Characterization of Visual Field Loss Over 4 Years in the Rate of Progression in USH2A-related Retinal Degeneration (RUSH2A) Study

Abstract

Purpose:

To report visual field loss using static perimetry (SP) and kinetic perimetry (KP) over 4 years in the Rate of Progression of USH2A-related Retinal Degeneration (RUSH2A) study.

Design:

Prospective, observational cohort study.

Subjects, Participants, and/or Controls:

Participants had USH2A-related rod-cone degeneration, visual acuity ≥ 20/80, and KP III4e ≥ 10° at baseline in the study eye. Preserved cohorts with baseline visual fields sufficient to detect progression were identified.

Methods:

Participants were examined annually through 4 years. Mixed effects models were used to estimate the annual, standardized rate, and percentage rates of change.

Main Outcome Measures:

SP measures included hill of vision (total: V_TOT, central 30°: V₃₀, and peripheral: V_PERIPH) and centrally weighted mean sensitivity (MScw). Percentages with 4-year progression exceeding the coefficient of repeatability (CoR) and with change meeting FDA-recommended criteria were estimated. KP seeing area (dB-steradian (sr)/degree) for I4e, III4e, and V4e isopters was calculated.

Results:

The average decline with SP (95% CI) was 1.94 (1.62, 2.25) dB-sr/year for V_TOT, 0.54 (0.45, 0.62) dB-sr/year for V₃₀, 1.37 (1.11, 1.63) dB-sr/year for V_PERIPH, and 0.56 (0.48, 0.64) dB/year for MS_cw. Average percentage decline per year was 8.6% (7.2, 10.0) for V_TOT, 6.4% (5.3, 7.5) for V₃₀, 13.6% (10.4, 16.7) for V_PERIPH, and 5.6% (4.7, 6.4) for MS_cw. The standardized rate of change was greatest at −1.35 for MS_cw. Rates were higher in the preserved cohorts. Progression exceeding the CoR was 18% (11, 28) for V_TOT, 21% (13, 31) for V₃₀, 21% (13, 31) for V_PERIPH and 17% (10, 27) for MS_cw. Progression exceeding an FDA-recommended threshold was 5% (2%, 12%) for all SP points and 45% (35%, 55%) for functional transition points. Average KP annual percentage decline was 13.1% (7.5, 18.5) for I4e, 12.1% (8.1,15.9) for III4e, and 9.2% (6.3,12.0) for V4e.

Conclusions:

All quantitative perimetry measures declined over 4 years. Progression was greater than the CoR in a relatively low percentage of eyes (17-21%); 45% exceeded the FDA-recommended threshold when only functional transition points were considered. Standardized rate of change was greatest for MS_cw. These measures are useful characterizations of vision loss in USH2A-related retinal degeneration.

Introduction

Disease-causing variants in the USH2A gene are among the most common causes of photoreceptor degeneration, either with congenital hearing loss (Usher syndrome type 2, USH2) or as non-syndromic autosomal recessive retinitis pigmentosa (ARRP). ^1-3 With recent advances in therapeutic approaches, such as antisense oligonucleotides (ASO) ^4-5 and gene editing using clustered regularly interspaced short palindromic repeats systems (CRISPR/Cas), ^6-9 several clinical trials have been initiated for USH2A-associated retinal degeneration with additional trials expected. However, several trials for inherited retinal degenerations have recently failed to demonstrate efficacy based on the pre-selected primary outcome measure, perhaps because prior natural history studies have not investigated quantitative measures of disease progression prospectively and longitudinally over sufficiently long periods to show meaningful change, or due to insufficient treatment effect. ^10-15

Natural history studies of disease are essential to understand the range of abnormalities that develop in affected individuals and the time course of disease progression. Information from natural history studies informs study design including eligibility criteria and the measures used to assess treatment effects. ¹⁶ Although there have been previous studies to characterize the course of disease in affected individuals, ^17-19 little information is available on the natural history derived from more modern methods of ocular imaging and assessment of retinal function.

The Rate of Progression of USH2A-related Retinal Degeneration (RUSH2A) Study was a multicenter, international, longitudinal, observational study to describe disease progression from data collected over 4 years on multiple visual, functional, and structural measures. The design of the RUSH2A study and baseline characteristics of participants have been reported previously. ^20-23 In addition, an interim report on 2-year changes in static perimetry (SP) measures and their association with other measures of retinal function and structure has been published. ⁶ Here we report annual changes in SP through 4 years and changes in kinetic perimetry (KP) between baseline and year 4, and their association with other measures of retinal function and structure.

Methods

Study Design

As described previously, ^{20, 6} the RUSH2A study enrolled 127 participants between August 2017 and December 2018 at 16 clinical sites in North America and Europe. The institutional review boards (IRBs) or ethics boards associated with each participating site approved the study, which adhered to the tenets of the Declaration of Helsinki including compliance with the associated federal regulations. Informed consent was obtained from all participants prior to enrollment. The RUSH2A protocol is listed on www.clinicaltrials.gov (NCT03146078) with registration completed prior to enrolling the first participant.

