IMAGING AND BASELINE EXAMINATIONS

Stereoscopic optic disc photographs were obtained by trained technicians using a digital fundus camera (Visucam 500, Carl Zeiss Meditec Inc). Each optic disc photograph was evaluated independently by 2 glaucoma experts (H.L.R. and N.K.P.) in a masked manner to determine the presence of glaucomatous changes (focal or diffuse neuroretinal rim thinning, localized notching, or RNFL defects) and the presence of disc hemorrhage (DH). The experts were masked to all of the clinical data, VF data, and the fellow eye data. Discrepancy in the classification between the 2 experts was adjudicated by a third glaucoma expert (Z.S.P.).

VF examination was performed using the Humphrey Field analyzer 3 (model 860, Carl Zeiss Meditec Inc) and the Swedish interactive threshold algorithm (SITA) standard 24-2 program. VFs were considered reliable if the fixation losses were <20% and the false-positive and false-negative response rates were <15%. The VF was considered glaucomatous if the glaucoma hemifield test result was outside normal limits, pattern SD was abnormal at the P < 0.05 level, or ≥3 test points in a cluster on pattern deviation probability plot were abnormal at P < 0.05 with at least 1 point abnormal at P < 0.01.

Visual sensitivity loss quantified by mean deviation (MD) was used to determine the severity of functional damage. In addition to the overall MD of the field, MD in both the superior and inferior hemifields, as provided on the Glaucoma Workplace 3.5 software (Carl Zeiss Meditec Inc), were used in the current analysis. Superior and inferior VF hemifields were evaluated separately for inclusion, and VF hemifields with severe glaucoma (defined as a hemifield MD of worse than −12 dB) were excluded to ensure the ability to detect progressive RNFL thinning without reaching the measurement floor.

OCT scanning of all patients was performed using the optic disc cube 200 × 200 and macula cube 200 × 200 scan protocols. These scan protocols have been explained in detail previously. From the optic disc cube scans, RNFL thickness was calculated along a circle 3.46 mm in diameter positioned evenly around the center of the optic disc. Average RNFL thickness over the entire circle as well as the 4 quadrants (temporal, superior, nasal, and inferior) of 90° each were determined. In the current study, RNFL thickness in the superior and inferior 90° quadrants were analyzed because these are the quadrants that demonstrate rapid progression.

From the macula cube scans, ganglion cell-inner plexiform layer (GCIPL) thickness measurements were calculated within a 14.13-mm elliptical annulus centered on the fovea with an inner vertical radius of 0.5 mm and outer vertical radius of 2 mm, stretched horizontally by 20%. The thickness parameters derived from the macula scan were the average GCIPL thickness across the entire elliptical annulus and the thickness at six 60° sectors of the elliptical annulus. The current study analyzed GCIPL thickness in the superior and inferior 60° sectors.

OCTA of the peripapillary and the macular region was performed by trained technicians at the baseline visit using Cirrus HD-OCT 10.0.0.14618 software. The procedure of OCTA imaging with Cirrus HD-OCT has been detailed previously. OMAG is the algorithm used to achieve blood vessel delineation on Cirrus HD-OCT. OMAG uses both the intensity and phase information from B scans repeated at the same position to delineate blood vessels. The peripapillary and macular regions were imaged using a 6- × 6-mm cube scan centered on the optic disc and the macula, respectively, as described previously. ^, The 6- × 6-mm scan pattern has 350 A-scans in each B-scan along both the horizontal and the vertical directions. Each B-scan is repeated 2 times in the 6- × 6-mm scan. The manufacturer’s retinal tracking technology was used to reduce motion artifacts. From the volume scans, retina and choroid were segmented into multiple slabs, and 2-dimensional angiographic images of each slab were generated.

