To determine the intra- and intervisit reproducibility of circumpapillary retinal nerve fiber layer (RNFL) thickness measures using eye tracking–assisted spectral-domain optical coherence tomography (SD OCT) in children with nonglaucomatous optic neuropathy.
Prospective longitudinal study.
Circumpapillary RNFL thickness measures were acquired with SD OCT using the eye-tracking feature at 2 separate study visits. Children with normal and abnormal vision (visual acuity ≥0.2 logMAR above normal and/or visual field loss) who demonstrated clinical and radiographic stability were enrolled. Intra- and intervisit reproducibility was calculated for the global average and 9 anatomic sectors by calculating the coefficient of variation and intraclass correlation coefficient.
Forty-two subjects (median age 8.6 years, range 3.9–18.2 years) met inclusion criteria and contributed 62 study eyes. Both the abnormal and normal vision cohort demonstrated the lowest intravisit coefficient of variation for the global RNFL thickness. Intervisit reproducibility remained good for those with normal and abnormal vision, although small but statistically significant increases in the coefficient of variation were observed for multiple anatomic sectors in both cohorts. The magnitude of visual acuity loss was significantly associated with the global (ß = 0.026, P < .01) and temporal sector coefficient of variation (ß = 0.099, P < .01).
SD OCT with eye tracking demonstrates highly reproducible RNFL thickness measures. Subjects with vision loss demonstrate greater intra- and intervisit variability than those with normal vision.
Optical coherence tomography (OCT) has become an invaluable and objective tool in the management of optic neuropathies. The evolution from time-domain OCT to spectral-domain OCT (SD OCT) not only has improved image resolution and acquisition time, but now includes mechanisms to improve test-retest reliability. Specifically, some current-generation SD OCT devices provide an eye tracking feature that uses a confocal scanning laser ophthalmoscope or line scanning ophthalmoscope to accommodate for eye movements and that improves the accuracy of subsequently acquired images to be captured in the same anatomic location. The within- and between-visit reproducibility of circumpapillary retinal nerve fiber layer (RNFL) thickness measures have been reported most frequently using the Spectralis OCT (Heidelberg Engineering, Inc, Heidelberg, Germany) and using the eye tracking feature.
Most studies describing the reproducibility of circumpapillary RNFL measures have examined adult subjects with glaucoma and/or healthy controls. Only 1 study has reported the intravisit reproducibility of circumpapillary RNFL thickness measures for children with glaucoma using the Spectralis with eye tracking. While the data presented to date is helpful when examining healthy subjects and adults with glaucoma, it is unclear whether reproducibility measures are comparable in children with optic neuropathies other than glaucoma, since the mechanism and pattern of RNFL loss differ.
The ability of SD OCT to monitor longitudinal circumpapillary RNFL changes in glaucoma and nonglaucomatous optic neuropathies could be especially beneficial in children, since their functional assessment of vision (ie, visual acuity and visual field) relies heavily on their cooperation and cognitive abilities. The goal of the current study is to determine the intra- and intervisit reproducibility of RNFL thickness measurements using eye tracking in children with nonglaucomatous optic neuropathies.
Children being cared for in the Neuro-Ophthalmology clinic at Children’s National Medical Center who underwent SD OCT imaging were eligible for study enrollment. Informed consent was obtained from the parent/legal guardian before study enrollment. The study adhered to the tenets of the Declaration of Helsinki and was approved by the Institutional Review Board at Children’s National Medical Center. All data collected were Health Insurance Portability and Accountability Act compliant.
Subjects determined to have the following nonglaucomatous optic neuropathies were eligible for enrollment: low-grade gliomas intrinsic to the visual pathway (ie, optic pathway gliomas), extrinsic tumors involving the anterior visual pathway (ie, craniopharyngioma, prolactinoma), and demyelinating disease. Healthy children with normal examinations and children with neurofibromatosis type 1 who did not have an optic pathway glioma were eligible to contribute a single eye to the healthy/unaffected cohort, since previous work has shown that RNFL thickness does not differ between these groups. Subjects were included if they met all of the following criteria: (1) 2 separate study visits within 12 months; (2) clinically stable disease without radiographic (ie, tumor growth, new/worsening signal characteristics, or new/worsening contrast enhancement) or clinical progression (ie, >0.1 logMAR worsening in visual acuity or progressive visual field loss); (3) absence of examination findings suggestive of glaucoma; (4) 3 or more acceptable circumpapillary RNFL scans acquired at the initial study visit and a minimum of 1 scan at the follow-up visit; (5) SD OCT scan signal strength >20 db; and (6) acquisition of the entire image volume without appreciable movement or acquisition artifact. Subjects who did not complete a second study visit or experienced clinical progression were excluded from the study. Subjects with tumors that could potentially cause vision loss in 1 or both eyes (ie, optic chiasm or optic tract lesions) who demonstrated clinical progression in only 1 eye were excluded.