Eligible participants were at least 8 years old with rod-cone degeneration and had at least 2 disease-causing USH2A sequence variants. Genetic reports were reviewed by a committee to confirm the variants as pathogenic or likely pathogenic. Variants in participants with ARRP were further documented as homozygous or heterozygous in trans based on segregation studies. The “study” eye was defined as the eye with better baseline best-corrected visual acuity (BCVA). Participants with a letter score of 54 or greater (20/80 or better) in the study eye, central visual field at least 10 degrees diameter to a III4e target based on KP, and stable fixation at baseline were enrolled in the primary cohort and followed annually over 4 years. Participants with worse visual function were enrolled in the secondary cohort that was studied only at baseline. The 105 participants in the primary cohort were scheduled for annual evaluations in the study eye over 4 years after the baseline visit. Whereas most testing including SP was performed annually, KP was performed only at baseline and 4 years in both eyes. Longitudinal data for participants in the primary cohort were included in this report. Follow-up visits were performed ideally within a ±4-week window of the target annual visit dates (i.e., 52, 104, 156 and 208 weeks from baseline visit date), but could occur up to 6 months after the target dates.

Perimetry Methods

SP was performed on the study eye using the Octopus 900 (Haag-Streit, Mason, OH) with the German Adaptive Thresholding Estimation (GATE) strategy and a custom centrally-weighted 186-point grid (historically called 185-point grid, see Supplementary Figure 1) to a size V stimulus; ²⁴ additional details of testing have been provided previously. ²⁰ SP results were graded by the Casey Reading Center (Casey Eye Institute, Oregon Health Sciences University, Portland, Oregon, USA). A topographic analysis of the SP values was used to generate a three-dimensional, quantitative surface model of the hill of vision.^22,23 The total volume, measured in decibel-steradians (dB-sr), beneath the surface of the thin-plate spline representation of the hill of vision within the external boundary of the grid was quantified as V_TOT. Additional SP measures included the central 30-degree hill of vision (V₃₀), peripheral hill of vision (V_PERIPH) defined as V_TOT minus V₃₀, and mean sensitivity (dB) of the centrally-weighted grid (MS_CW). ^25,26 The average of each SP measure from three repeated SP sessions at baseline was used as the baseline value for analysis.

Semi-automated KP was performed on both eyes for each participant, using the Octopus 900 (Haag-Streit, Mason, OH) with EyeSuite software to calculate seeing and non-seeing areas in degree² for each isopter and scotoma, respectively; additional details of testing have been provided previously. ²⁰ Six reaction-time vectors were presented within seeing areas, with 1 repetition horizontally, vertically, and diagonally, originating from 10° and 30° eccentricity. Scotomas were mapped at 2°/second angular velocity, with each vector originating from the assumed center. Seeing areas were calculated by subtracting scotomatous areas from the total area for each isopter. The three KP measures analyzed were I4e seeing area, III4e seeing area, and V4e seeing area.

Other Functional and Structural Measures

SP and KP were compared with additional measures of visual function or retinal structure. After protocol refraction, best corrected visual acuity (BCVA) testing was performed using either the electronic visual acuity (EVA) test protocol²⁰ or Early Treatment of Diabetic Retinopathy Study (ETDRS) charts with results recorded as the letter score. ^27,28 Full-field stimulus thresholds (FST) were determined after a 30-minute period of dark-adaptation using white, blue, and red stimuli (Espion E³ system, Diagnosys LCC, Lowell, MA). ^21,29 Fundus-guided mesopic (standard) microperimetry was performed using a Macular Integrity Assessment (MAIA-2) unit (iCare, Raleigh, NC) and summarized by mean sensitivity (MP MS). ^23,30 The ellipsoid zone (EZ) area and central subfield thickness (CST) were derived from optical coherence tomography (OCT) volume scans (Heidelberg Spectralis HRA+OCT, Heidelberg Engineering GmbH, Heidelberg, Germany) (described in RUSH2A-4 (MP-OCT)). ^23,30

Genetic Variant Analysis

USH2A variant analysis was performed by two reviewers independently who used the classification system recommended by the 2015 American College of Medical Genetics and Genomics (ACMG) and Association for Molecular Pathology (AMP) guidelines. ³¹ Each variant was classified as benign, likely benign, variant of unknown significance (VUS), likely pathogenic, or pathogenic. Discordant results were resolved by an independent adjudicator.

The number of USH2A truncating alleles was classified as 1, 2, or 3. ³² The following USH2A missense variants with cysteine substitutions were classified as “RP-enriched”: p.Cys759Phe, p.Cys3294Trp, and p.Cys3358Tyr. Participants with one truncating and one missense allele were defined as being in one of two groups: “RP-Enriched” or “Rest” if the missense allele was not RP-enriched, based on the original RUSH2A genetics study. ³²

Statistical Methods

The analysis cohort for this report was composed of study eyes that had test results for at least 2 of the 5 time points (baseline and 1 to 4 years) for SP, and/or that had results for both baseline and year 4 for KP. Data from the primary cohort (N=105) were included. A total of 103 study eyes met the above criterion for SP analyses and 80 study eyes were included for KP analyses. To mitigate floor effects, a preserved cohort for each SP measure was defined and analyzed separately in addition to the SP analyses using the entire cohort. The preserved cohort for each measure was formed by determining a threshold baseline value above which the slope of the measure over time was greater than 0: V_TOT > 5 dB-sr (N=91), V₃₀ > 3 dB-sr (N=91), V_PERIPH > 5 dB-sr (N=79) and SP MS_CW > 4 dB (N=93).

The distributions of SP measures at each annual visit and the distribution of KP measures at baseline and 4 years were summarized using means, standard deviations (SDs), medians, interquartile ranges (IQRs) and ranges.

