Progression of Atrophy as a Function of ABCA4 Variants and Age of Onset in Stargardt Disease

Abstract

Purpose

The purpose of this study was to assess the natural course of the retinal atrophy growth rate in patients with Stargardt disease (STGD1) with particular mutations in ABCA4, which may be eligible for mutation-specific therapy.

Methods

Fundus autofluorescence images (Heidelberg Spectralis) were gathered from 221 patients (436 eyes) in two centers: Radboud UMC and Ghent University Hospital. The area of definitely decreased autofluorescence and total decreased autofluorescence was measured using the Heidelberg RegionFinder software tool. Square root transformation was used to correct for two-dimensional growth. A mixed model was used to determine the atrophy growth rates. Atrophy growth rates were calculated for all eyes and were categorized into subgroups based on ABCA4 mutations potentially suitable for mutation-specific therapy (c.4539+2001G>A; c.5461-10T>C; c.5882G>A; c.768G>T), or subgroups based on age of onset.

Results

The mean square root–transformed growth rate of atrophy was 0.1446 mm/year (95% CI, 0.1382–0.1510 mm/year) for definitely decreased autofluorescence and 0.1459 mm/year (95% CI, 0.1402–0.1515 mm/year) for total decreased autofluorescence. Definitely decreased autofluorescence square root–transformed atrophy growth was slower in patients heterozygous for c.5882G>A (0.0821 mm/year) and c.4539+2001G>A (0.0686 mm/year) than c.768G>T (0.1299 mm/year) and c.5461-10T>C (0.1565 mm/year). Eyes of patients with late-onset STGD1 had the fastest atrophy growth (0.1782 mm/year), compared with eyes of patients with early-onset STGD1 (0.1655 mm/year) and patients with intermediate-onset STGD1 (0.1269 mm/year).

Conclusions

Atrophy growth rates vary among subgroups of patients with STGD1, depending on both specific mutations and age of onset. This pattern may have implications for the design of clinical trials for mutation-specific therapies.

Recessive Stargardt disease (STGD1, OMIM 248200) is an inherited retinal disease predominantly characterized by progressive macular degeneration that leads to severe visual impairment or even blindness.¹ The phenotypical expression of STGD1 covers a broad spectrum. The most severe forms occur in young children (early-onset STGD1) in whom STGD1 leads to central macular atrophy that ultimately affects the entire posterior pole, resulting in severe visual impairment in the first decade of life.² On the other end of the spectrum, adults (late-onset STGD1) with well-demarcated parafoveal atrophy and a coexisting preserved fovea, typically only start developing mild complaints after the fifth decade of life.^3–5 A group of patients with STGD1 with intermediate-onset disease make up the middle of the spectrum. These patients present with vision loss accompanied by yellow–whitish flecks and central macular atrophy with symptom onset in the second to fourth decades of life.⁶

The underlying pathophysiology of STGD1 is a dysfunctional adenosine triphosphate–binding cassette subfamily A member 4 (ABCA4) protein, caused by biallelic mutations in the ABCA4 gene.⁷ The ABCA4 protein is involved in recycling of the 11-cis and all-trans isomers of retinal by flipping it from the intradiscal space of the photoreceptor outer segment discs into the intracytoplasmic area, where all-trans-retinal is reduced to all-trans-retinol (vitamin A), by the enzyme all-trans-dehydrogenase (RDH). Subsequently, all-trans-retinol can be transported out of the photoreceptor cells toward the RPE.⁸ Dysfunction of the ABCA4 protein leads to accumulation of all-trans isomers forming bisretinoids as a part of toxic lipofuscin in both the photoreceptor outer segment discs and in the RPE cells, causing retinal degeneration.^8,9 Many new (deep) intronic ABCA4 variants have been discovered in recent years.^10,11 These deep intronic variants cause alternative splicing and lead to the inclusion of pseudo-exons that disrupt the reading frame resulting in diminished or absent ABCA4 protein function. In addition to the deep intronic variants, variants located in or near intron–exon boundaries can result in changed pre-mRNA splicing, either by exon skipping or exon elongation.¹² An improved understanding of these mechanisms has allowed for new molecular targets for potential treatments. In vitro studies have demonstrated the ability to prevent aberrant pseudo-exon inclusions in several deep intronic ABCA4 mutations through antisense oligonucleotide-based splicing modulation therapy.^10,12–17

