Subjects
This study included 135 pediatric subjects aged 4–14 years who presented to the Department of Ophthalmology at Eye & ENT Hospital of Fudan University from Jan 2021 to Jan 2022. Subjects were assigned into three groups according to their clinical diagnosis, including the amblyopia group (n = 45), recovered amblyopia group (n = 45), and control group (n = 45). All the subjects were Chinese. The study adhered to the tenets of the Declaration of Helsinki and was approved by the institutional review board of the Eye & ENT Hospital of Fudan University. Written informed consent was obtained from the parents/guardians of the subjects and assent was obtained from children ≥ 7 years of age prior to their participation.
All the subjects underwent a comprehensive eye examination, including visual acuity with the Standard Logarithm Visual Acuity Chart (arithmetic scaled high-contrast E optotype; the only type of chart available to us at the clinic), refractive errors with a cycloplegic refraction (1% cyclopentolate), ocular alignment with a simultaneous prism cover test and a prism and alternate cover test, anterior segment examination with the slit lamp, stereopsis with the Titmus Stereo Test (Stereo Optical Co, Inc), fundus examination, and eye movement functions (before cycloplegia) with the binocular paradigm proposed in this study. For data analysis, refractive errors were converted to spherical equivalent (SE), the sum of the spherical power and half of the cylindrical power. Best-corrected visual acuity (BCVA) was converted to logarithm of the minimal angle of resolution (logMAR) and approximate Snellen equivalent was provided. Anisometropia was defined as an interocular difference in SE of 1.00 diopters [D] or more. Refractive error was defined by the SE in the more ametropic eye (myopia < -0.50 D; emmetropia within ± 0.50 D; hyperopia > + 0.50 D).
Subjects were included only if they had no history of ocular trauma and/or ocular pathology (e.g., nystagmus, cataract, ptosis), no systemic disease (by-self report), no history of intraocular surgery, no measurable strabismus (≤ 5 PD at 6 m and 33 cm fixation with/without spectacle correction), and sufficient cooperation with the examinations. The inclusion criteria of each group were as follows:
The amblyopia group
Diagnosis of amblyopia at the most recent visit or unrecovered amblyopia with treatment less than 1 year, with BCVA of 20/30 or worse (20/50 for age 4 years; 20/40 for age 4 to ≤ 5 years) in the worse eye or an interocular difference in BCVA of two lines or more (≥ 0.2 logMAR) (according to the Amblyopia “PPP” guideline, 2017 [12]). Amblyopia associated with deprivation or uncorrected strabismus (> 5 PD at distance and/or near fixation) was excluded. BCVA in AE was used to classify the severity of amblyopia into mild to moderate (20/32–20/80) and severe (20/100 or worse) amblyopia.
The recovered amblyopia group
A history of amblyopia, with resolved visual acuity at the most recent visit after amblyopia treatment.
The control group
Normal or corrected-to-normal visual acuity in both eyes, stereoacuity ≥ 60 arc seconds, refractive errors within ± 6.00D sphere and ± 1.00D cylinder, absence or presence of anisometropia, and no history of amblyopia or other ocular diseases.
Eye movement assessment
Apparatus
The experiment took place in a quiet and private room with a natural and constant luminance. A 32-inch 3-dimension (3D) monitor (resolution 1920 × 1080 pixels at a refresh frequency of 120 Hz; LG Electronics, Seoul, Korea) was used to present stimuli at a viewing distance of 80 cm. Subjects were asked to wear 3D polarized glasses with spectacle correction (if any) and seated on a non-wheeled but height-adjustable chair with the eyes at the same level as the screen center. Gaze positions were recorded with a 120 Hz remote eye tracker (Tobii Eye Tracker 5). The presentation of stimuli was generated by MATLAB (MathWorks, Natick, MA).
Calibration
The subjects were briefly familiarized with the procedure by the experimenter, and were asked to adjust and maintain the head position until the eye tracker could catch his/her both eyes optimally. A 3-point (X, Y = 0°, + 13.5°; -13.5°, -13.5°; + 13.5°, -13.5°; presented for 4 s at each location) calibration and validation of the eye tracker was run at the beginning of the main experiment and whenever necessary during the experiment. The subjects were asked to fixate their gaze on the calibration stimulus, a bright blue dot on a black background which was dynamically shrinking from a normal size (diameter 0.3°), and the binocular data was collected at the moment the dot disappeared. The subjects did not have to keep the head completely still during calibration as long as their eyes remained focused on the stimulus, since the eye tracker was able to track and correct for head movements simultaneously. The gaze positions were measured as separate horizontal and vertical components by the eye tracker, simultaneously and respectively for both eyes. Blinks or partial blinks were automatically detected and removed from analysis. The following main experiment would be initiated when calibration and its subsequent validation were acceptable with adequate accuracy for each eye.