For SP outcomes, mixed effects models with a random intercept were used to estimate the annual rates of change and corresponding 95% confidence intervals (95% CI). Log transformed data were used to estimate percentage rates of change. Time was calculated as the number of days from baseline divided by 365.25. A model excluding unreliable test results (false positives ≥ 15%) and a model down-weighting outlier rates of change was also applied to the preserved cohort. For the outlier down-weighted model, the rate of decline for each participant was calculated from a simple linear regression model. A robust regression model using M estimation with a Huber weighting function^33,34 was then used to calculate the weight to be applied in the mixed effects model for each eye in the preserved cohort.

Analyses for estimating annual rates of change of the SP outcomes within subgroups were performed by including time, the baseline subgroup factor, and a term for the interaction between time and the baseline subgroup factor as covariates in the mixed effects models. Because the baseline level of the SP outcome was associated with the rate of change, it and its interaction term with time were also included in the models for the other baseline subgroup factors. The baseline factors investigated were the baseline value of the outcome, clinical diagnosis, age, disease duration, sex, smoking status, dietary supplement use, truncating group, and RP-enriched allele status. The coefficient of repeatability (CoR) for each SP measure was calculated using data from the 3 repeated SP sessions at baseline, and the proportion of eyes that had a decline exceeding the corresponding CoR from baseline to 4 years was reported. (Supplementary Table 1). Using the results from the mixed effects model applied to the preserved cohort for each measure, a standardized rate of change for each measure was estimated using the following formula: $\widehat{\beta }∕[{\sqrt{n}}^{\ast }\text{se}(\widehat{\beta })]$ where $\widehat{\beta }$ is the estimated slope of deterioration from the mixed model, $n$ is the number of subjects, and $\text{se}(\widehat{\beta })$ is the standard error of $\widehat{\beta }$.

The United States Food and Drug Administration (FDA) has provided guidance on clinically meaningful changes within an eye for SP (mean change of ≥ 7 dB in ≥ 5 prespecified points). ^35,36 The percentages of eyes with changes of these magnitudes were calculated at 4 years. ³⁵ The value of a point at baseline needed to be ≥ 8 dB to be considered a candidate for a prespecified point. Selection of prespecified points was motivated by results from Hood et al. of a large decrease in sensitivity between adjacent points along a pathway from the center to the periphery as the path crossed from within the EZ to outside the EZ. ³⁷ Functional transition points (FTPs) for static perimetry were identified by comparing each point to 1 to 4 (depending on the location of the point in the testing grid) more peripheral adjacent points on the testing grid. When there was a decrease in sensitivity of ≥7 dB from an inner point to the more peripheral adjacent point, the inner point was qualified as a candidate FTP. In an effort to maximize the likelihood of a candidate FTP losing sensitivity in the future, all candidate FTPs were ordered by the percentage of qualifying adjacent points, from 100% to 25%. Candidate points that were qualified by 100% of adjacent points were selected as an FTP. If the selected number of selected points was less than 5, then points with next highest percentage of qualifying points were included. Additional approaches for pre-specification of points included evaluation of the entire set of points on the testing grid and of the points in the central 30 degrees.

For KP outcomes, change from baseline to 4 years was calculated for each measure. Because distributions for KP measures were extremely skewed, percentage rates of change were calculated using similar mixed effects models as for the SP measures with natural log transformed KP data as the dependent variables. Similar subgroup analyses for estimating annual percentage rates of change were also performed for baseline factors.

Associations among change in SP and KP measures and with change in other measures (BCVA, FST, OCT and MP) from baseline to 4 years were assessed with Spearman correlation coefficients (r_s). FST testing was done after a 30-minute period of dark-adaptation. All analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC, USA) and reported P values are two-sided.

Results

Study Population

The number of participants completing baseline and 1-to-4-year annual visits is shown in Supplementary Figure 2. Among the 105 participants recruited into the primary cohort, one participant died before the 1-year visit, one participant died between the 3- and 4-year visits, and six additional participants dropped out after the 3-year visit. The number of participants who completed in-office visits for 1, 2, 3 and 4 years is 102, 88, 99 and 95, respectively, for assessment of visual functional and structural measures.

Among the 103 participants included in the SP analyses cohort, the clinical diagnosis was USH2 for 64 (62%) participants and ARRP for 39 (38%) participants. The mean age was 37 years (SD, 12), 58 (56%) were female, and 92 (89%) were white. The median duration of disease at enrollment was 12 years (IQR, 7 to 20). The baseline characteristics for the 83 participants in the KP analyses cohort were similar.

SP Outcomes

The distribution of four SP measures at each visit for the entire analysis cohort is shown in Table 1 and for the preserved cohort is shown in Supplementary Table 2. Figure 1 provides plots of the four SP measures by duration of disease at each visit, showing an overall downward trend over time. At baseline, 13% of participants had V_periph values less than 1 dB-sr, a percentage that increased to 17% by 4 years. Among the entire analysis cohort for SP measures [N=103], the average V_TOT was 32.9 (SD, 23.6) dB-sr at study baseline, which dropped to 29.1 (SD, 22.8) dB-sr at 2 years and 25.2 (SD, 22.3) dB-sr at 4 years. The average V₃₀ was 10.2 (SD, 5.6) dB-sr at study baseline, which dropped to 9.2 (SD, 5.4) dB-sr at 2 years and 7.9 (SD, 5.1) dB-sr at 4 years. The average V_PERIPH was 22.6 (SD, 18.8) dB-sr at study baseline, 19.9 (SD, 18.4) dB-sr at 2 years and 17.2 (SD, 17.9) at 4 years. The average SP MS_cw was 11.6 (SD, 5.8) dB at baseline, 10.6 (SD, 5.5) dB at 2 years and 9.4 (SD, 5.4) dB at 4 years (Table 1). At baseline, 13% of participants had V_PERIPH values less than 1 dB-sr, a percentage that increased to 17% by 4 years.