Mutation-specific therapies for STGD1 are emerging for several ABCA4 variants (c.4539+2001G>A; c.5461-10T>C; c.5882G>A; and c.768G>T).^{13,14,17–19} Testing the efficacy of a mutation-specific therapy in an orphan disease like STGD1 poses unique challenges, given the limited potential study population and the aforementioned variation of clinical phenotypes. To establish reliable sample size estimations and to ensure sufficient statistical power in a clinical trial, natural history studies have to be conducted to elucidate the natural progression of potential clinical trial end points.²⁰

One of the most frequently used biomarkers reflecting disease progression in STGD1 is abnormality of retinal autofluorescence owing to outer retinal atrophy on blue-light autofluorescence imaging (BAF). Using this technique, the size of the retinal atrophic area and the growth rate over time can be determined. Previous studies, which assessed the entire STGD1 population, have found highly variable atrophy progression rates.^21–24 Differences in atrophy growth rates between subgroups of patients with STGD1 may lead to the overestimation or underestimation of effect sizes in the sample size calculation and may potentially lead to underpowered clinical trials. Clinical trials for mutation-specific therapies, where included patient groups will potentially be more homogeneous owing to genetic inclusion criteria, will be served specifically by detailed analysis of growth rate of outer retinal atrophy in mutation specific subtypes of STGD1.

Therefore, this study aimed to assess the atrophy growth rate in patients with STGD1 carrying specific ABCA4 mutations (c.4539+2001G>A; c.5461-10T>C; c.5882G>A; and c.768G>T) that might be eligible for future molecular targeted therapies.

Methods

Patients and Subgroups

Patients were selected from the STGD1 database of the Radboud University Medical center in Nijmegen, the Netherlands, which comprises patients with a clinical diagnosis of STGD1 as assessed by a specialized ophthalmologist (C.H.). Only patients with genetically confirmed STGD1, that is, carrying two ABCA4 variants, were included in the study. Segregation analysis has not been performed in all cases; however, we considered patients with two ABCA4 variants and a clinical phenotype of STGD1 to have genetically confirmed STGD1. Subgroups were designed based on age of onset defined as early onset (≤10 years), intermediate onset (11–44 years), late onset (≥45) years⁶ or based on ABCA4 mutations eligible for future mutation specific therapy (c.4539+2001G>A; c.5461-10T>C; c.5882G>A; and c.768G>T). Onset of disease was defined as self-reported age of first symptoms or, in asymptomatic cases, age at diagnosis.²⁵ For the analysis of patients heterozygous or homozygous for the c.768G>T (p.(Leu257Valfs*17)) or the c.4539+2001G>A (p.[=Arg1514Leufs*36]) variant, the study population was enriched with patients with STGD1 from the genetic database of the national Belgian referral center for ophthalmic genetics at Ghent University Hospital, Ghent, Belgium. Approval for this study was waived by the Institutional Ethics Committee, CMO Radboud UMC (file number 2022-15718). The study was conducted in adherence to the provisions of the Declaration of Helsinki.

Clinical Data

Demographics and clinical data including age of onset and molecular diagnosis were collected from medical files. BAF imaging was acquired using the Spectralis (Heidelberg Engineering, Heidelberg, Germany) between 2004 and 2023 as a part of routine outpatient clinic visits. Because the first available imaging may not coincide with the actual first presentation, we defined first presentation as the first visit with available imaging within 5 years after onset of disease. We included all available consecutive imaging and clinical data.

Image Analysis

The area of atrophy was measured on BAF images using the semi-automatic grading tool ‘RegionFinder’ in Heidelberg (Heidelberg Engineering). The atrophy was determined to be either definitely decreased autofluorescence (DDAF) or questionably decreased autofluorescence (QDAF) depending on the darkness of the hypoautofluorescent area, respectively >90% (for DDAF) and 50% to 90% (for QDAF) of the darkness of the optic head.^26,27 The area of total decreased autofluorescence (TDAF) was defined as the sum of DDAF and QDAF. Measurements were performed on 30° or 55° scans. Initially, 30° scans were used. However, 55° scans were analyzed when atrophy extended beyond 30°. The BAF image was excluded from analysis if the atrophic area visible on the 55° scan extended beyond the borders of the scan, because the total extent of the atrophy could then not be determined.

