Natural History Study of Retinal Structure, Progression, and Symmetry Using Ellipzoid Zone Metrics in RPGR-Associated Retinopathy

Purpose This is a quantitative study of retinal structure, progression rates, and interocular symmetry in retinitis pigmentosa GTPase regulator gene (RPGR)-associated retinopathy using spectral-domain optical coherence tomography (OCT). Design Prospective, observational cohort study. Methods Thirty-eight subjects at Moorfields Eye Hospital in London were assessed with 2 spectral-domain OCT–derived ellipzoid zone (EZ) metrics with repeatability assessments. EZ width (EZW) measurements were made on transfoveal line scans. En face images of the EZ area (EZA) were generated from high-density macular volume scans and were quantified. Baseline size, progression rate, symmetry, associations with age and genotype, and baseline structure–function correlation were investigated. Results Baseline EZW and EZA measurements were 1963.6 μm and 3.70 mm2, respectively. The mean EZW progression rate was 233.6 μm per year, and the mean EZA rate was 0.67 mm2 per year. Relative interocular difference as an index of symmetry was 3% for both metrics, indicating good baseline symmetry in general—although significant variation existed across the cohort. Analysis of variance found a significant effect of age but not genotype on EZ dimension and progression rates. Larger EZ dimension and greater progression were seen in younger subjects. A positive correlation between EZ dimension and progression was evident. Overall exponential decline rates of 8.2% with EZW and 15.5% with EZA were obtained. Good functional correlation was found with EZW demonstrating stronger correlation; however, EZA correlation with function was also significant. Conclusions EZ metrics are sensitive structural biomarkers for measuring residual extent and progression in RPGR-associated retinopathy. Our elucidation of the natural history will provide clinicians and patients with more knowledge about the condition and inform the design and interpretation of interventional trials.

Spectral-domain optical coherence tomography (OCT) is in widespread use as an imaging modality to study retinal structure in a myriad of diseases. Previous OCT studies in RP have reported structural changes occurring at the transition zone as the scanned region of interest traverses from healthy central retinal tissue to diseased periphery. 8,9 Structural measurements of ellipzoid zone (EZ) width in RP, as a metric of disease severity and progression, have been shown to correlate well with retinal function. [10][11][12][13] Serial measurements of EZ width (EZW) to assess progression in XLRP have been studied [14][15][16][17] ; however, these studies were potentially limited by bias in eye selection, [14][15][16] do not distinguish between genetic causes of XLRP, 14,15 or are retrospective. 17 The quantification of ellipzoid zone area (EZA), made possible with the use of vendor software to construct en face images from spectral-domain OCT volume scans, has been shown to be feasible in quantifying progression in autosomal recessive (AR) RP. 18 Despite this, there have not been further studies on the use of EZA as a structural metric in RP.
We therefore investigated the following in this protocoldriven prospective spectral-domain OCT study comprised solely of RP subjects with molecularly confirmed pathogenic RPGR mutations: (1) intraobserver repeatability with EZW and EZA metrics; (2) characterize baseline retinal structure with both metrics; (3) characterize progression with both metrics; (4) characterize interocular symmetry at baseline and progression and establish indices to quantify symmetry with both metrics; (5) investigate correlations between baseline measurements, progression, and age; (6) investigate effects of age and genotype on baseline and progression; (7) determine overall exponential rates of progression with both metrics; and (8) investigate structure-function correlations at baseline with both EZW and EZA metrics.

APPROVAL WAS GRANTED BY THE ETHICS COMMITTEE AT
Moorfields Eye Hospital before conducting the study. The Declaration of Helsinki was adhered to throughout. All subjects were affected males with RP caused by molecularly confirmed disease-causing mutations in RPGR. Bidirectional sequencing to test for mutations in RPGR exons 1 to 14 and open reading frame 15 (ORF15) was performed at the Central Manchester University Hospitals Genomic Diagnostics Laboratory (Manchester, UK) before recruitment.