Table 1.

Static Perimetry (SP) Measures at Each Visit in the Entire Analysis Cohort

Measures	Baseline	Year 1	Year 2	Year 3	Year 4
VTOT (dB-sr)
N	103	96	86	99	87
Mean ± SD	32.9 ± 23.6	30.9 ± 23.4	29.1 ± 22.8	27.6 ± 22.8	25.2 ± 22.3
Median (IQR)	29.3 (11.8, 51.2)	24.5 (11.4, 50.1)	23.2 (8.5, 47.9)	19.7 (8.6, 47.8)	17.1 (7.9, 40.1)
V30 (dB-sr)
N	103	101	87	95	91
Mean ± SD	10.2 ± 5.6	9.7 ± 5.5	9.2 ± 5.4	8.9 ± 5.4	7.9 ± 5.1
Median (IQR)	9.8 (5.4, 14.3)	9.6 (5.3, 13.7)	8.2 (4.8, 13.5)	7.5 (4.5, 13.0)	6.5 (3.7, 10.7)
VPERIPH (dB-sr)
N	103	96	86	99	87
Mean ± SD	22.6 ± 18.8	21.0 ± 18.8	19.9 ± 18.4	18.8 ± 18.2	17.2 ± 17.9
Median (IQR)	18.9 (5.8, 37.7)	14.6 (4.4, 38.8)	14.5 (3.5, 33.7)	11.7 (2.5, 33.2)	9.0 (2.8, 29.4)
SP MSCW (dB)
N	103	96	86	99	87
Mean ± SD	11.6 ± 5.8	11.1 ± 5.7	10.6 ± 5.5	10.1 ± 5.6	9.4 ± 5.4
Median (IQR)	11.0 (6.7, 15.9)	10.5 (6.9, 14.7)	9.6 (6.4, 14.2)	8.6 (5.8, 14.2)	8.3 (5.2, 12.6)

Figure 1. Static Perimetry Measures by Duration of Disease for Study Eyes of Individual Patients (“Spaghetti Plots”).

A) V_TOT, B) V₃₀, C) V_PERIPH and D) SP MS_CW

The average annual decline (95% CI) in V_TOT was 1.94 dB-sr/year (1.62, 2.25; Table 2) or 8.6 %/year (7.2, 10.0; Table 3) in the entire cohort, but greater, at 2.16 (1.81, 2.51) dB-sr/year or 9.0 (7.5, 10.6) %/year in the preserved cohort (which consisted of participants with sufficient function at baseline for decline to be observed over time). The decline rate of V₃₀ was 0.54 (0.45, 0.62) dB-sr/year or 6.4 (5.3, 7.5) %/year in the entire cohort versus 0.63 (0.54, 0.71) dB-sr/year or 7.2 (6.2, 8.2) %/year in the preserved cohort. The decline of V_PERIPH was 1.37 (1.11, 1.63) dB-sr/year or 13.6 (10.4, 16.7) %/year in the entire cohort, and 1.72 (1.40, 2.04) dB-sr/year or 12.1 (9.3, 14.8) %/year in the preserved cohort. The decline rate for SP MS_CW was 0.56 (0.48, 0.64) dB/year or 5.6 (4.7, 6.4) %/year in the entire cohort and 0.61 (0.52, 0.70) dB/year or 6.0 (5.2, 6.9) %/year in the preserved cohort. Models excluding tests with ≥ 15% false positives, or down-weighting statistical outliers, reduced the estimated rates of decline compared to using all eyes in the preserved cohort.

Table 2.

Estimated Average Annual Change in Static Perimetry Measures

Outcomes	Entire cohort	-----------------Preserved cohort -----------------
		All	False positives <15% a	Outliers down- weighted b
VTOT (dB-sr/year)	N=103	N=91	N=89	N=91
Annual change c	−1.94	−2.16	−2.05	−2.02
95% CI	(−2.25, −1.62)	(−2.51, −1.81)	(−2.36, −1.75)	(−2.33, −1.71)
V30 (dB-sr/year)	N=103	N=91	N=90	N=91
Annual change c	−0.54	−0.63	−0.56	−0.54
95% CI	(−0.62, −0.45)	(−0.71, −0.54)	(−0.64, −0.49)	(−0.62, −0.47)
VPERIPH (dB-sr/year)	N=103	N=79	N=77	N=79
Annual change c	−1.37	−1.72	−1.68	−1.71
95% CI	(−1.63, −1.11)	(−2.04, −1.40)	(−1.97, −1.40)	(−2.00, −1.41)
SP MSCW (dB/year)	N=103	N=93	N=91	N=93
Annual change c	−0.56	−0.61	−0.56	−0.54
95% CI	(−0.64, −0.48)	(−0.70, −0.52)	(−0.63, −0.49)	(−0.61, −0.47)

^aTests with false positive rate <15%

^bOutliers were down-weighted using weighted mixed-effects model, weights were computed from robust regression modelling of estimate rate of decline from each participant

^cAll P values for testing the average annual change estimates against zero were <0.001

Table 3.