In addition, qualitative parameters as visualized in Figure 1, including the presence of flecks, foveal sparing atrophy (fovea surrounded by ≥180° of atrophy²⁸), and homogeneity of the background autofluorescence were graded. The background was considered homogeneous when it showed an even distribution of autofluorescence. A heterogeneous background was defined as the presence of widespread small foci of increased or reduced autofluorescence.

Figure 1.

Two examples demonstrating the grading criteria. (a) A 55° image of the retina of a 55-year-old woman carrying c.768G>T and c.5603A>T. The image shows presence of well-demarcated DDAF atrophy with foveal sparing, and a heterogeneous background. (b) A 30° image of the retina of a 24-year-old man carrying c.768G>T and c.[2588G>C;5603A>T]. The image shows presence of QDAF atrophy with a hyperfluorescent ring, absence of flecks, and a homogeneous background.

Images of patients heterozygous for either the c.768G>T or c.4539+2001G>A variant were double graded by two independent graders in each cohort. The limit of agreement was calculated to assess the intergrader variability. Given the decent limit of agreement between the graders, specifically 2.98 for DDAF and 7.76 for TDAF in the Ghent cohort (176 graded images), and 8.13 for DDAF and 12.96 for TDAF in the Nijmegen cohort (498 graded images), we decided to single grade the remaining patients. The corresponding Bland–Altman plots are shown in Supplementary Figure S1. We conducted t tests to assess whether intergrader differences influenced the atrophy growth rate.

Statistical Analysis

Statistical analyses were performed using RStudio version ‘2023.9.1.494’ (Posit Software, PBC, Boston, MA, USA), SPSS version 27 (IBM, Armonk, NY, USA), and Microsoft Office Excel version 2016 (Microsoft, Seattle, WA, USA). For descriptive analyses, the median with interquartile range was calculated for non-normally distributed variables.

To determine the atrophic growth rate, the relation between time and atrophic area was analyzed with both untransformed data in square millimeters and after square root transformation. In accordance with prior literature,²² a model using square root transformation of the atrophic area (accounting for two-dimensional growth) demonstrated a better fit in linear analysis compared with untransformed data. Higher-order root transformations yielded poorer fits compared with square root transformations. Therefore, square root transformed area of DDAF and TDAF was used in the analysis.

Mixed model analysis was performed using “lmerTest”²⁹ and “lme4”³⁰ packages in RStudio. Models were created for the subgroups based on age of onset and specific ABCA4 mutations. Measurements were nested within individual eyes, and eyes were nested within patients. This approach was chosen to control for the intrasubject correlation of observations, acknowledging that patients contributed data from two eyes. The mean growth rate of the atrophic area (either DDAF or TDAF) and appropriate confidence intervals were determined for each subgroup. No direct statistical testing between the mutation-specific subgroups was performed owing to multicollinearity; for instance, patients carrying both the c.768G>T and c.5882G>A variant are present in two subgroups. To test the effect of the second variant, all second variants were divided in subgroups based on Cornelis et al. 2017.³¹ The severity of the second variant was assessed using mixed model analysis. Graphs were designed using the “ggplot2”³² package in R and using Prism 9 (GraphPad Software, Boston, MA, USA).

Results

A total of 202 patients (399 eyes, 1767 images) from the Radboud UMC STGD1 database met the inclusion criteria of having a genetically confirmed clinical diagnosis of STGD1 and at least two gradable BAF images in their medical records. The study population was further enriched with 19 patients (38 eyes, 164 images) from the Ghent University Hospital genetic database who were heterozygous for either the c.768G>T or c.4539+2001G>A variant and had at least two gradable BAF images in their medical records. A total of 123 patients had two or three BAF images available for analysis, 98 patients had four or more images. Subgroups were generated based both on age of onset and on ABCA4 mutations. Table 1 provides an overview of the cohort. The age of onset was available for 212 patients. A total of 17 patients were heterozygous for the c.4539+2001G>A variant, 41 patients carried the c.5461-10T>C variant, and 41 patients had the c.768G>T variant on one ABCA4 allele (¹Fig. 2). One hundred eyes (63 patients) did not show DDAF during the most recent BAF image. Four eyes (two patients with intermediate-onset disease) did not develop outer retinal atrophy identifiable as DDAF or QDAF, despite their fundus flavimaculatus phenotype.

Table 1.