OCT ACQUISITION AND ANALYSIS: Dedicated clinical research ophthalmic technicians acquired protocol-driven OCT scans from both eyes of each subject at each visit, with the Spectralis OCT device (Heidelberg Engineering, Heidelberg, Germany). Horizontal high-resolution transfoveal line scans were obtained using automated real-time tracking (ART) with an average of 100 images. Automatic registration was used for all follow-up scans to ensure accurate correspondence of retinal locations. After line scan acquisition, bilateral macular high-resolution volume scans were acquired at the same visit. Each volume scan was comprised of 193 horizontal b-scans in high resolution mode with 30 mm distance between b-scans. Volume scans were purposefully acquired to allow the creation of en face images of the ellipzoid zone area (EZA). After a departmental upgrade to the Spectralis OCT imaging platform with OCT2 module (Heidelberg Engineering), volume scans acquired after June 2016 were captured with an average of 12 images (ART12) per horizontal b-scan, as allowed for by increased hardware capabilities and faster scanning speed with significant reduction in acquisition time.
Image analyses for both EZW and EZA metrics were performed using vendor software (Heidelberg Eye Explorer version 1.10.2.0; Heidelberg Engineering). EZW analyses were conducted on transfoveal line scan images with the following methodology: images were displayed in a 1:1 mm setting. The nasal and temporal extents of the EZ (point of EZ disappearance into the proximal retinal pigment epithelium border with loss of outer segment layer) were identified and marked with the arrow tool. EZW was measured with the caliper tool as a straight line tangential to the distal retinal pigment epithelium border ( Figure 1). Line scan images were analyzed in a random order for each subject.
En face images of the EZA were created from each macular volume scan with the following method using vendor software: images were created in a 3-dimensional mode and displayed in transverse view. The autosegmented Bruch membrane line was manually inspected for accurate placement along the outer border of the retinal pigment epithelium-Bruch membrane complex as this line was

VOL. 198
utilized as a reference for the slab contour. The reference line was displaced 25 mm inward from the Bruch membrane and a slab of 30 mm thickness was created. Slab settings were designed to capture the whole extent of the EZ layer in subjects. After image creation, the EZA was delineated and measured using minimum intensity projection with area value provided by the vendor software ( Figure 2).
STATISTICS: For the assessment of intraobserver repeatability, EZW and EZA baseline images of each eye were measured twice by a single observer, a minimum of 1 week apart. Intraobserver repeatability assessment was also conducted on EZA derived from ART12 images (EZA-ART12) acquired after the upgrade to OCT2. One image per eye per subject was used to maintain the ''independence-of-score'' and to avoid inducing systematic bias. The method popularized by Bland and Altman was used to calculate mean 6 standard deviation (SD) of testretest differences and corresponding 95% limits of agreement (LOA). Residuals were inspected for normality. Repeatability coefficient (1.96 3 SD) and the test-retest variability for both metrics were calculated and these are shown in Table 1.
The Bland-Altman method was further used to assess interocular symmetry at baseline and progression as observed with both metrics ( Table 2). An example is provided in Figure 3. Baseline EZW and EZA, EZW % rate, and EZA % rate residuals were inspected for normality. The mean 6 SD of interocular difference, together with 95% LOA, was calculated. Interocular differences were obtained from the subtraction of left from right eyes. The mean of interocular differences was expressed as a fraction of its corresponding cohort average (the cohort average in turn was calculated by obtaining the average of all interocular [mean of right and left eye] values of subjects), and this was presented as the relative interocular difference (RID). The interocular coefficient (1.96 3 SD), expressed as a fraction of the corresponding cohort average, was calculated and presented as the relative interocular variability (RIV). These indices were calculated to facilitate metric cross-comparison.
Progression rates for individual subjects were calculated using the following method: for each eye of each subject, EZ measurements obtained with each metric were plotted as a function of age on separate scatterplots. As shown in Figure 4, linear trendlines were fitted to data points using the least squares method in Excel (version 15.41; Microsoft, Inc., Redmond, WA, USA). Progression rates for each eye of each subject were obtained from trend line gradients.
Additional statistical analyses were performed with XLSTAT software (version 2018.1; Addinsoft, New York, New York, USA). Data were inspected for normality and log transformation performed before conducting tests of statistical analyses where required. Interocular correlations at baseline and progression were investigated with The number of eyes in the cohort for characterization of baseline EZW were 76 eyes of 38 subjects, baseline EZA with 62 eyes of 31 subjects, EZW rate with 56 eyes of 28 subjects, and EZA rate with 42 eyes of 21 subjects. Positive progression rates indicate constriction occurring over time.