Estimated Average Annual Percentage Change in Static Perimetry Measures Based on Log-transformed Data

Outcomes	Entire cohort	------------ Preserved visual field cohort ----------
		All	False positives <15% a	Outliers down- weighted b
VTOT (%/year)	N=103	N=91	N=89	N=91
Annual change c	−8.6	−9.0	−8.4	−8.0
95% CI	(−10.0, −7.2)	(−10.6, −7.5)	(−9.8, −6.9)	(−9.3, −6.6)
V30 (%/year)	N=103	N=91	N=90	N=91
Annual change c	−6.4	−7.2	−6.6	−6.3
95% CI	(−7.5, −5.3)	(−8.2, −6.2)	(−7.6, −5.6)	(−7.2, −5.3)
VPERIPH (%/year)	N=103	N=79	N=77	N=79
Annual change c	−13.6	−12.1	−11.3	−12.0
95% CI	(−16.7, −10.4)	(−14.8, −9.3)	(−14.0, −8.5)	(−14.6, −9.3)
MSCW (%/year)	N=103	N=93	N=91	N=93
Annual change c	−5.6	−6.0	−5.6	−5.4
95% CI	(−6.4, −4.7)	(−6.9, −5.2)	(−6.3, −4.8)	(−6.2, −4.7)

^aTests with false positive rate <15%

^bOutliers were down-weighted using weighted mixed-effects model, weights were computed from robust regression modelling of estimated rate of decline from each participant

^cAll P values for testing the average annual change estimates against zero were <0.001

The annual rates of decline for all four SP measures were significantly associated with their corresponding baseline values (all P values < 0.001) (Supplementary Table 3 for entire cohort and Supplementary Table 4 for preserved cohort). For all 4 SP measures and both cohorts, participants in the lowest tertile of baseline values experienced lower rates of decline than participants in the highest tertile of baseline values. For the entire cohort, SP V₃₀ had differing rates of decline by clinical diagnosis, with USH2 having a smaller rate of decline than ARRP in the entire cohort (−0.21 dB-sr/yr) (P value = 0.02). No significant difference in rates of decline was observed for SP V_TOT, V_PERIPH, and SP MS_CW by clinical diagnosis. No significant difference in rates of decline was observed for any of the SP measures by disease duration (Supplementary Tables 3 and 4).

The standardized rates of change per year from the mixed model were evaluated for the four SP measures in both the entire cohort and the preserved cohort (Table 4). Overall, the standardized rates of change per year for all four SP measures for the preserved cohort were of a greater magnitude than the entire cohort. The standardized rate of change per year for V_TOT was −1.18 for the entire cohort and −1.28 for the preserved cohort. The standardized rate of change per year for V₃₀ was −1.20 for the entire cohort and −1.51 for the preserved cohort. For V_PERIPH, the standardized rate of change per year for the entire cohort was −1.02, and −1.19 for the preserved cohort. For SP MS_CW, the standardized rate of change per year for the entire cohort was −1.35 and −1.45 for the preserved cohort (Table 4).

Table 4.

Evaluation of Change for Static Perimetry (SP) Measures

	VTOT	V30	VPERIPH	SP MSCW
Standardized rate of change
Entire cohort	−1.18	−1.20	−1.02	−1.35
Preserved cohort	−1.28	−1.51	−1.19	−1.45
Change at Year 4 > Coefficient of Repeatability
Entire cohort, n (%)	16 (18%)	19 (21%)	18 (21%)	15 (17%)
95% Confidence Interval	(11%, 28%)	(13%, 31%)	(13%, 31%)	(10%, 27%)
Preserved cohort, n (%)	16 (21%)	19 (24%)	18 (27%)	15 (19%)
95% Confidence Interval	(12%, 31%)	(15%, 35%)	(17%, 39%)	(11%, 29%)
Coefficient of Repeatability	14.0 dB-sr	3.4 dB-sr	11.4 dB-sr	3.5 dB

The CoR for V_TOT at baseline was 14.0 dB-sr and 16 (18%) of the study eyes in the entire cohort had V_TOT decline from baseline to 4 years exceeding this limit. The CoR for V₃₀ at baseline was 3.4 dB-sr and 19 (21%) of the study eyes in the entire cohort had V₃₀ decline from baseline to 4 years exceeding this value. For V_PERIPH, 18 (21%) had decline exceeding its CoR of 11.4 dB-sr, and for SP MS_CW, 15 (17%) had decline exceeding its CoR of 3.5 dB from baseline to 4 years (Table 4). The number of study eyes with decline exceeding the CoR for each SP measure was the same when restricting the data to the preserved cohort only, with corresponding percentages that were higher than in the entire cohort.

Outcomes for Prespecified Sets of SP Points:

The percentage of eyes meeting FDA criteria for meaningful progression at 4 years was 5% when the entire set of 185 points were prespecified, 9% when the 108 points in the central 30 degrees was prespecified; however, 45% of eyes met the FDA criteria for meaningful progression over between baseline and 4 years when the prespecified points included only the FTPs (Table 5). No eyes met FDA criteria for improvement when the FTPs were prespecified. The annual rate of change was −0.53 dB when the points in the entire grid or in the central 30 degrees were prespecified. The annual rate of change was considerably greater (−1.50 dB) considering only the FTPs as prespecified points. The standardized rate of change was similar when the points in the entire grid (−1.42) and the points in the central 30 degrees (−1.29) were prespecified but were considerably higher (−1.94) when the FTPs were prespecified.

Table 5.