Clinical and demographic characteristics were abstracted from the subject’s clinical record. Subjects were classified as having abnormal vision is their visual acuity was ≥0.2 logMAR above normal for age or had visual field loss. The magnitude of visual acuity loss was calculated by subtracting the current logMAR from the normal logMAR for age. When able to cooperate, visual field loss was determined by automated or kinetic perimetry.
Image Acquisition and Analysis
All images were acquired with the Spectralis SD OCT (Heidelberg Engineering GmbH, Heidelberg, Germany) using the “Nsite Analytics” (version 184.108.40.206) and “TruTrack” eye tracking. Once the patient fixates on an internal or external target, allowing the optic nerve to be visualized on the infrared image, this feature permits the operator to “freeze” the image and then manually align the 3.5 mm circular scan over the optic nerve head. If the subject blinks, stops fixating, or moves out of focus, the OCT acquisition will pause until proper alignment is again achieved. RNFL measures were acquired in high-speed mode (768 A-scans) with an automatic real-time (ART) setting of 16. RNFL thickness was measured in 9 sectors: temporal-superior, temporal, temporal-inferior, inferior, nasal-inferior, nasal, nasal-superior, superior, and papillomacular bundle. The global average of all 4 quadrants was also calculated. Scans and their segmentation were reviewed by 1 investigator (R.D.R.) who was masked to all clinical information. Each scan was inspected for segmentation errors and, when necessary, was manually adjusted using the manufacturer-supplied segmentation software. After the first study visit, the scan with the highest-quality score, absent of image artifacts, was chosen as the reference scan for the second-visit acquisitions.
Standard descriptive statistics were used to summarize clinical and demographic characteristics. Coefficient of variation and intraclass correlation coefficient (2-way mixed-effects model) were calculated for the intra- and intervisit analysis. The coefficient of variation, calculated as the standard deviation divided by the mean, represents the variability of a measure. A lower coefficient of variation value represents less variance, suggestive of more consistent measurements, whereas a higher coefficient of variation value represents greater variance among measurements. The intraclass correlation coefficient, another quantitative assessment of reproducibility, calculates the extent of congruence between individuals within a group. A higher intraclass correlation coefficient value represents greater reproducibility. Since this calculation is specific to the study population, intraclass correlation coefficient values cannot be compared across studies. Wilcoxon rank-sum test was used to compare the RNFL thickness, coefficient of variation, and image quality between subjects with and without vision loss. Paired t test using the Bonferroni adjustment for multiple comparisons (significance adjusted to P < .005) was used to compare the intervisit change of RNFL thickness and coefficient of variation within each cohort. Linear regression was used to evaluate the unadjusted and adjusted associations of age and tumor location (intrinsic vs extrinsic tumor) on coefficient of variation measures from the 4 anatomic quadrants (superior, nasal, inferior, temporal) and global average. Since only 1 eligible subject had demyelinating disease, that subject was included in the extrinsic tumor cohort. For subjects with normal vision in both eyes, 1 eye was selected for inclusion using a random number generator. Subjects with abnormal vision in both eyes could contribute 2 eyes, but a generalized estimating equation (GEE) was used instead of standard linear regression to account for the intereye correlation. The GEE model evaluated the unadjusted and adjusted associations of age, magnitude of visual acuity loss, and tumor location (intrinsic tumor vs others) on coefficient of variation measures in the abnormal vision cohort. Subjects with 1 eye with abnormal vision and 1 eye with normal vision could contribute to both vision cohorts.
Forty-two subjects met inclusion criteria and contributed 62 study eyes. Fifteen subjects contributed both eyes to the abnormal vision cohort while 4 subjects contributed 1 eye to the abnormal vision cohort and 1 eye to the normal vision cohort. Twenty subjects contributed 1 eye to the normal vision cohort. Table 1 lists the demographic and clinical characteristics for subjects with and without vision loss. Both groups had a similar interval between study visits.