the Pearson correlation coefficient (Table 3). Following this, right and left eye data were combined to investigate any correlation between age and baseline values, age and progression, and correlation between progression and baseline values ( Table 3). The effects of age and mutation on baseline and progression were assessed with a 2-way analysis of variance (ANOVA), as shown in Table 4. Subject age was calculated from birth to time at baseline visit and was further divided into 5 categories: category 1, <10 years of age; category 2, 10 to <15 years of age; category 3, 15 to <20 years of age; category 4, 20 to <25 years of age; and category 5, > _25 years of age. For the assessment of genotype, subjects were categorized into groups based on predicted effects of mutations: those will null allele mutations (premature stop codons or frameshifts leading to premature stop codons in exons [1][2][3][4][5][6][7][8][9][10][11][12][13][14] or those with mutations that were likely to result in translation of a variant protein product (missense mutations and mutations in ORF15). Splice-site mutations were separately grouped because of the unpredictability of their effects. 19 In instances of a significant ANOVA result, post hoc multiple pairwise comparisons were conducted with Tukey test.
Overall rates of progression for both metrics were modelled using a mixed-models method (Table 5). Analysis was performed with age (calculated from birth to time of OCT image acquisition) designated as a fixed effects quantitative explanatory variable. Each eye of each subject was designated as a random effects variable. Metric values were designated as dependent variables. Values were converted into natural log form to model an exponential decline. Previous studies have shown progression to be well characterized with an exponential decay model, [20][21][22] and this is further supported by evidence of exponential photoreceptor degeneration in animal models 23 and by inspection of our data ( Figure 5). Distribution of model residuals were inspected for normality.
The association between structure and function was investigated. EZW and EZA data as a representation of baseline structure were correlated with data obtained from tests of visual function performed at the same visit where available. Assessments of best corrected visual acuity (BCVA) were conducted with the Early Treatment Diabetic Retinopathy Study chart and contrast sensitivity (CS) assessments with the Pelli-Robson chart. BCVA was recorded in logarithm of minimum angle of resolution (logMAR) units and CS as logCS units. Automated static perimetry testing was performed on the Octopus 900 (Haag-Streit AG, Köniz, Switzerland) using a customized, radially oriented 185-point grid. Perimetry mean sensitivity (MS) values in decibel units were obtained from vendor software. The volumetric measures of V Total and V 30 in decibel-steradian units were obtained after analysis of perimetry data with third-party software (Visual Field Modeling and Analysis software). In essence, V Total and V 30 are metrics that characterize the total amount of sensitivity in the hill-of-vision as defined by the entire test grid (V Total ) or that which is contained within a central circle of a 30 8 radius (V 30 ). A comprehensive description was published by Tee and associates. 24 Significance level alpha for all statistical tests was set at 0.025 after Bonferroni correction for simultaneous analysis on 2 metrics (EZW and EZA).

ALL 38 SUBJECTS IN THIS STUDY POSSESSED BILATERAL EZS
visible at the time of baseline imaging. Mean 6 SD age for all subjects at baseline was 19.41 6 8.32 years, ranging from 8.37 to 42.12 years. Seventeen subjects possessed mutations in exons 1-14 and 21 in ORF15. Ten were predicted null allele mutations (all harboring exon 1-14 mutations), 25 predicted variant protein products (of which 4 were exon 1-14 mutations and 21 were ORF15 mutations), and 3 were splice-site mutations. Bilateral macular OCT volume scans for en face analysis were successfully acquired at baseline for a subset of 31 subjects. The mean 6 SD age for 31 subjects was 20.79 6 8.47 years (range 8.37-42.12 years).
PROGRESSION ANALYSIS: For the calculation of progression rates, linear trend lines were plotted for subjects with OCT follow-up spanning a period >1 year's duration, with a minimum of 3 imaging visits. Twenty-eight subjects had bilateral rates calculated with EZW metric. The mean 6 SD follow-up interval for these 28 subjects was 2.05 6 0.72 years (range 1.10-3.81 years). Twenty-one subjects had bilateral rates calculated with EZA metric. The mean 6 SD follow-up interval for these 21 subjects was 2.00 6 0.74 years (range 1.01-3.51 years).