Percentage of Eyes Meeting FDA Criterion for Change Between Baseline and 4 Years and Rates of Change over 4 Years for Different Choices of Prespecified Points on the Static Perimetry Testing Grid

Point Subset	Na	Number of Points	Meeting FDA Criterion (95% CI)	Nb	Annual Change (dB) (95% CI)	Standardized Rate of Change (95% CI)
Entire Grid	91	185	5% (2%, 12%)	102	−0.51 (−0.58, −0.44)	−1.42 (−1.49, −1.35)
Central 30°	91	108	9% (5%, 16%)	102	−0.53 (−0.61, −0.45)	−1.29 (−1.37, −1.21)
Functional Transition Points	91	5 – 30	45% (35%, 55%)	102	−1.50 (−1.65, −1.35)	−1.94 (−2.09, −1.79)

^aEyes with testing results at both baseline and 4 years

^bEyes with testing results at 2 or more times between baseline and 4 years, inclusive.

KP Outcomes

The distributions of the three KP measures at baseline and at 4 years, and the change from baseline to 4 years, for both study and non-study eyes are shown in Table 6. For study eyes, the median I4e seeing area was 132 (IQR, 42 to 930) deg² at baseline and 75 (IQR, 18 to 328) deg² at 4 years with median loss of 43 (IQR, 1 to 379) deg² from baseline to 4 years. The median III4e seeing area was 4982 (IQR, 772 to 9303) deg² at baseline and 2335 (IQR, 381 to 7310) deg² at 4 years with median loss of 900 (IQR, 84 to 2226) deg² from baseline to 4 years. The median V4e seeing area was 11038 (IQR, 5675 to 13035) deg² at baseline and 7445 (IQR, 2066 to 12552) deg² at 4 years with median loss of 1058 (IQR, 133 to 2904) deg² from baseline to 4 years. The estimated annual percentage rate of change (95% CI) was −13.1% (−18.5, −7.5) for I4e seeing area, −12.1% (−15.9, −8.1) for III4e seeing area, and −9.2% (−12.0, −6.3) for V4e seeing area among study eyes. The distribution of the three KP measures in non-study eyes and study eyes were similar, and the changes in the three KP measures from baseline to 4 years in study eyes were strongly correlated with the changes in non-study eyes (r_s ranged from 0.64 to 0.79, all P values < 0.001, Figure 2).

Table 6.

Kinetic Perimetry (KP) Measures at Baseline and 4 Years

Measures	Baseline	Year 4	Year 4 – Baseline
Study Eyes
I4e Seeing Area (deg2)
N	79	79	79
Mean ± SD	1104 ± 2051	653 ± 1554	−451 ± 1118
Median (IQR)	132 (42, 930)	75 (18, 328)	−43 (−379, −1)
Annual percent change (95% CI) a,b	−13.1 (−18.5, −7.5) %
III4e Seeing Area (deg2)
N	78	78	78
Mean ± SD	5374 ± 4366	4089 ± 4238	−1285 ± 2055
Median (IQR)	4982 (772, 9303)	2335 (381, 7310)	−900 (−2226, −84)
Annual percent change (95% CI) a,b	−12.1 (−15.9, −8.1) %
V4e Seeing Area (deg2)
N	75	75	75
Mean ± SD	9268 ± 4785	7489 ± 5089	−1779 ± 2694
Median (IQR)	11038 (5675, 13035)	7445 (2066, 12552)	−1058 (−2904, −133)
Annual percent change (95% CI) a,b	−9.2 (−12.0, −6.3) %
Non-study Eyes
I4e Seeing Area (deg2)
N	78	78	78
Mean ± SD	1004 ± 1980	685 ± 1621	−319 ± 1099
Median (IQR)	159 (32, 670)	76 (23, 335)	−48 (−204, −3)
Annual percent change (95% CI) a,b	−13.2 (−18.1, −8.0) %
III4e Seeing Area (deg2)
N	74	74	74
Mean ± SD	5540 ± 4465	4014 ± 4202	−1526 ± 2146
Median (IQR)	4676 (792, 9940)	2010 (375, 7358)	−750 (−2481, −118)
Annual percent change (95% CI) a,b	−14.0 (−18.1, −9.8) %
V4e Seeing Area (deg2)
N	78	78	78
Mean ± SD	9029 ± 4937	7314 ± 5047	−1714 ± 2848
Median (IQR)	10512 (3795, 13349)	7007 (2189, 12292)	−1031 (−2620, −258)
Annual percent change (95% CI) a,b	−9.1 (−11.8, −6.3) %

^aCalculated from mixed effects models using log-transformed data as dependent variables

^bAll P values for testing the average annual change estimates against zero were <0.001

Figure 2. Scatter Plots for Changes in 3 Kinetic Perimetry Measures from Baseline to 4 Years Among Study Eyes vs. Non-study eyes.

A) I4e seeing area, B) III4e seeing area and C) V4e seeing area.

The annual percentage rates of change in I4e seeing area were significantly associated with their corresponding baseline values (P = 0.002) (Supplementary Table 5). For the I4e stimuli, the lowest rates of decline in seeing area were in the category with the lowest baseline values while the opposite was true for V4e stimuli (Supplementary Table 5). The annual percentage rate of change for V4e seeing area was associated with sex, with female participants decreasing 6.5% more rapidly than male participants (P = 0.03) (Supplementary Table 5). No other measures showed a difference in rates of change based on sex.