Sustained fixation test
The sustained fixation test measured the deviation of sustained fixations under static binocular-viewing condition. The subjects were instructed to fixate their gaze on a target on a black background, which was a bright blue dot (1.4° diameter) with a black cross-shaped center. The target appeared in a fixed order at 9 locations, 8 locations on a peripherical circle (8.3° radius) and 1 location in the center (Fig. 1a). It remained for 3 s on each location and automatically switched to the next location; however, data recording at each location was manually started by the experimenter until ensuring that the subject’s fixation had changed and sustained on the target. Fixation was defined by an oculomotor behavior shown under the effort to maintain the gaze in a predefined region [13]. Gaze positions after the manual start of data recording were considered fixations (the initiation phase of fixation [14] was therefore removed from analysis). At each target location, the eye tracker cumulatively extracted 5 samples from all gaze positions it grabbed. The mean horizontal and vertical deviations of these 5 gaze positions relative to the target location were calculated for each eye individually and recorded as the binocular mean values.
Visually guided saccade test
The visually guided saccade test measured the deviation of post-saccadic fixations under dynamic binocular-viewing condition. A bright blue cross-shaped target (1.4° diameter) appeared at 8 locations on a black background in turn, ± 17.8° horizontally alternately and + 8.3°, + 2.8°, -2.8°, -8.3° vertically sequentially, requiring a wide-range ocular motion (saccade) and succeeding fixation (Fig. 1b). The target was presented for 3 s at each location and then switched to the next location, with data recording started automatically. The horizontal and vertical gaze positions were measured as the subject attempted to fixate the gaze on the target after saccades. Considering saccade latency and fixation initiation [14] during the fixation alternations, the eye tracker only grabbed gaze positions from 800 to 1800 ms and extracted 10 samples at each location. Likewise, at each target location, the mean horizontal and vertical deviations of these 10 gaze positions relative to the target location were calculated for each eye individually and recorded as the binocular mean values.
Subjects were asked to repeat the experiment after 20 min to validate the stability of this method and to rule out potential learning effects. In addition, since the results were displayed on the screen immediately at the end of each test, a re-test was required once the results seemed unreliable due to the subject’s distraction during the test.
Statistical analysis
The horizontal and vertical deviations (°) in the two tests were calculated for each subject by the equations below.
$$Horizontal\ deviation=\frac{1}{k}\sum_{1\le j\le k} \frac{1}{\mathrm{n}}\sum_{1\le i\le n}\frac{\left(\Delta {\mathrm{X}}_{Left}+\Delta {\mathrm{X}}_{Right}\right)}{2}$$
$$Vertical\ deviation=\frac{1}{k}\sum_{1\le j\le k} \frac{1}{\mathrm{n}}\sum_{1\le i\le n}\frac{\left(\Delta {\mathrm{Y}}_{Left}+\Delta {\mathrm{Y}}_{Right}\right)}{2}$$
In the equations, the parameter n was the number of gaze positions extracted at each target location, and the parameter k was the number of predefined locations where the target would appear during the test. As described above, n and k were 5 and 9 in the sustained fixation test, and 10 and 8 in the visually guided saccade test, respectively. ΔX and ΔY represented the horizontal and vertical deviations of each gaze position extracted relative to the target location, with subscripts Left and Right representing the left and right eye. The mean horizontal and vertical deviations from all target locations in each test were finally calculated as the main outcome measures.
For simplicity, the horizontal and vertical deviation were abbreviated as Fix-X and Fix-Y in the sustained fixation test, and Sac-X and Sac-Y in the visually guided saccade test, respectively. Small fixation deviations in the tests indicated accurate and controllable eye movements.
Normally distributed continuous data were presented as means with standard deviations (SD). Abnormally distributed continuous data were presented as median (25th percentile [P25], 75th percentile [P75]). Categorical variables were described as frequency counts and proportions (%). Group characteristics were compared using one-way analysis of variance (ANOVA) for normal-distributed continuous variables, nonparametric Kruskal–Wallis test for abnormal-distributed continuous variables, and Pearson’s χ2 test or Fisher’s exact test for categorical variables.
Since deviations were distributed right-skewed, Kruskal–Wallis test was used to compare the four deviations across the three groups, and pairwise comparisons were performed using Dunn test with the Bonferroni correction for multiple comparisons. Wilcoxon signed-rank test was used to compare repeated measures of deviations. Scheirer-Ray-Hare test, the non-parametric equivalent of ANOVA, was used to detect the interaction between subject groups and other clinical characteristics (age, refractive status, and anisometropia). Spearman’s rank correlation and Mann–Whitney U test was used for associations between clinical characteristics and deviations in the amblyopia group. P values less than 0.05 (two-sided) were considered statistically significant. Analyses were performed using the open-source statistical software R version 4.1.3 (R Foundation).