TEST-RETEST REPEATABILITY ANALYSIS: A comprehensive analysis of intraobserver test-retest repeatability is presented in Table 1. For both metrics, the mean of test-retest differences was small in comparison with baseline values. For the EZW metric, the mean test-retest difference was 21.3 mm compared with a mean EZW baseline of 1963.6 mm. Likewise, for the EZA metric, the mean test-retest difference was 0.025 mm 2 compared with a mean baseline value of 3.70 mm 2 . Calculated 95% LOA lay between À315.7 mm and 273.1 mm for EZW, between À1.06 mm 2 to 1.01 mm 2 for EZA, and between 0.73-0.57 mm 2 for EZA-ART12. Test-retest variability of 15% for EZW was comparable to test-retest variability of 17% for EZA-ART12, compared with a test-retest variability of 28% for EZA.  Table 2, together with indices of interocular symmetry. One subject (Figure 1) was excluded from symmetry analysis because the discrepancy between the right and left eyes was very large and resulted in an outlier effect, not at all in keeping with the entire cohort. Differences in mean EZW values at baseline for right eyes (1923.8 mm) and left eyes (2003.4 mm) were statistically insignificant (P ¼ .6750, paired samples t test). Differences in mean EZA values at baseline for right eyes (3.37 mm 2 ) and left eyes (4.01 mm 2 ) were also statistically insignificant (P ¼ .8177, paired samples t test).
Interocular differences in progression rates were statistically insignificant as assessed with both metrics. The mean EZW rate was 236.82 mm per year for right eyes and 230.28 mm per year for left eyes (P ¼ .9479, paired samples t test). The mean EZA rate was 0.61 mm 2 per year for right eyes and 0.73 mm 2 per year for left eyes (P ¼ .3346, paired samples t test). The equivalent annual progression rates in percentage for EZW was 13.6% for right eyes and 12.8% for left eyes. The equivalent annual progression rates in percentage for EZA was 16.0% for right eyes and 17.5% for left eyes.
A strong interocular correlation at baseline was seen with both EZW and EZA metrics (r > _ 0.94, P < .0001) and for EZA-derived progression (r ¼ 0.93, P < .0001), as shown in Table 3. Interocular correlation for EZWderived progression was also strong (r ¼ 0.65, P ¼ .0002), albeit of a lesser magnitude in comparison to EZAderived rates.
Further analysis of interocular symmetry was performed with the Bland-Altman method. Results are given in Table 2. A good level of interocular symmetry is evident with the use of both metrics to characterize baseline structure. This is reflected in RID values of 3.34% and 3.07% for EZW and EZA, respectively. The RIV, as an index of variability in interocular symmetry at baseline, was larger with EZW metrics at 31.87% compared with an RIV of 17.58% when baseline structure was characterized by the EZA metric.
A greater level of symmetry was seen with the use of EZW over EZA metric when characterizing progression. RID for EZW rates was 1.54% compared with 8.93% for EZA. There was, however, greater variability with the use of EZW over EZA. RIV was 102.60% for rates defined with EZW compared with 78.01% for rates defined with EZA metric.

ASSOCIATIONS OF AGE, BASELINE, AND PROGRESSION:
Correlation data are shown in Table 3. All correlations were statistically significant. A strong negative association between age and baseline is evident with both metrics (r ¼ À0.61, P < .0001 for EZW; r ¼ À0.64, P < .0001 for EZA). A moderate to strong negative correlation was seen between age and progression (r ¼ À0.58, P < .0001 for EZW; r ¼ À0.74, P < .0001 for EZA). A strong positive correlation between baseline and progression was evident with EZA (r ¼ 0.83, P < .0001), with a weaker positive correlation seen with the EZW metric (r ¼ 0.45, P ¼ .0005).  ANOVA ¼ analysis of variance; EZA ¼ ellipzoid zone area; EZW ¼ ellipzoid zone width; SD ¼ standard deviation. Significance level alpha was set at 0.025 following Bonferroni correction. Post hoc multiple pairwise comparisons between the age categories were performed using the Tukey test with those reaching statistical significance shown. Age categories: 1 ¼ <10 years, 2 ¼ 10 to <15 years, 3 ¼ 15 to <20 years, 4 ¼ 20 to <25 years, and 5 ¼ > _25 years of age.
The effects of age and mutation on baseline values and progression rates were further interrogated with 2-way ANOVA. Results together with mean 6 SD values for each category are shown in Table 4. The effect of age was significant on baseline and progression rates as determined by both EZW and EZA metrics. Post hoc pairwise comparisons revealed significant differences between subjects in age categories 1 and 2 versus age categories 4 and 5. In contrast, the effects of mutation on baseline and progression rates were insignificant.

MIXED-MODELS ANALYSES TO DETERMINE OVERALL
PROGRESSION: Overall rates of progression for the entire cohort was modelled using EZW and EZA data. As shown in Table 5, an annual exponential decline rate of 8.22% was obtained with EZW data. A greater exponential rate of 15.47% for the cohort was obtained with EZA data.

