurologic disease were recruited. Institutional review board approval of the protocol and an informed consent form were obtained before study initiation.

Using a standardized form, demographic and clinical data were collected from each patient’s clinical chart including: presence of NF1, history of optic pathway gliomas treatment (ie, chemotherapy or radiation), and VF testing results. Findings for the most recent MRI were abstracted from the formal reading and were classified as follows: no evidence of optic pathway glioma (tortuous optic nerves allowed), optic pathway glioma involving only the optic nerve, and optic pathway glioma involving the optic chiasm and posterior structures, including the hypothalamus.

Neuro-ophthalmologic Examination

Best-corrected VA was determined using the electronic visual acuity (EVA) system for pediatric patients. The subject was seated 3 m from the computer monitor in a windowless room. The examination chair was elevated to a level where the subject’s eyes were parallel to the center of the computer monitor. Each eye was tested separately, beginning with the right eye. First, the need for refraction was assessed. Optotypes equivalent to 20/16 were displayed. Children not able to read the 20/16 line accurately were refracted with trial lenses according to a standard protocol. If a refraction had been performed within the past 6 months, this was the subject’s starting refraction. High-contrast VA was tested according to the Amblyopia Treatment Study VA testing protocol with the EVA monitor placed at 3 m. The EVA computer monitor was calibrated for appropriate luminance before each testing session. Single-letter (H, O, T, or V) crowded optotypes were presented in 4 phases: screening, phase 1 (first threshold determination), reinforcement, and phase 2 (second threshold determination). Final VA was the smallest logarithm of the minimal angle of resolution (logMAR) level passed in phase 1 or phase 2.

Low-contrast letter acuity testing of each eye was performed in the same windowless examination room with the lights turned off using a retroilluminated cabinet. Sloan low-contrast letter acuity charts at 5%, 2.5%, and 1.25% contrast level (Precision Vision, LaSalle, Illinois, USA) were presented sequentially at 2 m. Low-contrast letter acuity score (logMAR) for each eye was determined to be the lowest line for which the subject was able to identify correctly at least 3 of the 5 letters.

Definition of Clinical Outcomes

High-contrast VA testing was used to classify eyes as having normal VA (logMAR, ≤ 0.1; 20/25 or better) or abnormal VA (logMAR, > 0.1). VF loss, determined by one coinvestigator (G.T.L.), was defined as any noncentral scotoma (ie, quadrantanopia, hemianopia, VF constriction, arcuate defect, or altitudinal defect) that was detected reliably with automated perimetry or confrontation testing. Our younger subjects, especially those with NF1, are not able to complete automated perimetry reliably; therefore, the VF results were not quantified, but rather categorized as either present or absent. Type of vision loss was classified into categories: normal VA and normal VF results, abnormal VA and normal VF results, normal VA and abnormal VF results, and abnormal VA and abnormal VF results. Patients with unilateral glioma of the optic nerve contributed that eye as an optic pathway glioma eye, whereas their fellow eye was classified as not having an optic pathway glioma. Both eyes were classified as having optic pathway glioma when the optic pathway glioma was present in the optic chiasm, posterior structures, or both.

Optical Coherence Tomography Imaging

RNFL thickness (in micrometers) of each eye was measured using the fast RNFL thickness protocol with a Stratus OCT (Carl Zeiss Meditec, Dublin, California, USA). Scan quality was optimized before imaging each eye by adjusting the polarization. Scanning commenced when the 3.4-mm diameter circle was centered over the optic nerve head as the subject fixated on an external or internal fixation light. Scans with a signal strength of less than 6 were discarded. Three scans were averaged to produce anatomic quadrant and average RNFL thickness in micrometers. If the pupil diameter was not sufficient to obtain OCT images successfully, mydriatic eye drops (1% tropicamide and 2.5% phenylephrine hydrochloride) were used.

Statistical Analysis

Demographic and clinical characteristics were summarized by standard descriptive summaries (eg, means and standard deviations for continuous variables such as age and percentages for categorical variables such as gender). RNFL thickness was categorized into quartiles from the optic pathway glioma patients resulting from the nonlinear association between RNFL thickness and the measures of vision. RNFL measurements from healthy controls and NF1 patients without an optic pathway glioma were not used to calculate the quartiles and were presented for descriptive comparison.

A one-way analysis of variance was used to determine if high-contrast VA (logMAR) and low-contrast letter acuity (logMAR) were associated with RNFL thickness quartiles in patients with optic pathway gliomas. The generalized estimating equation approach to variance estimation was used to account for the correlation between eyes of patients and to provide a robust estimator to accommodate for unequal variance in vision measures among RNFL thickness groups. Similarly, the generalized estimating equation approach also was used in multiple regression analyses to examine the influence of RNFL thickness quartiles, patient age, and tumor location on both high-contrast VA and low-contrast letter acuity in patients with optic pathway gliomas.

Because we were not aware of any previously published studies of RNFL thickness in children with optic pathway gliomas, sample size was calculated using RNFL data from studies of optic neuritis in multiple sclerosis. RNFL thickness differs by roughly 20% between patients with optic neuritis (80 ± 20 μm) and healthy controls (103 ± 12 μm). Setting the α value to 0.01 (based on multiple comparisons) to achieve a power of 0.9 necessitates recruitment of a total of 32 eyes (16 eyes with vision loss and 16 eyes without vision loss). The multivariate model contained 4 variables, and thus required a minimum of 80 study eyes (20 eyes per variable) to achieve sufficient precision.