|Abnormal Vision (N = 22)||Normal Vision (N = 24)|
|Age, y (mean/median)||10.7/10.2||8.5/8.2|
|Female sex, n (%)||14 (64)||16 (67)|
|Race, n (%)|
|White/Caucasian||18 (82)||19 (79)|
|Black/African American||4 (18)||4 (17)|
|Asian||0 (0)||0 (0)|
|Multiple races||0 (0)||1 (04)|
|Ethnicity, n (%)|
|Non-Hispanic||19 (86)||22 (92)|
|Hispanic||3 (14)||2 (8)|
|Time (mo) between visits (mean/median)||4.2/2.8||4.2/3.3|
|Diagnosis, n (%)|
|NF1 with OPG||6 (27)||9|
|Sporadic OPG||11 (50)||6|
|Langerhans cell histiocytosis||1 (5)||–|
|Category of vision loss, (eyes, n)|
|Abnormal visual acuity/normal visual field||5||–|
|Abnormal visual field/normal visual acuity||20||–|
|Abnormal visual acuity and visual field||12||–|
There was no statistical difference in scan quality between vision loss cohorts (Wilcoxon, P > .05). None of the sectors demonstrated a statistical change in RNFL thickness between visit 1 and visit 2 ( Table 2 ).
|Sector b||Abnormal Vision (N = 37) a||Normal Vision (N = 25) a|
|Visit 1||Visit 2||Visit 1||Visit 2|
|G||63 ± 16.7||64 ± 17.5||97 ± 14.9||97 ± 13.8|
|TS||92 ± 24.9||93 ± 25.4||137 ± 25.8||138 ± 24.1|
|T||41 ± 14.4||41 ± 15.6||67 ± 13.3||66 ± 13.0|
|TI||93 ± 36.1||94 ± 36.9||143 ± 23.7||143 ± 22.1|
|I||83 ± 25.2||85 ± 26.2||124 ± 20.4||125 ± 19.7|
|NI||74 ± 22.9||75 ± 24.3||106 ± 21.1||108 ± 22.2|
|N||46 ± 20.7||47 ± 21.1||73 ± 19.4||73 ± 19.1|
|NS||74 ± 31.2||76 ± 31.9||109 ± 26.5||109 ± 25.5|
|S||83 ± 23.1||85 ± 23.8||123 ± 21.5||123 ± 19.4|
|PMB||32 ± 16.6||32 ± 17.8||49 ± 9.6||49 ± 9.5|
For the intravisit analysis, both the abnormal and normal vision cohort demonstrated the lowest intraclass correlation coefficient (.961 and .992, respectively) and the highest coefficient of variation (6.2% and 2.2%, respectively) in the papillomacular bundle ( Table 3 ). Similarly, the abnormal and normal vision cohort demonstrated the lowest coefficient of variation (1.5% and 0.7%, respectively) for the global thickness. The normal vision cohort demonstrated statistically lower coefficient of variation measures compared to the abnormal vision cohort in the global, temporal, temporal-inferior, and nasal sectors ( Table 3 ).
|Sector||Abnormal Vision (N = 37) a||Normal Vision (N = 25) a||P Value|
|ICC (95% CI)||CV (%)||ICC (95% CI)||CV (%)|
|G||.998 (.99, .99)||1.5||.998 (.99, .99)||0.7||.04|
|TS||.994 (.98, .99)||2.4||.995 (.99, .99)||1.4||N/S|
|T||.984 (.97, .99)||3.1||.997 (.99, .99)||1.2||.03|
|TI||.998 (.99, .99)||2.5||.998 (.99, .99)||1.0||.01|
|I||.996 (.99, .99)||2.0||.998 (.99, .99)||1.0||N/S|
|NI||.998 (.99, .99)||2.1||.997 (.99, .99)||1.4||N/S|
|N||.997 (.99, .99)||3.7||.997 (.99, .99)||1.4||<.01|
|NS||.997 (.99, .99)||3.1||.995 (.99, .99)||2.0||N/S|
|S||.997 (.99, .99)||1.9||.995 (.99, .99)||1.4||N/S|
|PMB||.961 (.93, .97)||6.2||.992 (.98, .99)||2.2||N/S|
Both the abnormal and normal vision cohorts demonstrated lower intraclass correlation coefficient and slightly higher coefficient of variation values in the intervisit analysis ( Table 4 ). In the abnormal vision cohort, the lowest coefficient of variation was again demonstrated in the global measure (2.3%) and was highest in the papillomacular bundle (6.6%). For the normal vision cohort, the global measure also had the lowest coefficient of variation (1.3%), but the nasal and nasal-superior sectors (3.5% and 3.6%) demonstrated the highest coefficient of variation. The normal vision cohort demonstrated statistically lower coefficient of variation measures compared to the abnormal vision cohort in the global, temporal, temporal-inferior, nasal-inferior, nasal, and papillomacular bundle sectors ( Table 4 ).