ASSOCIATIONS OF STRUCTURE AND FUNCTION:
Correlations between structure and function at baseline are shown in Table 6. All correlations were statistically significant. EZW correlated strongly with functional metrics V Total , CS, and MS (r ¼ 0.64, 0.64, and 0.63, respectively). The same functional metrics (V Total , CS, and MS) also showed greatest correlations with EZA (r ¼ 0.48, 0.60, and 0.45, respectively). Correlations between EZW with V 30 and BCVA were moderate in strength (r ¼ 0.59 and À0.40, respectively). Correlations between EZA with V 30 and BCVA were weaker, albeit still significant (r ¼ 0.38 and À0.37, respectively).

HEREIN WE DESCRIBE THE FIRST PROTOCOL-DRIVEN SPEC-
tral-domain OCT study to characterize EZ changes in FIGURE 5. (Left) Scatterplots of ellipzoid zone width (EZW) and (right) ellipzoid zone area (EZA), both plotted against subjects' age. An exponential decline is evident for both. The reader is directed to Table 5 for further information on exponential decline rates calculated with the mixed-models method. Annual exponential decline rates together with half-lives were calculated from slope values obtained using the mixed-models method with age designated as a fixed effects variable. All values were converted into natural log form before analyses to model an exponential decline. Halflives with 95% CIs were calculated with the equation t 1/2 ¼ Àlog e (2)/k. The significance of age exerting an effect on the model is denoted by the corresponding P values. Significance level alpha was set at 0.025. subjects with molecularly proven RPGR-associated RP using both en face generated EZA and transfoveal EZW metrics.
The mean of individual progression rates in our cohort, as calculated by linear trend lines for each eye, was 234 mm per year with the EZW metric. This is comparable (albeit slightly less than) mean progression rates in XLRP reported by others using a similar metric: 248 mm, 14 270 mm, 15 and 289 mm 16 per year. However, subjects for these 3 studies were obtained from the same source (ie, the docosahexaenoic acid trial [NCT00100230]), 25 and therefore there is likely to be a degree of subject overlap in the 3 studies. Our equivalent mean annual progression rate is 13.2% relative to baseline values. This is greater than the annual rate of 9.6% previously reported in a study with mean baseline EZW of 3410 mm. 16 On average, subjects in our cohort possessed a smaller residual baseline EZ.
With the EZA metric, the mean of individual progression rates in our cohort, as calculated by linear trendlines for each eye, was 0.67 mm 2 per year (16.7%/year). This is comparable to a previously reported progression rate of 0.64 mm 2 per year (percentage rate equivalent was not provided 15 ) in a study where the EZA analysis was performed via layer segmentation and the use of third-party software because acquired volume scans of 31 b-scan density were not sufficiently dense to permit the construction of en face images with vendor software.
To our knowledge, the use of en face images to quantify EZA changes in RP has been carried out by just 1 other study, albeit on subjects with autosomal recessive RP. 18 A mean progression rate of 0.27 mm 2 per year (13%/year) was reported; however, subjects were followed for only 1 year and the progression rate was calculated from images acquired at just 2 time points. 18 The other studies on XLRP described above have also calculated progression with data from only 2 time points, taken an average of 2 years apart. 15,16 In these studies, only 1 eye of each subject was chosen for analysis; in the presence of multiple scans, the one with the clearest EZ band was chosen. 16 The possibility of selection bias influencing results of these studies cannot be excluded.
We have taken a different and arguably more robust approach in our study. Data were acquired from both eyes and analyzed separately and then together. Images from multiple time points (a minimum of 3 time points) were analyzed to plot individual trend lines, from which progression rates were obtained from the slopes of trend lines. A similar approach to calculating progression rates from slopes of linear trend lines has been used by Sujirakul and associates 26 and Cabral and associates 27 in their analyses of EZW, with a reported mean progression rate of 140 mm per year (5.2%/ year). Their cohort, however, comprised RP subjects with various inheritances (of which only 5% had XLRP) and therefore the rates are not directly comparable.