Oversight of the Cohort (N = 21)

Patient Characteristics
% Female	55.7
Mean age of onset of symptoms (years)*	29.7
Age of onset ≤10 years	N = 36	29 Nijmegen; 7 Ghent
Age of onset 11–44 years	N = 119	110 Nijmegen; 9 Ghent
Age of onset ≥45 years	N = 57	54 Nijmegen; 3 Ghent
Genetic Subgroups		Age of Onset ≤10 Years	Age of Onset 11–44 Years	Age of Onset ≥45 Years
c.768G>T	N = 41	17%	56%	27%
c.4539+2001G>A	N = 17	29%	47%	24%
c.5461-10T>C	N = 41	22%	44%	32%
c.5882G>A	N = 36	11%	69%	11%

^*Age of onset unknown for nine patients.

Figure 2.

Visualization of the patients included in the cohort.

Atrophy Growth in the Entire Nijmegen STGD1 Population

Figure 3 shows the natural course of the square-root transformed atrophy size for all eyes. Every line in the graph corresponds with the atrophic region in one eye of one patient. Consequently, there are two lines per patient if data of both eyes are available. A wide variability of the atrophy growth can be observed, yet TDAF and DDAF growth rate look similar. Figures 3b and 3d depict the square root–transformed DDAF and TDAF plotted against the time since the first measurement. Using this timeframe, the average slope or atrophy growth rate can be calculated. The mean growth rate of atrophy was 0.1446 mm/year (95% CI, 0.1382–0.1510 mm/year) for DDAF and 0.1459 mm/year (95% CI, 0.1402–0.1515 mm/year) for TDAF in the entire Nijmegen STGD1 population (399 eyes of 202 patients). Atrophy progression was faster in cases where the 55° image was used owing to the atrophy exceeding the borders of the 30° image (DDAF 30, 0.134 mm/year vs. 55, 0.265 mm/year and TDAF 30, 0.139 mm/year vs. 55, 0.233 mm/year).

Figure 3.

Atrophy progression of individual eyes in the Nijmegen cohort. (a) and (c) show the square root transformed DDAF (a) and TDAF (c) growth plotted against the time since age of onset, (b) and (d) show the square root transformed atrophy growth plotted against the time since first measurement.

No significant differences in the atrophy growth rate were observed between graders, with P values: DDAF P = 0.505 and TDAF P = 0.707 for FB vs. JP, and DDAF P = 0.272 and TDAF P = 0.895 for CL vs. JP.

Atrophy Growth in Age of Onset Subgroups in the Nijmegen Cohort

The atrophy growth rate of DDAF was 0.1655 mm/year (95% CI, 0.1490–0.1819 mm/year) for the early-onset disease patients, 0.1269 mm/year (95% CI, 0.1188–0.1350 mm/year) for intermediate-onset disease patients, and 0.1782 mm/year (95% CI, 0.1656–0.1908 mm/year) for late-onset disease patients. The atrophy growth rate of TDAF was 0.1897 mm/year (95% CI, 0.1753–0.2042 mm/year for the early-onset disease patients, 0.1271 mm/year (95% CI, 0.1200–0.1343 mm/year) for intermediate-onset disease patients, and 0.1681 mm/year (95% CI, 0.1571–0.1792 mm/year) for late-onset disease patients.

The DDAF growth rate was comparable between the early-onset and late-onset subgroups (P = 0.228). However, when looking at TDAF, the atrophy growth rate was significantly higher (P = 0.0201) for the early-onset disease group compared the with late-onset disease group. The progression of atrophy was statistically significantly slower in intermediate-onset patients with STGD1 as compared with both patients with early-onset STGD1 (P < 0.001) and patients with late-onset STGD1 (P < 0.001) (Fig. 4). Supplementary Figure S2 shows the visualization of our mixed effects model. Supplementary Table 1 shows patient characteristics and qualitative parameters per age of onset subgroup in the Nijmegen cohort. Qualitative parameters revealed several differences between the age at onset groups. Patients with early-onset disease show a lower prevalence of flecks, yet background autofluorescence is also less homogeneous.

Figure 4.

Atrophy growth rate per age of onset subgroup for square-root transformed DDAF (a) and TDAF (b), with 95% confidence intervals. The intermediate-onset group has a slower atrophy growth rate than the early-onset and late-onset group, both in DDAF as TDAF.