Correlation among Changes in Visual Measures

The changes in all four SP measures from baseline to 4 years were either moderately or highly correlated with each other, with r_s ranging from 0.48 (V₃₀ vs V_PERIPH) to 0.96 (V_TOT vs V_PERIPH) (all P < 0.001, Supplementary Table 6). Among the three KP measures, change in III4e seeing area moderately correlated with change in V4e seeing area (r_s = 0.56); however, change in I4e seeing area was not correlated with change in V4e seeing area (r_s = −0.01). Although change in V4e seeing area was at most weakly correlated with SP measures, change in III4e seeing area was weakly to moderately correlated with change in some SP measures with r_s ranging from 0.38 (V₃₀) to 0.51 (V_TOT) (all P < 0.001). Change in I4e seeing area was weakly correlated with change in V_TOT (r_s = 0.28) and moderately correlated with change in V₃₀ (r_s = 0.46) and SP MS_CW (r_s = 0.45), but not significantly with change in V_PERIPH.

The changes in the four SP measures were not significantly correlated with change in BCVA or FST measures. OCT EZ area was moderately correlated with V₃₀ (r_s = 0.45) and weakly correlated with SP MS_CW (r_s = 0.38) while OCT CST showed no significant correlation with SP measures. There were weak to moderate correlations between change in the four SP measures with change in MP MS (r_s range from 0.32 to 0.55). None of the three KP measures were significantly correlated with BCVA, FST or OCT measures, apart from a weak to moderate correlation of III4e seeing area with FST white (r_{s =} −0.34) and FST blue (r_s = −0.43), and a weak correlation of I4e seeing area with OCT CST (r_s = −0.36). There was moderate correlation between change in MP MS and change in I4e seeing area (r_s = 0.49).

Discussion

Although USH2A-related USH2 and ARRP are among the most common forms of retinal degeneration, there are few prospective, longitudinal natural history studies characterizing the rate of vision loss quantitatively in participants with this relentless cause of irreversible vision loss. ¹⁸ The RUSH2A study was designed to provide quantitative endpoint measures of progressive USH2A-related retinal degeneration to inform clinical trial design. ³⁸ Visual function was assessed in this prospective, longitudinal natural history study with the primary endpoint of SP using the hill of vision modeling algorithm; ²⁵ secondary endpoints analyzed KP, visual acuity, structural measures of EZ area, and possible risk factors (genotype, phenotype, environmental, and comorbidities) for progression. The main outcome measure, SP, demonstrated significant but modest change in several SP parameters at 2 years. ⁶ Here we present similar findings after 4 years. Overall, all perimetry measures, both SP and KP, decreased over time. Annual rates of change and percent rates of change at 4 years were similar to those observed at 2 years for all SP measures. ⁶ The estimates of the slopes (rates of visual field loss per year) for the SP measures were similar in both the 2-year and 4-year data. However, the confidence intervals for each measure at 4 years decreased in width by about 50%, indicating estimates of progression rates are more precise with increased follow-up.

Although the 2-year and 4-year estimates of the average annual rates of change were consistent, Figure 1 illustrates significant between-participant variability in the rate of change. Some participants experienced steady declines in SP MS_CW, for example, while others experienced large swings in mean sensitivity, and still others did not see any decline over four years. Figure 2 shows that even within-participant variability can be very high, with some large differences in seeing area change between the study eye and the non-study eye. In the clinical trial setting, some of the between-participant variability can be attenuated by requiring a sufficiently high mean sensitivity at baseline, thus reducing floor effects. However, the remaining variability in these perimetry measures is still a significant concern when designing clinical trials, as it necessitates more participants or longer follow-up or both.

Examination of three choices for pre-specification of SP points to assess meaningful change as recommended by the FDA, showed that <10% of eyes met the criteria for progression at 4 years when all points in the entire grid or in the central 30° were selected, but the percentage increased to 45% when FTPs were selected without any eyes meeting the criterion for improvement.

We analyzed risk factors predicting rate of progression and found the amount of change was correlated with visual field at baseline and was smallest in eyes in the lowest baseline tertile (i.e. smallest visual field area). For V_TOT and V_PERIPH, participants with the lowest baseline visual field showed the lowest rate of decline, likely because the central visual field is preserved until late in rod-cone degeneration. Participants with the best visual fields at baseline represented the earliest stages of progression and showed intermediate progression rates as compared with those having the smallest, more advanced visual fields, and participants in the intermediate group. Participants with intermediate visual field loss at baseline showed the greatest rate of change in V_TOT and V_PERIPH as midperipheral scotomas expanded during 4 years of progression, and those with the earliest stages of visual field loss showed the greatest rates of change in V₃₀ and SP MS_CW (Supplementary Tables 3 and 4).

Most of the changes in SP measures only were moderately correlated with the changes in KP measures (r_s = 0.25 to 0.51) indicating that these two measurements provide somewhat different information (Supplementary Table 6). Change in the smallest isopter (I4e) area was not significantly correlated with change in the other two KP measures, although the I4e area correlated most strongly with V₃₀ and MS_CW. MP mean sensitivity was correlated with all SP measures (r_s ranging from 0.32 to 0.55) and change in I4e seeing area (Supplementary Table 6). As reported in other studies, ^23,39-41 we also observed change in central retinal sensitivity, measured as V₃₀ and MS_CW, was correlated with change in EZ area (r_s = 0.45 and r_s = 0.38, respectively, Supplementary Table 6), and I4e seeing area was correlated with OCT CST (r_s = −0.36, Supplementary Table 6), although EZ area was not significantly correlated with any KP measures.