We were able to plot an exponential decline using a mixed-models approach, with data taken collectively, which afforded a wide age span. Here we found a progression rate of 8.22% per year with the EZW metric and 15.47% per year with the EZA metric. Our EZA rate, which is roughly twice the EZW progression rate, fulfils the mathematical prediction of a doubling of rate with progression tracked by area metrics (area of a circle ¼ p [d/2] 2 ). With the simultaneous use of both EZW and EZA metrics, we are able to prove this hypothesis which has been previously alluded to. 14,16 The exponential rates calculated may, in general, be more reflective of the average decline present in the population. Nonetheless, phenotypic heterogeneity is evident in this condition as demonstrated by our subjects, necessitating individual observations to be made.
INTEROCULAR SYMMETRY: There is a good level of interocular symmetry at baseline, with an RID of 3% for both EZW and EZA metrics. When characterizing progression, V Total and V 30 are volumetric metrics that characterize the total amount of sensitivity in the hill-of-vision as defined by the entire test grid (V Total ) or that contained within a central circle of 30 8 radius (V 30 ).
Significance level alpha was set at 0.025 after Bonferroni correction for simultaneous analysis with 2 structural metrics, EZW and EZA. Associations between EZW and BCVA/CS were studied on 76 eyes of 38 subjects. Associations between EZW and MS/V Total /V 30 were studied on 53 eyes of 28 subjects (28 right and 25 left eyes). Associations between EZA and BCVA/CS were studied on 62 eyes of 31 subjects. Associations between EZA and MS/V Total /V 30 were studied on 44 eyes of 23 subjects (23 right and 21 left eyes).
greater interocular symmetry was found with the use of EZW over the use of EZA metrics. RID for EZW rates was 1.54% compared with 8.93% for EZA rates.
Nevertheless, there was significant variation in the degree of interocular symmetry seen across the cohort, as typified by the RIV. RIV at baseline was 17.6% and 31.9%, respectively, for EZA and EZW metrics. RIV for progression rates were 78.0 % and 102.6% for EZA and EZW, respectively.
Despite this level of variation described, interocular correlation at baseline as assessed with both metrics were strong (r > _ 0.94, P < .0001). Interocular correlation for progression as characterized by EZA (r ¼ 0.93, P < .0001) and EZW (r ¼ 0.65, P ¼ .0002) were also strong and significant. In addition, interocular differences at baseline and for progression rates with both metrics were statistically insignificant when tested with the paired samples t test. This highlights the inadequacies of using correlation analyses as a sole method for assessing interocular symmetry.
For example, the mean baseline EZW in our cohort was 1964 mm. Despite a small mean interocular difference of 64 mm, the 95% LOA reached 669 mm. The implication is that observations are necessary for all subjects before inferring the presence of good interocular symmetry. As such, one can argue that in-depth natural history studies are a requirement before treatment trials for many reasons, including this one.
TEST-RETEST VARIABILITY: The 15% test-retest variability for EZW is comparable, albeit marginally smaller than the 17% test-retest variability found with EZA-ART12. In comparison, test-retest variability was 28% for EZA measurements derived from volume scans acquired without averaging. This finding indicates that both EZW and EZA-ART12 measurements are metrics with similar precision. Test-retest variability for both EZW and EZA-ART12 are equivalent to corresponding annual progression rates of 13% and 17%, respectively.
The greater precision seen with EZW and EZA-ART12 measurements is not unexpected. EZW measurements were made on transfoveal lines scans acquired with an average of 100 images each, rendering a high signal to noise ratio. This high average number of images is achievable as only 1 line scan is obtained per transfoveal image.
Following procurement of the Spectralis OCT2 module (Heidelberg Engineering), the significant reduction in acquisition time allowed high-resolution volume scans to be obtained with an average of 12 images per b-scan while maintaining the 193 b-scan density of each scan. EZA borders were more clearly visible on en face images generated from these ART12 volume scans (Figure 2), allowing for greater precision and better measurement repeatability, in contrast to EZA measurements made on images generated from earlier volume scans that were acquired without averaging.