Atrophy Growth in Mutation-Specific Subgroups in the Combined Nijmegen and Ghent Cohorts

The square root–transformed DDAF growth speed was fastest for c.5461-10T>C (0.1565 mm/year) and slowest for c.4539+2001G>A (0.0686 mm/year). The coefficient of determination (R2C) for the mixed effects model was 0.964. Similarly, the square root–transformed TDAF growth speed was fastest for c.5461-10T>C (0.1835 mm/year) and slowest for c.4539+2001G>A (0.0951 mm/year), with a coefficient of determination (R2C) of 0.965 for the model. An overview of all growth rates per mutations and the 95% confidence intervals can be found in Table 2.

Table 2.

Square-Root–Transformed TDAF and DDAF Growth Rate

	$\sqrt{\mathbf{DDAF}}$ (mm/Year)		$\sqrt{\mathbf{TDAF}}$ (mm/Year)
ABCA4 Mutation	Mean	95% CI	Mean	95% CI
c.768G>T	0.1299	0.1174–0.1423	0.1537	0.1422–0.1652
c.4539+2001G>A	0.0686	0.0539–0.0833	0.0951	0.0816–0.1086
c.5461-10T>C	0.1565	0.1448–0.1681	0.1835	0.1727–0.1943
c.5882G>A	0.0821	0.0676–0.0967	0.0953	0.0818–0.1088
other	0.1440	0.1353–0.1527	0.1363	0.1283–0.1443

From Figure 5, it can be derived that the mixed model analysis of the growth rate of DDAF and TDAF shows a clear separation of subgroups, with in the fast progressors group the c.768G>T, c.5461-10T>C, and “other” group (Nijmegen patients with other mutations in ABCA4 than those specifically mentioned), whereas c.5882G>A and c.4539+2001G>A tend to progress slower. The mixed effect model is visualized in Supplementary Figure S3.

Figure 5.

Atrophy growth rate per mutation subgroup for square-root transformed DDAF (a) and TDAF (b), with 95% confidence intervals.

Supplementary Table 2 shows that the atrophic area size also differs between the mutation subgroups. Patients carrying either c.5882G>A or those heterozygous for c.4539+2001G>A have a lower outer retinal atrophic surface area both at first presentation and at the last visit, although the disease duration is similar between subgroups. An analysis of the qualitative parameters shows that patients with c.5882G>A all have a homogeneous background at first presentation, with only one patient progressing to a heterogeneous background during the registered course of the disease, whereas in the other mutation subgroups this percentage is higher.

The effect of the second variant on the mutation-specific atrophy growth is shown in Supplementary Table 3. Mild second mutations had a small inhibiting effect on the atrophy growth rate in patients with c.768G>T and c.5461-10T>C, whereas severe mutation led to a faster disease progression in these patients. One and two homozygous cases were included in the c.768G>T and c.5461-10T>C groups, respectively. All cases had early-onset STGD1 (age of onset 6–8 years). The mean growth rates of the homozygous c.768G>T patient were 0.323 mm/year for DDAF and 0.395 mm/year for TDAF and homozygous c.5461-10T>C patients were 0.069 mm/year for DDAF and 0.070 mm/year for TDAF. Supplementary Figure S4 includes BAF images of a patient homozygous for c.768G>T and a patient homozygous for c.5461-10T>C.

Discussion

This retrospective natural history study demonstrated different atrophy progression rates between subgroups of patients with STGD1 based on age of onset and ABCA4 mutations. The data represent crucial information to consider when designing a clinical trial for a mutation-specific therapy, where the estimated disease progression used in the sample size calculations must reflect the true disease progression of the mutation-specific subgroup included in the trial.

With upcoming mutation-specific therapies, this study was designed to estimate disease progression with the presence of an allele that underlies a form of potentially treatable STGD1. Although we do understand that the natural course of disease will also depend on the second variant, this was not the scope of this research, because the second variant will also be variable in clinical trials for mutation-specific therapies.

Progression of Atrophy in STGD1

In the Nijmegen cohort, the mean growth rate of atrophy was 0.1446 mm/year (95% confidence interval, 0.1382–0.1510 mm/year) for DDAF. However, within the different mutation-based and age-of-onset subgroups, the mean growth rate varies between 0.0686-0.1782 mm/year. Several other studies have investigated atrophy growth rate in cohorts of patients with STGD1. In a large meta-analysis 1055 eyes of 689 patients, Bassil et al.³³ showed a mean square root–transformed RPE atrophy growth rate of 0.20 mm/year. In addition, Shen et al.²² conducted a meta-analysis in 2019, pooling data of 564 eyes from 7 studies. They reported a radius root–transformed atrophy growth rate of 0.104 mm/year. To facilitate comparisons, this value can be converted to square root–transformed growth rate by multiplying it by $\sqrt{\pi }$, resulting in an atrophy growth rate of 0.1843 mm/year. Both these meta-analyses showed atrophy growth rates slightly higher than our mean growth rate.