The current study analyzed the angiographic images of the superficial peripapillary and macular slabs. The inner boundary of the superficial retinal slab is the internal limiting membrane (ILM), and the outer boundary is the inner plexiform layer (IPL). The Cirrus HD-OCT angiometric software automatically calculates 2 parameters from the superficial retinal layer slab. Vessel density (VD), one of them, is defined as the total length of perfused vasculature per unit area in the region of measurement (measured as mm/mm ² ). Perfusion density (PD), the other one, is defined as the total area of perfused vasculature per unit area in the region of measurement (measured as percentage). The angiometric software, primarily designed to be used at the macula, calculates the VD and PD parameters in the various sectors of the Early Treatment Diabetic Retinopathy Study (ETDRS) grid placed over the macula ( Figure 1 , A). The same grid was placed on the peripapillary scan and manually centered on the optic disc as shown in Figure 1 , B. Angiometric parameters measured on the 4 outer sectors of the grid (ie, between the outer 2 circles [temporal, superior, nasal, and inferior sectors]) were determined. In the current study, the superior and inferior outer quadrant (each of 90°) OMAG measurements of the peripapillary and macular scans were analyzed.

Optical microangiography scan of the (A) macula and (B) optic disc shows the superficial layer vasculature. The images also show the Early Treatment Diabetic Retinopathy Study grid centered on the macula and disc and the perfusion density measurements calculated in each sector of the grid.

The dependent variable in the current analysis was the rate of change of quadrant RNFL thickness over time (defined in the Statistical Analysis) measured on the serial optic disc cube scans. The follow-up optic disc cube scans were referenced to the baseline scan using the manufacturer’s “track to prior scan” option from the time this feature was available. When the “track to prior scan” option is selected, the previously saved scanning laser ophthalmoscopy fundus image of the baseline scan is overlaid in the scan pattern box over the live scanning laser ophthalmoscopy fundus image matching for the blood vessel branchings, which allows the repeat scans to be tracked and acquired exactly on the baseline scan.

All of the baseline examinations for a particular individual were performed on the same day. Image quality was assessed for all OCT and OMAG scans. Poor-quality images, defined as those with a signal strength of <6, and images with motion artifacts and segmentation errors were excluded from the analysis. Also excluded were OMAG images with poor clarity or local weak signal.

STATISTICAL ANALYSIS

Descriptive statistics included mean and SD for continuous variables and percentages for categorical variables. The effect of clinical, VF, OCT, and OMAG parameters on rate of change of quadrant RNFL thickness (RNFL slope) was evaluated using linear mixed models with random intercepts and random slopes. ^, In this model, the change in the outcome variable (quadrant RNFL thickness) was explored using a linear function of time, and random intercepts and random slopes introduced patient-, eye-, and quadrant-specific deviations from the average value. The model accounts for the fact that different quadrants and eyes can have different RNFL slopes over the follow-up period, while accommodating correlations between the 2 quadrants and/or both eyes of the same individual. ^,

Because rate of RNFL change may depend on the disease severity, an unstructured covariance between random effects was assumed, allowing for correlation between intercepts and slopes. Effects of predictor variables were assessed on the baseline RNFL thickness (baseline severity) and on the change in RNFL thickness over time by introducing interaction terms between time and predictor variables. The clinical parameters (predictors) investigated for their association with baseline RNFL thickness and rate of RNFL thickness change were age, sex, presence of hypertension, diabetes, central corneal thickness (CCT), presence of DH, follow-up duration, mean IOP, and the IOP fluctuation (SD of IOP) during the follow-up. The baseline VF, OCT, and OMAG predictors investigated were the hemifield MD, quadrant RNFL thickness, sector GCIPL thickness, peripapillary PD and VD, and macular PD and VD.

Univariate models were constructed containing 1 predictor along with its interaction with time. Predictors associated with the rate of RNFL change at P < .10 in univariate analysis were introduced into multivariate analysis. Collinearity between predictor variables were evaluated, and predictors correlated with each other (correlation coefficient of >0.60) were evaluated in separate multivariate models. Rates of RNFL change were obtained from the linear mixed models using best linear unbiased prediction (BLUP). ^, Statistical analyses were performed using Stata 14.2 software (StataCorp LLC). A P value of ≤.05 was considered statistically significant for the final analysis.