TEST-RETEST VARIABILITY REPORTED IN OTHER STUDIES: None of the aforementioned studies assessed observer repeatability in patients with XLRP despite studying XLRP cohorts. One study selected 13 subjects with autosomal dominant RP (from 59 subjects including 26 with XLRP) with images measured twice over an unspecified time interval. Their intraobserver test-retest variability was estimated at 7.3% with a repeatability coefficient of 0.9 degrees (260 mm) and mean baseline EZW of 12.4 degrees (3584 mm). The mean annual constriction rate for their autosomal dominant RP cohort was 3.4%, indicating that their test-retest variability was twice the annual constriction rate for the tested subjects. 16 Two related studies 14,15 assessed repeatability in image acquisition but not image measurement. In both studies, a different group of RP patients (autosomal recessive or simplex RP) to those reported were imaged twice on the same day. Another study whose cohort comprised of RP of mixed inheritances assessed intraobserver repeatability with images measured twice several weeks apart, and reported a repeatability coefficient of 233 mm with testretest variability of 8.9%, which is almost twice that of their cohort annual constriction rate of 4.9%. 26

INDICES OF INTEROCULAR SYMMETRY:
The mean of intraobserver test-retest difference is small compared with the mean of interocular difference for both metrics. For EZW, the mean intraobserver test-retest difference is 21.3 mm compared with 63.5 mm for the mean of interocular difference. For the EZA metric, the mean intraobserver testretest difference is 0.03 mm 2 compared with 0.10 mm 2 for the mean of interocular difference. Both metrics are therefore suitable for use as structural OCT measures to quantify disease.
The 95% LOA for interocular symmetry with the EZW metric was À542 mm to 669 mm. In comparison, the corresponding 95% LOA for test-retest repeatability was À316 mm to 273 mm. Likewise, the test-retest repeatability coefficient of 294 mm for EZW is less than half its corresponding interocular coefficient of 605 mm. This finding of a test-retest repeatability that is under half that of expected interocular symmetry values further indicates that the EZW metric is reliable for use, especially where quantification of disease in the fellow eye is importantfor example, in cases where the fellow eye would be expected to act as a control to the eye undergoing the treatment trial.
With regard to the EZA metric, the 95% LOA for interocular symmetry of À0.49 mm 2 to 0.70 mm 2 is approximately similar to the corresponding 95% LOA for test-retest repeatability of À0.73 mm 2 to 0.57 mm 2 obtained with EZA-ART12. The test-retest repeatability coefficient of 0.65 mm 2 is also approximately similar to the corresponding interocular coefficient of 0.59 mm 2 . These findings again indicate that the EZA is a suitable metric for use, particularly when measurements are derived from dense volume scans obtained with good image averaging protocols.
The 95% LOA for EZA test-retest repeatability performed on scans acquired without averaging was larger at À1.06 mm 2 to 1.01 mm 2 . As mentioned, EZA measurements were not as precise when made on en face images derived from volume scans acquired without averaging on the OCT1 during the initial period of the study. Nevertheless, these earlier EZA findings are of value and can play an important role as a secondary OCT metric to corroborate and confirm findings made with the EZW metric.

ASSOCIATIONS AND EFFECTS OF AGE ON BASELINE AND
PROGRESSION RATES: We found a strong negative correlation between age and baseline, indicating that the EZ is smaller in older eyes. The moderate to strong negative correlation found between age and progression rates, together with the strong positive correlation between baseline and progression indicates that in general, progression is greater in younger eyes possessing a larger baseline.
Results from the ANOVA (Table 4) further demonstrate the significant effects of age on baseline and progression rates. The largest baseline values and progression rates are seen in the youngest subjects of the cohort. Post hoc comparisons confirm the biggest differences in baseline and rates were between subjects in younger age categories (categories 1 and 2) compared with those in the older age categories (categories 4 and 5). In contrast, the effects of mutation on baseline and progression were insignificant.
Our current findings substantiate those of our previous work. 17,22 CORRELATION OF STRUCTURE AND FUNCTION: We have demonstrated good functional correlation with the use of both EZW and EZA as structural metrics. Correlations between EZW and functional metrics were stronger overall; however, functional correlations with EZA were also significant. These findings provide further support for the use of both structural metrics as surrogate markers of disease in RPGR-RP, with the demonstrated functional correlation being of key importance to both patients and regulators alike.
In conclusion, we have provided and discussed prospectively acquired spectral-domain OCT data in a cohort of subjects with RPGR-RP with a particular focus on the EZ as a structural biomarker of disease. The use of 2 distinct EZ metrics in conjunction adds to the robustness of this study. Both EZW and EZA metrics provide sensitive and complementary parameters to characterize structure and progression in the condition. We anticipate our natural history findings to inform recently commenced RPGR treatment trials, both in the recruitment of trial subjects and in adjudicating treatment responses. Our findings will also be of use to clinicians caring for patients with RPGR-RP and other researchers in the expanding field of phenotyping inherited retinal conditions.