Sajovic et al.³⁴ 2023 reported a median DDAF growth rate of 0.354 mm²/year. However, direct comparison with our findings is challenging owing to the absence of square-root transformation in their analysis. Lambertus et al.⁴ calculated the square root RPE atrophy growth rate in late-onset patients with STGD1 as 0.22 mm/year, slightly exceeding the rate found in our study (0.1820 mm/year). Allegedly, a cohort consisting solely of late-onset patients shows faster progression as compared with other cohorts.

Heath Jeffery et al.³⁵ in 2021 looked into genotype-specific growth rates of DDAF, investigating the growth rate of five specific ABCA4 mutations using methods comparable to ours. They looked into the growth rate of c.5882G>A in 11 patients with STGD1 and found a mutation-specific growth rate of 0.106 mm/year, slightly higher compared with our growth rate of 0.0821 mm/year.³⁵ The difference between the growth rates in the two cohorts may be due to different ABCA4 mutations in trans with c.5882G>A. Yet, it seems atrophy grows more slowly if patients harbor the c.5882G>A variant, preserving the peripheral retina.

Clinical Trial Design

When designing a clinical trial to evaluate a therapy that potentially decreases disease progression with outer retinal atrophy expressed as growth rate of DDAF or TDAF as primary end point, the estimated effect size and consequent power calculation will be based on the specific atrophy growth rate of the target population.

When creating subgroups based on age of onset, we found the fastest atrophy growth rates in patients with early-onset and late-onset STGD1. With regard to trial design using atrophy as the primary end point, at least initially, we prefer the inclusion of patients with late-onset STGD1 over the inclusion of patients with early-onset STGD1 for two main reasons; first because of the ethical concerns about including children early in clinical trials for a condition that also affects adults. Lessons initially learned in an adult patient population can subsequently be tested in a pediatric cohort. Second, our study indicated that patients with late-onset STGD1 have relatively more DDAF compared with TDAF, as reflected by a larger DDAF/TDAF ratio than early-onset patients (0.59 vs. 0.89). Well-demarcated atrophy, specifically DDAF, is more reliable measured than QDAF/TDAF. Atrophy was well-demarcated in approximately 56% of the late-onset eyes as compared with approximately 7% of the early-onset eyes on the last visit, which implies that the inclusion of patients with late-onset STGD1 will lead to more accurate measurements. However, this does not mean that patients with early-onset STGD1 cannot be included in clinical trials; rather, it suggests that atrophy measured on BAF images may be a suboptimal end point for this group and that alternative end points should be considered. Surprisingly, the atrophy growth rate of patients with intermediate-onset STGD1 is slower than that of patients with both early-onset and late-onset disease. This difference may be due to sampling bias, excluding the patients with intermediate-onset STGD1 and larger lesions (which have faster atrophy growth), which have expanded beyond the 30° or 55° field of the BAF image. However, it might also be due to the high prevalence of the bull's eye maculopathy form of STGD1 in this group, characterized by a hyperfluorescent ring in approximately 45% of cases at first presentation. Moreover, the c.5882G>A p.(Gly1961Glu) variant was found in 25 of 119 patients (21%) with intermediate-onset STGD1. This variant was associated previously with bull's eye maculopathy and is considered to cause a milder disease expression.^36,37 Hence, a significant portion of the variability of disease progression may be explained by the impact of mutations on ABCA4 function.

In the analysis of mutation-specific subgroups, patients harboring either c.5461-10T>C or c.768G>T mutations exhibited the fastest atrophy growth rates, consistent with their known severe impact on mRNA splicing and ABCA4 protein function.^38,39 Interestingly, the homozygous case with c.768G>T exhibited a much faster disease progression compared with the heterozygous cases with the same mutation, whereas the homozygous cases with c.5461-10T>C showed slower disease progression. Because only three homozygous cases were analyzed, which aligns with the expected number based on prevalence,⁴⁰ it is difficult to determine whether these findings are representative of the mutations. Further research on homozygous cases is needed to confirm these observations.