Change in seeing area to the III4e target correlated with FST white (r_s = −0.34) and blue (r_s = −0.43) thresholds, but other changes in KP measures were not, or were at most moderately, correlated with changes in FST or visual acuity. FST white and blue thresholds were not correlated with any SP measures. FST white and blue thresholds reflect the most sensitive areas of retina and hence are non-localizing. ³⁸

At baseline, 37 eyes were rod-mediated and 29 were cone-mediated, with 10 (27%) rod-mediated eyes changing to cone-mediated over the course of the study and no eyes changing from cone-mediated to rod-mediated. The correlation between the change in V₃₀ and change in other SP and KP measures tended to be higher in eyes that were cone-mediated at baseline, while the correlation between V4e seeing area and other SP and KP measures tended to be higher in eyes that were rod-mediated at baseline (Supplementary Table 7).

We defined the preserved cohort for V_TOT in the same way as in the 2-yr study. However, in this 4-year analysis we customized the preserved cohort to each SP measure. We defined the primary cohort as participants with threshold baseline SP values above which the slopes of the measure over time in eyes was greater than 0, with the expectation that they still had room to deteriorate. Both our spaghetti plots (Figure 1) and the subgroup analyses (Supplementary Tables 3 and 4) demonstrate that eyes with low baseline values for SP do not show significant change, or change cannot be detected. We also provided analyses where we excluded unreliable fields and down-weighted outliers to provide more accurate estimates of the slopes. Such procedures may be considered for clinical trials with relatively small sample sizes in that they reduce the likelihood that unreliable fields or a few outliers in a treatment group will unduly influence the difference between treatment groups. However, these adjustments had minimal impact on the results or conclusions in this natural history study, so inclusion of participants with more advanced disease may not significantly affect study results over 2-4 years in future trials. The current study demonstrated the greatest rates of change in the middle tertile of baseline visual field area for the V_TOT, V_PERIPH, and III4e measures. Future trials that plan to use SP outcome measures may wish to enroll patients with those characteristics to increase the likelihood of observing significant change in 2–4-year trials.

Higher rates of change of the various measures need to be weighed against the variability of the change across eyes under study. To account for the variability in each measure, and to standardize across measures, we presented a standardized rate of change for each measure (Table 4). The results suggest that SP MS_CW shows the greatest change when the variability is taken into consideration for the entire cohort, while the rate of change was similar for SP MS_CW and V₃₀ in the preserved cohort. This finding has significance for design of future clinical trials because the calculation of SP MS_CW simply represents the mean of all the measured sensitivities, without requiring complex interpolation as provided by the hill of vision VFMA. ^25,42

The CoR provides another method to measure change in function that exceeds inter-test variability to provide information on real change, representing actual disease progression. Table 4 shows the CoR was greatest for V_TOT and lowest for V₃₀ and SP MS_CW. In addition, SP MS_CW showed the greatest standardized change in the entire cohort and may provide a sensitive measure of disease progression over 2-4 years for use as an outcome measure in clinical trials. However, since only up to 27% of participants in the preserved cohort, and only up to 21% in the entire cohort, showed change greater than the CoR over the 4-year RUSH2A study, it is possible that slope analysis of these endpoints could serve as a useful outcome measure over a shorter time period. The rates of decline in the ARRP group vs. the USH2 group were not statistically significantly different (Supplementary Table 3), except in V₃₀. Similarly, the genotype classification as “RP-enriched” was not associated with differences in rates of decline (Supplementary Table 3, 4). Although the number of truncating alleles was associated with differences in rates of decline, this association was not monotonic with the number of alleles (Supplementary Table 3, 4). These findings are intriguing as clinical diagnosis (ARRP vs USH2) and genotype (truncating number, RP-enriched missense variants) have been associated with differences in age of onset, structural, and functional measures of retinal degeneration in multiple independent studies. ^43-45 That these groups exhibit similar rates of progression in this study suggests that clinical endpoints may apply to USH2A-related retinopathies in a genotype-independent manner, or that the age of disease onset, but not subsequent rates of decline, affect the variables measured in the RUSH2A study.

Assessing the percentages of eyes meeting the current FDA recommended criteria for meaningful change showed that <10% of eyes met these criteria for progression over 4 years when either the entire field or central 30° were prespecified. ³⁵ Thus, testing an intervention intended to slow or stop progression using the percentage of eyes meeting FDA recommended criteria based on common visual field measures is not feasible because of the very large sample sizes required to demonstrate a reduction in these percentages even with 4 years of follow-up. However, analyzing rates of change may increase the numbers of participants meeting the current FDA recommended criteria. Furthermore, prespecifying only FTPs increased the percentage of eyes meeting the FDA criteria for progression to 45%, providing greater opportunity for an intervention to show a decrease in progression (Table 5). Similarly, when only the FTPs were used to assess the rate of change in mean sensitivity, the rate increased relative to SP MS_CW.

The RUSH2A study provides the largest prospective, longitudinal natural history study of subjects with USH2A-related USH2 and ARRP. Despite limitations including missed visits during the study period imposed by a global pandemic, the 2- and 4-year results show consistent, significant declines in visual field measured both by SP and KP in participants with USH2A-related retinal degeneration. The 4-year results showed lower variability but validated the findings reported after 2 years. ⁶ Taking inter-test variability into consideration, SP MS_cw, based on either the entire field or FTPs, had the greatest standardized rate of change and may serve as the measure that is most likely to demonstrate change during disease progression in this patient population. Although there was no control group of visually normal individuals, this study provides estimates of sensitivity decline over four years in this cohort. It is possible that some of this decline is natural and not due to Usher’s syndrome, and that is important for clinical trial planning. Ongoing studies of this cohort over longer periods of time should provide further information about risk factors associated with rates of visual field loss.