Patients carrying the c.5882G>A variant or the c.4539+2001G>A variant showed slower disease progression compared with patients carrying other mutations. Although this finding aligns with those in previous reports on c.5882G>A, it seems to contradict earlier reports on c.4539+2001G>A, which was previously linked to the cone–rod dystrophy form of STGD1, with the assumption that the additional rod involvement on ERG is a marker for faster disease progression.⁴¹ However, given the absence of imaging and ERG analyses in many of those studies, a direct comparison remains challenging.

The mean atrophy growth rate in a cohort is most likely dependent on the composition of different mutations in that cohort. Regarding our data, cohorts with a significant number of patients with either the c.5882G>A or the c.4539+2001G>A pathogenic variants will most likely have a lower mean progression speed than cohorts without these mutations. Analysis of the effect of the second mutation of the atrophy growth rate showed no large effect of the second mutation on the atrophy growth rate, suggesting that the differences in atrophy growth among mutation-specific subgroups may primarily result from the first mutation. However, it should be noted that segregation analysis was not conducted in this study, which prevents confirmation that the two identified variants are located on different alleles. In addition, the presence of an undetected third variant cannot be excluded, potentially altering the severity classification of the second variant. Future analyses should ideally focus on confirmed biallelic cases to improve the accuracy and generalizability of the findings.

When, for some reason, patients with mutations associated with slower atrophy growth rate are over-represented in the control group of a clinical trial, the control group will show a lower rate of disease progression and the therapeutic effect in the treatment group will be underestimated. This factor would lead to a preventable failure of the clinical trial. To avoid such failure of a clinical trial, it may be required to measure a patients atrophy growth rate before a clinical trial using natural history data, excluding patients with a slow atrophy growth rate from clinical trials that have the growth rate of atrophy as primary end point.

Limitations

Given the retrospective nature of the study, there was considerable heterogeneity of available data. Age of onset was defined as the age when the first symptoms were reported, although it is likely that structural changes preceded the manifestation of symptoms.

It has been reported previously that 55° images may overestimate the atrophy growth rate.⁴² In our cohort, patients for whom 55° images were used showed a higher growth rate compared with those assessed exclusively with 30° images. However, we believe this difference is attributable primarily to our methodology. Specifically, 55° images were used only when the atrophy extended beyond the borders of the 30° image, meaning that measurements in this group were taken predominantly from patients with larger atrophic areas and, consequently, a likely faster progression of atrophy. Another contributing factor to the difference in growth rates between 55° and 30° images may be the fact that no conversion factor was applied to account for the difference in image scale.

The exclusion of patients with extensive atrophy that exceeded the borders of the 30° or 55° BAF image may have induced a selection bias because more severe cases of STGD1 may have been omitted inadvertently. Such patients are also often lost to follow-up because of seemingly stable poor visual acuity with no available treatment options. However, upon examination with the relatively new technique of ultrawide-field imaging, the area of atrophy extends beyond the vascular arch, involving the peripheral retina, and the size of the atrophic area can be assessed (Fig. 6). The use of ultrawide-field images may play a crucial role in future clinical trials for assessing atrophy growth in patients with STGD1, as it allows for evaluation beyond the limitations imposed by the borders of 30° or 55° images.⁴³ It is important to recognize that STGD1 disease is not confined to the macula but continuously progresses toward the periphery over time.

Figure 6.

Widefield BAF and ultrawide field green laser AF of an end-stage STGD1 patient (age 70) carrying c.768G>T and c.5714+5G>A, with onset of symptoms at age 24. The circles on the ultrawide field images corresponds with the 55° field imaged on the regular BAF image. Although the 55° BAF image does not capture the border of the atrophic regions, ultrawide field green laser image reveals atrophy extending far beyond the vascular arches.

Another limitation is that this study only investigated DDAF and TDAF as potential end points for clinical trials. There are other structural and functional measures to determine disease progression in STGD1, - including OCT volume,^44,45 fleck size and numbers,⁴⁶ and microperimetry,⁴⁷ all of which can be used to identify progression even before hypofluorescence appears.

Conclusions and Future Directions

In this study, we found that atrophy growth rates vary among subgroups of patients with STGD1, depending on both age of onset and specific mutations. This underlines that awareness of the influence of specific mutations on progression of disease is important. The findings in this study support the need for natural history studies for specific ABCA4 mutations, before designing a clinical trial for a mutation-specific intervention.